Cognition 111 (2009) 345–355

Contents lists available at ScienceDirect

Cognition

journal homepage: www.elsevier.com/locate/COGNIT

Harnessing the wandering mind: The role of perceptual load

Sophie Forster *, Nilli Lavie

Institute of Cognitive Neuroscience and Research Department of Cognitive, Perceptual and Brain Sciences, University College London, Gower Street, London WC1E 6BT, UK article info abstract

Article history: Perceptual load is a key determinant of distraction by task-irrelevant stimuli (e.g., Lavie, N. Received 6 May 2008 (2005). Distracted and confused?: Selective attention under load. Trends in Cognitive Sci- Revised 26 January 2009 ences, 9, 75–82). Here we establish the role of perceptual load in determining an internal Accepted 19 February 2009 form of distraction by task-unrelated thoughts (TUTs or ‘‘mind-wandering”). Four experiments demonstrated reduced frequency of TUTs with high compared to low perceptual load in a visual-search task. Alternative accounts in terms of increased demands Keywords: on responses, verbal working memory or motivation were ruled out and clear effects of Task-unrelated thoughts load were found for unintentional TUTs. Individual differences in load effects on internal Mind-wandering Perceptual load (TUTs) and external (response-competition) distractors were correlated. These results sug- Attention gest that exhausting attentional capacity in task-relevant processing under high perceptual Response competition load can reduce processing of task-irrelevant information from external and internal Distractor interference sources alike. Ó 2009 Elsevier B.V. Open access under CC BY license.

1. Introduction significantly reduced with tasks of high (compared to low) perceptual load (e.g., Lavie, 1995; Rees, Frith, & Lavie, A main goal of attention research is to understand the 1997; see Lavie, 2005, for review). This conclusion has gen- determinants of successful focused attention that allows eralized across various manipulations of perceptual load for minimal distraction by goal-irrelevant information. (that have either involved a greater number of items This fundamental issue has stimulated much research over requiring identification in the high load conditions or a the past four decades (e.g., see Kahneman and Treisman greater complexity of the perceptual task, see Lavie, (1984), Lavie and Tsal (1994), Lavie (1995) for reviews) 2005, for review) and across various measures of distractor and a major determinant of focused attention that has processing. For example, perceptual load has been shown been highlighted is the level of perceptual load in a task. to reduce (and indeed typically eliminate) distractor inter- The role of perceptual load in attention has been eluci- ference effects measured with response competition dated within the perceptual load theory (Lavie, 1995; Lavie (Lavie, 1995; Lavie & Cox, 1997; Lavie, Ro, & Russell, & Tsal, 1994), which suggests that distractor processing 2003) and negative priming (Lavie & Fox, 2000). Perceptual critically depends on the availability of attentional capacity load has also been shown to modulate brain activity re- and can thus be prevented when the relevant-task process- lated to distractors whether these are visible (Rees et al., ing involves sufficient high perceptual load to engage full 1997; Schwartz et al., 2005; Yi, Woodman, Widders, attentional capacity. Marois, & Chun, 2004) or even invisible (Bahrami, Lavie, Evidence in support of this claim has been found in and Rees, 2007). many studies demonstrating that distractor processing is Recent studies have demonstrated that perceptual load effects can overcome individual differences in distractibil- ity (Forster & Lavie, 2007) and can also eliminate the ef- fects of highly salient distractors that, like many daily * Corresponding author. Tel.: +44 20 7679 5552; fax: +44 20 74364276. E-mail addresses: [email protected], sophie_forster@hotmail. life distractors, cause interference despite bearing no rela- com (S. Forster). tionship to the task currently being performed (Forster &

0010-0277 Ó 2009 Elsevier B.V. Open access under CC BY license. doi:10.1016/j.cognition.2009.02.006 346 S. Forster, N. Lavie / Cognition 111 (2009) 345–355

