mind wandering prospective memory

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here .

Loading metrics

Open Access

Peer-reviewed

Research Article

Thought probes during prospective memory encoding: Evidence for perfunctory processes

Roles Conceptualization, Data curation, Methodology, Project administration, Software, Supervision, Validation, Visualization, Writing – original draft, Writing – review & editing

* E-mail: [email protected]

Affiliation Baylor University, Department of Psychology & Neuroscience, Waco, TX, United States of America

Roles Conceptualization, Resources, Writing – review & editing

Affiliation Washington University in St. Louis, Department of Psychological and Brain Sciences, St. Louis, MO, United States of America

Roles Data curation, Formal analysis, Investigation, Software, Writing – review & editing

Roles Data curation, Investigation, Writing – review & editing

• Michael K. Scullin, 
• Mark A. McDaniel, 
• Michelle N. Dasse, 
• Ji hae Lee, 
• Courtney A. Kurinec, 
• Claudina Tami, 
• Madison L. Krueger

• Published: June 6, 2018
• https://doi.org/10.1371/journal.pone.0198646
• Reader Comments

For nearly 50 years, psychologists have studied prospective memory, or the ability to execute delayed intentions. Yet, there remains a gap in understanding as to whether initial encoding of the intention must be elaborative and strategic, or whether some components of successful encoding can occur in a perfunctory, transient manner. In eight studies (N = 680), we instructed participants to remember to press the Q key if they saw words representing fruits (cue) during an ongoing lexical decision task. They then typed what they were thinking and responded whether they encoded fruits as a general category, as specific exemplars, or hardly thought about it at all. Consistent with the perfunctory view, participants often reported mind wandering (42.9%) and hardly thinking about the prospective memory task (22.5%). Even though participants were given a general category cue, many participants generated specific category exemplars (34.5%). Bayesian analyses of encoding durations indicated that specific exemplars came to mind in a perfunctory manner rather than via strategic, elaborative mechanisms. Few participants correctly guessed the research hypotheses and changing from fruit category cues to initial-letter cues eliminated reports of specific exemplar generation, thereby arguing against demand characteristics in the thought probe procedure. In a final experiment, encoding duration was unrelated to prospective memory performance; however, specific-exemplar encoders outperformed general-category encoders with no ongoing task monitoring costs. Our findings reveal substantial variability in intention encoding, and demonstrate that some components of prospective memory encoding can be done “in passing.”

Citation: Scullin MK, McDaniel MA, Dasse MN, Lee Jh, Kurinec CA, Tami C, et al. (2018) Thought probes during prospective memory encoding: Evidence for perfunctory processes. PLoS ONE 13(6): e0198646. https://doi.org/10.1371/journal.pone.0198646

Editor: Sam Gilbert, University College London, UNITED KINGDOM

Received: April 19, 2018; Accepted: May 22, 2018; Published: June 6, 2018

Copyright: © 2018 Scullin et al. This is an open access article distributed under the terms of the Creative Commons Attribution License , which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Data Availability: All study data are available from the Open Science Framework database ( osf.io/63a7f ).

Funding: This work was supported by the National Institutes of Health R21-AG053161 supported MKS ( https://www.nih.gov/ ). Publication was made possible, in part, by support from the Open Access Fund sponsored by the Baylor University Libraries. The funder had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. There was no additional external funding received for this study.

Competing interests: The authors have declared that no competing interests exist.

Introduction

Prospective memory is an umbrella term that refers to remembering to execute goals, intentions, and chores in the future [ 1 , 2 ]. A prototypical prospective memory task is remembering to pick up milk at the grocery store, or, remembering to go to the grocery store at all. However, prospective memory encompasses a broader array of relationship-oriented tasks (e.g., returning a friend’s text message), household chores (e.g., take out the trash), health-oriented intentions (e.g., adhering to medication schedules), society-oriented goals (e.g., identifying missing or wanted persons), and workplace tasks and routines [ 3 – 5 ]. The goal of the present work was to advance understanding of how intentions are encoded.

Encoding processes in prospective memory

An intuitive view of prospective memory encoding is that intention formation is deliberate, elaborative, and strategic. Consider, for example, the Theory of Planned Behavior, which states that intention formation is the “ conscious plan or decision to exert effort to enact the behavior” (p. 1430 [ 6 ]; italics added). The more individuals draw upon working memory resources at encoding, the more likely they are to successfully complete their planned intention [ 7 – 10 ]. Furthermore, when studying word lists for later recognition or recall (“retrospective memory”), devoting greater working memory resources toward elaborative or organizational processing increases the probability of those items being retained [ 11 – 13 ]. Therefore, according to one view, the successful encoding of prospective memories will require strategic, controlled processes to elaborate on the intention (e.g., generating many retrieval cues). For convenience, we label this general position as the strategic/elaborative encoding view.

On the other hand, some information might be encoded quickly and with minimal cognitive effort, such as the associations amongst studied items [ 14 – 16 ]. According to this literature, it is plausible that some aspects of prospective memory encoding may be accomplished “in passing.” Anecdotally, one might remember to purchase several specific ingredients for a chicken curry dinner when only consciously encoding “curry dinner” as a general category (this specific example assumes the absence of strategic retrieval mode processes when arriving at the grocery store). Some researchers argue that prospective memory encoding can even be implicit, such as when one remembers to turn on their phone after a colloquium (after politely turning it off to listen), or when one remembers to resume drafting an e-mail after being interrupted by a phone call [ 17 , 18 ]. This general orientation anticipates that some components of prospective memory encoding may be cursory, transient, implicit, or otherwise engage minimal working memory resources. We label this position as the perfunctory/transient encoding view.

Encoding manipulations in prospective memory experiments

Some prospective memory research favors the strategic/elaborative encoding view [ 19 , 20 ]. When participants use an encoding strategy, they tend to generate more retrieval cues and perform better on tests of prospective memory [ 21 – 24 ]. In addition, neuroimaging studies suggest that greater activation during encoding (e.g., in motor regions) may predict better later retrieval [ 25 – 27 ]. Furthermore, when young and older adults encode complex prospective memory tasks, the older adults tend to show deficits in plan formation, possibly due to an age-related deficit in working memory resources [ 28 ].

However, not all studies have observed age differences in prospective memory planning [ 29 ] or that greater neural activation during encoding predicts later retrieval [ 30 ]. Strategic planning often diverges from prospective memory execution [ 31 ], and less elaborative planning can sometimes lead to better prospective memory [ 28 ]. Some intentions may even be implicitly formed, such as the intention to later put a wristwatch back on after being told to put it away; in observing that many participants could complete this implicit wristwatch task, Kvavilashvili and colleagues [ 18 ] concluded that “the conscious formation of intention may not always be necessary for successful remembering as stipulated in the prospective memory literature” (p. 873). To be clear, most prospective memory laboratory paradigms encourage, if not require, that the intention is consciously encoded. Whether some components of prospective memory encoding can still be perfunctory, even in a controlled laboratory environment, remains under-studied.

Overview of the current work

Across eight experiments, we used thought probes to gauge the processes operating during intention formation. There are many laboratory procedures for studying prospective memory, but the most common approach is the Einstein-McDaniel paradigm [ 32 ]. As shown in Fig 1 , participants practiced an ongoing task (lexical decision) and then were instructed to remember to press a specific key ( Q ) in response to a target stimulus (e.g., animal words). Immediately after encoding, participants reported what was currently on their mind and responded to questions targeted at identifying encoding processes. The encoding thought probe approach complements previous work that used thought probes during retrieval [ 33 – 35 ] as well as studies that inferred encoding processes from verbal plan descriptions, neuroimaging outcomes, later retrieval/performance, and simulations [ 21 , 25 , 28 , 36 ]. Given the number of experiments included, we summarize the research questions and results in Table 1 and Fig 2 . In overview, Experiments 1–7 were designed to address basic science questions about the processes operating at encoding. Experiment 8 was designed to test the consequences of these encoding processes for prospective memory retrieval.

• PPT PowerPoint slide
• PNG larger image
• TIFF original image

In Experiments 1–2, the target category was animals. This figure was adapted with permission [ 24 ].

https://doi.org/10.1371/journal.pone.0198646.g001

The figures depict the aggregate (A) free response data, (B) generation of specific exemplars, and (C) bias toward different encoding strategies.

https://doi.org/10.1371/journal.pone.0198646.g002

The reader is directed to the methods and results section of each study for research details and inferential statistics.

https://doi.org/10.1371/journal.pone.0198646.t001

Experiments 1–3

We investigated encoding processes by using categorical cues (animals, fruits [ 37 ]). One view is that participants will encode the prospective memory task exactly as the experimenter instructs them to: as a general, superordinate category [ 38 ]. An alternative view is that participants will generate specific category exemplars, such as apple [ 39 ]. If participants generate specific exemplars, then the critical theoretical question is whether they do so in a strategic/elaborative manner (as in category fluency neuropsychological tests [ 40 ]), or whether they generate exemplars “in passing” (e.g., via spreading activation in semantic networks [ 41 , 42 ]). To test whether we could bolster the exemplar-generation process, some participants were shown a prime word (e.g., apple) during a practice block.

Participants.

Washington University undergraduate students ( N = 68 in Experiment 1 and N = 61 in Experiment 2) and Baylor University undergraduate students ( N = 68 in Experiment 3) participated for partial class credit in the present protocol as well as an unrelated protocol on juror decision making. The unrelated protocol contained no animal or fruit stimuli and participants were told that they would perform a series of cognitive tasks (i.e., all procedures were described in one informed consent). Nevertheless, we ensured the generality of our findings in Experiments 4, 5, 6, and 8 by conducting the prospective memory procedures without an unrelated protocol. Table 1 foreshadows that the critical findings on perfunctory/transient processes replicated. Note that, in Experiment 2, one participant was excluded for inadvertently being run using an incorrect program (N = 60).

All experiments presented in this manuscript were approved by the local IRB (Baylor University, Washington University) and all participants provided written consent prior to participating. E-Prime 2.0 files and data are available at Open Science Framework (osf.io/63a7f).

As shown in Fig 1 , and following previous research [ 24 ], participants first learned the lexical decision task instructions (referred to as the word/nonword task) to respond as quickly and accurately as possible whether a string of letters formed a word or not (by pressing keys marked “Y” and “N” on the number pad). Then they practiced the lexical decision task for 10 trials, during which they received speed and accuracy feedback following each trial. The prime word fish was presented during the practice block in Experiment 1, but not in Experiment 2. In Experiment 3, we randomly assigned participants to prime and no-prime conditions that differed in whether the word apple was presented during the practice block (cf. [ 43 ]).

Participants were next given the following prospective memory task instructions (modifications for Experiment 3 are provided in brackets):

“In this experiment, we are also interested in your ability to remember to perform an action at a given point in the future. Therefore, during the word/nonword task, we would like you to perform a special action whenever you see a word that belongs to the category ANIMAL [FRUITS]. Whenever you see an animal [a fruit] word, you should remember to press the 'Q' key. Press Q to continue.”

On the next screen, participants typed whatever was on their mind at that moment, and then asked two yes/no questions about encoding specific examples of animals (fruits) versus keeping animals (fruits) in mind as a general, overarching category (order counterbalanced). They were further asked whether they were more focused on encoding specific examples, the general category, or if they hardly thought about this task at all (list order counterbalanced for specific/general options). Lastly, if participants previously indicated that they generated specific examples, they were asked to type which examples they thought of when they encoded the prospective memory task (and to avoid typing any new examples they just thought of). We used this thought probe procedure in every experiment, with the exception that in Experiment 1 participants were only asked to type what was on their mind, whether they thought of any specific animal words, and (if so) which animal words they encoded.

Statistical analysis.

To classify the free responses, three members of the research team independently rated the responses as “on-task,” “off-task,” or “both on and off task” [ 24 ]. They next rated the “on-task” responses according to whether they mentioned the target cue type, the ongoing task (contextual processing [ 44 ]), and the response key (motor planning [ 45 ]). The three raters were masked to experimental conditions and met to resolve any disagreements. In every experiment, ≥98% of the responses were reconcilable after discussion, and the remaining responses were listed as “unclassifiable.”

We conducted chi-square tests to determine whether there were significant differences in the distribution of encoding responses. Where a cell value was <5, we used Yates’ [ 46 ] correction. We also tested whether order counterbalance affected responses to the yes/no or encoding bias questions. In Experiment 3, we used t -tests to determine whether encoding durations (reading time on the encoding instructions screen) were associated with encoding thought probe responses (encoding duration data were not recorded by e-prime in the first two experiments).

On-mind free responses.

The free response data are presented in Table 2 and aggregated across all experiments in Fig 2 . We predicted that because the prospective memory procedure was brief and the encoding instructions are one of the most critical elements in prospective memory studies, that nearly all free responses would include on-task, experiment-relevant content. This prediction was clearly disfavored as there was a similar frequency of solely on-task and solely off-task responses (Experiment 1: χ 2 < 1; Experiment 2: χ 2 (1) = 1.21, p = .27; Experiment 3: χ 2 < 1). Participants’ thoughts often focused on food (“biscuits”), sleep (“I’m sleepy”), class (“I have an exam tomorrow”), relationships (“ex-boyfriend problems”), and current events (“world series win”). Table 2 further demonstrates that most on-task comments focused on the prospective memory cue type, with fewer encoding processes related to motor planning and very few to contextual processing.

On-task responses were further classified as mentioning the ongoing task (context), prospective memory response key, or cue words. The on-task specification numbers will not sum to 100% due to some participants providing only miscellaneous responses (e.g., “this experiment”) and others listing multiple components (e.g., response key and cue words).

https://doi.org/10.1371/journal.pone.0198646.t002

Yes/no question responses.

We next investigated the quality of encoding as the proportion of participants responding affirmative to general category encoding and specific exemplar encoding. The data are included in Table 3 and illustrated collapsed across all experiments in Fig 2 . Contrary to the view that participants never generate specific exemplars during categorical cue tasks, a significant proportion of participants reported to generating specific exemplars of animals/fruits at encoding in Experiment 1 (χ 2 (1) = 22.53, p < .001, Yates’ correction), Experiment 2 (χ 2 (1) = 20.26, p < .001, Yates’ correction), and Experiment 3 (χ 2 (1) = 37.60, p < .001, Yates’ correction). When participants were forced to choose whether they focused more on general category encoding or specific exemplar encoding, participants indicated a general category bias in Experiment 2 (χ 2 (1) = 16.81, p < .001), but not in Experiment 3 (χ 2 (1) = 2.69, p = .10).

https://doi.org/10.1371/journal.pone.0198646.t003

Frequency of cue words generated (priming effects).

When fish was a prime during a practice block (Experiment 1), it was the most commonly mentioned cue word ( n = 10); when it was not primed (Experiment 2), no participants reported encoding fish , χ 2 (1) = 7.64, p = .006 (Yates’ correction). In Experiment 3, apple was the most frequently generated fruit word in both the no-prime condition ( n = 9) and the prime condition ( n = 15), χ 2 (1) = 2.90, p = .09. Perhaps the magnitude of the priming effect depends on how typical the exemplar is to the encoded category (e.g., fish is a less typical exemplar of animals than apple is of fruits [ 47 ]).

Encoding duration.

If specific exemplar generation is the result of a strategic/elaborative encoding process, then encoding durations should be greater for individuals who reported having generated specific exemplars [ 48 ]. By contrast, Table 4 shows that there was no association between encoding duration and the likelihood of generating a specific exemplar, even when selecting only individuals who were not mind wandering (r(35) = -.17, p = .32). Fig 3 builds on this encoding duration null finding by presenting Bayesian prior and posterior distributions for effect size δ for Experiments 3–8. Collapsed across all studies, there was substantial evidence in favor of the null hypothesis that encoding duration was unrelated to specific exemplar generation (BF 10 = 0.21). Thus, exemplar generation seems perfunctory, perhaps the result of automatic, spreading activation processes [ 42 , 49 , 50 ].

The figure displays the prior and posterior distributions for effect size δ as a function of generation of specific exemplars. The sample size was limited to young adults in categorical prospective memory conditions. The BF 01 and BF 10 values from the Bayesian t-test both showed substantial evidence for the null hypothesis that encoding duration was similar for individuals who generated specific exemplars (n = 136; M = 21.62 sec, SD = 7.93) as those who did not generate exemplars (n = 212, M = 22.86 sec, SD = 11.76). The figures were produced using JASP software [ 56 ].

https://doi.org/10.1371/journal.pone.0198646.g003

Encoding duration data were not collected in Experiments 1–2. Positive correlations indicate that longer encoding durations were associated with more specific exemplar generation and more mind wandering.

https://doi.org/10.1371/journal.pone.0198646.t004

One view of prospective memory encoding emphasizes strategic/elaborative processes; however, at least 20% of participants reported that they hardly thought about the task at all. An even higher percentage of participants showed “off task” thoughts (mind wandering) immediately following the prospective memory instructions, even though the key to advance from the encoding screen to the thought probe screen was Q (i.e., the prospective memory response key). Therefore, in laboratory experiments, the encoding of prospective memories is conscious, but very short lived (transient). Interestingly, the participants who generated specific exemplars into their intention plan did not require additional time to do so (i.e., encoding duration), again indicating that some components of prospective memory encoding can be quick and cursory (perfunctory).

Experiment 4

We next tested for age effects on intention encoding processes. If encoding is strategic/elaborative, or otherwise cognitively-demanding, then older adults should show more frequent mind wandering [ 51 ] and generate fewer specific exemplars (as in category fluency tests; 40]. Alternatively, if intention encoding can be perfunctory, then there should be no age differences in intention encoding [ 18 ].

We recruited 128 adults who were living in the United States via Amazon’s Mechanical Turk (MTurk). Studies that compared data collection in the laboratory versus MTurk supported the validity of internet-based data collection [ 52 ]. Multiple prospective memory studies have been performed online [ 53 , 54 ]. Nevertheless, we restricted participation to MTurk workers with a 95%-100% approval rating, which increases data quality [ 55 ]. We excluded 13 participants whose ages diverged from the range specified during study advertisement for young adults (ages 18–30, 25.86 ± 2.65) and older adults (ages ≥60, 64.50 ± 4.97). Though the older adults reported to taking significantly more prescription medications ( M Older = 1.78 ± 2.13) than the young adults ( M Younger = 0.30 ± 1.06), t (86.48) = 4.74, p < .001 (corrected for unequal variances), the age groups were similar in years of education ( M Younger = 14.69 ± 2.18, M Older = 15.02 ± 2.98, t < 1), percentage of female participants (Younger: 32.73%; Older: 43.33%, χ 2 (1) = 1.37, p = .24), percentage of non-white participants (Younger: 23.64%; Older: 18.33%, χ 2 < 1), ratings of their health on a 1–5 scale ( M Younger = 3.60 ± 1.06, M Older = 3.57 ± 1.03, t < 1), reported number of hours slept the previous night ( M Younger = 7.27 ± 1.00, M Older = 7.03 ± 1.09, t (113) = 1.22, p = .23), and reported exercise frequency on a 1–4 scale ( M Younger = 2.73 ± 0.80, M Older = 2.45 ± 0.83, t (113) = 1.81, p = .07). Therefore, the older adults in the current study were generally very healthy.

All procedures mirrored Experiment 3’s no-prime condition except that participants completed questionnaires after the encoding thought probe procedure. The statistical analyses mirrored Experiments 1–3, with the addition of Bayesian analyses to statistically support the null hypothesis of no age effects. BF 10 < 1 is evidence in favor of the null hypothesis (i.e., no age differences in encoding) whereas BF 10 > 3 is substantial evidence for the alternative hypothesis (i.e., age differences in encoding). We conducted Bayesian analyses using JASP software [ 56 ].

As shown in Tables 2 and 3 , there were no significant differences between young and healthy older adults in specific exemplar generation (BF 10 = 0.32), off-task mind wandering (BF 10 = 0.42; less mind wandering overall in this MTurk sample), or any other aspect of prospective memory encoding (all χ 2 s < 2, p s > .10). The healthy older adult group (1.05 ± 2.08) generated nominally, but not significantly, more specific exemplars than the young adult group (0.62 ± 1.15; t (113) = 1.36, p = .18, d = .26, BF 10 = 0.46). Evidence in favor of the null was particularly strong when, based on the semantic fluency literature [ 40 ], the tested hypothesis was set to young adults being expected to generate more exemplars, BF 10 = 0.09. Table 4 shows that there were no significant associations between encoding duration and likelihood of generating specific exemplars in young or healthy older adults (see Fig 3 for encoding data across experiments). Therefore, the results of Experiment 4 suggested that prospective memory encoding need not always be cognitively demanding, but may instead be perfunctory/transient.

Experiment 5

One potential concern is that task demand characteristics cause participants to later say that they generated specific fruit words. For example, if participants believe the research hypothesis to be about specific exemplar encoding, then that would bias the results rather than indicate that some components of encoding can be perfunctory/transient. To investigate this demand-characteristic-view, we administered an established quantitative measure of demand characteristics [ 57 ] following the encoding thought probe procedure.

Adult participants (N = 59, ages 26.56 ± 3.61) living in the United States were recruited via MTurk according to the specifications described in Experiment 4. The procedure was identical to Experiment 4, with the addition of the Perceived Awareness of the Research Hypothesis scale (PARH [ 57 ]). The PARH requires participants to rate four statements on a 7-point scale (1 = Strongly Disagree, 7 = Strongly Agree), such as “I had a good idea about what the hypotheses were in this research.” If the mean score is below 4, then that indicates that participants were unclear about the hypotheses and that demand characteristics do not explain the study findings [ 57 ]. Following the rating scale, we also asked participants to free respond to the question “What do you think the researchers were trying to demonstrate with this study?”

In the free responses, a few participants showed partial knowledge of the hypotheses on encoding (e.g., “I honestly have no idea. Maybe trying to see if I thought of fruits as a general topic or more specifically? I really have no idea”). However, the most common response (23 of 55 provided responses) was a variant of “I honestly have no idea.” Importantly, PARH scores (2.70 ± 1.57) were significantly below the cutoff value of 4.0, t (58) = 6.34, p < .001, d = 1.66, indicating minimal demand characteristics. Individuals who reported generating specific exemplars (3.05 ± 1.15) showed similar PARH scores as individuals who did not (2.48 ± 1.77; t (56.97) = 1.52, p = .14, d = .39, Yates’ correction). There were outlier data points for encoding duration (<3 or >100 seconds), but regardless of whether these data points were excluded, encoding duration did not significantly differ across specific exemplar generators or non-generators (see Table 4 and Fig 3 ). Furthermore, there was no association between encoding duration and specific exemplar generation when only examining participants who were not mind wandering ( r (22) = -.14, p = .51). Thus, demand characteristics do not explain participants’ perfunctory/transient encoding of prospective memory intentions.

Experiment 6

In all preceding experiments we have assumed that the prospective memory intention was consciously encoded prior to assessing perfunctory/transient processes (cf. [ 18 ]). In Experiment 6, we experimentally confirmed conscious encoding by having participants verbally explain the prospective memory instructions to the experimenter. The idea here is that the verbal experimenter-check provides a strong test of the robustness of perfunctory/transient processes.

Sixty-two Baylor University undergraduate students participated in a cognitive laboratory setting. The procedure was identical to the no-prime condition in Experiments 3–5 except that participants were required to verbally explain the prospective memory task to the research assistant prior to completing the thought probe questions. Verbal explanation was not considered complete until participants had spoken the prospective memory cue (fruits) and response key (Q). Afterward, the experimenter advanced the screen so that participants could respond to the thought probe questions. Research assistants were masked to the study’s hypotheses.

Despite requiring participants to verbalize their general intention, Table 2 shows that mind wandering reports remained prevalent, demonstrating the transient nature of encoding processes. Furthermore, even though participants spent longer encoding their intention, including speaking their intention to the experimenter, specific exemplar generation occurred at similar rates as previous experiments (and was unrelated to encoding duration, even when off-task participants were excluded, r (27) = -.17, p = .39; see also Table 4 and Fig 3 ). These findings converge with the notion that specific exemplar encoding is more perfunctory than strategic.

Experiment 7

Better understanding of encoding processes will inform theoretical and methodological issues within the prospective memory field. According to the Multiprocess Framework [ 58 , 59 ], the overlap between how a target cue is encoded and how it is processed at retrieval determines the extent to which one must rely on strategic monitoring versus spontaneous retrieval processes ( cue focality hypothesis [ 60 ]). A typical example of a focal cue would be the target word “horse” during a task that requires processing of whole words (lexical decision task) whereas an example of a nonfocal cue would be detecting words that begin with the letter “h” during a lexical decision task. Fruit and animal category cues have nearly always been classified as nonfocal to ongoing tasks in review papers [ 61 ] and in meta-analysis articles [ 62 ]. However, in Experiments 1–6, many participants reported generating specific exemplars, which could transform a categorical intention from being a nonfocal cue into a focal cue. Therefore, it is pertinent to prospective memory theories to assess whether other cue types typically classified as “nonfocal” (i.e., during a lexical decision task) elicit similar variability in encoding processes.

In Experiment 7, we compared encoding processes for categorical cues relative to syllable cues and initial-letter cues. One hypothesis is that any cue type should encourage participants to generate specific exemplars (except for “exact” cue types, such as the specific cue word “table”), particularly if affirmative responses are due to task demand characteristics. An alternative hypothesis is that the superordinate, semantic (fruit) category triggers spreading activation to specific exemplars, and thus, participants may be less likely to generate specific exemplars of syllable and initial-letter cues in a perfunctory manner.

Ninety-nine Baylor University undergraduate students were randomly assigned to the fruits category, the syllable cue, and the initial-letter cue conditions. The practice block did not contain any prime words, prime letters, or prime syllables. The category cue procedure was identical to that used in the no-prime condition in Experiment 3 ( Fig 1 ). The instructions for the initial-letter condition were as follows (syllable cue condition in brackets):

In this experiment, we are also interested in your ability to remember to perform an action at a given point in the future. Therefore, during the word/nonword task, we would like you to perform a special action whenever you see an item that BEGINS with the letter T [item that includes the syllable "tor"]. Whenever you see an item that begins with the letter T [includes the syllable tor], you should remember to press the 'Q' key. Press 'Q' to continue.

For free response and forced-choice response data, we conducted planned comparisons between the categorical cue, initial-letter cue, and syllable cue conditions individually. For the encoding duration data, we conducted a series of between-subjects analyses of variance (ANOVAs) to evaluate whether condition and/or encoding type (specific) related to encoding duration.

As shown in Table 2 , mind wandering (off-task responses) did not significantly differ across conditions (all χ 2 < 1.3, p s > .10).

Specific exemplar generation.

Specific exemplar generation occurred in the categorical cue condition, χ 2 (1) = 10.91, p < .001 (Yates’ correction), and the syllable cue condition, χ 2 (1) = 7.00, p = .008 (Yates’ correction), but not significantly in the initial-letter cue condition, χ 2 (1) = 2.39, p = .12 (Yates’ correction; Table 3 ). The direct comparison between proportion of specific exemplar generators in the categorical cue and initial-letter cue conditions was less definitive, χ 2 (1) = 3.33, p = 0.07 (Yates’ correction). However, when measuring the total number of fruits generated, a large reduction was clearly evident from the categorical cue condition (1.06 ± 1.71) to the initial-letter condition (0.18 ± 0.72), t (42.60) = 2.74, p = .009, d = 0.84 (corrected for unequal variances). The mean number of specific exemplars generated did not differ between the syllable cue condition (0.59 ± 1.41) and the other two conditions ( p s > .10). The initial-letter cue participants were overall less likely to respond affirmative than the categorical cue participants for the general category question, χ 2 (1) = 5.81, p = .02, but importantly, when forced to choose whether they focused more on generating specific exemplars or on the overarching category, participants in the initial-letter cue condition were less likely to be biased toward specific exemplar generation than those in the categorical cue condition, χ 2 (1) = 4.30, p = .04 (Yates’ correction; no significant differences relative to the syllable condition, p s > .10).

Some readers may be surprised that specific exemplar generation was not also reduced in the syllable cue condition. We identified a counterbalance effect in the syllable cue condition regarding whether participants were first asked if they generated specific exemplars or first asked if they thought of cues as a general category (no counterbalance effects in the initial-letter condition, p s > .10). When the specific exemplar question was asked first, there was not a statistical difference in specific exemplar generation between the syllable cue (50.0%) and categorical cue (33.3%) conditions (χ 2 < 1). When the general category question was asked first, on the following screen, none of the syllable cue participants stated that they generated specific exemplars. This 0% of syllable cue participants was significantly lower than the 33.3% of categorical cue participants who were in the same counterbalance order, χ 2 (1) = 4.13, p = .04. These counterbalance patterns might be spurious (Type I error), they might reflect differential difficulty understanding the questions asked, or they might simply indicate that syllable cues are less likely to trigger specific exemplar generation under some conditions.

Mean encoding duration was similar across the three cue conditions (all t s < 1; Table 4 ), implying that the group differences in specific exemplar generation were not explained simply by alterations in strategic/elaborative encoding processes. Interestingly, there was a significant interaction between cue condition and whether participants indicated that they generated specific exemplars, F (2, 93) = 4.07, MSE = 76.03, p = .02, η p 2 = .08 (the main effect of specific exemplar generation was not significant, F <1). In the categorical condition, specific exemplar generation was unrelated to encoding duration, as in the previous experiments ( Fig 3 ; t < 1; Specific-Yes = 20.90 ± 5.36; Specific-No = 23.46 ± 11.57). For the syllable cue condition, participants who spent longer encoding the prospective memory instructions were significantly more likely to generate specific exemplars, t (30) = 2.21, p = .03, d = 0.81 (Specific-Yes = 28.46 ± 5.04; Specific-No = 22.20 ± 7.41). The reverse pattern was observed in the initial-letter condition, but there were only four exemplar-generators in this condition t (32) = 2.11, p = .04, d = 0.75 (Specific-Yes = 13.16 ± 11.79; Specific-No = 23.14 ± 8.54). These data suggest that whether exemplar generation is strategic/elaborative versus perfunctory/transient depends on the prospective memory cue type.

Some intentions may be more easily formed “in passing” than others. Relative to categorical cues, other initial-letter and syllable cue conditions elicited fewer specific exemplars. This experimental effect converges with Experiment 5 in showing that demand characteristics do not lead participants to respond affirmative to the specific exemplar generation question. Interestingly, the relationship between encoding duration and specific exemplar generation differed across cue types: Exemplars of category cues may be encoded in a perfunctory manner whereas exemplars of syllable cues require strategic/elaborative processing. The theoretical implication is that encoding processes not only vary across individuals, but also across different cue types, even for cue types that have historically been classified together as nonfocal .

Experiment 8

A remaining question is whether encoding processes predict later retrieval. Prospective memory researchers distinguish between top-down monitoring processes, and bottom-up spontaneous retrieval processes [ 59 ]. For example, one might effortfully maintain a prospective memory intention in working memory (pick up groceries) and monitor for potential retrieval cues (grocery store signs). Because monitoring is a controlled process that requires working memory resources that would normally be devoted to ongoing activities (e.g., driving), monitoring incurs a cost to ongoing task performance (e.g., slowed response times [ 63 ]).

Monitoring is a cognitively demanding process, and therefore, individuals tend not to monitor continuously across long retention intervals [ 64 – 68 ]. In the absence of monitoring, prospective memories can still sometimes be spontaneously retrieved. For example, we [ 69 ] instructed participants to remember to press the Q key if they ever saw the word crossbar (focal condition) or a word beginning with the letter c (nonfocal condition), and then had them perform 500 lexical decision trials before presenting crossbar. Monitoring costs were absent by trial 501, yet approximately ¾ of participants in the focal condition still remembered to press the Q key, relative to fewer than ¼ in the nonfocal condition (see also [ 70 ]). Thus, cue focality is considered a discriminating factor between whether an individual can successfully rely on spontaneous retrieval versus needing to monitor for cues.

In Experiment 8, after the thought probe procedure, participants performed a 500-trial lexical decision block, with the first target event on trial 501. We predicted that specific exemplar generators would outperform non-generators (Hypothesis 1) because categorical cue studies have observed greater prospective memory performance when highly-typical versus atypical categorical cues were presented [ 37 , 39 , 43 , 71 , 72 ].

We also included a retrospective-memory comparison group that did not encode the prospective memory task. This comparison group allowed us to determine monitoring costs for the prospective memory group [ 69 ]. The cue focality hypothesis would predict monitoring cost to be present in individuals focused on fruits as a general category, but reduced or absent in individuals focused on specific examples of fruits (Hypothesis 2).

Several design challenges emerge with directly connecting thought-probe encoding processes to later performance (cf. [ 73 ]). For example, participants might generate specific exemplars that are not later presented, and doing so would be expected to trigger retrieval-induced forgetting [ 74 ]. We avoided this pitfall by selecting 10 highly-typical exemplars of fruits to be successively presented (beginning on trial 501). Another challenge is that the encoding thought probes might change how participants approached the task, for example, by instilling more importance to the prospective memory task (for discussion, see Kliegel et al.’s plan aloud procedure [ 75 ]). To address the general issue of the thought probes increasing task importance, we included a “standard” prospective memory comparison condition in which encoding processes were not assessed, but all other procedural elements were maintained. If the encoding thought probes increased strategic processing (cf. importance effects [ 76 ]), then the group with encoding thought probes should outperform the standard prospective memory condition.

Baylor University undergraduate students ( N = 149) participated for partial class credit. Participants were randomly assigned to the following conditions: retrospective memory control ( n = 30), standard prospective memory ( n = 30), and PM-Encoding-Probes ( n = 89). A larger sample size was recruited for the PM-Encoding-Probes condition to ensure reasonable subgroup sizes (i.e., given the frequencies in Table 3 , we expected a minimum of n = 20 to generate specific exemplars).

Lexical decision task filler items were the same as used in a previous study [ 69 ]. Highly typical fruit prospective memory words were selected using semantic norm databases [ 47 , 77 ].

After being introduced to the lexical decision task and performing a practice block, participants completed a pre-encoding, control block of 100 lexical decision trials. No fruit prime words appeared during practice or baseline/control blocks.

Participants in the prospective memory conditions were next instructed that they would perform another lexical decision block, but to remember to press the Q key if they ever see any fruit words. Participants in the retrospective -memory control condition were instructed:

“In this experiment, we are also interested in your ability to remember certain "target" keys and categories. Your target key is "Q" and your target category is "fruits." At the end of the experiment, we will ask you to recall your target key and target category. Press 'Q' to continue.”

In the PM-Encoding-Probes condition, we then presented the free response and yes/no questions shown in Fig 1 . All participants then completed 510 lexical decision task trials. Apple was presented on trial 501, followed by the following fruit words: cherry, orange, peach, banana, berry, pear, plum, kiwi, and apple (Experiment 3 showed apple to be the most commonly generated fruit exemplar, and so we presented it twice to maximize the probability of a retrieval). We selected the procedure of having all targets at the end of the block rather than early to minimize strategic monitoring processes; if a target cue is presented early it will trigger more monitoring, and perhaps additional attempts at cue generation for the remainder of the block [ 68 ]. Though participants were allowed to press the Q key immediately upon seeing the target, or after making their ongoing task response, pressing the Q key advanced the screen, so functionally, participants could make the Q response instead of an ongoing task response. After the prospective memory experiment, a subset of participants ( n = 116) completed the automated reading span task to estimate working memory capacity [ 78 ].

Statistical analyses.

For prospective memory performance, we calculated the proportion of fruit target trials in which the Q key was pressed. For ongoing task cost, we used the same analysis of covariance (ANCOVA) approach as in our previous study [ 69 ]: We calculated mean response times to all trials with correct responses and covaried response times from the pre-encoding, control block. To complement the “untrimmed” response time analyses, we also trimmed response times ±2 standard deviations from each individual’s (sub)block mean, because trimmed response times are sometimes considered to be more sensitive to group differences (lower variance). Wherever trimmed response times led to a different statistical conclusion (alpha = .05) than untrimmed response times, we present those data. We planned to compare prospective memory performance and ongoing task cost as a function of encoding subgroups (yes/no, encoding bias questions), and further planned to compare these subgroups against the retrospective-memory control group. Because we identified pre - experimental group differences in the standard prospective memory condition relative to the other conditions, we report those data separately.

Encoding thought probe responses.

The thought probe data converged with Experiments 1–7 and are shown in Tables 2 and 3 ( Fig 2 shows the data collapsed across experiments). Though theories of planning emphasize the role of working memory capacity [ 79 ], reading span scores were not associated with specific exemplar generation (Specific-Yes: 56.64 ± 9.80, Specific-No: 55.35 ± 11.32, t < 1, BF 10 = 0.30) or encoding the fruit cue as a general category question (General-Yes: 56.98 ± 9.65, General-No: 52.75 ± 13.00, t (57) = 1.36, p = .18, d = .36, BF 10 = 0.61). Moreover, reading span scores did not significantly distinguish on-task participants (57.28 ± 9.12) from participants who were mind wandering (52.79 ± 13.23; t (26.42) = 1.34, p = .19, d = .52, BF 10 = 0.72, corrected for unequal variances). These data converge with the view that prospective memory encoding can be perfunctory.

Frequency of cue words generated.

Nearly all the specific-exemplar-generator participants (93.9%) encoded a fruit word that would be a prospective memory target word. The most frequently generated fruits were banana, apple, and orange. Of participants who generated specific fruits, participants listed 2.45 ± 1.28 fruit words.

According to the strategic/elaborative view, because prospective memory is future oriented, it may prompt greater imaginal-enactive processes at encoding than retrospective memory encoding [ 80 ]. However, as shown in Table 4 , encoding duration did not significantly differ across the PM-Encoding-Probes and retrospective memory control conditions ( t < 1; cf. [ 81 ]). There were also no associations between encoding duration and encoding thought probe responses (all p s > .10; see Fig 3 ). Furthermore, if successful intention encoding requires strategic/elaborative processing, then longer encoding durations should predict better prospective memory performance; however, encoding duration correlated negatively (nonsignificantly) with later performance ( r p (116) = -.14, p = .14, controlling for condition). Thus, forming a category-cue intention does not require more strategic processing than reading a similar length instruction screen, and even perfunctory encoders can be successful prospective memory performers.

Standard condition showed pre-experimental differences.

Despite random assignment to conditions, and identical instructions, the standard condition took significantly longer to encode the prospective memory task than the PM-Encoding-Probes condition, t (35.47) = 2.47, p = .02, d = .83 (corrected for unequal variances). Moreover, during the control lexical decision block (Tables 5 and 6 ), the standard condition showed slower response times than the retrospective-memory condition, t (42.01) = 2.59, p = .01, d = 0.79 (corrected for unequal variances) and PM-Encoding-Probes condition, t (117) = 1.85, p = .07, d = 0.34. For prospective memory responses, in the standard condition, 90% of participants remembered to press Q at least once and there were significantly more overall Q responses to fruit words ( M = .73) than in the PM-Encoding-Probes condition, t (66.21) = 3.24, p = .002, d = 0.79 (corrected for unequal variances). It is unclear why this condition was so aberrant, but the direction of the results was opposite of the prediction that the thought probe questions would increase the importance of the prospective memory task.

Prospective memory performance relative to encoding processes.

In the PM-Encoding-Probes condition, one hypothesis was that specific exemplar generation would increase prospective memory performance. As illustrated in Fig 4 , participants who reported generating specific exemplars performed significantly better than those who did not, t (72.41) = 2.68, p = .009, d = 0.63 (corrected for unequal variances). Moreover, participants who generated specific exemplars and indicated that they were biased toward specific encoding (0.69 ± 0.35) significantly outperformed those who did not generate specific exemplars and reported being biased toward categorical processing (0.41 ± 0.43), t (36.85) = 2.41, p = .02, d = 0.79 (corrected for unequal variances).

Error bars reflect standard errors and ** indicates p < .01.

https://doi.org/10.1371/journal.pone.0198646.g004

If successful encoding always requires the engagement of strategic/elaborative processes, then participants who reported that they hardly thought about the prospective memory task (at encoding) should perform very poorly. By contrast, performance did not differ as a function of responses to the encoding bias question (Hardly Thought About It = 0.53 ± 0.40; Exemplar Bias = 0.54 ± 0.42; Category Bias = 0.46 ± 0.43; p s > .10).

Ongoing task performance.

A second hypothesis was that encoding biases might alter subsequent retrieval processes (monitoring versus spontaneous retrieval), as measured by ongoing task performance. Typically, ongoing task accuracy is not a sensitive measure of monitoring, and Table 5 shows that accuracy cost did not significantly differ across the PM-Encoding-Probes condition and the retrospective-memory control condition ( F < 1) or as a function of encoding thought probe responses (largest F (1, 63) = 2.17, MSE = .006, p = .15, η p 2 = .03, for encoding bias question).

https://doi.org/10.1371/journal.pone.0198646.t005

Table 6 presents the unadjusted and untrimmed mean response times on correct, non-target lexical decision trials. Response time cost did not differ across the PM-Encoding-Probes condition and the retrospective-memory control condition, or as a function of individuals’ responses to the specific exemplar and general category questions (all F s < 1). However, as illustrated in Fig 5 , separating participants based on the encoding bias question demonstrated that participants who focused on fruits as a general category tended to show greater cost than those who focused on specific fruit exemplars (trimmed response times: F (1, 62) = 4.02, MSE = 8393.10, p < .05, η p 2 = .06; untrimmed: F (1, 62) = 3.73, MSE = 11538.33, p = .06, η p 2 = .06). Furthermore, there was evidence for a greater group difference in response time cost late in the prospective memory block (trials 401–500; trimmed response times: F (1, 62) = 4.36, MSE = 18070.19, p = .04, η p 2 = .07; untrimmed: F (1, 62) = 3.52, MSE = 22459.00, p = .07, η p 2 = .05) relative to early in the prospective memory block (trials 1–100; F (1, 62) = 1.69, MSE = 6970.64, p = .20, η p 2 = .03; untrimmed: F (1, 62) = 2.36, MSE = 9105.61, p = .13, η p 2 = .04), though the direct test for the block by group interaction was nonsignificant ( F (1, 62) = 1.08, MSE = 8394.04, p = .30, η p 2 = .02).

Baseline-adjusted mean trimmed responses times across quintiles of the prospective memory test block in Experiment 8. The cost results are separated by individuals focused on fruits as a general category and individuals focused on specific fruit exemplars. Error bars represent standard errors.

https://doi.org/10.1371/journal.pone.0198646.g005

https://doi.org/10.1371/journal.pone.0198646.t006

Inter-individual variability in encoding was associated with prospective memory performance (Hypothesis 1) and retrieval processes (Hypothesis 2). Consistent with the Multiprocess Framework, participants who generated specific exemplars at encoding (focal cues) showed significantly greater prospective memory performance than those who did not [ 37 , 39 , 43 , 71 , 72 ]. However, because the specific exemplar feature was quasi-experimental (cf. [ 75 ]), we cannot rule out that “participants who show good prospective memory are also good planners” (p. 1737 [ 75 ]). For example, perhaps participants who generated specific exemplars were more motivated to perform the prospective memory task. If so, then based on previous work [ 76 ], specific-exemplar encoders should have shown more ongoing task costs, higher working memory scores, or altered encoding durations. By contrast, individuals who focused on specific fruit cues (focal cue) demonstrated fewer monitoring costs than those that focused on fruits as a general category (nonfocal cue), with no group differences in encoding duration or working memory scores. Relative to the retrospective-memory control condition, specific-exemplar encoders showed no ongoing task costs, indicating that spontaneous retrieval processes supported their prospective remembering [ 58 ]. Though additional research is warranted, the collective findings are more consistent with the cue focality account than a motivation account.

Consistent with the perfunctory/transient view, there was minimal-to-no evidence that prospective memory performance suffered in participants who were mind wandering, who had low working memory capacity, or who reported to hardly thinking about the prospective memory task. These results distinguish prospective memory encoding from theoretical views in the planning literature [ 79 ] and the retrospective memory encoding literature [ 82 – 85 ]. Even the literature on goal fulfillment, which argues that many individuals form general intentions (with minimal cognitive effort), predicts that strategic/elaborative processes are beneficial, if not necessary, for later goal execution [ 86 ]. Prior to conducting the current work, we would have assumed that categorical prospective memory encoding constitutes “deep” processing [ 19 ], but the totality of findings on mind wandering, brief encoding durations, and null associations between mind wandering and prospective memory performance converge on the conclusion that at least some components of intention encoding can be perfunctory/transient.

Conclusions

We investigated the encoding of prospective memory intentions using a thought probe procedure that has previously been useful in examining retrieval processes [ 33 – 35 ]. As a theoretical orientation, we contrasted two general views. The elaborative/strategic view, which emanates from the literature on planning and retrospective memory and emphasizes the functional importance of effortful, working memory resources. By contrast, the perfunctory/transient view emphasizes that some components of prospective memory intentions might be encoded with minimal effort. The consistent theme across eight experiments was that there exists substantial quantitative and qualitative variability in the manner in which participants encode laboratory prospective memory intentions. Whereas quantitative differences in encoding duration seemed to have minimal functional value, differences in encoding quality clearly mattered: Intentions that were encoded more specifically were more likely to be later remembered with lower or no cost (Experiment 8). In other words, the most effective form of encoding occurred in a perfunctory manner.

Transience of prospective memory encoding

Task disengagement, or mind wandering, is common in classrooms and during psychology experiments [ 87 , 88 ]. It is surprising, however, that over 40% of free responses were solely off-task ( Fig 2 ). Our procedure was not a long, monotonous task, as is the case in many mind wandering studies. Furthermore, the prospective memory instructions are arguably the most important stage of a prospective memory experiment. Obviously, this stage is more important to scientists than to most participants. A potential caveat is that some participants who were classified as “off-task” may have initially been engaged. But, it seems highly unlikely that all of the participants categorized as off-task were engaging strategic/elaborative encoding processes: Nearly one-quarter of participants reported that they hardly thought about the prospective memory task at all ( Fig 2 ).

Similar levels of hardly-thinking-about-encoding have been reported in naturalistic studies. For example, in a naturalistic study of eight participants, Holbrook and Dismukes [ 89 ] found that for 23% of intentions that participants “did not think very much about the intention, just assumed [they] would remember to perform it” (see also, Marsh and colleagues’ [ 31 ] study of “recorders” and “nonrecorders”). Such participants performed poorly in their study [ 89 ], but in other naturalistic research, participants who only implicitly formed an intention to put their watch back on their wrist were able to successfully remember that intention [ 18 ].

Categorical cues: Focal, nonfocal, neither, or both?

Even when participants were “on-task,” they differed in how they encoded the prospective memory cue. Some researchers have acknowledged that participants might generate specific exemplars during category prospective memory encoding [ 39 , 43 , 90 ], but many scientific reports that used categorical cues have dismissed or otherwise ignored this possibility. Our review papers and others’ meta-analysis papers have always classified categorical cues as “nonfocal” to ongoing tasks [ 61 , 62 ]. Therefore, a salient finding from the encoding thought probe procedure was the robustness of specific exemplar generation in all experiments ( Fig 2 ). Particularly relevant to prospective memory’s cue focality hypothesis [ 60 ], in Experiment 8, we observed that the variability in encoding specificity mattered to prospective memory accuracy and ongoing task cost: The more specifically a categorical cue was encoded, the more likely it was to elicit performance akin to a focal-cue condition. Thus, encoding variability may explain why categorical cues can sometimes trigger spontaneous retrieval [ 91 ] and be associated with minimal age differences in prospective memory performance [ 92 ]. Indeed, in Experiment 4, we found that healthy older adults were as likely as young adults to encode specific exemplars.

The methodological implication for future research on cue focality may be to use initial-letter cues. Perceptual identification studies indicated that initial-letters were as easily identifiable as whole words, which are the prototypical focal cue [ 69 ]. In addition, in Experiment 7, specific exemplar generation was reduced with initial-letter cues relative to categorical cues, possibly because superordinate categories (animals, fruits) cause spreading activation in semantic networks to a category’s exemplars [ 49 , 50 ]. To be clear, we are not arguing that researchers should never use categorical cues. Instead, we recommend using categorical cues to investigate encoding variability, encoding—retrieval interactions, and similar questions (but not to investigate cue focality).

Strategic versus perfunctory: Dichotomy or continuum?

In the current work, we described strategic/elaborative processing and perfunctory/transient processing as a dichotomy. We selected this “either/or” approach to provide straightforward exposition that allowed for competing research hypotheses. Moreover, the dichotomy conceptualization builds on Searle’s [ 93 ] philosophical distinction between prior intentions and intentions-in-action, as well as Kvavilashvili and colleagues’ [ 18 ] empirical isolation of implicit intentions. Nevertheless, when considering the Dynamic Multiprocess Framework’s proposal that bottom-up and top-down processes are both engaged for individual intentions [ 59 ], it may be more realistic (albeit less parsimonious) to expect that every time one encodes an intention that some aspects of encoding will be perfunctory (e.g., specific cues related to an overarching intention) and other aspects of encoding will be strategic/elaborative (e.g., the sequence of planned actions). If we conceptualize strategic/elaborative and perfunctory/transient encodings as part of a continuum, then the summed degree of strategic/elaborative processing likely depends on whether the intention is self-generated or other-generated [ 94 ], whether the content is important and complex [ 58 ], and whether the retrieval context is predictable and controllable [ 75 ]. Mapping the degrees of strategic-to-perfunctory processing during individual encodings seems a worthy, albeit challenging, goal for future research.

Practical implications

From a translational perspective, our findings emphasize the importance of specifically encoding intentions [ 75 ]. Implementation intention encoding [ 86 ] is one strategy to improve goal fulfillment via re-phrasing a general intention into specific exemplars. For example, instead of “I need to get gas” one might state “When I see the red gas station sign , then I will remember to fill up my car with gas.” We previously found that implementation intention encoding increased the number of specific exemplars generated during a category prospective memory task, particularly when a structured “When…then” statement was paired with visual imagery of the intention [ 24 ]. Thus, even though specific exemplar encoding can occur via perfunctory processes, it can also be stimulated strategically via an implementation intention strategy. Increasing the probability of spontaneous retrievals via encouraging specific exemplar generation is likely to be one mechanism by which implementation intentions improve remembering of laboratory and naturalistic prospective memory tasks [ 95 , 96 ].

Some prospective memory research has indicated that strategic/elaborative encoding, a view adapted from theories of planning [ 79 ], is required to successfully encode an intention [ 19 , 26 , 28 ]. The results of other prospective memory studies, however, indicate that aspects of encoding can be perfunctory/transient [ 18 , 29 , 30 ]. Our findings of the commonality of mind wandering, brief encoding durations, similarities across young and healthy older adults, and null associations between mind wandering and prospective memory performance, converge with the perfunctory view. In other words, some prospective memory encoding may be done “in passing.”

Acknowledgments

Portions of this project were presented at the International Conference on Prospective Memory (Naples, Italy, 2014), the Meeting of the Psychonomic Society (Chicago, IL, 2015), and the Cognitive Aging Conference (Atlanta, GA, 2016). We are appreciative to Khuyen Nguyen, Mericyn Daunis, Hannah Ballard, Kiersten Scott, Stacy Nguyen, Mary High, Taylor Terlizzese, Sarah Thomas, and Chenlu Gao for their assistance.

• View Article
• Google Scholar
• PubMed/NCBI
• 7. Miller GA, Galanter E, Pribram ICH. Plans and the structure of behavior. New York: Holt, Rinehart and Winston; 1960.
• 17. Dismukes RK. Prospective memory in everyday and aviation settings. In Kliegel M, McDaniel MA, Einstein GO, editors. Prospective memory: Cognitive, neuroscience, developmental, and applied perspectives. Taylor and Francis; 2008. pp. 411–431.
• 20. Moscovitch M. Commentary: A perspective on prospective memory. In Kliegel M, McDaniel MA, Einstein GO, editors. Prospective memory: Cognitive, neuroscience, developmental, and applied perspectives. Taylor and Francis; 2008. pp. 309–320.
• 21. Cohen AL, Gollwitzer PM. The cost of remembering to remember: Cognitive load and implementation intentions influence ongoing task performance. In Kliegel M, McDaniel MA, Einstein GO, editors. Prospective memory: Cognitive, neuroscience, developmental, and applied perspectives. Taylor and Francis; 2008. pp. 367–390.
• 49. Anderson JR, Bower GH. Human associative memory. Washington, D.C.: Winston; 1973.
• 56. Love J, Selker R, Marsman M, Jamil T, Dropmann D, Verhagen AJ, et al. JASP (Version 0.7)[computer software]. Amsterdam, the netherlands: Jasp project. 2015.
• 58. McDaniel MA, Einstein GO. Prospective memory: An overview and synthesis of an emerging field. Thousand Oaks, CA: Sage; 2007.
• 61. Einstein GO, McDaniel MA, Scullin MK. Prospective memory and aging: Understanding the variability. In Naveh-Benjamin M, Ohta N, editors. Memory and aging: Current issues and future directions. Philadelphia, PA: Psychology Press; 2012.
• 79. Morris R, Ward G, editors. The cognitive psychology of planning. New York, NY: Psychology Press; 2004.
• 89. Holbrook JB, Dismukes RK. Prospective memory in everyday tasks. In Proceedings of the Human Factors and Ergonomics Society 53rd Annual Meeting. Santa Monica, CA: Human Factors and Ergonomics Society; 2009. pp. 590–594.
• 93. Searle JR. Intentionality: An essay in the philosophy of mind. Cambridge university press; 1983.
• 94. Kvavilashvili L, Ellis JA. Varieties of intention: some distinctions and classifications. In Brandimonte M, Einstein GO, McDaniel MA, editors. Prospective memory: theory and applications. Erlbaum Mahwah, NJ; 1996. pp. 23–51.
• 96. Cohen AL, Hicks JL. Prospective memory: Remembering to remember, remembering to forget. Springer; 2017. https://doi.org/10.1007/978-3-319-68990-6

• Original article
• Open access
• Published: 26 September 2018

How the stimulus influences mind wandering in semantically rich task contexts

• Myrthe Faber   ORCID: orcid.org/0000-0002-6972-9962 1 , 2 , 3 &
• Sidney K. D’Mello 1 , 4 , 5  

Cognitive Research: Principles and Implications volume  3 , Article number:  35 ( 2018 ) Cite this article

6372 Accesses

26 Citations

Metrics details

What do we think about when we mind wander and where do these thoughts come from? We tested the idea that semantically rich stimuli yield patterns of mind wandering that are closely coupled with the stimuli compared to being more internally triggered. We analyzed the content of 949 self-reported zone outs (1218 thoughts) and 519 of their triggers from 88 participants who read an instructional text and watched a film for 20 min each. We found that mind wandering associated with memory retrieval was more frequent than prospection and introspection across both stimuli. Over 70% of autobiographical and semantic memory retrievals were triggered by the content of the stimuli, compared to around 30% for prospective and introspective thoughts. Further, latent semantic analysis revealed that semantic and unspecific memories were more “semantically” similar to their triggers than prospective and introspective thoughts, suggesting that they arise from spontaneous associations with the stimulus. These findings suggest a re-evaluation of how internal concerns and the external world give rise to mind wandering and emphasize the importance of studying mind wandering in semantically rich contexts akin to much of the real world.