Lavie, 2008a; Forster & Lavie, 2008b). However, in daily 1970; Smallwood, Baracaia, Lowe, & Obonsawin, 2003; life sources of distraction may not only be found in the Smallwood, O’Connor, Sudberry, Haskell, & Ballantyne, external environment, but also in the form of internally 2004; Smallwood, McSpadden, & Schooler, 2007; see also generated distractions such as task-unrelated thoughts Smallwood & Schooler, 2006; Smallwood, Fishman, & (TUTs). For example, a person may be distracted from Schooler, 2007 for review). reading this article by the intrusion of a thought about Despite these apparent distracting effects of mind-wan- an unrelated issue - perhaps some salient recent event dering, however, no study as yet has demonstrated any in his or her daily life. The purpose of this paper is to clar- causal role for focused attention in determining levels of ify whether the established role of perceptual load in TUT. A number of studies have directly manipulated task determining task-irrelevant processing would also apply factors that were found to decrease the level of task-unre- to internal sources of potential distraction, such as dis- lated thoughts,1 but none has yet used an established traction by TUTs. manipulation of focused attention. Moreover, in almost all We reasoned that, similarly to the role of load in pro- cases these manipulations are likely to have interfered di- cessing task-irrelevant information from external sources, rectly with the very ability to produce or maintain a thought. the processing of task-irrelevant information from internal Thus the role of focused attention in TUTs (or mind-wander- sources such as mind-wandering may also be determined ing) remains unclear. by perceptual load. The load research we previously men- For example, it has been shown that working memory tion has established that processing of task-irrelevant load can reduce TUT (Teasdale et al, 1995; Teasdale, stimuli with very different contents (e.g., verbal vs. visual Proctor, Lloyd, & Baddeley, 1993), yet working memory is and complex stimuli such as words or visual scenes vs. clearly needed in order to provide a mental work space simple stimuli such as letters or contrast and motion, for thought (for example one would need to maintain the Brand-D’Abrescia & Lavie, 2007; Lavie, 2006; Macdonald start of the thought in order to develop its semantics in a & Lavie, 2008; Rees et al., 1997; Schwartz et al., 2005) coherent manner in which the end relates to the start, and sources (e.g., different sensory modalities, conscious see Baddeley (1986)). It has also been shown that perform- as well as unconscious processes, e.g., (Bahrami, Carmel, ing a task relative to no task at all or increasing stimulus Walsh, Rees, & Lavie, 2008; Cartwright-Finch & Lavie, presentation rates can reduce the rates of TUT reports 2006; Dalton, Lavie, and Spence, 2008; Macdonald & Lavie, (e.g., Giambra, 1995; McKiernan, D’Angelo, Kaufman, & 2008) is modulated by the level of load in the task-relevant Binder, 2006). However, not only do conditions of task per- processing. These findings provide support for the Load formance (compared to no task performance) and faster Theory claim that the processing of any task-irrelevant (vs. slower) presentation rates involve a higher working information (be it verbal or visual, simple or complex, con- memory load, but they also involve increased demands scious or unconscious, in one sensory modality or another, on responses (as with higher presentation rate the re- and so forth) requires limited-capacity attention and can sponse rate is also higher). The simple act of making a re- therefore occur only when the processing of task-relevant sponse has been shown to directly interfere with the rate information leaves some spare attentional capacity (under of TUT reports (e.g., see Antrobus (1968)). By contrast, here conditions of low perceptual load). We thus reasoned that we ask whether it is possible to reduce task-unrelated higher perceptual load that would engage more attentional thoughts by engaging more attentional capacity in a task capacity in the processing of task-relevant information with high (compared to low) perceptual load, without di- may reduce the processing of task-irrelevant information rectly drawing on thought or response components that not only from external sources (e.g., visual distractors) are clearly part and parcel of the production or report of but also from internal sources (e.g., mind-wandering) in any thought. the present study. The suggestion of a recent review of the mind-wander- ing literature that the rate of TUTs reported is higher in 1.1. Mind-wandering measures and previous findings tasks that require only ‘‘superficial engagement” compared to those that require moderate or deeper levels of engage- In comparison to the extensive body of research exam- ment (Smallwood, Fishman et al., 2007) is encouraging for ining the attention principles that govern distraction by our present load hypothesis. However, the broad term external stimuli, the topic of attention and distraction by ‘‘task-engagement” may encompass a number of factors mind-wandering has been relatively understudied, per- in addition to attention (e.g., changes in the overall level haps due to lack of any truly objective method for directly of motivation, interest and arousal, as well as the engage- measuring the occurrence of such a highly subjective phe- ment of processes such as working memory and thought) nomenon. However, laboratory measures of mind-wander- and the tasks presumed in the review to involve differing ing have been developed (e.g., measuring TUT occurrence levels of task engagement (e.g., signal detection tasks were with probe-caught methods, whereby on the appearance assumed to involve superficial engagement, whereas read- of a probe the subject has to report whether or not they ing tasks were assumed to always involve deep engage- had just experienced TUT), and studies using these ment) also differed in terms of many of these factors. methods suggest that mind-wandering is a ubiquitous Therefore any variation in the rate of TUTs reported during and potent source of distraction: TUTs often occur uninten- tionally and interfere with performance on a range of tasks from signal detection tasks to encoding and reading tasks 1 Note that we restrict our review only to manipulations that produced (Schooler, Reichle, & Halpern, 2004; Segal & Fusella, significant effects on the level of TUT reported. S. Forster, N. Lavie / Cognition 111 (2009) 345–355 347 these different tasks cannot be safely