Significance

Mind wandering frequently occurs during everyday activities such as reading a book or watching TV, but where do these thoughts come from and are they influenced by the ongoing activity? We analyzed the content of thoughts, their triggers, and thought trains during reading and film comprehension. We found that: (1) mind wandering associated with memory retrieval is highly common when processing semantically rich content; and (2) much of spontaneous thinking is driven by the stimulus itself. We propose that research should consider semantic information in the environment to better understand how internal concerns and the external world give rise to mind wandering.

When reading a text, listening to a conversation, or watching a film, some of our thoughts are focused on the content of the task at hand whereas others wander off towards past memories, introspections, prospections, and even fantasies. Experience sampling studies tell us that mind wandering is ubiquitous, occurring as much as 50% of the time in everyday life (Killingsworth & Gilbert, 2010 ). Although numerous studies have focused on analyzing the frequency of mind wandering across tasks and contexts (see Randall, Oswald, & Beier, 2014 for an overview), few have explored from where it arises and how it influences subsequent thinking. In fact, a recent review of mind wandering research identified “[characterizing] the environmental conditions and internal concerns that tend to initiate [it]” as a key issue (Smallwood & Schooler, 2015 , p. 511). The emphasis on internal states poses an important challenge because much of psychological research focuses on establishing a relationship between a stimulus or manipulation and behavior, somewhat neglecting that much of our thinking is self-generated and often off-task. It is also possible that the task or stimulus itself is the driver of mind wandering, raising interesting questions about certain experimental effects (cf. Faber, Mills, Kopp, & D’Mello, 2017 ; Krasich et al., 2018 ). Here, we analyze the content of thoughts to provide insight into how task stimuli drive mind wandering.

We begin by considering the nature of mind wandering. The term itself captures one of its key characteristics: “wandering” meaning to “move hither and thither without fixed course or certain aim” (Christoff, Irving, Fox, Spreng, & Andrews-Hanna, 2016 , p. 719). Indeed, mind wandering is less deliberate than goal-directed thinking, as its content mostly arises outside of cognitive control (Christoff et al., 2016 ). This is not to say that “undeliberate” thoughts are completely unconstrained – some spontaneous thoughts, such as rumination and obsessive thoughts, are strongly driven and automatically constrained by their affective salience (Christoff et al., 2016 ).

Importantly, sensory salience can lead to strongly constrained spontaneous remindings, for example, seeing a professor in a green tie and vest can automatically trigger memories about a pre-school teacher playing tricks on his or her students on St. Patrick’s Day (example from Ball & Little, 2006 p. 1173). In line with this, one study found that people mind wandered more about the past than the present or future while completing a vigilance task with emotionally valenced words (Plimpton, Patel, & Kvavilashvili, 2015 ), suggesting that emotional words trigger past memories. Similarly, another study (unexpectedly) found that when reading a text on membranes, people who had more experience with the topic of a text (e.g. years of formal biology education) mind wandered more about the past than about the future (Smallwood, Nind, & O’Connor, 2009 ). These findings suggest the content of mind wandering might be to some extent triggered by the stimulus, likely through activations of memory traces (Faber & Mills, 2018 ).

Of course, it is unlikely that all mind wandering arises from stimulus Footnote 1 processing. People also frequently mind wander in situations that are devoid of much sensory or affective salience. Studies that have used semantically impoverished tasks, such as the Sustained Attention to Response (SART) task, have found a strong bias towards prospective mind wandering (e.g. Baird, Smallwood, & Schooler, 2011 ; McVay & Kane, 2013 ; Smallwood et al., 2009 ; Stawarczyk, Cassol, & D’Argembeau, 2013a ; Stawarczyk, Majerus, Maj, Van der Linden, & D’Argembeau, 2011 ). According to the current concerns hypothesis (Klinger, 1978 , 1999 ), personally relevant information, such as unfulfilled goals, is the source of much of spontaneous cognition. Indeed, when participants are asked to focus on their personal concerns or needs before engaging in a task, rates of mind wandering increase (Klinger, 2013 ; Kopp, D’Mello, & Mills, 2015a ; Masicampo & Baumeister, 2011 ; Rummel & Nied, 2017 ; Stawarczyk, Majerus, & D’Argembeau, 2013b ; Stawarczyk et al., 2011 ), suggesting that this type of mind wandering might to a large extent be driven by the importance of these current concerns rather than spontaneous associations elicited by the environment. This does not imply that prospective mind wandering is solely driven by internal concerns because a semantically rich environment might also provide cues that align with an individual’s active goals or concerns, triggering prospective thoughts (Klinger, 2013 ; McDaniel, Einstein, Guynn, & Breneiser, 2004 ; Stawarczyk et al., 2011 ).

Hence, to what extent is the content of mind wandering associated or triggered by the task stimulus versus more internally driven? We hypothesize that when people engage in semantically rich task contexts that give rise to spontaneous associations, such as reading or watching a film, we expect to find a greater propensity of mind wandering thoughts related to autobiographical and semantic memory retrieval compared to thoughts pertaining to more internally driven current concerns and feelings. We further hypothesize that the source of mind wandering varies systematically, such that autobiographical and semantic memories should be triggered by and align with the stimulus to a greater extent than prospective and introspective thoughts. This would suggest that the former are more likely to arise from stimulus processing whereas the latter are driven primarily (but not exclusively) by internal concerns of importance to the individual.

We leverage code and count techniques as well as computational linguistics approaches (latent semantic analysis [LSA]; Landauer, Folt, & Laham, 1998 ) to test these predictions when people engage in two real-world semantically rich activities: reading a text and watching a film. We asked 88 participants to report the verbatim content of their thoughts and what, if anything, triggered them whenever they caught themselves mind wandering during these 20-min long tasks. We then categorized the content of the reported thoughts using rubrics based on previous mind wandering (see below), diary, and mental time travel studies (e.g. Ball & Little, 2006 ; Berntsen & Hall, 2004 ; D’Argembeau, Renaud, & Van Der Linden, 2011 ; Hintzman, 2011 ; Kvavilashvili & Mandler, 2004 ; Mace, 2005 ; Miles & Berntsen, 2015 ; Schank, 1983 ; Seilman & Larsen, 1989 ). We compared the proportion of thoughts across categories, and for each category, computed the proportion of thoughts that were triggered by the stimulus, and importantly, whether the thought content was meaningfully related to the trigger.

Our study builds on previous studies that have compared the occurrence of task-related interferences and task-unrelated thoughts (e.g. Baird et al., 2011 ; Sarason, Sarason, Keefe, Hayes, & Shearin, 1986 ; Smallwood, Obonsawin, & Heim, 2003 ; Stawarczyk et al., 2011 ) and those that have distinguished among thoughts pertaining to sensory and emotional states, the self, current concerns, prospective memory, stimuli, environmental distractions, and fantasies (e.g. Baumeister, Vohs, & Oettingen, 2016 ; Krawietz, Tamplin, & Radvansky, 2012 ; Schooler, Reichle, & Halpern, 2004 ; Smallwood et al., 2016 ; Song & Wang, 2012 ). However, in most of these studies, participants did not report their thoughts verbatim (with the exception of Baird et al., 2011 ), but either rated them on a Likert-scale (e.g. using a 1–5 scale, to what extent a thought was related to a plan) or selected a thought category out of a number of options (e.g. “school-related,” “yourself,” “text-related,” “fantasies,” etc.; Krawietz et al., 2012 ), limiting an in-depth analysis of the thought content like in the present work.

We also collected data on people’s verbatim thought triggers and used LSA (Landauer et al., 1998 ) to test associations between the trigger and the subsequent mind wandering thought. For instance, the trigger “all the talk about water” from the text stimulus and the mind wandering memory-based thought, “[a] beach nearby me at home that I always go to,” are meaningfully related, because water and beach share associations like the sea and swimming. In contrast, the relationship between the prospective thought “what I am going to wear to class tomorrow” and its trigger “the red balloon” is less apparent, as clothes and balloons have less semantic overlap. The goal of this study is to tease apart these relationships to provide insight into how mind wandering emerges in semantically rich tasks contexts.

Participants

Participants were 88 undergraduate students from a medium-sized private U.S. university ( N  = 65) and a large public U.S. university ( N  = 23) who participated for course credit (69% female). Participants were on average 19 years old; 63% were Caucasian/White, 22% African-American/Black, 6% Hispanic, Latino, or of Mexican origin, 4% Asian, 4% American Indian or Native Alaskan, and 1% reported “other.” Because the primary goal of the study was to collect verbatim thought content, and our previous studies with similar stimuli used suggested that the number of self-caught mind wandering reports varied considerably across participants (Faber, Radvansky, & D’Mello, 2018 ; Kopp et al., 2015a ; Kopp, Mills, & D’Mello, 2015b ), we sought to collect as much data within the subject pool schedule. As such, we did not conduct an a priori power analysis.

Ethics, consent, and permissions

Before the study, participants read and signed an agreement to participate and a (voluntary) data release form permitting the use of their data for publication. They were informed that they were free to withdraw at any time. All materials, procedures, and forms were approved by the Institutional Review Board for the Protection of Human Subjects at both universities.

The text excerpt was taken from a book entitled Soap-bubbles and the Forces which Mould Them (Boys, 1890 ), which is an instructional text on a relatively unfamiliar topic (surface tension). The text is rich in semantic content as it describes a series of experiments that the reader has to visualize (e.g. “I have in my hand a common camel’s-hair brush. If you want to make the hairs cling together and come to a point, you wet it, and then you say the hairs cling together because the brush is wet”) (Boys, 1890 , p. 15). To resemble naturalistic computerized reading, an average of 650 words were presented per screen resulting in ten screens of text. Participants read at their own pace and read on average 6.1 screens in the allotted 20 min. For the film, we used the first 20 min of the movie Le Ballon Rouge (‘The Red Balloon’), a 32.5-min French film with English subtitles about a young boy in Paris who finds a red balloon that follows him wherever he goes (Lamorisse, 1956 ).

All experimental procedures were delivered on a computer. Participants were informed that the primary task was to read/watch an excerpt from the book/film (order counterbalanced) for 20 min each. They were instructed to report mind wandering whenever they found themselves zoning out while completing the primary task. Participants received the following instructions:

While you are [reading/ watching the film], you may find yourself thinking about something other than what you are [reading/watching]. This is called “zoning out.” We are interested in what types of things people think about during a task like this (and during other kinds of tasks). In order to examine this, if you catch yourself zoning out at any time during reading, simply press the key labeled “ZONE OUT” on the keyboard. Please locate the “ZONE OUT” key now. When you indicate that you are zoning out the computer will ask you what you were just thinking about. It is perfectly normal to think about things that are not related to the task and to have different kinds of thoughts during different kinds of tasks. Please try your best to honestly assess your thoughts at the time when we ask.

Whenever participants reported zoning out, they were further instructed as follows:

In the space below please tell us what you were thinking about when you zoned out. Was there something in the [text/video] that triggered this thought? If yes then please describe what it was and if no then leave blank.

The task paused while participants were reporting thoughts. Participants were informed that the reading and film comprehension phases of the study would last 20 min each and that reporting the content of their thoughts would not increase the amount of time in the study. They were encouraged to be as complete as possible when reporting their thoughts. After both phases of the study, participants were interviewed by the experimenter about whether the instructions were clear, what triggered their thoughts during both phases and whether they had any issues reporting thoughts.

Number of thoughts and triggers

Participants reported a total of 949 instances of mind wandering (557 during text comprehension, 392 while watching the film; an average of 10.8 [ SD  = 6.08] per participant). Whereas most instances (77.1%) contained one mind wandering thought, 217 instances were associated with two or more thoughts (e.g. “[When I was Facetiming with my mom yesterday] THOUGHT1 [and] [step team tryouts] THOUGHT2 ”), resulting in a total of 1215 thoughts. We identified 43 thoughts that were explicitly aimed at gaining a deeper understanding of the text or film as on-task thoughts and excluded them (e.g. “The camel’s hair experiment performed in the text,” “Whether the man in the window was the one who wrote down something in the previous scene,” “How capillary action in plants carry water up the stem. I have an image in my head of water going up the stem of a plant”).

We obtained participants’ responses for 1082 thought triggers (we did not systematically obtain triggers for 136 thoughts from 10 participants due to experimental error). A total of 524 (48.4%) thoughts were accompanied with a trigger (302 in the text condition, 222 in the film condition). Upon closer inspection, 46 triggers contained content associated with the thought (e.g. “[The soap] TRIGGER [reminded me that] [I need to give my dog a bath and make an appointment for him to get his nails clipped] CONTENT ”). We separated the thought content from the trigger and counted it as a thought unless it fully overlapped with the reported thought. Eight triggers consisted of only thought content (e.g. “Fortunate to live in the country and have gone down to play by the brook”) and were therefore not counted as triggers but instead as thoughts. For three thoughts, no trigger was reported but the thought content explicitly mentioned the trigger. In those cases, we removed the trigger from the thought content and counted it as a trigger (e.g. “[Reading the name “Lear”] TRIGGER [made me think of] [King Lear] CONTENT ”). This resulted in a total of 1218 thoughts (732 from the text condition, 486 from the film condition; on average 13.8 mind wandering thoughts ( SD  = 8.04) per participant) and 519 triggers (298 in the text condition, 221 in the film condition). Out of these triggers, only 24 (4.58%) were related to internal states of the participants (e.g. “I’m just really bored”). All other triggers were related to the content of the stimuli.

Thoughts per category

Two researchers—one of whom was naïve about the aims of the study— coded the following content categories: autobiographical memories; semantic memories; fantasies; prospection (including current concerns); task-related interferences; thoughts about the stimulus itself; environmental distractions; and introspection (see Appendix and Table  1 for the full rubric along with examples). When it was unclear whether a memory was semantic or autobiographical, it was categorized as an unspecific memory. Thoughts that did not clearly fit in a category were categorized as vague. The coders first coded 100 randomly selected thoughts and discussed their categorizations until they reached full agreement. They then independently coded the remaining instances, achieving fair agreement (Cohen’s κ = 0.71). Finally, the coders resolved all disagreements through discussion.

Table 1 gives an overview of percentages and number of thoughts per category. We used Wilcoxon’s signed rank test for paired samples at the participant level (non-parametric testing due to zero-inflated distributions and overdispersion; Bonferroni corrected for multiple comparisons; Table  2 ) to establish whether some categories were more frequent than others. Introspection, prospection, autobiographical and semantic memories, and thoughts about the stimulus and task occurred more frequently than fantasies and environmental distractions. Prospection and introspection were more frequent than task-related interferences, but not autobiographical and semantic memories. When we combined autobiographical, semantic, and unspecific memories into one memory category, we found that memories ( M  = 4.58, SD  = 4.18) were significantly more frequent than prospection ( Z  = 4.22, p  < 0.001) and introspection ( Z  = 4.10, p  < 0.001). This suggests that the prospective bias observed in previous studies (e.g. Baird et al., 2011 ; McVay & Kane, 2013 ; Smallwood et al., 2009 ) might be limited to contexts relatively devoid of semantic content.

We conducted two follow-up analyses. First, we used thought-level mixed-effects logistic regressions to explore whether thoughts from each category (coded as 1 or 0) were more likely to occur when participants (added as a random intercept) completed the text or film comprehension task first. These analyses yielded no significant differences (all p values > 0.287). We therefore did not distinguish between the different orders in the subsequent analyses.

We also used thought-level mixed-effects logistic regressions to investigate whether thoughts from certain categories were more likely to occur during the text or film comprehension task (fixed effect) with participant as an intercept-only random effect. We found one significant difference (all other p values > 0.168) – thoughts about the stimulus were more likely during film comprehension (e.g. “Whether the balloon is real or picture animated,” “I wish this balloon made sound or something”) (odds ratio [OR] = 13.5, SE  = 2.57, p  = 0.006), which is unsurprising due to the comparatively stronger audiovisual information in the film compared to reading text (Fig.  1 ).

Proportions of thoughts per category per task pooled across thoughts ( light gray : film comprehension, dark gray : text comprehension)

Triggers per category

Where do mind wandering thoughts come from? It is clear that thoughts about external distractors, the task, and stimuli are cued by the environment. But what about thinking of “a cat I had that used to follow me around everywhere” or “what if the kitty of Alice in Wonderland is actually who is leading the balloon?” In particular, we asked whether memories were more likely to be triggered by the stimulus compared to introspective and prospective thoughts that are considered to arise primarily from feelings and current concerns, respectively. We used thought-level mixed-effect logistic regressions to predict whether a thought was triggered (1) or not (0) from the thought category, using participant as a random intercept. To test whether patterns were the same across reading and film comprehension, we added task as an interaction term. This led to convergence issues, so we repeated the analysis for each task separately. In a preliminary analysis, we ascertained that order (text or film comprehension first) did not affect the likelihood of a thought being triggered (OR = 0.867, SE  = 1.37, p  = 0.652), so we did not distinguish between orders here.

We found that the likelihood of a thought being triggered differed across thought categories (main effects: Wald χ 2 (9) = 90.2, p  < 0.001 for text comprehension, Wald χ 2 (4) = 47.9, p  < 0.001 for film comprehension) (Fig.  2 ). Planned comparisons (estimated marginal means; Lenth, 2018 ) indicated that semantic memories were indeed more likely to be triggered than prospection (text/film: OR = 15.4/18.3, SE  = 6.34/11.7, both p  < 0.001) and introspection (text/film: OR = 13.6/14.6, SE  = 5.64/8.92, both p  < 0.001). We found the same pattern for autobiographical memories (text/film: OR = 8.85/12.3, SE  = 3.31/7.67, both p  < 0.001 for introspection; OR = 7.83/9.82, SE  = 3.00/6.12, both p  < 0.001 for prospection). Unspecific memories followed a similar pattern, but differences were only significant for the text condition, likely because there were only 11 unspecific memories for the film condition (text/film: OR = 5.38/5.93, SE  = 2.61/6.07, p  < 0.001/ p  = 0.082 for introspection; OR = 4.76/4.73, SE  = 2.33/4.66, p  = 0.002/ p  = 0.115 for prospection). In separate analyses for each thought category, we ascertained that the likelihood of a thought being triggered did not vary as a function of task for any of these categories (all p values > 0.360). Together, these findings suggest that memories might arise from processing semantically rich information, whereas prospection and introspection might primarily (albeit not exclusively) be driven more by internal factors.

Proportion of triggered thoughts per category

Relationship between thoughts and triggers

We then asked how thought content is related to the reported triggers. We hypothesized that the relationship between thoughts and triggers should be stronger for thoughts associated with memory retrieval compared to thoughts that primarily arise from current concerns, such as prospection or introspection. We used an open source implementation (Olney, 2009 ) of LSA (Landauer et al., 1998 ) – a computational technique to measure the semantic similarity between two texts based on a reference semantic space – to obtain a measure of semantic overlap between each thought and its trigger. We used the Touchstone Applied Science Associates (TASA) corpus (with 300 dimensions with log entropy weighting) for the semantic space. We then used linear mixed-effects regression to test whether semantic overlap differed across content categories and tasks, controlling for the number of words in the thoughts and triggers. Participant identity was added to the model as a random intercept. We excluded thoughts pertaining to environmental distractors as there were insufficient instances for modeling ( N  = 2). In a separate analysis, we found that task order did not affect semantic overlap (Wald χ 2 (1) = 0.008, p  = 0.930) and was therefore not included in these analyses.

Semantic overlap between thoughts and triggers varied marginally across content categories (Wald χ 2 (8) = 15.0, p  = 0.059) and significantly across tasks (Wald χ 2 (1) = 4.20, p  = 0.040) (Fig.  3 ). Overlap was on average higher for thoughts reported while watching the film ( M  = 0.225, SD  = 0.203) than during text comprehension ( M  = 0.172, SD  = 0.185). There was no interaction between task and content (Wald χ 2 (8) = 10.3, p  = 0.243). Planned comparisons between the memory-related categories, prospection, and introspection revealed that semantic memories were more similar to their triggers than prospective thoughts ( b  = 0.079, SE  = 0.032, p  = 0.014) and introspection ( b  = 0.069, SE  = 0.031, p  = 0.024). Unspecific memories displayed a similar pattern ( b  = 0.123, SE  = 0.047, p  = 0.010 compared to prospection and b  = 0.113, SE  = 0.046, p  = 0.014 compared to introspection). Autobiographical memories did not differ significantly from either category ( b  = 0.039, SE  = 0.032, p  = 0.232 for prospection and b  = 0.029, SE  = 0.031, p  = 0.353 for introspection). As an additional check, we confirmed that thoughts about the stimulus displayed a strong relationship with their triggers compared to thoughts about prospection and introspection ( b  = 0.084, SE  = 0.036, p  = 0.022 for prospection and b  = 0.074, SE  = 0.034, p  = 0.031 for introspection).

Average LSA score for the relationship between the thoughts and triggers ( dark gray ) and thoughts and random triggers ( light gray ) at the thought-trigger level

If the content of thoughts is driven and constrained by the stimulus that triggered them, then breaking those links should lead to significantly weaker relationships. To test this hypothesis, we randomly shuffled the thoughts within each participant’s reports for a category to obtain a measure of overlap between a trigger and a randomly shuffled thought from the same category while accounting for the individual’s “thought space” (thoughts in that category for a given participant). Participants who only reported one trigger–thought pair for a category were excluded from the analysis for that category shuffling was not feasible. We tested whether LSA overlap differed significantly between the actual and the shuffled thoughts using the linear mixed-effects modeling approach above. We found that, as expected, the relationship between thoughts and shuffled triggers was significantly weaker for semantic, autobiographical, and unspecific memories (Wald χ 2 (1) = 19.7, p  < 0.001, Wald χ 2 (1) = 5.92, p  = 0.015, and Wald χ 2 (1) = 5.91, p  = 0.015, respectively) and for thoughts about the stimulus (Wald χ 2 (1) = 29.4, p  < 0.001) but not for introspection (Wald χ 2 (1) = 1.07, p  = 0.301), and prospection (Wald χ 2 (1) = 0.311, p  = 0.577). Thus, the results suggest that the relationship between thoughts and their triggers is meaningful for memory-related thoughts compared to introspective and prospective thoughts.

Train of thoughts

If thoughts are driven by associations, then the content of one thought might trigger another (i.e. the experience of a train of thought ) within the same mind wandering episode. For example, consider the following thought train from a participant: “beach nearby me at home that I always go to” → “my job as a beach tagger during high school” → “a guy that I used to like.” Or: “Louvre TRIGGER ” → “the Louvre” → “haha last time I was in the Louvre I threw up in front of the Mona Lisa” → “I wonder how strange the people looking at this data will think I am” → “Maybe I should have admitted this after all.”

In our sample, 217 thoughts were followed by (at least) a second thought. For these instances, we computed the LSA score between the first and second thought; there were insufficient data to go to the third thought and beyond. We compared these scores to random surrogates obtained by pairing the same first thought with a second thought from a randomly selected episode from the same participant (e.g. the thought “beach nearby me at home that I always go to” would be paired with “my friend’s parents being here last weekend”). Participants who only reported one mind wandering episode were excluded from the analysis as thought pairs could not be shuffled. Using the linear mixed-effects modeling approach above (ignoring content category due to sample size), we found that the relationship between consecutive thoughts ( M  = 0.422, SD  = 0.255) was stronger than between shuffled thoughts ( M  = 0.367, SD  = 0.244); (Wald χ 2 (1) = 4.89, p  = 0.027). There was no significant interaction with task ( p  = 0.513), suggesting that the relationship between consecutive thoughts was stronger than between random thoughts for both reading and film comprehension. These findings suggest that the content of one thought can trigger another related thought to produce semantically related thought trains.

Our aim was to test the idea that the task stimulus itself might trigger certain types of mind wandering. We found that during real-world, semantically rich, reading and film comprehension tasks, memories (pooled across autobiographical, semantic, and unspecific memories) were almost twice as frequent as prospective (and also introspective) thoughts. Furthermore, approximately half of the mind wandering thoughts were triggered from the stimulus, a conservative estimate which relies on participants recalling the trigger (see below). Thoughts pertaining to memories were more likely to be triggered from the stimulus than prospective and introspective thoughts. Importantly the content of the semantic and unspecific memories was more strongly semantically related to their reported triggers than prospective and introspective thoughts, suggesting that the stimulus can drive and constrain the content of mind wandering that arises from memory associations.

The pattern of thought content observed here differs from many laboratory studies that have found that mind wandering thoughts tend to be focused on the future (e.g. Baird et al., 2011 ; McVay & Kane, 2013 ; Smallwood et al., 2009 ). The high prevalence of memories across both tasks, combined with the fact that we found no differences in the frequencies of prospective and memory-related thoughts across tasks, supports the idea that the content of mind wandering varies as a function of whether a task requires processing semantically rich information. Thus, the prospective bias observed in task contexts that are relatively devoid of semantic content might not generalize to real-world semantically rich tasks like those studied here, but would apply to other real-world tasks, such as vigilance tasks (Giambra, 1993 ).

Our findings also differ from experience sampling studies (e.g. asking people to report their thoughts throughout the day) which have suggested that much of mind wandering is future related (Song & Wang, 2012 ; Spronken, Holland, Figner, & Dijksterhuis, 2016 ). However, the relatively uncontrolled nature of these studies makes it difficult to investigate the relationship between what a person is doing and thinking as the data lack contextual detail and temporal precision. Based on our study, we would predict that mind wandering thoughts would more likely consist of memory-based retrievals when people engage in semantically richer activities like reading the newspaper or watching television, whereas prospection would be more frequent during more repetitive task like doing the dishes or vacuuming.

We do not claim that prospection is by definition stimulus-unrelated as around 30% of prospective thoughts were triggered by the stimulus content. This finding aligns with previous studies that have shown that cueing a person’s current concerns, for instance by asking them to make a to-do list (Kopp et al., 2015a ) or read words that are related to the current concerns (McVay & Kane, 2013 ) can increase mind wandering. Examples from our data support this conclusion – we observed that “the passage [..] continuously talking about math” can trigger thoughts about “my math test at 11:20,” and seeing the letters “ AB carved in the brick wall” can lead to “thinking about plans for the weekend” with a “friend [whose] initials are AB .” The current study suggests that stimulus processing can give rise to prospection if its content is related to the person’s concurrent goals. This also resonates with findings from the prospective memory literature, which suggest that cues that are related to a prospective memory (e.g. an unfulfilled task) may reflexively trigger spontaneous retrieval of that task or goal (McDaniel et al., 2004 ; Scullin, McDaniel, Shelton, & Lee, 2010 ).

A distinction based on whether a thought is directly triggered by the stimulus somewhat overlaps with the distinction between “stimulus-independent and task-unrelated thoughts” and “task-related interferences” (Frank, Nara, Zavagnin, Touron, & Kane, 2015 ; Stawarczyk et al., 2011 ; Zavagnin, Borella, & De Beni, 2014 ). However, our findings suggest that stimulus-dependence and task-relatedness are distinct dimensions in semantically rich task contexts. In particular, a task-unrelated thought (e.g. thinking about homework while reading a text) can be stimulus-dependent (e.g. triggered by the text) or stimulus-independent (e.g. arising from salient internal concerns). Similarly, task-related interferences can be more (e.g. wondering how long the passage of text would be) or less (e.g. wondering how many minutes have passed) stimulus-dependent. Although it might seem counterintuitive, thoughts can also be stimulus-independent, yet task-dependent. For example, reading a text on cell biology can lead one to deliberate on a previously studied genetics text – here the stimuli are different but the thought space is conceptually connected and such integration lies at the heart of deep learning (McNamara, Oreilly, & Vega, 2012 ).

The finding that mind wandering is to some extent driven by stimulus context prompts a definition that captures this quality. As we have shown here, defining mind wandering as “stimulus-unrelated thought” (Smallwood & Schooler, 2015 ) misses an important part of the phenomenon. Stimulus processing gives rise to mind wandering through spontaneous associations, which are relatively constrained by the semantic relationship with stimulus-based triggers. This is in line with the idea that mind wandering is a type of spontaneous thought that is relatively unconstrained by cognitive control (although some mind wandering might be intentional; see Seli, Risko, & Smilek, 2016 ), but varies in how strongly it is constrained by sensory (and affective) salience (Christoff et al., 2016 ).