attributed solely to of search non-targets2 (often termed ‘‘set size”) is very well the level of attentional engagement, rather than to these established (see Lavie (2005) for review) and has been other factors. shown to reduce distractor perception and related activity Recent findings of an association between the speeding in the visual cortex in many previous studies (see of responses in a go/no go task (which may reflect less Macdonald and Lavie (2008) for a recent demonstration careful task-performance stemming from reduced task- and review). engagement) and a higher rate of TUT reports (Smallwood Notice in particular that the controlled aspects of the et al., 2004; Smallwood, McSpadden et al., 2007) also pro- task (e.g., maintaining the target template in working vide some suggestive evidence that the level of attentional memory) remain constant across the levels of perceptual engagement in a task may determine TUT. However, load, only the number of search items is varied. As such, changes in the speed of responses may be driven by a num- this increase in the number of items places demands on ber of factors other than attentional engagement in the perceptual search processes (e.g., serial spatial scanning task (e.g., practice, arousal, motivation). Furthermore, in of a greater number of search items, see Treisman (1988), the absence of any manipulation of attention it is impossi- Treisman and Gelade (1980)) rather than control pro- ble to determine whether it was the speeding of responses cesses. Indeed, previous research has indicated that (and putative task-disengagement it reflects) that led to a increasing perceptual load in this task does not interfere higher rate of TUT or whether the occurrence of TUT led with working memory task performance, nor does it inter- to the putative attention disengagement and hence the act with the effects of working memory load on distraction speeding of the task responses. (e.g., Lavie et al., 2004; Logan, 1978). Thus, any effects of The present study is thus the first to examine whether perceptual load on TUT in this task are unlikely to reflect focused attention plays a causal role in determining direct interference with thought components, or with their mind-wandering, by directly manipulating the level of de- executive control (e.g., by blocking working memory). mand on attentional capacity (through our manipulation of perceptual load) in visual search: a paradigmatic and 2. Experiment 1 widely used attention task. Finally, it is important to note that some effects of a sec- In Experiments 1a and 1b participants performed a let- ondary task or working memory load on mind-wandering ter search task with high and low perceptual load, during have previously been interpreted as indicating that mind- which they responded to thought probes. In the original wandering draws on central executive resources (e.g., perceptual load paradigm on which our task is based (Lavie Smallwood & Schooler, 2006; Teasdale et al., 1995) and & Cox, 1997) feedback was given for errors in the form of a in some cases these have been referred to as ‘‘executive brief ‘‘beep” tone. As the error rate is typically higher in the attention” (Smallwood & Schooler, 2006). However recent high load condition this means that the high load condition advances in attention research demonstrate that ‘atten- would be associated with more frequent feedback tones. It tion’ cannot be used as an umbrella term encompassing was therefore important to establish that any reduction in all limited-capacity processes, instead it is important to the level of TUT reports under high perceptual load did not distinguish between the effects on focused attention of simply reflect an effect of the greater number of feedback capacity limits in perception vs. in executive control pro- tones in this condition. For example, as abrupt onsets have cesses. Whereas exhausting perceptual capacity in tasks been shown to capture attention during the performance of high perceptual load reduces distractor processing and of an external task (e.g., see Jonides and Yantis (1988), thus improves focused attention, load on executive control Yantis and Jonides (1984)), the greater number of abrupt processes such as working memory during selective atten- sound onsets in the high load condition might capture tion tasks has the opposite effect of increasing rather than attention towards the external environment and task and decreasing distractor processing. Since load on working therefore away from any task-unrelated thoughts. In addi- memory reduces its availability to actively maintain fo- tion, the higher level of feedback may have increased moti- cused attention in line with current stimulus processing vation to remain on task in the high (compared to low) priorities the irrelevant distractor stimuli gain more pro- load condition. Indeed feedback tones can be used to im- cessing access (e.g., De Fockert, Rees, Frith, & Lavie, 2001; prove participants’ ability to maintain sustained attention Lavie, 2000; Lavie, Hirst, de Fockert, & Viding, 2004). to a task (e.g., Manly et al., 2004). For this reason, we ran We consider the implications of this dissociation for two versions of Experiment 1, with and without the error mind-wandering later (in the General Discussion). For feedback. Participants in Experiment 1a were presented now we note that in the present study, given our focus with error feedback, but those in Experiment 1b were gi- on perceptual load rather than executive control, our ven feedback only during the practice trials and not during manipulation of perceptual load was designed to minimize the main task performance. any confounding involvement of higher levels of executive control resources with the increase in perceptual load. To this end we incorporated thought probes into a typ- ical perceptual load paradigm involving a visual search 2 Note that the effects of perceptual load have previously been demon- task in which the number of search non-targets was varied strated using both manipulations of set size and manipulations of the complexity of processing demands (e.g., see Lavie (1995) for review see (from one to six items) to produce high or low perceptual Lavie (2005)). We have chosen to use the search set size manipulation for load (e.g., Lavie, 1995; Lavie, 2005; Lavie & Cox, 1997). this study as this manipulation has been shown not to interfere with The manipulation of perceptual load through the number working memory task performance, as discussed above. 348 S. Forster, N. Lavie / Cognition 111 (2009) 345–355

2.1. Method possible while also maintaining a high level of accuracy. They were told that there would be no need to prepare 2.1.1. Participants for the thought probes (e.g., by keeping their fingers on All participants in the reported experiments were vol- the response keys) before they appear on screen, as unteers recruited from University College London subject instructions for the responses to the thought probes would pool and had normal or corrected to normal vision. Ten be presented with each probe. They were also reassured participants (three males, 20–35 years old) participated that there was no right or wrong answer to the probe ques- in Experiment 1a. Fourteen participants (three males, 8– tion and that their probe responses were not being timed, 35 years old) participated in Experiment 1b. The data from so they should therefore simply respond honestly to the one additional participant in Experiment 1b was not in- question. Participants then completed eight high load and cluded in the analysis as this participant performed at eight low load blocks of 48 trials in the order ABBAABBA- chance level on the task. ABBAABBA (counterbalanced between participants), with a thought probe presented at the end of each block. 2.1.2. Stimuli and procedure E-Prime was used to program and run the experiment. 2.2. Results and discussion All stimuli were presented in white on a black background, on a 15 in. monitor and at a viewing distance of approxi- All RT analyses in the experiments reported here were mately 60 cm. Each trial in the visual search task began conducted only on trials to which a correct response was with 500 ms black screen interval, followed by a fixation made. Analyses of RTs and percentage error (PE) rates in point presented for 500 ms in the centre of the screen. This the high and low load conditions replicated previous find- was followed immediately by the stimulus display, con- ings (e.g., Lavie & Cox, 1997), confirming that the increase sisting of six letters arranged to form a circle (with the ra- in the relevant search set size (through the number of dius subtending 1.6° degrees of visual angle), presented for angular letters) was effective in increasing load. Partici- 100 ms. Participants were required to search the display pants’ RTs were significantly longer under high load (for for a target letter (either X or N, subtending 0.6° Â 0.4°) Experiment 1a M = 644 ms; for Experiment 1b M = and respond using the numerical keypad by pressing the 706 ms) than under low load (for Experiment 1a M = 0 key if the target was an X and the 2 key if the target 486 ms; for Experiment 1b M = 490 ms, t (9) = 14.11, was an N. In Experiment 1a, a 90 ms beep sounded on SEM = 11.20, p < .001, d = 9.41 for the difference in Experi- incorrect responses or if the participant failed to respond ment 1a; t (13) = 11, SEM = 19.66, p < .001, d = 6.10 for the within a 3000 ms time window. In Experiment 1b, partici- difference in Experiment 1b). pants were given this same feedback during the practice The high load condition was also associated with a sig- trials, but were given no feedback during the main exper- nificantly higher PE rate (for Experiment 1a, M = 16.4%; for imental trials. In the low load condition the five non-target Experiment 1b, M = 9.9%) than the low load condition (for letters were all small Os (subtending 0.15°), whereas in the Experiment 1a, M = 7.2%, for Experiment 1b, M = 4.7%), t high load the non-target letters were heterogeneous angu- (9) = 5.18, SEM = 1.78, p = .001, d = 3.45 for the difference lar letters (selected at random from H, K, M, Z, W, V). All in Experiment 1a; t (13) = 4.99, SEM = 1.03, p < .001, combinations of target identity and position were fully d = 2.77 for the difference in Experiment 1b. counterbalanced. Across the load conditions, the percentage of thought Participants performed 3 slower low and high load probes to which a TUT was reported varied between 25% example trials before receiving definitions and examples and 88% (M = 52.63%, S.D. = 16.19). Importantly, paired t- of task-related and task-unrelated thoughts. Task related tests confirmed our main prediction that perceptual load thoughts were defined as being thoughts about any aspect would reduce TUTs: As can be seen in Fig. 1, the percentage of the task being performed at that moment, with exam- of TUTs reported was significantly reduced in the high load ples being ‘‘where is the target letter?” or ‘‘oops, I’ve condition (for Experiment 1a, M = 48.9%, for Experiment pressed the wrong button”. Task-unrelated thoughts were 1b, M = 44%), compared to the low load condition (for defined as being thoughts about anything other than the Experiment 1a, M = 61.6%, for Experiment 1b, M = 57.5%), task being performed, with examples being ‘‘my lecture t (9) = 2.40, SEM = 5.29, p = .04, d = 1.60 for the difference this morning was boring.” or ‘‘I must stop by the supermar- in Experiment 1a, t (13) = 2.16, SEM = 6.24, p = .05, ket on the way home.” Participants then completed twelve d = 1.20 for the difference in Experiment 1b. Clearly then, practice trials of each level of load with a practice thought Experiment 1b establishes that the load effects on TUT per- probe after each set of twelve trials. The thought probe sist in the absence of any feedback and so cannot be attrib- consisted of the question ‘‘What were you thinking just uted to the greater likelihood of error feedback in the high now?”. Participants were instructed to report the thought (compared to low) load condition. that had been passing through their mind in the moment A mixed model ANOVA with the between-subject fac- just before the probe appeared. Onscreen instructions tors of experiment (Experiments 1a and 1b) and the with- asked the participant to press A if they were thinking about in-subject factor of load (low, high) revealed that the the task that they were performing, and to press Z if they somewhat surprising numerical trend for an increase were thinking about something unrelated to the immedi- (rather than decrease) in the rate of TUT reports in Exper- ate task. Participants were told that during the letter task iment 1a (M = 55.25%) compared with Experiment 1b they should keep their fingers ready on the relevant keys (M = 50.75%) was not significant and neither was the inter- and try to focus on the task in order to respond as fast as action of experiment and load (both Fs < 1). The finding S. Forster, N. Lavie / Cognition 111 (2009) 345–355 349