We also observed that consecutive thoughts were more strongly semantically related than random thoughts sampled from an individual’s thought space, suggesting that the content of one thought triggers and constrains the next. It is also possible that another source (e.g. a potentially unreported thought or trigger) triggered both of them somewhat independently. Thoughts might also become more loosely associated over time, as one would expect during generation of new mental content (e.g. creative thinking) (Mills, Herrera-Bennett, Faber, & Christoff, 2018 ) but we could not analyze thought trains beyond the second thought due to a limited amount of data. Further research could shed light on how trains of thought unfold, elucidating the underlying principles of how the content of mind wandering arises.

It is important to consider some caveats with the present study. Because we focused on the self-caught method of reporting, mind wandering instances that did not reach meta-cognitive awareness might have been missed (Smallwood & Schooler, 2006 ). However, a benefit is that reports can occur at any time, independent of whether and when a participant received a thought probe, which is the more common way to track mind wandering (Giambra, 1995 ; Schooler et al., 2004 ). This is important for the purpose of the present study as it aims to elucidate the relationship between mind wandering and stimulus content without being limited to specific probe locations. An open question pertains to the systematic relationships between awareness of mind wandering, its content, and associated triggers and how the task context modulates these relationships. Although these aspects are beyond the scope of the present paper, establishing how internal and external triggers interact and compete to influence mind wandering and meta-cognitive awareness is an important step towards understanding the dynamics of spontaneous thought.

People also require awareness of the mind wandering triggers. Previous work has suggested that people might not actually be aware of how a stimulus influences their behavior but will still report a relationship when asked (Nisbett & Wilson, 1977 ), suggesting that people might infer a relationship based on causal theories or expectations. Therefore, it is possible that some triggers were inferred rather than remembered. That said, we aimed to avoid this by making reporting of the trigger voluntary. Specifically, we ask participants whether there was “something in the [text/video] that triggered this thought,” and explicitly gave instruction for both options (“If yes then please describe what it was and if no then leave blank”). The fact that participants reported triggers for around half of the thoughts suggests that they indeed did not feel compelled to report a trigger for every thought. Furthermore, the ~ 10% of the thought–trigger pairs for which thoughts and triggers were (inadvertently) reported together provide some insight into the validity of reported triggers. Examples suggest that these thoughts are triggered by a specific aspect of the stimuli, rather than post-hoc inference of the relationship: “The word “good” reminded me of my philosophy homework I haven’t finished,” “The boy was headed somewhere with his briefcase and it reminded me of what I have to do,” “I read the words “don’t know” and it reminded me of Socrates basically saying we don’t know wisdom, only God does.” Furthermore, there is also the possibility that it might be easier to remember and therefore report semantically associated triggers. These are known limits of verbal protocols and we are unaware of any alternative to obtain the contents of consciousness.

In addition, our self-caught approach ostensibly requires participants to divide attention between the primary task (reading text/watching film) and thought monitoring. It is possible that due to the demanding nature of the primary tasks, participants missed some instances of mind wandering. It is also possible that the extraneous load of simultaneous thought monitoring influenced how deeply participants processed the text or film. If processing was shallow, the frequency of associations triggered by the stimulus might be relatively low compared to when participants focus only on reading or watching the film. It might also be the case that constant thought monitoring resulted in an on average earlier termination of trains of thoughts. If the mind wanders further away from the stimulus as the train of thoughts continues, then our sample might be biased towards thoughts that are more closely related to the stimulus. Further research could shed light on these questions by exploring the relationship between thought content and the stimulus in a probe-caught or retrospective paradigm, although each has its limitations with regard to sampling frequency, probe placement, and the veracity of memories.

Conclusions

In sum, our findings suggest that an analysis of mind wandering in semantically rich task contexts should account for multiple thought categories and associated triggers. Spontaneous associations that arise from stimulus processing are expected due to the associative nature of memory. These associations can be relevant to the task at hand and even enhance performance on the primary task as in the case of inference generation and creative ideation. However, as illustrated here, stimulus processing can also lead to retrieval of content that is irrelevant to the current task, such as memories, fantasies, and prospection. Importantly, the semantic-richness of the task context moderates (among other factors) the extent to which the stimulus activates different mind wandering thoughts. Semantically light environments should trigger a relatively high proportion of thoughts that arise from internal concerns, whereas semantically rich environments should trigger more stimulus-driven mind wandering. Whereas the present research has shown a higher propensity towards memory associations compared to thoughts arising from current concerns in semantically rich environments, further research is needed to make a more direct comparison between task contexts. We suggest that mind wandering research should move towards a comprehensive framework of when, why, and how the mind wanders when people engage in real-world tasks with varying degrees of semantic content.

We distinguish between the content of the task at hand (i.e. the stimulus) from the context in which it occurs (i.e. the environment). For example, if the task is to read a text in the lab, the stimulus is the text and the environment is the lab, the desk, etc.

Abbreviations

Latent semantic analysis

Baird, B., Smallwood, J., & Schooler, J. W. (2011). Back to the future: Autobiographical planning and the functionality of mind-wandering. Consciousness and Cognition , 20 (4), 1604–1611. https://doi.org/10.1016/j.concog.2011.08.007 .

Article   PubMed   Google Scholar  

Ball, C. T., & Little, J. C. (2006). A comparison of involuntary autobiographical memory retrievals. Applied Cognitive Psychology , 20 (9), 1167–1179. https://doi.org/10.1002/acp.1264 .

Article   Google Scholar  

Baumeister, R. F., Vohs, K. D., & Oettingen, G. (2016). Pragmatic prospection: How and why people think about the future. Review of General Psychology , 20 (1), 3–16. https://doi.org/10.1037/gpr0000060 .

Berntsen, D., & Hall, N. M. (2004). The episodic nature of involuntary autobiographical memories. Memory & Cognition , 32 (5), 789–803. https://doi.org/10.3758/BF03195869 .

Boys, C. V. (1890). Soap-bubbles, and the forces which mould them . Ithaca, NY: Cornell University Library Project Euclid.

Christoff, K., Irving, Z. C., Fox, K. C. R., Spreng, N., & Andrews-Hanna, J. R. (2016). Mind-wandering as spontaneous thought: A dynamic framework. Nature Reviews Neuroscience , 17 (11), 718–731. https://doi.org/10.1038/nrn.2016.113 . 

D’Argembeau, A., Renaud, O., & Van Der Linden, M. (2011). Frequency, characteristics and functions of future-oriented thoughts in daily life. Applied Cognitive Psychology , 25 (1), 96–103. https://doi.org/10.1002/acp.1647 .

Faber, M., & Mills, C. (2018). The critical role of the Hippocampus in mind wandering. The Journal of Neuroscience , 38 (29), 6439–6441. https://doi.org/10.1523/JNEUROSCI.0995-18.2018 .

Faber, M., Mills, C., Kopp, K., & D’Mello, S. (2017). The effect of disfluency on mind wandering during text comprehension. Psychonomic Bulletin and Review , 24 (3) 914-919. https://doi.org/10.3758/s13423-016-1153-z .

Faber, M., Radvansky, G. A., & D’Mello, S. K. (2018). Driven to distraction: A lack of change gives rise to mind wandering. Cognition , 173 , 133-137. https://doi.org/10.1016/j.cognition.2018.01.007 .

Frank, D. J., Nara, B., Zavagnin, M., Touron, D. R., & Kane, M. J. (2015). Validating older adults’ reports of less mind-wandering: an examination of eye movements and dispositional influences. Psychology and Aging , 30 (2), 266–278. https://doi.org/10.1037/pag0000031 .

Giambra, L. M. (1993). The influence of aging on spontaneous shifts of attention from external stimuli to the contents of consciousness. Experimental Gerontology , 28 (4–5), 485–492. https://doi.org/10.1016/0531-5565(93)90073-M .

Giambra, L. M. (1995). A laboratory method for investigating influences on switching attention to task-unrelated imagery and thought. Consciousness and Cognition , 4 (1), 1–21. https://doi.org/10.1006/ccog.1995.1001 .

Hintzman, D. L. (2011). Research strategy in the study of memory: Fads, fallacies, and the search for the “coordinates of truth”. Perspectives on Psychological Science , 6 (3), 253–271. https://doi.org/10.1177/1745691611406924 .

Killingsworth, M. A., & Gilbert, D. T. (2010). A wandering mind is an unhappy mind. Science , 330 (6006), 932–932. https://doi.org/10.1126/science.1192439 .

Klinger, E. (1978). Modes of normal conscious flow. In K. Pope (Ed.) The stream of consciousness . (pp. 225–258). Boston: Springer.

Chapter   Google Scholar  

Klinger, E. (1999). Thought flow: Properties and mechanisms underlying shifts in content. In J. A. Singer & P. Salovey (Eds.), At play in the fields of consciousness: Essays in honor of Jerome L. Singer (pp. 29–50). Mahwah: Lawrence Erlbaum Associates Publishers.

Klinger, E. (2013). Goal commitments and the content of thoughts and dreams: Basic principles. Frontiers in Psychology , 4 , 415. https://doi.org/10.3389/fpsyg.2013.00415

Kopp, K., D’Mello, S., & Mills, C. (2015a). Influencing the occurrence of mind wandering while reading. Consciousness and Cognition , 34 , 52–62. https://doi.org/10.1016/j.concog.2015.03.003 .

Kopp, K., Mills, C., & D’Mello, S. (2015b). Mind wandering during film comprehension: The role of prior knowledge and situational interest. Psychonomic Bulletin & Review , 23 (3), 842-848. https://doi.org/10.3758/s13423-015-0936-y

Krasich, K., McManus, B., Hutt, S., Faber, M., D’Mello, S. K., & Brockmole, J. R. (2018). Gaze-Based Signatures of Mind Wandering During Real-World Scene Processing. Journal of Experimental Psychology: General , 147 (8), 1111-1124. https://doi.org/10.1037/xge0000411

Krawietz, S. A., Tamplin, A. K., & Radvansky, G. A. (2012). Aging and mind wandering during text comprehension. Psychology and Aging , 27 (4), 951–958. https://doi.org/10.1037/a0028831 .

Kvavilashvili, L., & Mandler, G. (2004). Out of one’s mind: A study of involuntary semantic memories. Cognitive Psychology , 48 (1), 47-94. https://doi.org/10.1016/S0010-0285(03)00115-4

Lamorisse, A. (1956). The Red Balloon . New York, NY, US: Doubleday Books for Young Readers.

Landauer, T. K., Folt, P. W., & Laham, D. (1998). An introduction to latent semantic analysis. Discourse Processes , 25 (2), 259–284. https://doi.org/10.1080/01638539809545028 .

Lenth, R. (2018). Emmeans: Estimated marginal means, aka least-squares means. R Package Version 1.2.

Mace, J. H. (2005). Priming involuntary autobiographical memories. Memory , 13 (8), 874–884. https://doi.org/10.1080/09658210444000485 .

Masicampo, E. J., & Baumeister, R. F. (2011). Consider it done! Plan making can eliminate the cognitive effects of unfulfilled goals. Journal of Personality and Social Psychology , 101 (4), 667–683. https://doi.org/10.1037/a0024192 .

McDaniel, M. A., Einstein, G. O., Guynn, M. J., & Breneiser, J. (2004). Cue-focused and reflexive-associative processes in prospective memory retrieval. Journal of Experimental Psychology: Learning Memory and Cognition , 30 (3), 605–614. https://doi.org/10.1037/0278-7393.30.3.605 .

McNamara, D. S., Oreilly, T., & De Vega, M. (2012). Comprehension skill, inference making, and the role of knowledge. In F. Schmalhofer & C.A. Perfetti (Eds.) Higher Level Language Processes in the Brain: Inference and Comprehension Processes , (pp. 233–254). London, UK: Psychology Press. https://doi.org/10.4324/9780203936443

Book   Google Scholar  

McVay, J. C., & Kane, M. J. (2013). Dispatching the wandering mind? Toward a laboratory method for cuing “spontaneous” off-task thought. Frontiers in Psychology , 4 , 570. https://doi.org/10.3389/fpsyg.2013.00570

Miles, A. N., & Berntsen, D. (2015). The forgotten remindings: Personal remindings examined through self-probed retrospection during reading and writing. Consciousness and Cognition , 33 , 67–77. https://doi.org/10.1016/j.concog.2014.11.010 .

Mills, C., Herrera-Bennett, A., Faber, M., & Christoff, K. (2018). Why the mind wanders: How spontaneous Thought’s default variability may support episodic efficiency and semantic optimization. In K. Christoff, & K. C. R. Fox (Eds.), The Oxford handbook of spontaneous thought: Mind-wandering, creativity, and dreaming . Oxford, UK: Oxford University Press. https://doi.org/10.1093/oxfordhb/9780190464745.013.42 .

Nisbett, R. E., & Wilson, T. D. (1977). Telling more than we can know: Verbal report on mental process. Psychological Review , 84 (3), 231–259.

Olney, A. M. (2009). GnuTutor: An open source intelligent tutoring system based on AutoTutor. In AAAI Fall Symposium - Technical Report , (vol. FS-09-02, pp. 70–75).

Google Scholar  

Plimpton, B., Patel, P., & Kvavilashvili, L. (2015). Role of triggers and dysphoria in mind-wandering about past, present and future: A laboratory study. Consciousness and Cognition , 33 , 261–276. https://doi.org/10.1016/j.concog.2015.01.014 .

Randall, J. G., Oswald, F. L., & Beier, M. E. (2014). Mind-wandering, cognition, and performance: A theory-driven meta-analysis of attention regulation. Psychological Bulletin , 140 (6), 1411–1431. https://doi.org/10.1037/a0037428 .

Rummel, J., & Nied, L. (2017). Do drives drive the train of thought?—Effects of hunger and sexual arousal on mind-wandering behavior. Consciousness and Cognition , 55 , 179–187. https://doi.org/10.1016/j.concog.2017.08.013 .

Sarason, I. G., Sarason, B. R., Keefe, D. E., Hayes, B. E., & Shearin, E. N. (1986). Cognitive interference: Situational determinants and traitlike characteristics. Journal of Personality and Social Psychology , 51 (1), 215–226. https://doi.org/10.1037/0022-3514.51.1.215 .

Schank, R. C. (1983). Dynamic memory: A theory of reminding and learning in computers and people . New York: Cambridge University Press.

Schooler, J. W., Reichle, E. D., & Halpern, D. V. (2004). Zoning Out While Reading: Evidence for Dissociations Between Experience and Metaconsciousness. In D. T. Levin (Ed.), Thinking and Seeing: Visual Metacognition in Adults and Children , (pp. 203–226). Cambridge, MA: MIT Press.

Scullin, M. K., McDaniel, M. A., Shelton, J. T., & Lee, J. H. (2010). Focal/nonfocal cue effects in prospective memory: Monitoring difficulty or different retrieval processes? Journal of Experimental Psychology: Learning Memory and Cognition , 36 (3), 736–749. https://doi.org/10.1037/a0018971 .

Seilman, U., & Larsen, S. F. (1989). Personal resonance to literature: A study of remindings while reading. Poetics , 18 (1–2), 165–177. https://doi.org/10.1016/0304-422X(89)90027-2 .

Seli, P., Risko, E. F., & Smilek, D. (2016). On the necessity of distinguishing between unintentional and intentional mind wandering. Psychological Science , 27 (5), 685–691. https://doi.org/10.1177/0956797616634068 .

Smallwood, J., Karapanagiotidis, T., Ruby, F., Medea, B., De Caso, I., Konishi, M., … Jefferies, E. (2016). Representing representation: Integration between the temporal lobe and the posterior cingulate influences the content and form of spontaneous thought. PLoS One , 11 (4). https://doi.org/10.1371/journal.pone.0152272 .

Smallwood, J., Nind, L., & O’Connor, R. C. (2009). When is your head at? An exploration of the factors associated with the temporal focus of the wandering mind. Consciousness and Cognition , 18 (1), 118–125. https://doi.org/10.1016/j.concog.2008.11.004 .

Smallwood, J., Obonsawin, M., & Heim, D. (2003). Task unrelated thought: The role of distributed processing. Consciousness and Cognition , 12 (2), 169–189. https://doi.org/10.1016/S1053-8100(02)00003-X .

Smallwood, J., & Schooler, J. W. (2006). The restless mind. Psychological Bulletin , 132 (6), 946–958. https://doi.org/10.1037/0033-2909.132.6.946 .

Smallwood, J., & Schooler, J. W. (2015). The science of mind wandering: Empirically navigating the stream of consciousness. Annual Review of Psychology , 66 (1), 487–518. https://doi.org/10.1146/annurev-psych-010814-015331 .

Song, X., & Wang, X. (2012). Mind wandering in Chinese daily lives--an experience sampling study. PLoS One , 7 (9), e44423. https://doi.org/10.1371/journal.pone.0044423 .

Article   PubMed Central   PubMed   Google Scholar  

Spronken, M., Holland, R. W., Figner, B., & Dijksterhuis, A. (2016). Temporal focus, temporal distance, and mind-wandering valence: Results from an experience sampling and an experimental study. Consciousness and Cognition , 41 , 104–118. https://doi.org/10.1016/j.concog.2016.02.004 .

Stawarczyk, D., Cassol, H., & D’Argembeau, A. (2013a). Phenomenology of future-oriented mind-wandering episodes. Frontiers in Psychology , 4 , 425. https://doi.org/10.3389/fpsyg.2013.00425

Stawarczyk, D., Majerus, S., & D’Argembeau, A. (2013b). Concern-induced negative affect is associated with the occurrence and content of mind-wandering. Consciousness and Cognition , 22 (2), 442–448. https://doi.org/10.1016/j.concog.2013.01.012 .

Stawarczyk, D., Majerus, S., Maj, M., Van der Linden, M., & D’Argembeau, A. (2011). Mind-wandering: Phenomenology and function as assessed with a novel experience sampling method. Acta Psychologica , 136 (3), 370–381. https://doi.org/10.1016/j.actpsy.2011.01.002 .

Zavagnin, M., Borella, E., & De Beni, R. (2014). When the mind wanders: Age-related differences between young and older adults. Acta Psychologica , 145 (1), 54–64. https://doi.org/10.1016/j.actpsy.2013.10.016 .

Download references

Acknowledgements

We would like to thank Dr. Arthur Graesser for invaluable discussions that led to this study. We also thank past and current members of the Emotive Computing Lab for their assistance in data collection and rubric development.

This research was supported by the National Science Foundation (NSF) (DRL 1235958 and IIS 1523091). Any opinions, findings and conclusions, or recommendations expressed in this paper are those of the authors and do not necessarily reflect the views of NSF.

Availability of data and materials

To maintain confidentiality of our participants, we will not share the full dataset that contains verbatim thought content. The content-coded dataset supporting the conclusions of this article will be made available to any interested individuals upon request.

Author information

Authors and affiliations.

Department of Psychology, University of Notre Dame, Notre Dame, IN, 46556, USA

Myrthe Faber & Sidney K. D’Mello

Donders Institute for Brain, Cognition and Behaviour, Centre for Cognitive Neuroimaging, Radboud University, Nijmegen, The Netherlands

Myrthe Faber

Department of Cognitive Neuroscience, Radboud University Medical Centre, Nijmegen, The Netherlands

Institute of Cognitive Science, University of Colorado Boulder, Boulder, CO, 80309, USA

Sidney K. D’Mello

Department of Computer Science, University of Colorado Boulder, Boulder, CO, 80309, USA

You can also search for this author in PubMed   Google Scholar

Contributions

Concept of experiment: SDM. Category coding rubric development: MF and SDM. Data analysis: MF and SDM. Preparation of manuscript: MF and SDM. Both authors read and approved the final manuscript.

Corresponding author

Correspondence to Myrthe Faber .

Ethics declarations

Ethics approval and consent to participate.

Before the study, participants read and signed an agreement to participate. They were informed that they were free to withdraw at any time. All materials and forms were approved by the Institutional Review Board for the Protection of Human Subjects of both universities.

Consent for publication

Participants read and signed a data release form permitting the use of their data for publication as examples of data output. This form was voluntary. All materials and forms were approved by the Institutional Review Board for the Protection of Human Subjects of both universities.

Competing interests

The authors declare that they have no competing interests.

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Mind wandering content coding rubric

Overview of mind wandering content coding rubric

Thoughts pertaining to the immediate environment

Environmental distractions.

Thoughts pertaining to the immediate environment decoupled from the task. This includes thoughts about:

Time, e.g. “I wonder what time it is.”

Temperature and weather, e.g. “It is warm in this room,” “The weather is so gloomy.”

Room, building, and items in it, e.g. “These eye sensors blink a lot,” “The fact that there are no windows in here.”

External distractors, e.g. “Something making a faint sound near the desk,” “I hear bagpipes.”

Task-related interferences

Thoughts related to the task itself (rather than to the stimulus). This includes thoughts about:

The experiment, e.g. “Why the experimenters would choose this movie,” “What zoning out really is and how I can tell if I’m doing it.”

Progression of the experiment, e.g. “I was wondering what the next part of the study is,” “I wondered how much of the text I would have to read,” “Wondering how much time is left.”

Apparatus, e.g. “The eye tracking device,” “What does this wrist band do.”

Thoughts related to the content or presentation of the text/film, but not associations going beyond the content. This includes thoughts about:

The text/film content, e.g. “I wish this balloon made sound or sound or something,” “The name Simple Simon is ridiculous,” “The fakeness of the balloon,” “Whether the balloon is real or picture animated.”

The text/film presentation, e.g. “These sentences are abnormally long,” “It is strange watching a video with this aspect ratio.”

Thoughts, reflections, and inferences related to the immediate content of the film, e.g. “What is so special about this red balloon,” “I hope his balloon pops.”

Thoughts pertaining to memories

Semantic memory.

General knowledge (including facts, meaning, and concepts), e.g. “Math and physics formulae different equations math problems,” “Uptown Girls movie where the little girl gets a pet pig from her nanny,” “Song lyrics.”

Autobiographical memory

These are thoughts about events that are (1) related to the person themselves and (2) have actually happened, e.g. “When I rode on a bus in Colorado,” “Painting as a child.” They are anchored in time (and space). In some cases, the “when” can be inferred (e.g. “My high school bio class”).

Unspecific memory

Thoughts that require retrieval of information not immediately relevant to prospection/current concerns, but that do not meet the requirements for autobiographical memory, e.g. “My dog,” “Thinking about my grandpa.”

Introspection

Thoughts related to meta-cognition or feelings or thoughts regarding the self. This includes thoughts about:

Bodily feelings, e.g. “How tired I am,” “My knee itches,” “I am sleepy,” “Getting really hungry.”

Emotional states, e.g. “Kind of confused,” “Frustrated,” “Worrying about my exams.”

Mental states and reflections, e.g. “I miss my dog,” “It’s been a long week.”

Reflections on task performance, e.g. “I am thinking about how difficult it is to focus on this page,” “I read really slow[ly],” “How I need to focus on reading,” “How much I zone out while studying,” “I’m not absorbing anything.”

Prospection

Thoughts about the future, including current concerns. This includes thoughts about:

Future plans, e.g. “My plans for the weekend,” “Assignments due tomorrow,” “Practice later,” “Where I’m going for dinner.”

Prospective memory, e.g. “I need to put gas in my truck,” “Having to respond to a text telling work I did the laundry today,” “I need to get my nails done.”

Planning, e.g. “Goals for the future,” “If I’m going to run and shower before [dinner] or wait until later and just run on the treadmill,” “Trying to make a plan of what I am going to do tonight.”

Current concerns, e.g. “The word good reminded me of my philosophy homework I have not finished.”

Fantasies and counterfactual situations, e.g. “What it would be like if there were a fire drill right now,” “I was imagining football players trying to stay on top of their school work,” “What would have happened if I had crossed the street and got hit. I would have had to tell someone I had Medex.”

Thoughts that are none of the above. This includes thoughts about:

No content, e.g. “I just started thinking,” “Nothing in particular,” “Do not remember.”

Sleep, e.g. “Sleeping,” “A nap,” “Napping.”

Unclear whether prospection or memory, or, e.g. “Traveling to India,” “Going to work,” “Soccer practice.”

Unclear content, e.g. “Bese.”

Too generic, e.g. “Lunch,” “Eating,” “Starbucks Coffee.”

Not mind wandering

Thoughts that are immediately relevant to understanding the text or film: e.g. “How capillary action in plants carry water up the stem. I have an image in my head of water going up the stem of a plant,” “Hydrogen bonds and the chemical properties of water,” “Predicting a conclusion of water’s cohesive properties,” “How this has nothing to do with bubbles right now. Or maybe it does, but I can’t make the connection,” “Is letting go of the balloon and getting on the bus symbolic of growing? Of losing color in adulthood?”, “Is this balloon supposed to be a metaphor for something?”, “Trying to figure out the plot of this movie.” This includes wondering what a word means (e.g. “A water butt? I have never heard of that before,” “What’s a tumbler?”, “What duckweed looks like”) or what is on the screen (e.g. “I can’t tell what kind of animal that is”) and reflections on what was on the screen (e.g. “Yay! He got his balloon back!”, “Words!” [in response to the boy speaking in the movie]), as this is on-task behavior.

Rights and permissions

Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License ( http://creativecommons.org/licenses/by/4.0/ ), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.

Reprints and permissions

About this article

Cite this article.

Faber, M., D’Mello, S.K. How the stimulus influences mind wandering in semantically rich task contexts. Cogn. Research 3 , 35 (2018). https://doi.org/10.1186/s41235-018-0129-0

Download citation

Received : 15 January 2018

Accepted : 16 August 2018

Published : 26 September 2018

DOI : https://doi.org/10.1186/s41235-018-0129-0

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

• Mind wandering
• Comprehension

Concurrent prospective memory task increases mind wandering during online reading for difficult but not easy texts

• Published: 01 March 2022
• Volume 51 , pages 221–233, ( 2023 )

Cite this article

• Teresa Schurer 1 ,
• Bertram Opitz 1 , 2 &
• Torsten Schubert 1 , 3  

1811 Accesses

1 Altmetric

Explore all metrics

Many prior theories have tried to explain the relationship between attentional processes and mind wandering. The resource-demand matching view argues that a mismatch between task demands and resources led to more mind wandering. This study aims to test this view against competing models by inducing mind wandering through increasing the level of demands via adding a prospective memory task to cognitively demanding tasks like reading. We hypothesized that participants with a second task still in mind (unfinished group) engage more in task-unrelated thoughts (TUTs) and show less text comprehension compared to participants who think a second task is finished (finished group). Seventy-two participants had to study 24 items of a to-do list for a recall test. After a first cued recall of ten items, participants were either told that a second task was finished or that the recall was interrupted and continued later. All participants then started reading an easy or difficult version of the same unfamiliar hypertext, while being thought probed. Text comprehension measures followed. As expected, participants in the unfinished group showed significantly more TUTs than participants in the finished group when reading difficult texts, but, contrary to our assumptions, did not show better text comprehension measures when reading difficult text. Nevertheless, participants compensate for the influence of the second task by reading longer, which in turn has a positive effect on their reading knowledge. These findings support the resource-demand-matching model and thus strengthen assumptions about the processing of attention during reading.

Similar content being viewed by others

The effect of disfluency on mind wandering during text comprehension.

The relationship between mind wandering and reading comprehension: A meta-analysis

The impact of mind wandering on the recall of central ideas

Avoid common mistakes on your manuscript.

Introduction

Mind wandering is a pervasive daily phenomenon of mental activity in which attention engages with thoughts that are unrelated to external demands (Smallwood, 2013 ), and plays a critical role for performance in many cognitive tasks (e.g., reading, listening to lectures, driving). For example, you can think of your shopping list when solving a text task or conducting other acitivies in parallel. In the past, different perspectives have been proposed to explain the relationship between attentional processes and mind wandering (McVay & Kane, 2010 ; Schurer et al., 2020 ; Smallwood & Schooler, 2006 ).

To begin with, Smallwood and Schooler ( 2006 ) proposed the resource-demand theory , which assumes that attention is distracted from the primary task to task-unrelated thoughts (TUTs) and therefore mind wandering consumes executive resources that compete with the main task. They argue further that individuals with high resources are less likely to exhibit impaired performance when experiencing mind wandering than those with low resources because they can easily allocate these resourcs between TUTs and task performance. In support, Smallwood et al. ( 2008 ) showed that readers exhibiting more mind wandering episodes during reading were less able to build situation models (integrated mental representations) from the text. Furthermore, a study by Feng et al. ( 2013 ) showed that attentional resources devoted to the successful construction of a situation model help to suppress TUTs. Feng et al. ( 2013 ) showed that difficult texts lead to problems in building a situation model and therefore bind fewer attentional resources, which in turn lead to mind wandering. This model implies that mind wandering occurs when it is too demanding or, looked at the other way around, more mind wandering leads to lower performance (e.g., to a low build-up of a situation model). However, this model does not really make any statements about the available resources, like differences in working memory capacity (WMC). As an extension of this view, the context regulation hypothesis (Smallwood & Andrews-Hanna, 2013 ) further suggests that the occurrence of mind wandering can be actively regulated depending on the task context. The relationship between psychological well-being in terms of mood and self-generated thoughts is argued to depend on an individual's ability to regulate the content of their thoughts (Smallwood & Andrews-Hanna, 2013 ). That is, in easy tasks more mind wandering is allowed because it is conducive to well-being and in difficult tasks it tends to be inhibited because it interferes with task processing. Context seems to play an important role for understanding different thinking patterns and it causes abilities such as WMC to help suppress thoughts outside of the task when a subject is motivated to focus on an ongoing task, which is also in accordance with recent neural evidence suggesting the context-dependent occurrence of mind wandering (Turnbull et al., 2019 ).

In contrast, the control failure x concerns hypothesis (McVay & Kane, 2010 ) postulates that executive control capabilities prevent mind wandering by keeping the attention on the primary task and suppressing interference from rather spontaneously occurring TUTs that are activated by environmental cues and do not consume executive resources. Consequently, individuals with greater executive attention abilities (i.e., with more WMC) will have more attentional control resources to stay focused on a task while suppressing TUTs (McVay & Kane, 2009 ). Individual differences in the propensity to mind wander are jointly determined by cognitive ability and by context, i.e., by the presence of personally salient concerns that interfere with task focus. A person's current concerns may therefore automatically activate off-task thoughts, and failure to maintain attention on a current task then leads to mind wandering. This correspondends to an additional cognitive load that, like the actual task, ties up resources and therefore leads to more TUTs as well as determines the content of these TUTs. In support of their view, McVay and Kane ( 2012a , 2012b ) showed that participants with greater WMC reported less mind-wandering episodes during reading tasks and showed better reading comprehension performance than participants with lower WMC. They proposed that with a greater WMC, participants are more able to adjust their attention to the task demands, whereas participants with a lower WMC were less able to create a mental model of the text. This was also supported and extended by the cognitive-flexibility hypothesis of Rummel and Boywitt ( 2014 ), according to which cognitive control is not only needed to inhibit TUTs when another task is performed. Instead, the authors proposed that the relationship between WMC and mind wandering is dependent on task demands: individuals with a greater WMC engage in TUTs when task demands are low and reduce TUTs when tasks demands are attention-demanding. People with a greater WMC tend to be more effective in their allocation of cognitive control abilities (Rummel & Boywitt, 2014 ) – they can better afford to let their mind wander from a task when attentional demands are low but are also more able to stay focused on a task when they need to. The control failure x concerns hypothesis suggests that more resources (e.g., WMC) lead to less mind wandering or, on the other hand, few free resources lead to more mind wandering. In addition, the cognitive-flexibility hypothesis states that individuals with more available resources are better able to regulate TUTs. However, this view is not explicit about individuals with low available resources.

Capitalizing on certain aspects to explain mind wandering, all the above views propose that mind wandering affects executive control but differ in the way mind wandering is assumed to be related to executive control. While the resource-demand theory (Smallwood & Schooler, 2006 ) assumes that mind wandering requires executive resources and therefore impedes task performance, the control failure x concerns hypothesis (McVay & Kane, 2010 ) and the cognitive-flexibility hypothesis (Rummel & Boywitt, 2014 ) consider mind wandering as a result of executive-control failures or adaptation processes. Both theories thus make opposing predictions and cannot comprehensively explain the findings that support the respective other theory. Crucially, each of these models is based on findings regarding different fractions of potential factors influencing mind wandering.

To better account for the existing data, we took a more comprehensive approach considering multiple factors influencing mind wandering like task demands and available cognitive resources, but also performance aspects, and proposed in a prior study an extension of the existing models, the so-called resource-demand-matching view (Schurer et al., 2020 ). We suggested that mind wandering occurs whenever the available cognitive resources of an individual (WMC, prior knowledge, etc.) do not match the task demands. We assume that cognitive resources are primariliy allocated to the main task first. Crucially, we assume mind wandering to be a spontaneously occurring but resource-demanding process that individuals may engage with if their cognitive resources exceed the current task demands or cannot avoid if task demands surpass the available resources. In other words, our resource-demand-matching view states that low availability of cognitive resources, for instance due to low WMC with high demands imposed by high text difficulty in a reading comprehension task leads to more mind wandering as does high availability of these cognitive resources exceeding the demand of an easy-to-comprehend text. Unlike previous models (McVay & Kane, 2010 ; Rummel & Boywitt, 2014 ; Smallwood et al., 2021 ; Smallwood & Schooler, 2006 ) assuming a more or less uniform relationhip between task demands and available cognitive resources to induce mind wandering, although individual differences in factors like WMC are assumed to moderate this relationship, our model predicts an interaction between the two.

To test the predictions made by different models, the present study increased the cognitive demands capitalising on findings demonstrating that unfulfilled tasks and goals remain persistent in the mind and, thus, present additional cognitive demands for ongoing task processing (Zeigarnik, 1927 ). In the original study (Zeigarnik, 1927 ), participants were asked to complete a series of separate, simple tasks like solving puzzles, making clay figures, or completing math problems. Half of the tasks were interrupted before the participants could complete them. Interestingly, participants recalled details of the interrupted tasks 90% better than details of uninterrupted tasks. It was also suggested that failing to finish a task leads to an underlying cognitive tension, which leads to more mental effort and rehearsal to keep focused on the task. Only when the task is finally finished does the mind let the effort go.

A former study included mind-wandering aspects in their investigations of the Zeigarnik effect (Steindorf & Rummel, 2017 ). The authors implemented this task as a prospective memory task (PM task). In their study, participants had to study and remember a grocery shopping-list for a prospective memory (PM task) test. After the first recall phase (half of the shopping list items) participants in a finished condition (FC) were told that they did not urgently need the remaining items, that the PM task was finished, and that they would now work on a different task (two-back task). In contrast, participants in an interrupted condition (IC) were told that they now must work on a different task and, therefore, need buy the remaining items later (i.e., to finish the PM task). Both groups of participants were thought probed during the two-back task to assess mind wandering and had then to complete the second recall phase. The group of participants interrupted during the recall of a shopping list showed more TUTs related to this secondary recall task during the performance of a two-back task than a group of participants who were told that the recall task is finished before commencing the two-back task. According to Steindorf and Rummel ( 2017 ), mind wandering can be controlled by second tasks in mind, which could be beneficial. This implies an additive effect and shows that mind wandering should not depend on the difficulty of the primary task. In contrast, Rummel et al. ( 2017 ) showed lower TUT rates when a PM intention (to respond to members of a presented semantic category) was embedded during an ongoing task (lexical decision task) as opposed to processing the ongoing task alone. In this case, PM requirements did not add cost to the ongoing task, but instead promoted thoughts to be focussed on the ongoing task. Taken together, these studies nicely demonstrate a strong connection between prospective memory requirements and mind wandering (see Kvavilashvili & Rummel, 2020 , for review).

As in our previous study (Schurer et al., 2020 ), we also manipulated text difficulty based on Kintsch and Van Dijk’s ( 1978 ) model of text comprehension allowing for a finer grained approach to disentangling the views introduced above. Crucially, text cohesion is an essential factor for successfully constructing the situation model from the text, and thus for text comprehension, which is why we manipulated text cohesion by increasing or decreasing text cohesion (see Schurer et al., 2020 ). Previous studies showed that a higher level of text cohesion improves text comprehension (Graesser & McNamara, 2011 ; McNamara et al., 1996 ; Schurer et al., 2020 ). We wanted to investigate the induction of mind wandering by a PM task in a more complex primary-task situation than Steindorf and Rummel ( 2017 ). Accordingly, we chose a similar task to tie up resources and combined it with easy and difficult texts. Thus, participants in the current study were first asked to study to-do list items for a following recall test. Participants were told either that the recall test was over or would be continued at a later time after ten items had been queried. After the first recall, the participants read a high- or a low-cohesive version of an expository text about the copyright law, with the goal to answer questions about the text after reading. We assessed the occurrence of mind wandering by presenting probes asking participants to indicate the occurrence of different types of thoughts during text reading. After text reading, participants answered reading comprehension questions about the text. Next, the second recall of another ten to-do-list items followed. Lastly, participants completed WMC measures.

In accordance with the assumption of our model (Schurer et al., 2020 ), we assume that an interaction of the PM task and text difficulty has an impact on TUTs and reading comprehension, making our model the only one that predicts the interaction. We predicted an interaction of text difficuty (easy vs. difficult) × PM task (finished vs. unfinished). It was hypothesized that participants having a second task in mind (prospective memory task; PM task) experience more TUTs than participants who think a second task is finished when reading difficult texts . We also predicted that participants show a worse text comprehension when they have a second task still in mind when reading difficult texts. Further, we assume that there are no differences between the two groups (unfinished, finished) with respect to WMC. According to our model, if the demands are below the available resources (see also Schurer et al., 2020 ), a matching of resources is more likely to occur, but if the demands are above the available resources, the equilibrium is shifted and more demands than resources are available. Therefore, in the difficult text condition, there should be a mismatch between resources and demands, and, based on our model, more mind wandering.

Participants

Participants were 72 students (49 female) across several courses at the Martin-Luther-University Halle-Wittenberg. We determined the size of our participants’ sample according to previous PM studies investigating mind wandering with total sample sizes from N = 55 (Steindorf & Rummel, 2017 ), N = 68 and N = 61 (Scullin et al., 2018 ), N = 73 (Masicampo & Baumeister, 2011 ) up to N = 104 (Rummel et al., 2017 ) participants. To achieve a comparable statistical power, we followed previous studies with comparable participant sample sizes (Rummel et al., 2017 ; Scullin et al., 2018 ; Steindorf & Rummel, 2017 ) for a comparison between two groups (see below for further details). Footnote 1

Detailed sample characteristics can be found in Table 1 . All participants were between 18 and 32 years old and were native speakers of German. The mean age of the participants was 23.86 years ( SD = 3.29). All particiants were pseudo-randomly assigned to the different groups/conditions to ensure equal group sizes and gender distributions. Thus, there were 36 participants each in the finished and the unfinished group. The experimental protocol conformed tenets of the Declaration of Helsinki and written informed consent was obtained from each participant before the commencement of the study. The study was approved by the ethics committee of the Deutsche Gesellschaft für Psychologie (DGPs). Participants received 12 Euro as compensation for their time.

Materials and procedure

We adopted the procedure of our previous study (Schurer et al., 2020 ) with some deviations. The entire study was conducted in a single 2.0-h session. Participants signed a consent form and provided demographic information including gender, age, study course, and semester. Then, they completed a short content knowledge test to assess their prior knowledge of the content domain, followed by a PM task. In the PM task, participants were presented with the following scenario: The semester break is due, and they are to complete 24 tasks that they have put on a to-do list. They should now remember the tasks of the to-do list. After the study phase, ten items of the to-do list were recalled. Half of the participants were informed that they had to move on to the next task and that another ten items of the to-do list would be recalled at a later point in time (unfinished group) or that the recall had ended, and they had to move on to the next task (finished group). Afterwards, participants received detailed instructions for the reading task and the thought-probing procedure. Then, they read a hypertext about the copyright law and were presented with thought probes while reading, answered reading comprehension questions based on the text, and took part in a memory test. Now the second recall of the remaining ten to-do list items followed for all participants. Finally, the participants completed two working memory tasks (Ospan and Rspan) to assess their working memory performance. Mind wandering was not collected in the WMC tasks because their duration was too short to generate meaningful statements about TUTs. Furthermore, WMC was only included as a covariate in the analyses and not as a whole experiment as previous studies (e.g., Mrazek et al., 2012 ).

Prospective memory task

In the PM task, participants studied 24 to-do list items; four of them were used as buffer items (two items as primacy, two items as recency buffer items during both recall phases). The items included typical activities that students engage in during their semester breaks (e.g., paying semester fees, creating a timetable for the upcoming semester). Participants were asked to learn the to-do list items in a random order for 2,000 ms each followed by a 500-ms inter-stimulus interval. Secondly, we created two recall phases consisting of ten test items each, which the participants had to recall in the two recall phases on a sheet of paper (the buffer items were excluded in the recall). All items had different first letters. The first two initial letters were presented and were provided as cues in the recall phases, which the participants then had to remember and complete. Recall performance was defined as the number of correctly recalled items for each recall phase.

Working memory capacity tasks

After some initial practice with the corresponding tasks, participants completed two complex span tasks (operation span and reading span) to assess individual working memory performance. A total memory span score was computed as the overall mean proportion correct responses from the Ospan and the Rspan task (see Schurer et al., 2020 , for details). WMC was not analyzed further in this study, but was nevertheless included in the analyses as a control variable.

Content knowledge test

To investigate their general knowledge about the copyright law, participants completed a paper-pencil content knowledge test about general copyright law aspects with a total of five single-choice questions. For each question, participants had to choose one answer out of four possible alternatives. The correct answers were added together to obtain a total score of prior content knowledge.

Hypertext reading

Participants read an expository hypertext about the copyright law (see Schurer et al., 2020 , for details). A version with high cohesion (easy condition) and low cohesion (difficult condition) was created. Both versions were identical to Schurer et al. ( 2020 ). We used the same cohesion manipulations at the local and global levels as McNamara et al. ( 1996 ). There was a total of 68 manipulations. On average, one to two manipulations appeared on a global level and about five on a local level in a text segment of 500 words. The length of the highly cohesive version was 4,870 words and the length of the low cohesive version was 4,620 words. The text differed in length (e.g., by removing headings or connectors), but not in text content. The average Flesch Reading Ease score was 35 in an easy and 38 in a difficult state, indicating a medium level of difficulty (Schöll, 2015 ). The text was displayed on a computer screen as several pages in black on a white background. One page contained about 500 words. Participants continued to the next page by clicking the "Next" button in the lower right corner of the screen. They could not return to read a page again after clicking on the next page but were given as much time as they needed to read the text. Participants were informed before reading that they had to take a reading comprehension test after the reading process.

Mind-wandering probes

Before participants started reading the hypertext, they were presented with an instruction, which contained a definition of mind wandering, which was used in previous studies (Schurer et al., 2020 ; Smallwood et al., 2007 , p. 533). During reading the hypertext, participants were asked at random intervals of 2–4 min with an average duration of 3 min what they were thinking about immediately before the thought probe appeared. This question appeared in a pop-up window at the bottom of the screen with a beep (Stawarczyk et al., 2011a , b ; Unsworth & McMillan, 2013 ). With the appearance of the thought probe, participants had to select an answer from four answer categories in line with their instructions by pressing the corresponding number on their keyboard. The participants` thoughts could be directed either on: (1) the text; (2) how well I understand the text; (3) the current state of being; (4) a memory in the past or something in the future (Unsworth & McMillan, 2013 ). After responding to a category, the participants continued reading the text. For this response category the proportion of TUT-responses was computed, with a higher proportion indicating more mind wandering. Furthermore, we did not define the mind-wandering categories as goal-related (TUTs that are used for the maintenance of future task goals) or goal-unrelated (TUTs that are related to personal issues and concerns; see Rummel et al., 2017 ; Steindorf & Rummel, 2017 ). To ensure comparability with our previous study (Schurer et al., 2020 ), we used category classifications that are based on Smallwood et al.’s ( 2007 ) definition and include similar categorizations as those in McVay and Kane ( 2012b ). In addition, we recorded the times for reading the entire hypertext.

Reading comprehension test

To test participants’ understanding of the text, they completed a paper pencil reading comprehension task with a total of 12 single-choice questions. Each question contained four possible answers, from which the participants had to choose one. During this task, the participants had no access to the hypertext. The results were the sum of the correctly answered questions.

Memory test

The situational model of the text develops a mental representation of the text content and organizes it in memory (Ericsson & Kintsch, 1995 ). In the memory test, participants had to distinguish whether a sentence presented on a screen appeared in the hypertext or not. There were 16 sentences in total, eight sentences were original text sentences, and eight sentences were manipulated either on surface or textbase structure. Manipulations on surface structure consisted of the shifting of a clause within the base sentence to a new position, so that the surface sentence structure changed, whereas manipulations on textbase structure consisted of replacing a proposition in the base sentence, so that the meaning of the text altered (see Schurer et al., 2020 , for details). For statistical analyses, correct percentage answers were calculated.

Statistical analyses

All statistical analyses were carried out using IBM SPSS Statistics 23.0. An alpha value of .05 was adopted for all significance testing. Estimated effect sizes are reported using partial eta squared (η p 2 ). Post hoc tests were adjusted using Bonferroni correction. Analyses of covariance (ANCOVAs), analyses of variance (ANOVAs), and paired t-tests were conducted for the analyses of mind wandering, reading comprehension performances, as well as PM task performance. In a first analysis, we examined the potential influence of different factors on mind wandering and conducted a two-way ANCOVA to analyse the interaction of the factors PM task group and text difficulty. WMC was included as a control variable in the analyses, in order to control for potential confounding influences of WMC on the findings (e.g., Schurer et al., 2020 ). To follow up, significant interaction of task group and text difficulty post hoc t-tests were conducted. In a second analysis, we examined potential influencing factors on reading comprehension (text comprehension and text memory) and conducted two-way ANCOVAs to analyse the interaction of the PM task group and text difficulty. In a third analysis, we examined PM task performance via a 2 (unfinished, finished) × 2 (recall phase: first, second) mixed-factorial design. As a subsidary analysis, we looked at the correlations between the TUT rates, reading comprehension measures and reading time.

In addition, we used a Bayesian model-selection approach because of the extended possibilities for model testing and model selection compared to the more commonly used null-hypothesis testing approach (see, e.g., Gelman et al., 2013 ; Kruschke, 2014 ). Note that the Bayesian approach allows for proper model selection, which is based on an assessement of the strength of evidence associated with the null model or its rejection as the alternative model. In the following sections, Bayes factors are interpreted according to the proposal of Jeffreys ( 1961 ), with BF < 3 indicating anecdotal evidence, BF > 3 indicating moderate evidence, BF > 10 indicating strong evidence, and BF > 100 indicating overwhelming evidence for a model assuming or rejecting the null hypothesis (see also Lee & Wagenmakers, 2013 ). Furthermore, we report the common information for ANCOVAs and t-tests, but use the additionally reported Bayes factors to draw conclusions concerning our hypotheses. To this end, for all our analyses, we included WMC as the only factor in the null model and report BF 10 as the Bayes factor in favour of the alternative model. Thus, BF 10 values < 0.3 are strong evidence in favour of the null model.

Two-back task performance

Table 1 presents descriptions of sample characteristics. A series of ANOVAs revealed no significant differences between the experimental PM task groups (finished, unfinished) concerning gender, age, prior knowledge, WMC, and PM task performance in both recall phases (all p s > .05). This finding indicates that the participants were comparably engaged in the reading task setting, regardless of the particular PM task condition they received.

First analysis: Mind wandering

Looking at the mean proportion of the unfinished and finished PM task group, participants experienced more TUTs (33.9%) in the unfinished than in the finished PM task group (25.8%). A two-way ANCOVA was run to examine the effect of PM task group, and text difficulty, with WMC as a control variable on overall mind wandering and mind-wandering categories “current state of being.” For the total amount of mind-wandering rates, we found significant main effects for PM task group ( F (1,67) = 5.457, p = .022, η 2 = .075), and text difficulty ( F (1,67) = 7.971, p = .006, η 2 = .106), with strong evidence (BF 10 = 14.33) for both effects also in the Bayesian analysis. We could not find any significant interaction (all p s > .05). The ANCOVAs conducted on the mind-wandering category “something in the past/future” did not reveal any significant main or interaction effects (all p s > .05). Looking at the mind wandering category “current state of being”, another ANCOVA found a significant main effect for the PM task group ( F (1,67) = 16.933, p = .000, η 2 = .202), and text difficulty ( F (1,67) = 16.059, p = .000, η 2 = .193). Participants in the unfinished group showed significantly more TUTs concerning the current state of being ( M = .259, SD = .114) than the finished group of participants ( M = .140, SD = .132; see Table 2 ). The significant main effect of the factor text difficulty revealed an increase in TUTs when reading difficult texts ( M = .259, SD = .165) compared to easy texts ( M = .141, SD = .107). As expected, the two-way interaction between PM task group and text difficulty was significant ( F (1,67) = 5.841, p = .018, η 2 = .080). The results of a Bayesian ANCOVA were consistent with the results of the classical ANCOVA. Further support for a significant interaction comes from a Bayes factor of BF 10  = 3.08.

Further, a paired t-test was run to determine whether there was a difference between mind-wandering rates when participants read the easy text compared to the difficult text between the PM task group.

For the difficult text condition the analysis showed significantly more TUTs for participants in the unfinished ( M = .35, SD = .12) than in the finished group (M = .16, SD = .15 ; t (34) = -5.073, p = .000; see Fig. 1 ). Additional Bayesian t -tests supported these results with BF 10  = 116.61 for the difficult text condition. In the easy text condition, there was no significant difference between the unfinished ( M = .16, SD = .10) and the finished group ( p > .05; see Fig. 1 ), which was paralleled by BF 10  = 0.687. However, this result represents only anecdotal evidence for the null hypothesis.

Proportion of task-unrelated thoughts (current state of being) between the groups for easy and difficult texts

Second analysis: Reading comprehension and PM task performance

For the easy text condition, the mean sum of correct answers in the reading comprehension test amounted to M = 6.94 ( SD = 1.55) in the unfinished and to M = 7.61 ( SD = 1.46) in the finished group. For the difficult text condition, the mean sum of correct answers in the reading comprehension test amounted to M = 6.61 ( SD = 1.54) in the unfinished and to M = 7.83 ( SD = 1.72) in the finished group. Using working memory as a covariate, the ANCOVA analysis revealed that there was a significant main effect for the PM task group on reading comprehension ( F (1,67) = 6.409, p = .014, η 2 = .087; see Fig. 2 ). Participants in the finished PM task group showed a better reading comprehension score ( M = 7.72; SD = 1.58) than participants in the unfinished PM task group ( M = 6.78; SD = 1.53). We could not find a significant main effect of text difficulty on reading comprehension (all p s > .05). The two-way interaction was not significant ( p > .05). This result makes it clear that despite possible differences in WMC, the effect of the PM task on reading comprehension would remain.

Mean differences in the reading comprehension score between the groups for easy and difficult texts

We also assessed memory for the text by the recognition of sentences that were manipulated on the surface or at text base level. Participants in the unfinished PM task group correctly recognized more textbase manipulations (76%) than participants in the finished PM task group (71%). A two-way ANCOVA was run to examine the effect of PM task group, and text difficulty with WMC as a control variable on correctly recognized sentences. The ANCOVA conducted on correctly recognized surface manipulations did not reveal any significant main effects or interaction (all p s > 0.05). The ANCOVA conducted on correctly recognized textbase manipulations did not reveal a significant main effect of PM task group, or text difficulty (all p s > .05). In addition, no significant interaction between PM task and text difficulty was revealed ( p > .05) (see Fig. 2 ). The memory performance was not affected by the presence versus absence of a second task still in mind.

Further analysis

As an additional result and in order to assess the relationship between the amount of mind wandering and reading performance, we conducted Pearson’s correlation analysis between the related values. We did not observe a significant correlation between mind-wandering rates and reading comprehension score (all p s > .05), except for a low significant negative correlation between the mind-wandering category “something in the past/future” and correctly recognized original sentences ( r = -.255; p = .031). However, the analysis indicated a low significant positive correlation between overall reading times and reading comprehension score ( r = .238, p = .044), which indicates that participants performed better on the reading comprehension test if they read longer. The Bayesian factors for the models were BF 10 = 1.74 when correlating TUTs about the past/future with correctly recognized original sentences and BF 10 = 1.28 for the correlation of reading time with reading comprehension score. Thus, the evidence was not very strong but anecdotal (Lee & Wagenmakers, 2013 ). Nevertheless, this relationship could explain the lack of correlations between mind wandering (current state of being) and reading comprehension. However, due to the weak evidence, these results should be interpreted with caution.

Further, we investigated differences between participants in the recall performance . A 2 × 2 ANOVA with PM task group (unfinished vs. finished) as between-participants’ and recall phase (first, second) as within-participants’ factor showed that participants performed better in the first ( M = 6.03, SD = 1.94) than in the second recall phase ( M = 4.28, SD = 1.78; F (1,70) = 77.156, p = .000, η 2 = .524). We could not find a significant main effect for PM task group on performance ( F (1,70) = .382, p = .539, η 2 = .005) and no significant interaction between PM task group and recall phase on performance ( F (1,70) = .011, p = .917, η 2 = .000).

The present study aimed to examine whether mind wandering can be induced by an additional demand and how this induction is modulated by text difficulty in a complex reading situation. In addition, the study aimed to test different models of mind wandering against each other with the use of an additional demand. Therefore, we chose a PM task in a reading task setting, which differed in text difficulty. This should show that participants having a second task in mind experience more TUTs than participants who think a second task is finished when reading difficult texts . We found that participants in the unfinished group experienced more TUTs (overall TUTs and category “current state of being”) than participants in the finished group when reading difficult but not easy texts. Although we found significant effects on mind wandering, we could only find a significant main effect of the PM task on reading comprehension, but no significant interactions on reading comprehension. A further analysis showed that participants probably compensate the influence of the second task by reading longer, which in turn has a positive effect on their reading comprehension performance.

• Mind wandering

We hypothesized that mind wandering would be influenced by an interaction of two factors: text difficuty (easy vs. difficult) × PM task group (finished vs. unfinished). We found a greater difference in mind wandering between the finished and the unfinished group when reading the difficult text compared to the easy text, with more mind wandering occurring in the difficult text version. Thus, these results demonstrated that the frequency of mind wandering (thoughts about the current concerns) can be influenced by increasing task demands when reading difficult texts, which needs to be discussed in the light of previous studies on mind wandering and PM task situations.

The current findings are related to results of the study of Steindorf and Rummel ( 2017 ), in which participants in the unfinished group and the finished group did not differ in the general amount of off-task processing during an ongoing task. However, the results showed a change in the content of off-task processing with larger amount of thoughts goal-related to the fulfillment of PM task requirements in subjects of the unfinished task condition.