75 Low load 70 High Load 65 60 55 50 % TUT % TUT reports 45 40 35 Experiment 1a Experiment 1b Experiment 2 Experiment 3

Fig. 1. Percentage of probes at which a task-unrelated thought (TUT) was reported under high and low perceptual load in Experiments 1–3. that TUT reports were clearly not reduced with the pres- set under high load. To rule out this possibility, in Experi- ence of feedback (in Experiment 1a) does not rule out the ment 2 we used Hebrew letters for the search stimuli possibility that feedback can reduce TUT reports when and recruited English participants who were unfamiliar subjects are explicitly instructed to use the feedback as a with Hebrew, so that our search stimuli now appeared as reminder to stay on task (cf. Manly et al., 2004). However, meaningless shapes in both of the low and high load by demonstrating clear effects of load on TUT in paradigms conditions. both with and without feedback we have ruled out the any alternative interpretation of these load effects as being dri- 3.1. Method ven by the greater levels of feedback in the high load condition. 3.1.1. Participants In addition, we note that, as each trial was triggered by Twelve new participants (two males, 8–23 years old) the response to the previous trial, the greater RTs in the participated in Experiment 2. None of the subjects were high load condition would have also resulted in a slower able to read Hebrew or name any Hebrew letters. stimulus presentation rate in this condition. However, as previous studies (e.g., Antrobus, 1968; Giambra, 1995) 3.1.2. Stimuli and procedure have consistently reported increases in TUT as stimulus All stimuli and procedure were identical to those used presentation rate decreases, our finding of a decrease in in Experiment 1a, with the exception that Hebrew letters TUT reports under high load cannot be accounted for by were used in place of the search letters of the previous the slower stimulus presentation rate in this condition. experiments (see Fig. 2 for an example stimulus display). Thus, Experiment 1 establishes that perceptual load, al- The target letters were ‘ ’ (mapped to the response key 2) ready known to be a powerful determinant of the process- and ‘ ’ (mapped to the response key 0). In the low load con- ing of task-unrelated stimuli in the external world, may dition the non-targets were all small dots (subtending also determine the level of internal distractors in the form 0.2° Â 0.2°). In the high load condition three Hebrew of task-unrelated thoughts. As no error feedback was given non-target letters, selected at random from the set , , , in Experiment 1b, these effects cannot be attributed to any were presented with equal likelihood in three of the five effect of feedback (e.g., in capturing attention towards the positions not occupied by the target, with the remaining task or in increasing the overall level of motivation in the two positions occupied by small dots. All letters subtended high load compared to the low load condition). between 0.6° and 0.9° vertically by 0.4° horizontally and were presented in gray for 100 ms exposure duration as 3. Experiment 2 before. A set size of four was used in the high load condition In Experiment 2 we sought to rule out any component as pilot testing had indicated that, using the Hebrew of verbal working memory load in our task. As we used let- stimuli with non-Hebrew speaking participants, this set ters as search stimuli in Experiment 1 it could be argued size was sufficient to induce an increase in RTs and error that, in addition to increased perceptual demands, the ver- rate of a similar magnitude to that associated with the bal working memory demands in the high load condition increase in load in Experiment 1. In addition, in order (in which participants were required to discriminate the to ensure that the high load condition did not involve target letter from amongst five different non-target letters) a greater verbal processing load than the low load condi- were also greater than those in the low load condition (in tion (due to any verbal processing of the added non-tar- which the five non-target letters were all Os and therefore get symbols) we asked three independent judges to only two letters could potentially be verbalized). The re- complete the eight high load blocks used in the actual duced level of TUTs under high load therefore could have experiment (following the same experimental procedure) simply reflected the suppression of verbal TUTs by partic- and, immediately following the final trial, to answer the ipants’ internal verbalization of the larger letter search following questions: 350 S. Forster, N. Lavie / Cognition 111 (2009) 345–355

Fig. 2. Example stimulus displays with low and high perceptual load in Experiment 2.

1. Did you, at any point during the experiment, assign ver- flects an effect of load on deliberate TUTs or on uninten- bal names to any of the symbols that you saw on tional TUTs. Previous perceptual load research suggests screen? If so please could you write these down on that the effects of perceptual load would not be confined the sheet of paper beside you, and also indicate whether just to deliberate TUTs: Increasing perceptual load has pre- these were target symbols or non-target symbols. viously been found to eliminate interference even from po- 2. Did you, at any point in the experiment, have verbal tent task-irrelevant distractors that interfere with task thoughts relating to specific symbols? If so please could performance in conditions of low perceptual load despite you write them down, indicating whether each thought participants’ attempts to ignore them (e.g., Forster & Lavie, was about a target or non-target symbol. 2008a; Lavie, 1995) This leads us to predict that high per- ceptual load would be capable of reducing the rate of unin- Although two out of the three judges indicated that tentional intrusions by TUTs. Alternatively, however, one they had assigned a verbal label to one of the two target might argue that for internal sources of irrelevant distrac- symbols (‘‘shoe” or ‘‘boot”), none of the judges had as- tion high perceptual load could merely reduce the deliber- signed any verbal names or reported any verbal thought ate intention to engage in TUT (for example, because the associated with the non-target symbols. Note that any ver- high load task may be less boring) but would have no effect bal demands associated with the target letters remained on those TUTs that intrude into the mind unintentionally. constant across the two load conditions. To examine this issue, in Experiment 3 participants were probed as to whether any task-unrelated thought they 3.2. Results may have had was deliberate or unintentional.