On the basis of these findings and of a further study (Rummel et al., 2017 ) the authors argued that under high task demands, i.e., when all resources are committed, less mind wandering might also occur but that the content of the off-task thoughts might even be more focussed on the primary task processing (see also Rummel et al., 2017 ). This view would predict that in the current study participants in the unfinished group should not show more TUTs than participants in the finished group when both groups are already engaging in a demanding reading of a difficult text. Neither predictions are consistent with the present results. Nevertheless, this discrepancy can be explained by differences in the overall task requirements between the studies of Rummel et al. ( 2017 ) and the current study.

In more detail, we used a far more complex reading task than the ongoing primary tasks in the studies of Steindorf and Rummel ( 2017 ), i.e., an N-back task, and of Rummel et al. ( 2017 ), i.e., a lexical decision task, which, probably, left sufficient space for subjects’ mental resources to cope with the requirements of the ongoing task. Therefore, the fact that we used a much more demanding task than Steindorf and Rummel ( 2017 ) would suggest that the degree of complexity of the primary ongoing task should be considered as an important factor when determining the factors potentially affecting the occurrence of TUTs (and of their content) in finished and unfinished task conditions. In the present study, we found an influence of the PM requirements on the occurrence of mind wandering in the difficult reading condition but not in the easy reading condition. This is consistent with the assumption that the mental resources in the easy reading condition might have already exhausted mental resources to a degree that allowed just for proper task fulfilment of the ongoing task together with a sufficient control of off-task processing in the PM task. This would also be consistent with the findings of Konu et al. ( 2021 ), who demonstrated that the emergence of thoughts can vary depending on task complexity.

This explanation would also relate to the findings by Smallwood and Schooler ( 2006 ); see also Smallwood & Andrews-Hanna, 2013 ), who demonstrated more mind wandering when tasks were easy, so that more executive resources remained available for mind wandering. On the other hand, when task demands were high, fewer resources were available and less mind wandering occurred. Interestingly, the drop in mind wandering with increasing task difficulty has been argued to be steeper for people with higher WMC (Smallwood & Andrews-Hanna, 2013 ). Our results are rather opposite to this view: Mind wandering increased in the most difficult condition (difficult text and PM demands), while it was lowest in the easy condition with the finished PM task and an easy text. Nevertheless, this latter observation would be consistent with the assumption that the difficulty of the general task situation, differing accross studies, affects the degree of participants’ involvement in task processing and consequently the occurrence of mind wandering. Neither of the above views fully explains the interaction between cognitive resources and task demands as observed in the present study, although each model would partially be consistent with parts of our results. Instead, the overall pattern of findings is suggestive for a rather complex interaction of the demands related to the main task, the task aimed at inducing TUTs and also the individual resources of subjects allowing more or less task involvement, which is consistent with the present resource-demand-matching view.

In more detail, given that WMC did not differ between groups in our study (see Table 1 ), a matching of resources takes place in the easy and difficult text condition. The additional PM task adds an additional demand, which then leads to a breakdown of the existing cognitive resources and therefore to a mismatch in the difficult text condition due to the assumption of our model. Therefore, the results show more mind wandering in the difficult condition with an unfinished task in the mind. Furthermore, it seems that the combination of two demanding tasks, i.e., reading a difficult text while memorising items from a to-do-list, was sufficient to create additional demands, which far exceed the resources. When task demands are high, executive control capabilities are decreased and fewer resources are available to perform a primary task (McVay & Kane, 2010 ). On the other hand, unfulfilled tasks and goals in mind represent additional cognitive demands. For this reason, the interaction of higher demands and unfulfilled tasks in the mind leads to mind wandering. In future studies, more integration should occur across theories to indicate the complexity of TUTs. Our model aims to contribute to this by looking at the interaction of demands and resources. Notably, issues of power also need to be considered when discussing the pattern of PM-related findings observed by different studies, which will be accomplished later below.

Interestingly, our results were significant only with respect to the mind-wandering category “current state of being”. A potential explanation might be offered by the personal relevance of the PM task, which encompassed a to-do list of students’ daily activities. It is conceivable that a self-relevant PM task could tap similar resources to self-relevant worries, thereby inducing mind wandering (see also McVay & Kane, 2013 ).

Reading comprehension and recall performances

We further hypothesized that participants reading a difficult text with a second task in mind (unfinished group), would show less text comprehension results compared to participants reading an easy text version. Participants who did not remember a second task (finished group) showed significantly better results in the reading comprehension test than participants with a second task in mind (unfinished group). The control for potential differences in working memory capacity by the ANCOVA results showed little or a negligible impact of WMC on the comprehension scores. This indicates that working memory does not account for a significant amount of variance in the current study. We further tested the impact of the PM task group (unfinished, finished) and recall phases (first, second) on recall performance. The observation of a significant main effect of recall phases on PM task performance indicates that the PM task manipiulation was effective. We found further that the participants performed worse in the second recall phase than in the first recall phase. In contrast, PM task performance decreased over time in both groups (unfinished, finised). In the first recall phase, the performance was identical between the two groups (unfinished, finished) as expected, and the observed significant effect of PM task groups on TUTs was therefore not caused by a potential memory difference between groups. Contrary to the study of Steindorf and Rummel ( 2017 ), the performance of the reading task was affected by the PM task group (unfinished and finished group). It seems that the second task distracts participants when reading attentively, even if no significant correlation between reading comprehension and TUTs can be found. While in the study by Steindorf and Rummel ( 2017 ) the PM task manipulation (unfinished, finished) had an influence on the performance in the second recall phase, we found an influence on the performance in the primary task (reading comprehension test). One reason for this result could be that we used a more complex primary task in contrast to the n-back task. This suggests that the relationship between the different demands of the primary and the secondary tasks (text difficulty, PM task) plays an important role. Due to the more demanding reading task, the forgetting rate in the second recall phase seems to be the same in both groups (unfinished, finished). In the study of Steindorf and Rummel ( 2017 ), however, the participants of the unfinished group still remembered the items of the second recall.

Contrary to our expectations, we did not observe a negative association between comprehension measures and the frequeny of mind wandering. In the present study, an increased amount of mind wandering did not reduce text comprehension. An additional test revealed a significant negative correlation between the mind-wandering category “something in the past/future” on correctly recognized original sentences. This could be an indication that the participants think more about the second task, i.e., the additional demand, which in turn might affect the more surface text comprehension, and thus the results here are to be interpreted with caution due to the lack of moderate-sized correlations ( r = .030; Cohen, 1992 ). Moreover, the present experimental study was not designed to rigorously investigate correlational patterns, and therefore the related findings in the current study do not represent the main vein of our argumentation. Nevertheless, a potential explanation could be that mind wandering can be compensated by longer reading times, which retains task performance (in terms of reading comprehension), but is still less efficiant because more time is needed for the same performance level. In fact, the current findings showed a positive correlation of reading time with reading comprehension, which might have meant that a prolongation of the the reading time led to better understanding of the text and, thus, possibly could compensate for the influence of the PM task, as the significant main effect of the PM task group on reading comprehension might indicate (participants showed a better reading comprehension score in the finished than in the unfinished condition). This could be evidence that participants were less successful in construction situation models when reading these texts, which might lead to more mind wandering, and which does indeed reduce text understanding. However, due to the low correlation, this result should be interpreted with caution. Noteworthy, we observed a similar correlation in a previous study by our group (see Schurer et al., 2020 ) and, therefore, consider reading time as a mediator, which needs to be investigated in more detail in future studies. Furthermore, our findings require further replication in larger samples. Likewise, this supports the assumption that the reading task, unlike the primary task of Steindorf and Rummel ( 2017 ), was so demanding that the participants in the unfinished group had no chance to remember better. However, reading time was not associated with any variation in mind wandering. The precise role of such compensatory processes remains to be elucidated in future studies. Another explanation could be that few studies of mind wandering included expository texts. For example, Kane and McVay ( 2012 ) found a positive relationship between mind-wandering measures and narrative text comprehension, but this was not found for expository text comprehension. More studies investigating mind wandering and expository text comprehension would be needed to find out if findings from studies of mind wandering in narrative text comprehension could be transfered to expository texts. Former studies showed that the frequency of mind wandering and its influence on text comprehension is strongly associated with topic interest and motivation to do well in the task (Unsworth & McMillan, 2013 ). A limitation of our study is that we did not control interest and motivation.

Limitations

Finally, it has been argued that studies, despite an achieved power of 80%, sometimes have insufficient sample sizes to ensure scientific conclusions that meet the requirements of strong and optimal power considerations (Brysbaert, 2019 ). Such power issues should be discussed with respect to the current findings and, in particular, to the seemingly diverging pattern of findings about the potential impact of PM manipulations on the occurrence of mind wandering as mentioned before. In calculating the sample size for the current study, we assumed a rather large effect size based on the calculation of the effect sizes in selected but comparable studies (see Footnote 1). Undoubtedly, an increase in subject numbers would have been desirable in order to ensure valid detection of significant findings under conditions of smaller expected effect sizes, just as the application of a Bayesian approach would do if this approach was applied in order come to an evidence-based assessement of the rejection or acceptance of the null hypothesis (see Brysbaert, 2019 ). Therefore, the current findings need to be considered with caution and need to be carefully compared with the overall pattern of related studies. Based on our own findings, a power of .82 is expected when calculating an ANCOVA (see G*Power, Faul et al., 2007 ), which is comparable to the expected power in other studies applying similar designs of inducing mind wandering with unfinished versus finished PM task-related manipulations. For example, the calculated power in the cited studies ranges from .52 to .90 (Masicampo & Baumeister, 2011 ; Rummel et al., 2017 ; Scullin et al., 2018 ; Steindorf & Rummel, 2017 ). In more detail, a power of values between .63 and .95 was expected in the study by Rummel et al. ( 2017 , Experiments 1–3), a power of .67 in the study by Steindorf and Rummel ( 2017 ), and a power of .80 in Scullin et al. ( 2018 , Experiment 2). While this pattern of power values would suggest considerable validity of the observed findings when speaking about the isolated studies, an overall conclusion based on a summarized view across several studies just at the cutoff for sufficient power might suffer from the the occurrence of effects just reaching significance by incidence. As had been noted, this is an issue for many studies in cognitive psychology and often relates, in particular, to studies with more complex experimentals designs investigating the effects of several factors and their interaction on cognition (see also Brysbaert, 2019 ). For the current experimental design combining a two-factorial design with a factor on participants’ individual WM performance, a sample size of N = 210 participants would be required on the basis of an a priori sample size calculation with G*Power assuming an expected power of 95% and a medium effect size of f = .25 (and of 1,302 participants when assuming a small effect size; Brysbaert, 2019 ). While complex factorial designs seem necessary to investigate the current research question, the recruitment of large sample-sized studies might be challenging for future studies. Therefore, further studies with larger sample sizes are definitely needed to validate the conclusions of the current study together with alternative approaches to ensure sufficient validity of the perceived conclusions, such as the conduction of independent replication studies (LeBel et al., 2017 ) and obtaining converging evidence with related designs or coming from other methodological perspectives. This would allow for broader methodological approaches when evaluating the validity of the proposed model to explain the effects of task difficulty, cognitive resources and task requirements on the distraction of attention from task processing (Stawarczyk, Majerus, Maquet, & D'Argembeau, 2011b ; see also LeBel et al., 2017 ).

The present study gives insights into attention processes when reading digital texts. Moreover, the present study showed what happens if mind wandering was induced by a second task in mind when task demands were high. Furthermore, the study provided partial support for the resource-demand-matching view of mind wandering, although the model needs to be examined more closely in future studies. The interaction between text difficulty and PM task group had an impact on mind wandering. Participants who read the difficult text can only absorb the additional demands without a second task still in mind. Based on these results, it is important to think about the task itself, since too-easy tasks can lead to falsifying the results. Therefore, it is essential to match the task requirements to the available resources. The resource-demand-matching view could be further investigated by further studies in other contexts, for example during online reading, and here in particular, during reading of hypertexts. In particular, the deliberate distraction of the reading process, for example by pop-up windows or messenger messages that pop up when reading could be of interest for the model mentioned above (e.g., Levy et al., 2016 ).

The adopted size of our participant sample would be consistent with an a priori power analysis with G*Power (Faul et al., 2007 ) for ANCOVAs assuming a medium to large effect size ( f = .34). This effect sized was based on collapsing the individual effect sizes observed in comparable previous studies, for example, Masicampo and Baumeister ( 2011 ), f = .36; Steindorf and Rummel ( 2017 ), f = .46; Rummel et al. ( 2017 ), f = .22, and assuming a standard two-tailed alpha value ( p < .05) at 80% power, which results in a required sample size of N = 68. We return to the discussion of the sample size and corresponding power issues of the present study compared to other studies in the Limitation section of the General discussion .

Brysbaert, M. (2019). How many participants do we have to include in properly powered experiments? A tutorial of power analysis with reference tables. Journal of Cognition, 2 (1), 16. https://doi.org/10.5334/joc.72 .

Article   PubMed   PubMed Central   Google Scholar  

Cohen, J. (1992). Statistical Power Analysis. Current Directions in Psychological Science, 1 (3), 98–101. https://doi.org/10.1111/1467-8721.ep10768783

Article   Google Scholar  

Ericsson, K. A., & Kintsch, W. (1995). Long-term working memory. Psychological Review, 102 (2), 211–245. https://doi.org/10.1037/0033-295X.102.2.211

Article   PubMed   Google Scholar  

Faul, F., Erdfelder, E., Lang, A. G., & Buchner, A. (2007). G*Power 3: a flexible statistical power analysis program for the social, behavioral, and biomedical sciences. Behavior Research Methods, 39 (2), 175–191. https://doi.org/10.3758/bf03193146

Feng, S., D'Mello, S. K., & Graesser, A. ( 2013 ). Mind wandering while reading easy and difficult texts. Psychonomic Bulletin & Review, 20 (1), 586–592. https://doi.org/10.3758/s13423-012-0367-y

Gelman, A., Carlin, J. B., Stern, H. S., Dunson, D. B., Vehtari, A., & Rubin, D. B. (2013). Bayesian Data Analysis (3rd ed.). Chapman and Hall/CRC. https://doi.org/10.1201/b16018

Book   Google Scholar  

Graesser, A. C., & McNamara, D. S. (2011). Computational analysis of multilevel discourse comprehension. Topics in Cognitive Science, 3 , 371–398. https://doi.org/10.1111/j.1756-8765.2010.01081.x

Jeffreys, H. (1961). Theory of probability (3rd ed.). Oxford, UK: Oxford University Press.

Google Scholar  

Kane, M. J., & McVay, J. C. (2012). What Mind Wandering Reveals About Executive-Control Abilities and Failures. Current Directions in Psychological Science, 21 (5), 348–354. https://doi.org/10.1177/0963721412454875

Kintsch, W., & Van Dijk, D. (1978). Toward a model of text comprehension and production. Psychological Review, 85 , 363–394. https://doi.org/10.1037/0033-295X.85.5.363

Konu, D., Mckeown, B., Turnbull, A., Ho, N., Karapanagiotidis, T., Varnderwal, T., McCall, C., Tipper, S., Jefferies, B., & Smallwood, J. (2021). Exploring patterns of ongoing thought under naturalistic and conventional task-based conditions. Consciousness and Cognition, 93 . https://doi.org/10.1016/j.concog.2021.103139

Kruschke, J. K. (2014). Doing Bayesian data analysis: A tutorial with R, JAGS, and Stan (2nd ed.). https://doi.org/10.1016/B978-0-12-405888-0.09999-2

Kvavilashvili, L., & Rummel, J. (2020). On the nature of everyday prospection: A review and theoretical integration of research on mind-wandering, future thinking, and prospective memory. Review of General Psychology, 24 ( 3 ), 210–237. https://doi.org/10.1177/1089268020918843

LeBel, E. P., Berger, D., Campbell, L., & Loving, T. J. (2017). Falsifiability is not optional. Journal of Personality and Social Psychology, 113 (2), 254–261. https://doi.org/10.1037/pspi0000106

Lee, M. D., & Wagenmakers, E. J. (2013). Bayesian cognitive modeling: A practical course . Cambridge University Press. https://doi.org/10.1017/CBO9781139087759

Levy, E. C., Rafaeli, S., & Ariel, Y. (2016). The effect of online interruptions on the quality of cognitive performance. Telematics and Informatics, 33 , 1014–1021. https://doi.org/10.1016/j.tele.2016.03.003

Masicampo, E. J., & Baumeister, R. F. (2011). Consider it done! Plan making can eliminate the cognitive effects of unfulfilled goals. Journal of Personality and Social Psychology, 101 (4), 667–683. https://doi.org/10.1037/a0024192

McNamara, D. S., Kintsch, E., Songer, N. B., & Kintsch, W. (1996). Are good texts always better? Text coherence, background knowledge, and levels of understanding in learning from text. Cognition & Instruction, 14 , 1–43. https://doi.org/10.1207/s1532690xci1401_1

McVay, J. C., & Kane, M. (2009). Conducting the Train of Thought: Working Memory Capacity, Goal Neglect, and Mind Wandering in an Executive-Control Task. Journal of Experimental Psychology . Learning, Memory, and Cognition, 35 , 196–204. https://doi.org/10.1037/a0014104

McVay, J. C., & Kane, M. J. (2010). Does mind wandering reflect executive function or executive failure? Comment on Smallwood and Schooler (2006) and Watkins (2008). Psychological Bulletin, 136 (2), 188–197. https://doi.org/10.1037/a0018298

McVay, J. C., & Kane, M. J. (2012a). Drifting from slow to “D’oh!”: Working memory capacity and mind wandering predict extreme reaction times and executive control errors. Journal of Experimental Psychology: Learning, Memory, and Cognition, 38 , 525–549. https://doi.org/10.1037/a0025896

McVay, J. C., & Kane, M. J. (2012b). Why does working memory capacity predict variation in reading comprehension? On the influence of mind wandering and executive attention. Journal of Experimental Psychology: General, 141 (2), 302–320. https://doi.org/10.1037/a0025250

McVay, J. C., & Kane, M. (2013). Dispatching the wandering mind? Toward a laboratory method for cuing “spontaneous” off-task thought. Frontiers in Psychology, 4 , 570. https://doi.org/10.3389/fpsyg.2013.00570

Mrazek, M., Smallwood, J., Franklin, M., Chin, J., Baird, B., & Schooler, J. (2012). The Role of Mind-Wandering in Measurements of General Aptitude. Journal of Experimental Psychology General, 141 (4). https://doi.org/10.1037/a0027968

Rummel, J., & Boywitt, C. D. (2014). Controlling the stream of thought: Working memory capacity predicts adjustment of mind-wandering to situational demands. Psychonomic Bulletin & Review, 21 (5), 1309–1315. https://doi.org/10.3758/s13423-013-0580-3

Rummel, J., Smeekens, B. A., & Kane, M. J. (2017). Dealing with prospective memory demands while performing an ongoing task: Shared processing, increased on-task focus, or both? Journal of Experimental Psychology - Learning, Memory, and Cognition, 43 (7), 1047–1062. https://doi.org/10.1037/xlm0000359

Schöll, P. (2015). Berechnung des Lesbarkeitsindex Flesch nach Rudolf Flesch . Fleschindex.  http://fleschindex.de . Accessed 20 June 2020

Schurer, T., Opitz, B., & Schubert, T. (2020). Working Memory Capacity but Not Prior Knowledge Impact on Readers' Attention and Text Comprehension. Frontiers in Education, 5 , 1–12. https://doi.org/10.3389/feduc.2020.00026

Scullin, M. K., McDaniel, M. A., Dasse, M. N., Lee, J. H., Kurinec, C. A., Tami, C., & Krueger, M. L. (2018). Thought probes during prospective memory encoding: Evidence for perfunctory processes. PLoS ONE, 13 (6). https://doi.org/10.1371/journal.pone.0198646

Smallwood, J. (2013). Distinguishing how from why the mind wanders: a process-occurrence framework for self-generated mental activity. Psychological Bulletin, 139 (3), 519–535. https://doi.org/10.1037/a0030010

Smallwood, J., & Andrews-Hanna, J. (2013). Not all minds that wander are lost: the importance of a balanced perspective on the mind-wandering state. Frontiers in Psychology, 4 , 441. https://doi.org/10.3389/fpsyg.2013.00441

Smallwood, J., & Schooler, J. W. (2006). The restless mind. Psychological Bulletin, 132 (6), 946–958. https://doi.org/10.1037/0033-2909.132.6.946

Smallwood, J., McSpadden, M., & Schooler, J. W. (2007). The lights are on but no one's home: meta-awareness and the decoupling of attention when the mind wanders. Psychonomic Bulletin Review, 14 , 527–533. https://doi.org/10.3758/BF03194102

Smallwood, J., McSpadden, M., & Schooler, J. W. (2008). When attention matters: The curious incident of the wandering mind. Memory & Cognition, 36 (6), 1144–1150. https://doi.org/10.3758/MC.36.6.1144

Smallwood, J., Turnbull, A., Wang, H. T., Ho, N., Poerio, G. L., Karapanagiotidis, T., Konu, D., Mckeown, B., Zhang, M., Murphy, C., Vatansever, D., Bzdok, D., Konishi, M., Leech, R., Seli, P., Schooler, J. W., Bernhardt, B., Margulies, D. S., & Jefferies, E. (2021). The neural correlates of ongoing conscious thought. iScience, 24 (3), 102132. https://doi.org/10.1016/j.isci.2021.102132

Stawarczyk, D., Majerus, S., Maj, M., Van Der Linden, M., & D’Argembeau, A. (2011a). Mind-wandering: phenomenology and function as assessed with a novel experience sampling method. Acta Psychologica, 136 , 370–381. https://doi.org/10.1016/j.actpsy.2011.01.002

Stawarczyk, D., Majerus, S., Maquet, P., & D'Argembeau, A. (2011b). Neural correlates of ongoing conscious experience: both task-unrelatedness and stimulus-independence are related to default network activity. PloS one, 6 (2), e16997. https://doi.org/10.1371/journal.pone.0016997

Steindorf, L., & Rummel, J. (2017). “I Should not Forget the Apples!”-Mind-Wandering Episodes Used as Opportunities for Rehearsal in an Interrupted Recall Paradigm: Adaptive mind-wandering. Applied Cognitive Psychology, 31 (4), 424–430. https://doi.org/10.1002/acp.3328

Turnbull, A., Wang, H. T., Murphy, C., Ho, N. S. P., Wang, X., Sormaz, M., Karapanagiotidis, T., Leech, R. M., Bernhardt, B., Margulies, D. S., Vatansever, D., Jefferies, E., & Smallwood, J. (2019). Left dorsolateral prefrontal cortex supports context-dependent prioritisation of off-task thought. Nature Communications, 10 (3816), 1–10. https://doi.org/10.1038/s41467-019-11764-y

Unsworth, N., & McMillan, B. D. (2013). Mind wandering and reading comprehension: Examining the roles of working memory capacity, interest, motivation, and topic experience. Journal of Experimental Psychology: Learning, Memory, and Cognition, 39 (3), 832–842. https://doi.org/10.1037/a0029669

Zeigarnik, B. (1927). Über das Behalten von erledigten und unerledigten Handlungen. Psychologische Forschung, 9 , 1–85 Codelab.  http://codeblab.com/wp-content/uploads/2009/12/On-Finished-and-Unfinished-Tasks.pdf . Accessed 20 June 2020

Download references

Acknowledgements

We thank Melanie Müller, Antonia Küttner, Jost Eisenmenger, and Laura Petermann for supporting the data collection. We also thank Sebastian Kübler for his contribution during various stages of the paper preparation.

Availability data

All datasets generated for this study are included in the article.

Code availability

Not applicable.

TSchur and BO were supported by a Grant of the German Federal Ministry of Research and Education, No. 01PL17065, Quality Pact for Teaching. We acknowledge the financial support within the funding programme Open Access Publishing by the German Research Foundation (DFG), VAT DE 811353703. TS was supported by a grant of the German Research Foundation (Schu 1397/7-2).

Author information

Authors and affiliations.

Center of Multimedia Teaching and Learning, Martin Luther University Halle-Wittenberg, Hoher Weg 8, 06120, Halle, Germany

Teresa Schurer, Bertram Opitz & Torsten Schubert

School of Psychology, University of Surrey, Guildford, UK

Bertram Opitz

Institute of Psychology, Martin Luther University Halle-Wittenberg, Halle, Germany

Torsten Schubert

You can also search for this author in PubMed   Google Scholar

Contributions

Teresa Schurer: conceptualization, methodology, data curation, formal analysis, visualization, writing – original draft. Bertram Opitz: conceptualization, methodology, data curation, writing – review and editing. Torsten Schubert: conceptualization, methodology, writing – review and editing.

Corresponding authors

Correspondence to Teresa Schurer or Torsten Schubert .

Ethics declarations

Conflict of interest.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Ethics statement/Consent to participate

The study was approved by the ethics committee of the Deutsche Gesellschaft für Psychologie (DGPs). The experimental protocol conformed to tenets of the Declaration of Helsinki and written informed consent was obtained from each participant before the commencement of the study.

Consent for publication

Additional information, publisher’s note.

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary Information

(XLSX 20 kb)

(PDF 491 kb)

(PDF 463 kb)

Rights and permissions

Reprints and permissions

About this article

Schurer, T., Opitz, B. & Schubert, T. Concurrent prospective memory task increases mind wandering during online reading for difficult but not easy texts. Mem Cogn 51 , 221–233 (2023). https://doi.org/10.3758/s13421-022-01295-1

Download citation

Accepted : 15 February 2022

Published : 01 March 2022

Issue Date : January 2023

DOI : https://doi.org/10.3758/s13421-022-01295-1

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

• Find a journal
• Publish with us
• Track your research

An official website of the United States government

The .gov means it’s official. Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

The site is secure. The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

• Publications
• Account settings

Preview improvements coming to the PMC website in October 2024. Learn More or Try it out now .

• Advanced Search
• Journal List
• HHS Author Manuscripts

Adaptive constructive processes and the future of memory

Daniel l. schacter.

Department of Psychology, Harvard University, Cambridge, MA 02138

Memory serves critical functions in everyday life, but is also prone to error. This article examines adaptive constructive processes , which play a functional role in memory and cognition but can also produce distortions, errors, or illusions. The article describes several types of memory errors that are produced by adaptive constructive processes, and focuses in particular on the process of imagining or simulating events that might occur in one’s personal future. Simulating future events relies on many of the same cognitive and neural processes as remembering past events, which may help to explain why imagination and memory can be easily confused. The article considers both pitfalls and adaptive aspects of future event simulation in the context of research on planning, prediction, problem solving, mind-wandering, prospective and retrospective memory, coping and positivity bias, and the interconnected set of brain regions known as the default network.

In 1932, Sir Frederic Bartlett published his landmark volume, Remembering: A Study in Experimental and Social Psychology , which drew on evidence of memory distortions to refute the idea that remembering is a literal or exact reproduction of the past. Bartlett (1932) argued instead that remembering “is an imaginative reconstruction or construction (p. 213)” that depends heavily on the operation of a schema, a concept that he borrowed from the British neurologist Henry Head. Bartlett defined a schema as “an active organisation of past reactions, or of past experiences, which must always be supposed to be operating in any well-adapted organic response (1932 , p. 201)”. He further emphasized the importance of “the organism’s capacity to turn round upon its own ‘schemata’ (p. 213)” during acts of remembering.

The somewhat opaque idea of an organism “turning round upon its own schemata” became sufficiently controversial that Bartlett later tried to clarify the concept in some unpublished notes that have been made available by the Sir Frederic Bartlett Archives at the University of Cambridge ( http://www.ppsis.cam.ac.uk/bartlett/NotesOnRemembering.htm ). “There is probably no other phrase in Remembering that has received as much attention as the expression ‘turning round on one's own schemata’” wrote Bartlett. He went on to explain that “turning round” refers to cognitive activities that occur “whenever remembering demands more than the production of a fully learned response” – that is, strategic, voluntary, and constructive activities that are required to respond to a current environmental demand when automatic, learned responses are not elicited. Bartlett argued further that such activities are of great functional importance: “when some current situation demands an adaptive reaction, selection from, or reconstruction of, the organised past must be effected.” Despite the adaptive value of this constructive activity, however, “turning round” also has a downside, often resulting in “rationalisation, condensation, very often in a considerable rearrangement of temporal relations, in invention and in general in an exercise of constructive imagination to serve whatever are the operating interests at the time at which the turning round takes place.”

Adaptive Constructive Processes and Memory Distortion

Bartlett’s elaboration of what he meant by “turning round” on one’s own schemata reveals an important property of human memory: some processes that contribute to adaptive responding also result in error. In this article, I call them adaptive constructive processes , which I define as processes that play a functional role in memory and cognition but produce distortions, errors, or illusions as a consequence of doing so . Adaptive constructive processes are not uniquely characteristic of memory. For example, in one of their classic papers on judgment and decision making, Tversky and Kahneman (1974 , p. 1124) observed that the heuristics people use when making judgments about the likelihood of uncertain events “are quite useful, but sometimes they lead to severe and systematic errors”, thus falling under the rubric of adaptive constructive processes. Students of perception have long argued that visual illusions result from the operation of constructive processes that contribute to the efficient functioning of the visual system (e.g., Gregory & Gombrich, 1973 ; Roediger, 1996 ).

Although the idea that memory distortions sometimes reflect the operation of adaptive processes can be traced to Bartlett’s (1932) work, and has been embraced by other researchers from time-to-time (e.g., Brainerd & Reyna, 2005 ; Howe, 2011 ; Howe, Garner, Charlesworth, & Knott, 2011 ; Neisser, 1967 ; Newman & Lindsay, 2009 ; Schacter, 1999 , 2001 ), in general memory distortions have been viewed as indications of defects or flaws in memory. Consistent with this view, there is evidence that increased incidence of memory distortions is associated with various indicators of suboptimal processing. For example, people who are especially prone to disruptions in consciousness or dissociative experiences have also shown increased rates of susceptibility to various kinds of memory distortions (e.g., Clancy, Schacter, McNally, & Pitman, 2000 ). More recent studies have linked memory distortion to low intelligence ( Zhu et al., 2010 ) and also to symptoms of post-traumatic stress disorder ( Goodman et al., 2011 ).

Such findings may appear to cast doubt on the adaptive perspective. However, Scott Guerin, Peggy St. Jacques and I ( Schacter, Guerin, & St. Jacques, 2011 ) recently marshaled emerging evidence in favor of the view that some memory distortions do indeed reflect the operation of what I call here adaptive constructive processes (note that I use the term “adaptive” in this article to refer to a beneficial characteristic of an organism, and make no claim about the evolutionary origins of adaptive constructive processes; for discussion of this issue, see McKay & Dennett, 2009 ; Schacter, 2001 ; Schacter, Guerin, & St. Jacques, 2011 ). In our review, we focused on three memory distortions that we believe reflect the operation of such processes: 1) post-event misinformation, 2) gist-based and associative memory errors, and 3) imagination inflation. In the present article, I briefly summarize arguments concerning adaptive aspects of the first two kinds of memory distortions, and then elaborate on the adaptive constructive processes associated with the third.

The misinformation effect pioneered by Loftus and colleagues (for review, see Loftus, 2005 ) occurs when misleading information presented after an event results in distorted memory for the original event. Though misinformation-based memory errors have important practical consequences ( Loftus, 2005 ), Schacter, Guerin, and St. Jacques (2011) suggested that they can be viewed as a consequence of adaptive updating processes that are crucial for the operation of a dynamic memory system that flexibly incorporates relevant new information (for recent evidence and related ideas, see Edelson, Sharot, Dolan, & Dudai, 2011 ; Hardt, Einarsson, & Nader, 2010; St. Jacques & Schacter, in press ).

Gist-based memory errors occur when people falsely remember a novel item that is similar to an item that they encountered previously, making their memory decision based on the gist of what happened ( Brainerd & Reyna, 2005 ; Koustaal & Schacter, 1997). Associative memory errors occur when people falsely remember a novel item that is associated with previously studied items, as in the well known Deese/Roediger-McDermott or “DRM” memory illusion, where presentation of a series of words (e.g., candy, sour, sugar, bitter, good, taste, tooth, nice, honey, soda, chocolate, heart, cake, eat, pie ) that are all associated to a nonpresented “critical lure” word (e.g., sweet ) results in a high level of false recall or recognition of the critical lure on a later memory test ( Deese, 1959 ; Roediger & McDermott, 1995 ; for review, see Gallo, 2010 ). Such responses are rightly classified as memory distortions – people claim to remember items that they did not study – but these errors also reflect retention of useful information concerning the general themes or meanings that participants did encounter. Retention of such information can facilitate generalization and abstraction (e.g., Brainerd & Reyna, 2005 ; McClelland, 1995 ; Schacter, 1999 , 2001 ) and in that sense can be considered adaptive. Recent evidence links associative false memories with creativity. Dewhurst, Thorley, Hammond, and Ormerod (2011) showed that susceptibility to DRM false recognition is predicted by performance on a remote associates task, which measures convergent thinking – a component of creativity that taps an individual’s ability to generate broad and numerous associations, and can thus be considered an adaptive cognitive process (for related evidence, see Howe et al., 2011 ).

Additional evidence consistent with an adaptive interpretation of gist-based and associative memory distortions comes from neuroimaging studies that have documented that a) many of the same brain regions are active during both associative/gist-based false recognition and true or accurate recognition, and b) regions that are active when people engage in semantic elaboration during encoding, which serves the adaptive function of promoting long-term retention, support both subsequent true and false recognition (for discussion, see Schacter, Guerin, & St. Jacques 2011 ; Schacter & Slotnick, 2004). Thus, both cognitive and neuroimaging evidence supports an adaptive interpretation of gist-based and associative memory errors.

Imagination Inflation and the Simulation of Future Events

The third kind of memory distortion that Schacter, Guerin, and St. Jacques (2011) discussed within an adaptive framework is known as imagination inflation: imagining events can lead to false memories that the event actually occurred (e.g., Garry, Manning, Loftus, & Sherman, 1996 ; Loftus, 2003 ). Imagination inflation is typically viewed as a consequence of a failure in source monitoring operations that allow us to distinguish between events that actually happened and events we only imagined (e.g., Johnson, Hashtroudi, & Lindsay, 1993 ). There is little doubt that source monitoring failure does play a key role in imagination inflation. Arguing from an adaptive perspective, however, we suggested that imagination inflation also results in part from the role of a constructive memory system in imagining or simulating future events. The capacity to simulate experiences that might occur in one’s personal future is potentially adaptive because it allows individuals to mentally “try out” different versions of how an event might play out ( Buckner & Carroll, 2007 ; Ingvar, 1979 ; Gilbert & Wilson, 2007 ; Schacter & Addis, 2007 ; Suddendorf & Corballis, 2007 ; Tulving, 2005 ). During the past few years, research in my lab and others has documented striking similarities between remembering the past and imagining the future (for reviews, see Schacter, Addis, & Buckner, 2007 , 2008 ; Szpunar, 2010 ). For example, neuroimaging studies have revealed extensive overlap in the neural processes that are engaged when people remember past events and imagine future events or novel scenes (e.g., Addis, Wong, & Schacter, 2007 ; Addis, Pan, Vu, Laiser, & Schacter, 2009 ; Hassabis, Kumaran, & Maguire, 2007 ; Okuda et al., 2003 ; Spreng & Grady, 2010 ; Szpunar, Watson, & McDermott, 2007 ). Similarly, behavioral studies have documented striking similarities in the corresponding cognitive processes associated with remembering the past and imagining the future (e.g., D'Argembeau & Van der Linden, 2006 ; D’Argembeau & Mathy, 2011 ; Szpunar & McDermott, 2008 ). Moreover, deficits in remembering the past are often accompanied by parallel deficits in imagining the future in various populations, including several patients with amnesia (for review, see Addis & Schacter, 2012 ), older adults and patients with Alzheimer’s disease (for review, see Schacter, Gaesser, & Addis, 2011 ), and patients with depression ( Williams et al., 1996 ) or schizophrenia ( D’Argembeau, Raffard, & Van der Linden, 2008 ). These similarities can help to explain why memory and imagination are easily confused: they share common neural and cognitive underpinnings (see also Johnson et al., 1993 ).

Even more important from the perspective of adaptive constructive processes, Donna Addis and I have argued that these observations provide clues about the adaptive functions of a constructive memory system. Specifically, Schacter and Addis (2007) have put forward the constructive episodic simulation hypothesis , which holds that past and future events draw on similar information stored in memory (episodic memory in particular) and rely on similar underlying processes. Episodic memory, in turn, supports the construction of future events by extracting and recombining stored information into a simulation of a novel event. Schacter and Addis (2007) claimed that such a system is adaptive because it enables past information to be used flexibly in simulating alternative future scenarios without engaging in actual behaviors, but it comes at a cost of vulnerability to errors and distortions that result from mistakenly combining elements of imagination and memory (for related ideas, see Suddendorf & Corballis, 2007 ).

In the remainder of this article, I will discuss further the process of imagining or simulating future events from the perspective of adaptive constructive processes, considering both the vulnerabilities and adaptive functions of future event simulation.

Future Event Simulation: Some Pitfalls

A central tenet of the constructive episodic simulation hypothesis ( Schacter & Addis, 2007 ) and related perspectives ( Suddendorf & Corballis, 2007 ) is that the ability to flexibly recombine elements of past experience into simulations of novel future events is an adaptive process, sufficiently beneficial to the organism that it is worth the concomitant cost in memory errors that result from occasionally mistakenly combining those elements. From this perspective, simulating future events ought to confer discernable advantages on the organism.

Mispredicting the Future and the Planning Fallacy

One problem with this view, however, is that considerable research indicates that future event simulations are themselves error prone. Consider, for example, predictions that people make about their future happiness and related hedonic experiences. People frequently overestimate or underestimate their future happiness across a range of situations, which Gilbert and Wilson (2007) attribute to the properties of the simulations that people use as a basis for predictions. Specifically, Gilbert and Wilson (2007) point out that simulations of future experiences are frequently unrepresentative , often capturing the most salient but not the most likely elements of an experience; essentialized , omitting some nonessential details that can impact future happiness; abbreviated , often overemphasizing the initial part of an event; and decontextualized , ignoring aspects of a future context that affect the experience of an event.

Similarly, Dunning (2007) has highlighted the limitations of simulation (what Dunning refers to as “scenario building”) in the context of planning for the future. For example, the well-known planning fallacy ( Buehler, Griffin, & Peetz, 2010 ; Kahneman & Tversky, 1979 ) occurs when people tend to underestimate the time that will be needed to complete a future task, ranging from an undergraduate senior thesis to income tax returns and holiday shopping (for review, see Buehler et al., 2010 ). Dunning (2007) summarizes evidence that people depend on simulations of how they will go about completing a task that are incomplete in critical respects and therefore contribute to the occurrence of the planning fallacy. Dunning (2007) argues that simulations can result in poor planning outcomes for a variety of reasons, including that people often rely too heavily on a few abstract features of the simulated scenarios, neglect alternative outcomes to the ones they simulate, highlight positive aspects of simulated scenarios while overlooking their negative aspects, and fail to take into account the reliability and validity of the information that is included in simulations (for related ideas, see Buehler et al., 2010 ; Kahneman & Tversky, 1979 ).

While these observations clearly indicate that there are situations in which simulations can lead us astray (see Mathieu & Gosling, 2012 , for circumstances in which predictions show relative accuracy), such errors may reflect, at least in part, the tight connection between memory and simulation ( Gilbert & Wilson, 2007 ; Schacter et al., 2008 ). Considering the planning fallacy, for example, Roy, Christenfeld, and McKenzie (2005) discuss evidence that predictions about future task duration tend to be based on memories of past event duration. Critically, these memories sometimes underestimate the actual duration. If one mistakenly remembers, for instance, that completing one’s income taxes took an hour rather than an entire afternoon, then one may be unpleasantly surprised to discover that the task cannot be completed during the time one predicted would be sufficient to complete it. Morewedge, Gilbert, and Wilson (2005) found that people often make predictions of their future happiness based on atypical past experiences that are highly memorable to them. However, these atypical experiences do not accurately predict what is likely to occur in the future, and thus can lead to prediction errors.

Instability of Future Simulations

In addition to this evidence that future simulations are error prone, other studies indicate that the act of imagining a future event can alter the subjective likelihood that an event will occur, even though there is no corresponding change in objective circumstances that would warrant a change in subjective perception. This effect was first demonstrated when Carroll (1978) showed that participants who imagined that Jimmy Carter would win the 1976 presidential election were more likely to predict that Carter would win the election over Gerald Ford, whereas participants who imagined that Ford would win were more likely to predict a Ford victory. Subsequent studies extended this basic finding to other kinds of events, such as imagining winning a contest or contracting a disease (for review, see Koehler, 1991 ).

More recently, Karl Szpunar and I showed that repeatedly imagining specific, everyday future experiences – interpersonal interactions comprised of familiar people, locations, and objects – increases the subjective plausibility that the simulated experiences would actually occur ( Szpunar & Schacter, in press ). However, this increased plausibility was observed only for positive or negative emotional events (not for neutral events; see Szpunar and Schacter, in press , for discussion of possible cognitive mechanisms). While it is difficult to know whether an initial simulation or a repeated simulation provides a more accurate assessment of future likelihood or plausibility, these experiments illustrate that a critical aspect of simulating an emotionally arousing future event – its subjective plausibility – can change significantly even when there are no changes in objective circumstances that correspond to the changes in subjective plausibility. These and earlier findings raise the possibility that instability in future simulations could undermine their usefulness as a guide to predicting or planning the future.

The Default Network: An Antagonist of Goal-Directed Cognition?

As noted earlier, neuroimaging studies have shown that remembering the past and imagining the future engage many of the same brain regions. This common core network ( Schacter et al., 2007 ), also known as the default network (e.g., Raichle et al., 2001 ; for review, see Buckner, Andrews-Hanna, & Schacter, 2008 ), includes medial prefrontal cortex, retrosplenial cortex, posterior cingulate, medial temporal lobe, and lateral temporal and lateral parietal cortices. The default network was initially identified in neuroimaging studies as increased activity in the foregoing brain regions during passive rest states compared with conditions in which individuals performed attention demanding, goal-directed cognitive tasks ( Raichle et al., 2001 ; Shulman et al., 1997 ). In other words, default network activity showed a relative decrease during goal-directed cognitive tasks compared with passive rest states. These passive rest states were not themselves targets of experimental investigation, but instead were included as control or comparison conditions for the goal-directed cognitive tasks of interest ( Buckner et al., 2008 ). In light of more recent research showing default network activity when people remember the past or imagine the future, it seems likely that during passive rest states, participants’ thoughts drifted off to past experiences or possible future experiences. Most critical for the present purposes, the observation that the default network was less active during goal-directed cognitive tasks than during passive rest led a number of subsequent investigators to propose that the default network does not contribute to goal-directed cognitive processing and that its activity might even be antithetical to goal-directed cognition (for discussion, see Spreng, Stevens, Chamberlain, Gilmore, & Schacter, 2010 ).

Consistent with these observations, Mason et al. (2007) reported fMRI evidence that activation in default network regions can indicate the occurrence of mind-wandering during task performance: default network activity increased when participants performed well-practiced, goal-directed working memory tasks that were characterized by frequent incidents of mind-wandering, compared with novel task conditions in which mind-wandering occurred less frequently. Moreover, increased activity in several default network regions during practiced (versus novel) tasks was positively correlated with self-reported tendencies for mind-wandering.

These observations do not directly question the adaptive value of future simulations, and indeed hypotheses have been advanced concerning possible adaptive functions of the default network (see Buckner et al., 2008 ). Nonetheless, since future simulations are thought to be important for goal-directed tasks, the foregoing considerations may raise questions concerning their utility because they indicate that the brain network most closely linked with future simulation is also associated with mind-wandering activity that increases when individuals stray from performing a goal-directed task. Moreover, in most studies that have linked default network activity with simulation of future experiences, the simulated future events are not linked to formulating a plan, solving a future problem, or any other kind of goal-directed cognitive activity. Instead, they represent imaginary scenes or scenarios that might or might not occur to the individual within a particular future time frame (e.g., Addis et al., 2007 ; Addis, Pan, et al., 2009 ; Hassabis et al., 2007 ; Okuda et al., 2003 ; Spreng & Grady, 2010 ; Szpunar et al., 2007 ). Therefore, these studies do not indicate whether the default network can contribute to goal-directed cognitive activity.

Future Event Simulation: The Case for Adaptive Function

The evidence considered in the previous section indicates that future simulations can be error prone, unstable, and associated with a brain network that supports off-task mental activity, thereby casting doubt on the adaptive value of the ability to simulate future events. Let us now consider evidence that supports an adaptive role for future simulations.

The Default Network Can Support Goal-Directed Cognition

In light of evidence linking default network activity to off-task mind-wandering, it is important that recent studies show that, contrary to early ideas, the default network can indeed support certain kinds of goal-directed cognition (e.g., Andrews-Hanna, Reidler, Sepulcre, Poulin, & Buckner, 2010 ). Consider a recent fMRI study from my lab led by Nathan Spreng ( Spreng et al., 2010 ) that examined brain activity associated with two forms of planning. Visuospatial planning was assessed by the well-established Tower of London task (e.g., Shallice, 1982 ), where participants are shown a configuration of discs on a vertical rod in an initial position. Participants attempt to determine the minimum number of moves needed to match the configuration of discs shown in a goal position on another vertical rod, while following rules that constrain the kinds of moves they can make. Autobiographical planning was assessed by a novel task that was visually matched to the Tower of London task but required participants to devise plans in order to meet specific goals in their personal futures. For example, freedom from debt constituted one of the goals in the autobiographical planning task. Participants viewed the goal and then saw two steps they could take toward achieving that goal (good job and save money) as well as an obstacle they needed to overcome in order to achieve the goal (have fun). They were instructed to integrate the steps and obstacles into a cohesive personal plan that would allow them to achieve the goal.

The fMRI results showed clearly that goal-directed autobiographical planning engaged the default network. Importantly, during the autobiographical planning task, activity in the default network coupled with a distinct network, known as the frontoparietal control network (e.g., Vincent, Kahn, Snyder, Raichle, & Buckner, 2008 ) that has been linked to executive processing. By contrast, visuospatial planning during the Tower of London task engaged a third network – the dorsal attention network, which is known to increase its activity when attention to the external environment is required (e.g., Corbetta & Shulman, 2002) – that also coupled with the frontoparietal control network. These results suggest that the default network can support goal-directed cognition of a particular kind – autobiographical planning – and that it does so by working with the frontoparietal control network, which appears capable of flexibly coupling with distinct networks depending on task demands (for further discussion and additional data with older adults, see Spreng & Schacter, in press ).

A related study led by Kathy Gerlach provides additional evidence on this point ( Gerlach, Spreng, Gilmore, and Schacter, 2011 ). Gerlach et al. (2011) conducted fMRI scans while participants performed a goal-directed task in which they generated mental simulations in order to solve specific problems that arose in imaginary scenarios. For example, participants were asked to imagine being left alone in a friend's dorm room, and trying on their friend’s ring, which they could not remove. They were then given the cue word “soap” to help them imagine a solution to the problem. Gerlach et al. (2011) found that, relative to a control task that involved semantic processing but not mental simulation, the problem-solving task engaged several key regions within the default network, including medial prefrontal cortex and posterior cingulate, as well as a region of dorsolateral prefrontal cortex that has been linked with executive processing. Converging nicely with these results and those of Spreng et al. (2010) , Ellamil, Dobson, Beeman, and Christoff (2012) examined the generation and evaluation of creative ideas, using a fMRI-compatible tablet that allowed participants to draw and write ideas during the fMRI scan. Ellamil and colleagues reported that during creative generation, the medial temporal lobes showed increased activity; during creative evaluation, default network regions coupled with executive regions, including dorsolateral prefrontal cortex.

The foregoing evidence that the default network can support certain kinds of goal-directed activity also fits well with recent cognitive evidence concerning the adaptive value of mind-wandering. Contrary to the prevalent idea that mind-wandering represents a kind of cognitive failure, explorations of the content of mind-wandering by Baird, Smallwood, and Schooler (2011) reveal that people typically focus on the future and engage in extensive autobiographical planning during mind-wandering episodes (for similar findings, see Stawarczyk, Majerus, Maj, Van der Linden, & D’Argembeau, 2012 ). Critically, individuals with high working memory capacity, a measure of executive processing skills, engaged in more future-oriented thought during mind-wandering than did individuals with low working memory capacity. These findings further support the idea that mind-wandering serves adaptive functions and are consistent with fMRI observations that both default network and executive regions are active during mind-wandering ( Christoff, Gordon, Smallwood, Smith, & Schooler 2009 ).

Future Simulations Can Benefit Goal-Directed Cognition

The preceding evidence shows clearly that the default network, which underpins future event simulation, supports internally-directed cognitive activities that are associated with adaptive, goal-directed processing. Consistent with this view, behavioral evidence also links future simulations with planning, problem-solving, and related forms of goal-directed processing. Taylor, Pham, Rifkin, and Armor (1998) pointed out that mental simulations are well-suited to support planning and problem solving activity because they: 1) include specific information about people, places, and social roles that can be helpful to generating problem solutions; 2) frequently contain a causal structure that resembles an actual situation, and 3) may provide access to information that would be otherwise overlooked but is critical to planning. Studies by Taylor and her colleagues have shown that simulations can help college students to plan and prepare for upcoming exams when their simulations include specific information about the steps they need to take to prepare for the exam (see Taylor et al., 1998 , for review). More recent evidence indicates that simulations are useful when attempting to solve open-ended social problems, where different possible solutions to a problem need to be explored and evaluated. Sheldon, McAndrews, and Moscovitch (2011) reported that older adults, who tend to provide less detailed autobiographical memories and simulations of future events than younger adults (e.g., Addis, Wong, & Schacter, 2008 ; for review, see Schacter, Gaesser, & Addis, 2011 ) also generated fewer relevant steps than controls when simulating solutions to ill-defined problems, suggesting that without an ability to generate detailed simulations, the effectiveness of problem solving is reduced. The tight linkage between simulations and goal-directed processing has been emphasized by D’Argembeau and Mathy (2011) , who reported that when people simulated future events, cuing participants with their personal goals facilitated access to episodic details. These observations led the authors to conclude that: “knowledge about personal goals plays an important role in the construction of episodic future thoughts (p. 258)”.

Future simulations can also have beneficial consequences on decisions about the future as well as the likelihood of carrying out future actions. Consider the phenomenon of temporal discounting: people tend to devalue a reward according to the extent of delay until the reward is delivered ( Green & Myerson, 2004 ). Boyer (2008) argued that a key adaptive function of future simulation is to allow individuals to represent emotional aspects of distant future rewards in a way that overcomes temporal discounting, producing less impulsive and more farsighted decisions. Consistent with this view, recent research has shown that when people imagine experiencing a reward in the future, they show an increased tendency to favor rewards that produce greater long-term payoffs, thereby countering the normal tendency to devalue delayed rewards ( Benoit, Gilbert, & Burgess, 2011 ; Peters & Büchel, 2010 ). For example, Benoit et al. (2011) instructed participants to imagine specific episodes of spending money in a pub at particular times in the future. Compared with a control condition in which they estimated what the money would purchase, simulating the future rewards biased participants toward accepting a larger delayed reward (e.g., $70 in 90 days) rather than a smaller immediate reward (e.g., $50 now). Benoit et al. (2011) scanned participants during this procedure, and showed that effects of episodic simulation on temporal discounting are associated with increased coupling between activity in the hippocampus and prefrontal regions involved in reward representation (see also, Peters & Büchel, 2010 ).

Future Simulations Can Enhance Prospective and Retrospective Memory

Simulating future events can also increase prospective memory or the probability of carrying out intended actions in the future. This point has been demonstrated in studies of implementation intentions : plans that link an intention with a specific anticipated situation in which the plan is to be executed ( Gollwitzer, 1999 ). Implementation intentions benefit subsequent prospective memory performance by increasing the probability that when the future context is encountered, the intended action is triggered (e.g., Chasteen, Park, & Schwarz, 2001 ), and recent evidence indicates that mental simulations contribute significantly to the effectiveness of implementation intentions ( Brewer & Marsh, 2010 ; Papies, Aarts, & De Vries, 2009 ). These findings documenting beneficial effects of simulations on prospective memory complement a large research literature demonstrating that imagining carrying out various kinds of skills – ranging from athletic acts to surgical procedures – can produce significant benefits on their later performance (e.g., Arora et al., 2011 ; Taylor et al., 1998 ; van Meer & Theunissen, 2009 ).

Recent evidence indicates that simulating future events can also aid performance on traditional tests of retrospective memory. Several decades ago, Ingvar (1985) argued that “memory of the future” – that is, remembering the contents of simulated future events – constitutes an important adaptive function because remembering what we have planned to do or say in an upcoming episode can increase the effectiveness and efficiency of future behavior. Although little is known about memory for future simulations, recent studies by Klein, Robertson, and Delton (2010 , 2011) have shown that constructing simulations of possible future events constitutes a highly effective form of memory encoding. Their studies addressed research by Nairne and colleagues that had shown that encoding information with respect to its potential survival value results in greater subsequent recall and recognition than a variety of well-established encoding procedures (for review, see Nairne, 2010). Klein and colleagues demonstrated that much or possibly all of the benefit of such “survival encoding” is attributable to planning processes. For example, when participants imagine scenarios in which they are stranded in grasslands without food, and encode a list of words with respect to their survival relevance, survival scenarios that involve planning produce superior subsequent memory to survival scenarios that do not involve planning; superior recall is also observed for scenarios that involve planning but not survival (e.g., planning a dinner party; Klein et al., 2011 ). This encoding benefit is specific to future scenarios: it is not observed when people encode information by calling up past scenarios or imagining “atemporal” scenarios ( Klein et al., 2010 ).

Although next to nothing is known about the neural processes that support encoding of future scenarios, a recent fMRI study by Martin, Schacter, Corballis, and Addis (2011) indicates that the hippocampus plays an important role. During fMRI scanning, participants imagined future scenarios comprised of people, locations, and objects that were extracted from autobiographical memories provided by each participant prior to the scan, and had been randomly recombined by the experimenters. Memory for simulations was tested shortly after the scan by providing two elements of the simulated episode (e.g., person and object) and probing recall of the third element (e.g., location); a simulation was classified as “remembered” when participants recalled the third element correctly and as “forgotten” when they did not. Greater hippocampal activity was observed during construction of subsequently remembered than forgotten simulations, even when controlling for the amount of detail associated with each simulation (for discussion of related findings, see Addis & Schacter, 2012 ; Buckner, 2010 ; Maguire, Vargha-Khadem, & Hassabis, 2010 ; Schacter & Addis, 2009 ; Squire et al., 2010 ).

Future Simulations Can Enhance Psychological Well-Being

The adaptive value of future simulations is also supported by research that has established that they can contribute to psychological well-being. For example, college students who simulated details and emotions associated with an ongoing stressful event reported using more effective coping strategies one week later compared with control groups ( Taylor et al., 1998 ). Similarly, in a study where women with first-time pregnancies were asked to simulate going into labor and arriving at the hospital on-time, more detailed and coherent simulations were correlated with increased subjective probability of a positive outcome and decreased amounts of worry related to the future event ( Brown, Macleod, Tata, & Goddard, 2002 ).

These findings are of interest with respect to the positivity bias that frequently characterizes future thinking ( Sharot, 2011 ), because such biases have been linked to a number of adaptive processes, including emotional well-being, forming social bonds, productivity at work, and coping with stress effectively (e.g., Taylor, 1989 ). Further, recent research has shown that positivity biases are observed when people remember simulations of positive, negative, and neutral future events: details associated with negative simulations were more difficult to remember over time than details associated with positive or neutral simulations, thus promoting recollection of a rosy simulated future ( Szpunar, Addis, & Schacter, 2012 ).