As in Experiment 1, the RT and error data showed the 4.1. Method expected effects of perceptual load. RTs under high load were significantly longer (M = 719 ms) than under low load 4.1.1. Participants (M = 547 ms), t (11) = 7.87, SEM = 21.89, p < .001, d = 4.75 Eighteen new participants (five males, 18–33 years old) for the difference. The PE rate was also significantly higher participated in Experiment 3. under high load (M = 10.25%) than under low load (M = 6.41%), t (11) = 2.34, SEM = 1.64, p = .039, d = 1.41 for 4.1.2. Stimuli and procedure the difference. The stimuli and procedure were the same as those used The key finding of this experiment was that the partic- in Experiment 1a, with the exception that following the re- ipants again reported less TUTs in the high perceptual load sponse to the thought probe the participants were pre- condition (M = 53.15%), compared to low load condition sented with an additional probe asking them to indicate (M = 65.62%), t (11) = 2.43, SEM = 5.10, p = .033, d = 1.47 whether the reported TUT was deliberate or unintentional for the difference (see Fig. 1). Thus, Experiment 2 estab- by pressing either the X or C keys respectively (or by press- lishes that the effect of perceptual load on the rate of ing the A key if no TUT was reported). The following expla- TUT reports does not require the use of familiar verbal nation of deliberate and accidental TUTs was given to the material that could serve to load verbal working memory participants before they started the experiment: ‘‘Some- and thus directly block the mental workspace for thought. times people may, for one reason or another, deliberately allow their mind to wander from a task that they are per- 4. Experiment 3 forming. This is a deliberate TUT. Other times a person may be really trying to concentrate on a task but somehow acci- In Experiment 3 we examined whether the reduction in dentally their mind wanders off to think about something the rate of TUTs reported under high perceptual load re- else, unrelated to the task. This is an unintentional or acci- S. Forster, N. Lavie / Cognition 111 (2009) 345–355 351 dental TUT.” Participants were informed that there was no 5. Experiment 4 right or wrong response to this question and that they should simply be honest. Having established the effects of perceptual load on mind-wandering, in Experiment 4 we sought to relate the 4.2. Results effects of perceptual load on the processing of task-irrele- vant and potentially distracting information from internal As in the previous experiments, RT and PE data showed and external sources. We thus included external task-irrel- the expected effects of perceptual load. RTs were longer in evant distractors, adding flanker letters that could be the high load condition (M = 679 ms) compared to the low incongruent or congruent with the target to the paradigm load condition (M = 516 ms), t (17) = 14.07, SEM = 11.95, used in Experiments 1–3, and assessed the effects of per- p < .001, d = 6.82 for the difference. PE rates were also sig- ceptual load on both the internal (rates of TUT) and exter- nificantly higher in the high load condition (M = 10.17%), nal (response-competition effects) forms of potential compared to the low load condition (M = 4.61%), t distraction. (17) = 5.76, SEM = 0.97, p < .001, d = 2.79 for the difference. The TUT reports were entered into a within-subject AN- 5.1. Method OVA with the factors of perceptual load (low, high) and intention (deliberate, unintentional). As can be seen in 5.1.1. Participants Fig. 3, the ANOVA showed a main effect of perceptual load Twenty new participants (six males, 19–30 years old) that replicated the results of previous experiments: The participated in Experiment 4. rate of TUTs reported under low perceptual load (M = 60.06%) was significantly reduced under high percep- 5.1.2. Stimuli and procedure tual load (M = 44.87%), F (17) = 13.05, MSE = 81.65, All stimuli and procedure were the same as in Experi- 2 p = .002, gp = .356. There was also a main effect of inten- ment 1a, with the exception that the distractor letters X tion: unintentional TUTs were reported on a significantly or N, each subtending 0.8° Â 0.5°, were presented to the greater percentage of trials (M = 38.28%) than deliberate left or the right of the letter circle (1.4° from the nearest TUTs (M = 13.89%), F (17) = 9.41, MSE = 1083.95, p = .007, circle letter). All combinations of distractor position, dis- 2 gp = .434. The interaction between load and intention was tractor identity, target position and target identity and load not significant, F (17) = 2.33, MSE = 222.20, p = .145, were fully counterbalanced. 2 gp = .120. However, the low rate of deliberate TUTs found in this experiment makes this measure insensitive to show 5.2. Results any effect or interaction. For this reason we confine our conclusion to the unintentional TUTs only. With respect Table 1 presents the results. A 2 Â 2 within-subjects to these, the results of Experiment 3 clearly show that high ANOVA of the mean RT with the factors of distractor con- perceptual load can reduce intrusions by unintentional gruency and load revealed main effects of load, F (1, 2 task-irrelevant thoughts. This finding has important impli- 19) = 82.56, MSE = 10140.6, p < .001, gp = .813, and distrac- cations both for predicting the ability to pay attention to a tor congruency, F (1, 19) = 16.11, MSE = 539.30, p = .000, 2 task with minimal distraction by unintentional TUTs in gp = .449. Of most importance was the interaction between daily life and for clinical populations (e.g., ADHD). We dis- load and distractor congruency, F (1, 19) = 17.655, 2 cuss these in greater detail in Section 6. MSE = 319.6, p < .001, gp = .482, indicating that the distrac-

60 Low load High Load 50

40

30

20 % TUT reportsTUT %

10

0 Intentional Unintentional

Fig. 3. Percentage of probes at which intentional and unintentional task-unrelated thoughts (TUT) were reported under high and low perceptual load in Experiment 3. 352 S. Forster, N. Lavie / Cognition 111 (2009) 345–355