Finally, recent studies have revealed a benefit of future simulations with potentially important social implications: mentally simulating positive encounters with members of an outgroup, including individuals of a different race, age, or sexual orientation, results in more positive attitudes toward, and less stereotyping of, the outgroup represented in the simulated contact ( Crisp & Turner, 2009 ). Simulated contact reduces anxiety associated with outgroup encounters ( Crisp & Turner, 2009 ) and thereby positively impacts psychological well-being.

Concluding Comments

Future event simulation clearly plays a functional role in memory and cognition, but also can produce distortions or errors, and in that sense constitutes a paradigmatic case of an adaptive constructive process. How can we reconcile the contrasting patterns of evidence reviewed in the previous sections? A key point arises from the observation made by such researchers as Gilbert and Wilson (2007) and Dunning (2007) that simulations of future events can result in inaccurate predictions regarding the future, and often provide an inadequate basis for planning, because they are incomplete in various ways. As a result, when a future scenario involves features or properties that are not represented when people imagine that scenario, but are relevant to how they will feel or perform when the scenario actually unfolds, individuals are very likely to be led astray by their incomplete, essentialized, or unrepresentative simulations. By contrast, simulations tend to be useful when they do represent critical features of an upcoming situation.

This point is illustrated nicely by studies from Taylor et al. (1998) referred to earlier. In one study, college students who simulated the specific steps that were important for success on an exam ( process simulation ; e.g., simulating themselves in the act of studying) began studying earlier, spent more time studying, and achieved a higher grade than did students who simulated how good they would feel if they received a high grade ( outcome simulation ). In a related study, students who constructed process simulations for an upcoming project that contained the steps critical to executing the project (e.g., imagining themselves gathering relevant materials and beginning to work on the project) were less prone to the planning fallacy than were students who constructed outcome simulations that did not contain the critical steps (e.g., how pleased they would be with the completed project). In both examples, simulations were useful only when they contained features that were critical to later task execution.

These considerations suggest that our understanding of both the benefits and foibles of future simulations will be improved by attempting to specify the conditions that promote a match or mismatch between the elements of a simulation and critical features of an upcoming event. The simulation elements and event features could entail steps necessary to perform a task or plan for its execution, an unresolved personal problem, or feelings about pleasant or unpleasant personal outcomes. Understanding the factors that promote match or mismatch between simulation elements and event features will, in turn, depend on better understanding how people retrieve and recombine information from memory to represent a future event ( Gilbert & Wilson, 2007 ; Schacter & Addis, 2007 ; Tulving, 2005 ). More generally, studying adaptive constructive processes should help to provide a deeper understanding of the functions of memory, in line with the agenda set forth by Bartlett (1932) eighty years ago. Bartlett emphasized not only the constructive nature of memory, but also the functions that memory serves in such diverse processes as interpretation, problem solving, and social cognition. A combined emphasis on constructive and functional processes should broaden our understanding of how memory links the past with the future.

Acknowledgments

I am deeply grateful to receive the APA Distinguished Scientific Contribution Award. The receipt of this award reflects the combined efforts of many individuals, including too many students and collaborators to name, and it is to them that I am the most grateful. The present article describes ideas and findings that have emerged mainly during the past few years, and to which many members of my lab and several collaborators have contributed. In particular, I wish to thank for discussion of the issues considered in the article, and comments on previous drafts that have helped to improve it, Donna Addis, Felipe de Brigard, Randy Buckner, Brendan Gaesser, Kathy Gerlach, Daniel Gilbert, Scott Guerin, Demis Hassabis, Kevin Madore, Susan McGlynn, Morris Moscovitch, Cliff Robbins, Nathan Spreng, and Karl Szpunar. Research from my lab described in the article was made possible by grants from the National Institute of Mental Health and National Institute on Aging, whose support has been essential to our work.

• Addis DR, Pan L, Vu MA, Laiser N, Schacter DL. Constructive episodic simulation of the future and the past: Distinct subsystems of a core brain network mediate imagining and remembering. Neuropsychologia. 2009; 47 :2222–2238. [ PubMed ] [ Google Scholar ]
• Addis DR, Schacter DL. The hippocampus and imagining the future: Where do we stand. Frontiers in Human Neuroscience. 2012; 5 :173. [ PMC free article ] [ PubMed ] [ Google Scholar ]
• Addis DR, Wong AT, Schacter DL. Remembering the past and imagining the future: Common and distinct neural substrates during event construction and elaboration. Neuropsychologia. 2007; 45 :1363–1377. [ PMC free article ] [ PubMed ] [ Google Scholar ]
• Addis DR, Wong AT, Schacter DL. Age-related changes in the episodic simulation of future events. Psychological Science. 2008; 19 :33–41. [ PubMed ] [ Google Scholar ]
• Andrews-Hanna JR, Reidler JS, Sepulcre J, Poulin R, Buckner RL. Functional-anatomic fractionation of the brain's default network. Neuron. 2010; 65 :550–602. [ PMC free article ] [ PubMed ] [ Google Scholar ]
• Arora S, Aggarwal R, Sirimanna P, Moran A, Grantcharov T, Kneebone R, et al. Mental practice enhances surgical technical skills: A randomized controlled study. Annals of Surgery. 2011; 253 :265–270. [ PubMed ] [ Google Scholar ]
• Baird B, Smallwood J, Schooler JW. Back to the future: autobiographical planning and the functionality of mind-wandering. Consciousness and Cognition. 2011; 20 :1604–1611. [ PubMed ] [ Google Scholar ]
• Bartlett FC. Remembering. Cambridge: Cambridge University Press; 1932. [ Google Scholar ]
• Benoit RG, Gilbert SJ, Burgess PW. A neural mechanism mediating the impact of episodic prospection on farsighted decisions. Journal of Neuroscience. 2011; 31 :6771–6779. [ PMC free article ] [ PubMed ] [ Google Scholar ]
• Boyer P. Evolutionary enonomics of mental time travel? Trends in Cognitive Sciences. 2008; 12 :219–224. [ PubMed ] [ Google Scholar ]
• Brainerd CJ, Reyna VF. The science of false memory. New York, NY: Oxford University Press; 2005. [ Google Scholar ]
• Brewer GA, Marsh RL. On the role of eposidic future simulation in encoding of prospective memories. Cognitive Neuroscience. 2010; 1 :81–88. [ PubMed ] [ Google Scholar ]
• Brown GP, MacLeod AK, Tata P, Goddard L. Worry and the simulation of future outcomes. Anxiety, Stress, and Coping. 2002; 15 :1–17. [ Google Scholar ]
• Buckner RL. The role of the hippocampus in prediction and imagination. Annual Review of Psychology. 2010; 61 :27–48. [ PubMed ] [ Google Scholar ]
• Buckner RL, Andrews-Hanna JR, Schacter DL. The brain's default network: Anatomy, function, and relevance to disease. The Year in Cognitive Neuroscience, Annals of the New York Academy of Sciences. 2008; 1124 :1–38. [ PubMed ] [ Google Scholar ]
• Buckner RL, Carroll DC. Self-projection and the brain. Trends in Cognitive Sciences. 2007; 11 :49–57. [ PubMed ] [ Google Scholar ]
• Buehler R, Griffin D, Peetz J. The planning fallacy: Cognitive, motivational, and social origins. In: Zanna MP, Olson JM, editors. Advances in experimental social psychology. Vol. 43. San Diego: Academic Press; 2010. pp. 1–62. [ Google Scholar ]
• Carroll JS. The effect of imagining an event on expectations for the event: An interpretation in terms of the availability heuristic. Journal of Experimental Social Psychology. 1978; 14 :88–96. [ Google Scholar ]
• Chasteen AL, Park DC, Schwarz N. Implementation intentions and facilitiation of prospective memory. Psychological Science. 2001; 12 :457–461. [ PubMed ] [ Google Scholar ]
• Christoff K, Gordon AM, Smallwood J, Smith R, Schooler JW. Experience sampling during fMRI reveals default network and executive system conributions to mind wandering. Proceedings of the National Academy of Sciences of the United States of America. 2009; 106 :8719–8724. [ PMC free article ] [ PubMed ] [ Google Scholar ]
• Clancy SA, Schacter DL, McNally RJ, Pitman RK. False recognition in women reporting recovered memories of sexual abuse. Psychological Science. 2000; 11 :26–31. [ PubMed ] [ Google Scholar ]
• Corebetta M, Shulman GL. Control of goal-directed and stimulus-driven attention in the brain. Nature Reviews Neuroscience. 2002; 3 :201–215. [ PubMed ] [ Google Scholar ]
• Crisp RJ, Turner RN. Can imagined interactions produce positive perceptions? Reducing prejudice through simulated social contact. American Psychologist. 2009; 64 :231–240. [ PubMed ] [ Google Scholar ]
• D'Argembeau A, Mathy A. Tracking the construction of episodic future thoughts. Journal of Experimental Psychology: General. 2011; 140 :258–271. [ PubMed ] [ Google Scholar ]
• D'Argembeau A, Raffard S, Van der Linden M. Remembering the past and imagining the future in schizophrenia. Journal of Abnormal Psychology. 2008; 117 :247–251. [ PubMed ] [ Google Scholar ]
• D'Argembeau A, Van der Linden M. Individual differences in the phenomenology of mental time travel. Consciousness and Cognition. 2006; 15 :342–350. [ PubMed ] [ Google Scholar ]
• Deese J. On the prediction of occurrence of particular verbal intrusions in immediate recall. Journal of Experimental Psychology. 1959; 58 :17–22. [ PubMed ] [ Google Scholar ]
• Dewhurst SA, Thorley C, Hammond ER, Ormerod TC. Convergent, but not divergent, thinking predicts susceptibility to associative memory illusions. Personality and Individual Differences. 2011; 51 :73–76. [ Google Scholar ]
• Dunning D. Prediction: The inside view. In: Higgins ET, Kruglanski A, editors. Social psychology: Handbook of basic principles. 2nd ed. New York: Guilford; 2007. pp. 69–90. [ Google Scholar ]
• Edelson M, Sharot T, Dolan RJ, Dudai Y. Following the crowd: brain substrates of long-term memory conformity. Science. 2011; 333 :108–111. [ PMC free article ] [ PubMed ] [ Google Scholar ]
• Ellamil M, Dobson C, Beeman M, Christoff K. Evaluative and generative modes of thought during the creative process. NeuroImage. 2012; 59 :1783–1794. [ PubMed ] [ Google Scholar ]
• Gallo DA. False memories and fantastic beliefs: 15 years of the DRM illusion. Memory and Cognition. 2010; 38 :833–848. [ PubMed ] [ Google Scholar ]
• Garry M, Manning C, Loftus EF, Sherman SJ. Imagination inflation: Imagining a childhood event inflates confidence that it occurred. Psychonomic Bulletin and Review. 1996; 3 :208–214. [ PubMed ] [ Google Scholar ]
• Gerlach KD, Spreng RN, Gilmore AW, Schacter DL. Solving future problems: Default network and executive activity associated with goal-directed mental simulations. NeuroImage. 2011; 55 :1816–1824. [ PMC free article ] [ PubMed ] [ Google Scholar ]
• Gilbert DT, Wilson T. Prospection: Experiencing the future. Science. 2007; 317 :1351–1354. [ PubMed ] [ Google Scholar ]
• Gollwitzer PM. Implementation intentions: Strong effects of simple plans. American Psychologist. 1999; 54 :493–503. [ Google Scholar ]
• Goodman GS, Ogle CM, Block SD, Harris LS, Larson RP, Augusti EM, et al. False memory for trauma-related Deese-Roediger-McDermott lists in adolescents and adults with histories of child sexual abuse. Development and Psychopathology. 2011; 23 :423–438. [ PMC free article ] [ PubMed ] [ Google Scholar ]
• Green L, Myerson J. A discounting framework for choice with delayed and probabilistic rewards. Psychological Bulletin. 2004; 130 :769–792. [ PMC free article ] [ PubMed ] [ Google Scholar ]
• Gregory RL, Gombrich EH. Illusion in nature and art. London: Duckworth; 1973. [ Google Scholar ]
• Hassabis D, Kumaran D, Maguire EA. Using imagination to understand the neural basis of episodic memory. Journal of Neuroscience. 2007; 27 :14365–14374. [ PMC free article ] [ PubMed ] [ Google Scholar ]
• Howe ML. The adaptive nature of memory and its illusions. Current Directions in Psychological Science. 2011; 20 :312–315. [ Google Scholar ]
• Howe ML, Garner SR, Charlesworth M, Knott L. A brighter side to memory illusions: False memories prime children's and adults' insight-based problem solving. Journal of Experimental Child Psychology. 2011; 108 :383–393. [ PubMed ] [ Google Scholar ]
• Ingvar DH. Hyperfrontal distribution of the cerebral grey matter flow in resting wakefulness: On the functional anatomy of the conscious state. Acta Neurologica Scandinavica. 1979; 60 :12–25. [ PubMed ] [ Google Scholar ]
• Ingvar DH. "Memory of the future": An essay on the temporal organization of conscious awareness. Human Neurobiology. 1985; 4 :127–136. [ PubMed ] [ Google Scholar ]
• Johnson MK, Hashtroudi S, Lindsay DS. Source monitoring. Psychological Bulletin. 1993; 114 :3–28. [ PubMed ] [ Google Scholar ]
• Kahneman D, Tversky A. Intuitive prediction: Biases and corrective procedures. TIMS Studies in Management Science. 1979; 12 :313–327. [ Google Scholar ]
• Klein SB, Robertson TE, Delton AW. Facing the future: memory as an evolved system for planning future acts. Memory and Cognition. 2010; 38 :13–22. [ PMC free article ] [ PubMed ] [ Google Scholar ]
• Klein SB, Robertson TE, Delton AW. The future-orientation of memory: planning as a key component mediating the high levels of recall found with survival processing. Memory. 2011; 19 :121–139. [ PMC free article ] [ PubMed ] [ Google Scholar ]
• Koehler DJ. Explanation, imagination, and confidence in judgment. Psychological Bulletin. 1991; 110 :499–519. [ PubMed ] [ Google Scholar ]
• Koutstaal W, Schacter DL. Gist-based false recognition of pictures in older and younger adults. Journal of Memory and Language. 1997; 37 :555–583. [ Google Scholar ]
• Loftus EF. Make-believe memories. American Psychologist. 2003; 58 :867–873. [ PubMed ] [ Google Scholar ]
• Loftus EF. Planting misinformation in the human mind: a 30-year investigation of the malleability of memory. Learning & Memory. 2005; 12 :361–366. [ PubMed ] [ Google Scholar ]
• Maguire EA, Vargha-Khadem F, Hassabis D. Imagining fictitious and future experiences: evidence from developmental amnesia. Neuropsychologia. 2010; 48 :3187–3192. [ PMC free article ] [ PubMed ] [ Google Scholar ]
• Martin VC, Schacter DL, Corballis MC, Addis DR. A role for the hippocampus in encoding simulations of future events. Proceedings of the National Academy of Sciences of the United States of America. 2011; 108 :13858–13863. [ PMC free article ] [ PubMed ] [ Google Scholar ]
• Mason MF, Norton MI, Van Horn JD, Wegner DM, Grafton ST, Macrae CN. Wandering minds: The default network and stimulus-independent thought. Science. 2007; 315 :393–395. [ PMC free article ] [ PubMed ] [ Google Scholar ]
• Mathieu MT, Gosling SD. The accuracy or inaccuracy of affective forecasts depends on how accuracy is indexed: A meta-analysis of past studies. Psychological Science. 2012; 23 :161–162. [ PubMed ] [ Google Scholar ]
• McClelland JL. Constructive memory and memory distortions: A parallel-distributed processing approach. In: Schacter DL, editor. Memory distortion: How minds, brains and societies reconstruct the past. Cambridge, MA: Harvard University Press; 1995. pp. 69–90. [ Google Scholar ]
• McKay R, Dennett D. The evolution of misbelief. Behavioral and Brain Sciences. 2009; 32 :493–561. [ PubMed ] [ Google Scholar ]
• Morewedge CK, Gilbert DT, Wilson TD. The least likely of times: How rememering the past biases forecasts of the future. Psychological Science. 2005; 16 :626–630. [ PubMed ] [ Google Scholar ]
• Narine JS. Adaptive memory: Evolutionary constraints on remembering. In: Ross BH, editor. The psychology of learning and motivation. Vol. 53. London: Academic Press; 2010. pp. 1–32. [ Google Scholar ]
• Neisser U. Cognitive psychology. New York: Appleton-Century-Crofts; 1967. [ Google Scholar ]
• Newman EJ, Lindsay SD. False memories: What the hell are they for? Applied Cognitive Psychology. 2009; 23 :1105–1121. [ Google Scholar ]
• Okuda J, Fujii T, Ohtake H, Tsukiura T, Tanji K, Suzuki K, et al. Thinking of the future and the past: The roles of the frontal pole and the medial temporal lobes. NeuroImage. 2003; 19 :1369–1380. [ PubMed ] [ Google Scholar ]
• Papies EK, Aarts H, de Vries NK. Planning is doing: Implementation intentions go beyond the mere creation of goal-directed associations. Journal of Experimental Social Psychology. 2009; 45 :1148–1151. [ Google Scholar ]
• Peters J, Büchel C. Episodic future thinking reduces reward delay discounting through an enhancement of prefrontal-mediotemporal interactions. Neuron. 2010; 66 :138–148. [ PubMed ] [ Google Scholar ]
• Raichle ME, MacLeod AM, Snyder AZ, Powers WJ, Gusnard DA, Shulman GL. A default mode of brain function. Proceedings of the National Academy of Sciences of the United States of America. 2001; 16 :676–682. [ PMC free article ] [ PubMed ] [ Google Scholar ]
• Roediger HL., III Memory illusions. Journal of Memory and Language. 1996; 35 :76–100. [ Google Scholar ]
• Roediger HL, III, McDermott KB. Creating false memories: Remembering words not presented in lists. Journal of Experimental Psychology: Learning, Memory, and Cognition. 1995; 21 :803–814. [ Google Scholar ]
• Roy MM, Christenfeld NJS, McKenzie CRM. Underestimating the duration of future events: Memory incorrectly used or memory bias? Psychological Bulletin. 2005; 131 :738–756. [ PubMed ] [ Google Scholar ]
• Schacter DL. The seven sins of memory: Insights from psychology and cognitive neuroscience. American Psychologist. 1999; 54 :182–203. [ PubMed ] [ Google Scholar ]
• Schacter DL. The seven sins of memory: How the mind forgets and remembers. Boston: Houghton Mifflin; 2001. [ Google Scholar ]
• Schacter DL, Addis DR. The cognitive neuroscience of constructive memory: Remembering the past and imagining the future. Philosophical Transaction of the Royal Society of London B. 2007; 362 :773–786. [ PMC free article ] [ PubMed ] [ Google Scholar ]
• Schacter DL, Addis DR. On the nature of medial temporal lobe contributions to the constructive simulation of future events. Philosophical Transactions of the Royal Society (B) 2009; 364 :1245–1253. [ PMC free article ] [ PubMed ] [ Google Scholar ]
• Schacter DL, Addis DR, Buckner RL. Remembering the past to imagine the future: The prospective brain. Nature Reviews Neuroscience. 2007; 8 :657–661. [ PubMed ] [ Google Scholar ]
• Schacter DL, Addis DR, Buckner RL. Episodic simulation of future events: Concepts, data, and applications. The Year in Cognitive Neuroscience, Annals of the New York Academy of Sciences. 2008; 1124 :39–60. [ PubMed ] [ Google Scholar ]
• Schacter DL, Gaesser B, Addis DR. Age-related changes in the episodic simulation of past and future events. In: Benjamin AS, editor. Successful remembering and successful forgetting. New York: Psychology Press; 2011. pp. 505–525. [ Google Scholar ]
• Schacter DL, Guerin SA, St Jacques PL. Memory distortion: an adaptive perspective. Trends in Cognitive Sciences. 2011; 15 :467–474. [ PMC free article ] [ PubMed ] [ Google Scholar ]
• Schacter DL, Slotnick SD. The cognitive neuroscience of memory distortion. Neuron. 44 :149–160. [ PubMed ] [ Google Scholar ]
• Shallice T. Specific impairments of planning. Philosophical Transactions of the Royal Society of London B. 1982; 298 :199–209. [ PubMed ] [ Google Scholar ]
• Sharot T. The optimism bias. New York: Pantheon Books; 2011. [ Google Scholar ]
• Sheldon S, McAndrews MP, Moscovitch M. Episodic memory processes mediated by the medial temporal lobes contribute to open-ended problem solving. Neuropsychologia. 2011; 49 :2439–2447. [ PubMed ] [ Google Scholar ]
• Shulman GL, Fiez JA, Corbetta M, Buckner RL, Miezin FM, Raichle ME, et al. Common blood flow changes across visual tasks: II. Decreases in cerebral cortex. 1997; 9 :648–663. [ PubMed ] [ Google Scholar ]
• Spreng RN, Grady CL. Patterns of brain activity supporting autobiographical memory, prospection, and theory-of-mind and their relationship to the default mode network. Journal of Cognitive Neuroscience. 2010; 22 :1112–1123. [ PubMed ] [ Google Scholar ]
• Spreng RN, Schacter DL. Default network modulation and large-scale network interactivity in healthy young and older adults. Cerebral Cortex. (in press). [ PMC free article ] [ PubMed ] [ Google Scholar ]
• Spreng RN, Stevens WD, Chamberlain JP, Gilmore AW, Schacter DL. Default network activity, coupled with the frontoparietal control network, supports goal-directed cognition. NeuroImage. 2010; 53 :303–317. [ PMC free article ] [ PubMed ] [ Google Scholar ]
• Squire LR, van der Horst AS, McDuff SGR, Frascino JC, Hopkinse RO, Mauldin KN. Role of the hippocampus in remembering the past and imagining the future. Proceedings of the National Academy of Sciences of the United States of America. 2010; 107 :19044–19048. [ PMC free article ] [ PubMed ] [ Google Scholar ]
• St. Jacques PL, Schacter DL. Modifying memory: Selectively enhancing and updating personal memories for a museum tour by reactivating them. Psychological Science. (in press). [ PMC free article ] [ PubMed ] [ Google Scholar ]
• Stawarczyk D, Majerus S, Maj M, Van der Linden M, D'Argembeau A. Mind-wandering: Phenomenology and function as assessed with a novel experience sampling method. Acta Psychologica. 2011; 136 :370–381. [ PubMed ] [ Google Scholar ]
• Suddendorf T, Corballis MC. The evolution of foresight: What is mental time travel and is it unique to humans? Behavioral and Brain Sciences. 2007; 30 :299–313. [ PubMed ] [ Google Scholar ]
• Szpunar KK. Episodic future thought: An emerging concept. Perspectives on Psychological Science. 2010; 5 :142–162. [ PubMed ] [ Google Scholar ]
• Szpunar KK, Addis DR, Schacter DL. Memory for emotional simulations: Remembering a rosy future. Psychological Science. 2012; 23 :24–29. [ PMC free article ] [ PubMed ] [ Google Scholar ]
• Szpunar KK, McDermott KB. Episodic future thought and its relation to remembering: Evidence from ratings of subjective experience. Consciousness and Cognition. 2008; 17 :330–334. [ PubMed ] [ Google Scholar ]
• Szpunar KK, Schacter DL. Get real: Effects of repeated simulation and emotion on the percieved plausibility of future interpersonal experiences. Journal of Experimental Psychology: General. (in press). [ PMC free article ] [ PubMed ] [ Google Scholar ]
• Szpunar KK, Watson JM, McDermott KB. Neural substrates of envisioning the future. Proceedings of the National Academy of Sciences of the United States of America. 2007; 104 :642–647. [ PMC free article ] [ PubMed ] [ Google Scholar ]
• Taylor SE. Positive illusions. New York: Basic Books; 1989. [ Google Scholar ]
• Taylor SE, Pham LB, Rivkin ID, Armor DA. Harnessing the imagination: Mental simulation, self-regulation, and coping. American Psychologist. 1998; 53 :429–439. [ PubMed ] [ Google Scholar ]
• Tulving E. Episodic memory and autonoesis: Uniquely human? In: Terrace HS, Metcalfe J, editors. The missing link in cognition: Origins of self reflective consciousness. Oxford: Oxford University Press; 2005. pp. 3–56. [ Google Scholar ]
• Tversky A, Kahneman D. Judgment under uncertainty: Heuristics and biases. Science. 1974; 185 :1124–1131. [ PubMed ] [ Google Scholar ]
• van Meer JP, Theunissen NCM. Prospective educational applications of mental simulation: a meta-review. Educational Psychology Review. 2009; 21 :93–112. [ Google Scholar ]
• Vincent JL, Kahn I, Snyder AZ, Raichle ME, Buckner RL. Evidence for a frontoparietal control system revealed by intrinsic functional connectivity. Journal of Neurophysiology. 2008; 100 :3328–3342. [ PMC free article ] [ PubMed ] [ Google Scholar ]
• Williams JM. Depression and the specificity of autobiographical memory. In: Rubin DC, editor. Remembering our past: Studies in autobiographical memory. Cambridge: Cambridge University Press; 1996. pp. 244–267. [ Google Scholar ]
• Zhu B, Chen C, Loftus EF, Lin C, He Q, Li H, et al. Individual differences in false memory from misinformation: cognitive factors. Memory. 2010; 18 :543–555. [ PubMed ] [ Google Scholar ]

Concurrent prospective memory task increases mind wandering during online reading for difficult but not easy texts

Affiliations.

• 1 Center of Multimedia Teaching and Learning, Martin Luther University Halle-Wittenberg, Hoher Weg 8, 06120, Halle, Germany. [email protected].
• 2 Center of Multimedia Teaching and Learning, Martin Luther University Halle-Wittenberg, Hoher Weg 8, 06120, Halle, Germany.
• 3 School of Psychology, University of Surrey, Guildford, UK.
• 4 Center of Multimedia Teaching and Learning, Martin Luther University Halle-Wittenberg, Hoher Weg 8, 06120, Halle, Germany. [email protected].
• 5 Institute of Psychology, Martin Luther University Halle-Wittenberg, Halle, Germany. [email protected].
• PMID: 35233743
• DOI: 10.3758/s13421-022-01295-1

Many prior theories have tried to explain the relationship between attentional processes and mind wandering. The resource-demand matching view argues that a mismatch between task demands and resources led to more mind wandering. This study aims to test this view against competing models by inducing mind wandering through increasing the level of demands via adding a prospective memory task to cognitively demanding tasks like reading. We hypothesized that participants with a second task still in mind (unfinished group) engage more in task-unrelated thoughts (TUTs) and show less text comprehension compared to participants who think a second task is finished (finished group). Seventy-two participants had to study 24 items of a to-do list for a recall test. After a first cued recall of ten items, participants were either told that a second task was finished or that the recall was interrupted and continued later. All participants then started reading an easy or difficult version of the same unfamiliar hypertext, while being thought probed. Text comprehension measures followed. As expected, participants in the unfinished group showed significantly more TUTs than participants in the finished group when reading difficult texts, but, contrary to our assumptions, did not show better text comprehension measures when reading difficult text. Nevertheless, participants compensate for the influence of the second task by reading longer, which in turn has a positive effect on their reading knowledge. These findings support the resource-demand-matching model and thus strengthen assumptions about the processing of attention during reading.

Keywords: Attention; Mind wandering; Reading.

© 2022. The Psychonomic Society, Inc.

Publication types

• Research Support, Non-U.S. Gov't
• Comprehension
• Memory, Episodic*
• Mental Recall