Table 1 of load, Pearson r (20) = .571, p = .009 (see Fig. 4). This cor- Experiment 4: Mean RTs (SE in parentheses) and percentage error rates as a relation was not due to variations in the extent to which function of distractor conditions and load, and mean percentage of task- load has increased the demand of the search task as it re- unrelated thoughts (TUTs) as a function of load. mained significant when individual variations in the mag- Low load High load nitude of the main effect of load on RTs were partialed out, I r (17) = .491, p = .033 (for the same correlation with dis- RT (ms) 605 (29) 793 (36) tractor response-competition effects calculated as percent- % Error 717 ages r = .599, p = .007). C The clear relationship between the magnitude of the RT (ms) 567 (24) 789 (36) load effects on internal (TUTs) and external (response com- % Error 4 16 TUTs 41.25% 51.90% petition) distractors across individuals suggests that these decreases in both internal and external distractors reflect Note. I = incongruent distractor, C = congruent distractor. one underlying focused attention ability driven by the level of perceptual load in the task. Notice that the finding of a correlation between individ- tor response-competition effect under low load (distractor ual differences in the magnitude of the effect of perceptual effect M = 38 ms, t (19) = 5.13, SEM = 7.33, p < .001, load on internal and external distractors does not contra- d = 2.35) was significantly reduced with high load (distrac- dict previous load research by suggesting that perceptual tor effect M = 4 ms, t < 1). These results replicate previous load is less effective in reducing interference by external findings (e.g., Lavie & Cox, 1997), demonstrating a signifi- distractors for some individuals: Consistent with previous cant reduction in interference from external distractors studies (e.g., Forster & Lavie, 2007), high perceptual load under high perceptual load. also eliminated distractor response-competition effects A2 2 ANOVA on the PE rates also revealed main ef- for all subjects in the present study (see Table 1), but the fects of load, F (1, 19) = 50.61, MSE = 46.73, p < .001, correlation indicates that perceptual load was most effec- g2 = .727, and distractor congruency, F (1, 19) = 32.90, p tive in reducing not only distractor effects but also TUTs MSE = 2.88, g2 = .634, p < .001. Although the load  con- p for those individuals with the greatest levels of distractor gruency interaction was not significant, F (1, 19) = 1.30, 2 effects in the conditions of low perceptual load. MSE = 8.09, p = .269, gp = .064, the numerical trends seen in Table 1 were in the same direction as the RT. As in Experiments 1–3, high perceptual load has also 6. General discussion significantly reduced the rate of TUTs reported in this experiment (for the high load M = 41.25%, for low load The present experiments establish that, like task-irrele- M = 51.90%), t (19) = 3.43, SEM = 3.10, p = .003, d = 1.57 vant external distractors, internal sources of distraction in for the difference. Furthermore, there was a significant cor- the form of task-unrelated thoughts can also be modulated relation between the effect of perceptual load on TUTs and by the level of perceptual load in the task performed. This on the distractor response-competition effects both when finding implies that, despite the obvious differences be- these were calculated as the mean I–C under each level tween TUTs and external distractors (e.g., TUTs are inter- of load, Pearson r (20) = .489, p = .029, and when individual nally generated and may be drawn from long term differences in the baseline RT were accounted for by com- memory processes, see Christoff, Ream, and Gabrieli puting the percentage RT increase in the incongruent (vs. (2004)), the process of the distraction of attention by congruent) distractor conditions under the different levels task-irrelevant stimuli may involve common mechanisms

15%

10%

5%

0%

Load effect on RC (% low –high load) Load effect -5% -20% -10% 0% 10% 20% 30% 40%

Load effect on TUT (%TUT low - high load)

Fig. 4. Correlation between load effects on internal distractors (TUT reports) and external (response competition) distractors. Response-competition (RC) effects were calculated as the percentage increase in RTs on incongruent trials compared to congruent trials. S. Forster, N. Lavie / Cognition 111 (2009) 345–355 353

(regardless of the nature of the source of distraction). The therefore be an interesting avenue for future research to experiments further clarify that the reduction in the level examine whether the effects of perceptual load on task- of TUTs reported under high perceptual load is not simply unrelated thoughts extend to unconscious thoughts. due to a change in the demands on verbal working mem- As the first study of mind-wandering to directly manip- ory (Experiment 2), rate of responses and response feed- ulate task demands on attentional capacity, our conclusion back or the level of motivation (Experiment 1) or that perceptual load plays a causal role in determining deliberate intention to engage in TUTs (Experiment 3). mind-wandering can both accommodate previous sugges- As we have ruled out any alternative account for the ef- tions and guide future frameworks of mind-wandering. For fects of perceptual load in terms of a change in the extent example, it has been recently suggested that mind-wan- to which the visual search task directly recruits compo- dering depends on the extent to which a task demands nents that are needed either for the actual production or engagement with the external environment, and that the the report of any thought (e.g., working memory, cf. smaller rate of TUTs found in some studies using reading Smallwood & Schooler, 2006; Teasdale et al., 1995), these tasks compared to other studies that have used signal results suggest the sharing of a central perceptual process- detection tasks is due to the greater level of task engage- ing resource (attention) between task-unrelated thoughts ment in reading vs. signal detection tasks (Smallwood, and the high load search task in the manner predicted by Fishman et al., 2007). the Load Theory. Notice that in this respect the effects of Our present application of Load Theory to mind-wan- perceptual load demonstrated here are rather different to dering can clarify the concept of task engagement by sug- the effects of other task manipulations that have been pre- gesting that the level of task engagement would depend on viously found to reduce TUT reports. As we discuss earlier the level of perceptual load in the task. Notice that this pro- (in the General Introduction) the previous task manipula- posal does not rely on a taxonomy of tasks that are sup- tions shown to reduce TUT reports (dual task vs single task posed to always demand greater or lesser levels of conditions, fast vs. slow stimulus presentation rates or engagement. Instead we propose that any type of task conditions of high vs low working memory load) have di- may be prone to more or less TUT depending on the level rectly drawn on either thought or response components. of perceptual load it involves. For instance a signal detec- The reduction in TUT reports in those cases may therefore tion task conducted under conditions of high perceptual be due to direct interference with thoughts or the pro- load would be predicted to be less prone to mind-wander- cesses involved in reporting thoughts rather than to the ing than either the same task with low perceptual load or a enhanced focusing of attention on the task. reading task of a low perceptual load (e.g., repeated read- The correlation between the effects of perceptual load ing of the same words). on TUTs and distractor response-competition effects In this way, Load Theory offers a theoretical framework (Experiment 4) further demonstrates the role of a common for determining levels of task-engagement and (con- focused attention mechanism (driven by the level of per- versely) task-unrelated thought. A fruitful direction for fu- ceptual load in the task) in determining the levels of both ture research would be to explore this suggestion by internal and external distractor interference. The present measuring TUT whilst employing a previously established findings therefore demonstrate that the Load Theory of manipulation of perceptual load during a signal detection attention can be generalized to determine another potent task (e.g., see Macdonald & Lavie, 2008). form of distraction due to mind-wandering. By demonstrating the effect of perceptual load on task- Our use of a probe-caught method that did not require unrelated thought, the present study represents a first the participants to continuously monitor for the occur- stage in the application of load theory to internal distrac- rence of a task-unrelated thought, instead asking the sub- tors. Load theory makes a clear distinction between the ef- jects to report what they were thinking ‘‘just now”, fects of perceptual load and of executive control load on makes it unlikely that the results could be attributed to distractor processing, Whereas high perceptual load re- alternative accounts in terms of load effects on either duces distractor processing, high levels of cognitive load meta-awareness of thoughts (see Smallwood and Schooler increase distraction due to reduced ability to suppress (2006)) or on the encoding of thoughts into memory. As irrelevant information (e.g., see Lavie et al. (2004)). With our self-report measure did require participants to be suf- respect to TUTs this would lead to an interesting counter- ficiently conscious of their thoughts to be able to report intuitive prediction that high cognitive load would lead to them, however, it remains possible that the effects of load an increased rate of TUT. A major obstacle, however, to are restricted to conscious task-unrelated thoughts but testing this second prediction of load theory would be that, that unconscious thought processes are unaffected by load. as has been discussed above, various components of exec- Interestingly, a recent study demonstrates that percep- utive control appear likely to be involved in the production tual load can modulate unconscious processing of external of any thought (indeed it has been shown that increasing task-irrelevant stimuli. Bahrami et al. (2008) examined tilt working memory load decreases TUT, Teasdale et al., after effects following adaptation to oriented gratings that 1993; Teasdale et al., 1995, presumably by directly inter- were rendered invisible during adaptation with continuous fering with the production of thought). flash suppression. Participants performed either a high or a Finally, by identifying perceptual load as means by low perceptual load task at fixation during adaptation to which to reduce levels of task-unrelated thoughts, the the invisible oriented gratings and showed reduced tilt after present findings also suggest that perceptual load may be effects following the high compared to low load task, despite useful in applied and clinical settings. As discussed above, having no conscious perception of the gratings. It would task-unrelated thoughts have been shown to interfere with 354 S. Forster, N. Lavie / Cognition 111 (2009) 345–355 performance on a wide range of tasks – future research Lavie, N. (2000). Selective attention and cognitive control: Dissociating should investigate the possibility that perceptual load attentional functions through different types of load. In S. Monsell & J. Driver (Eds.), Attention and performance XVIII (pp. 175–194). may protect against such interference by assessing the ef- Cambridge, Massachusetts: MIT press. fect of TUTs on the performance of tasks modified to have Lavie, N. (2005). Distracted and confused?: Selective attention under load. high and low perceptual load. Another important implica- Trends in Cognitive Sciences, 9, 75–82. Lavie, N. (2006). The role of perceptual load in visual awareness. Brain tion of these results would be that higher levels of percep- Research, 1080(1), 91–100. tual load could temporarily alleviate the unusually high Lavie, N., & Cox, S. (1997). On the efficiency of visual selective attention: levels of mind-wandering and task-unrelated thoughts Efficient visual search leads to inefficient distractor rejection. Psychological Science, 8, 395–398. associated with ADHD (Shaw & Giambra, 1993) or the Lavie, N., & Fox, E. (2000). The role of perceptual load in negative priming. intrusive task-unrelated thoughts associated with clinical Journal of Experimental Psychology-Human Perception and Performance, disorders such as post-traumatic stress disorder (e.g., see 26, 1038–1052. Lavie, N., Hirst, A., de Fockert, J. W., & Viding, E. (2004). Load theory of Falsetti, Monnier, and Resnick (2005) and Obsessive Com- selective attention and cognitive control. Journal of Experimental pulsive Disorder (e.g., see Clark and O’Conner (2005)). Psychology: General, 133, 339–354. We are currently testing these hypotheses. Lavie, N., Ro, T., & Russell, C. (2003). The role of perceptual load in processing distractor faces. Psychological Science, 14, 510–515. Acknowledgments Lavie, N., & Tsal, Y. (1994). Perceptual load as a major determinant of the locus of selection in visual-attention. Perception and Psychophysics, 56, This research was supported by a Welcome Trust award 183–197. Logan, G. D. (1978). Attention in character-classification tasks: Evidence (WT080568MA) to Nilli Lavie. for the automaticity of component stages. Journal of Experimental We thank Sarah Hurcomb for her assistance in data Psychology: General, 107, 32–63. collection. Macdonald J., & Lavie N. (2008). Load induced blindness. Journal of Experimental Psychology: Human Perception and performance, 34(5), 1078—1091. References Manly, T., Heutink, J., Davison, B., Gaynord, B., Greenfield, E., Parr, A., et al. (2004). An electronic knot in the handkerchief: ‘Content free cueing’ Antrobus, J. S. (1968). Information theory and stimulus-independent and the maintenance of attentive control. Neuropsychological thought. British Journal of Psychology, 59, 423–430. Rehabilitation, 14(1–2), 89–116. Baddeley, A. (1986). Working memory. Oxford: Clarendon. McKiernan, K. A., D’Angelo, B. R., Kaufman, J. N., & Binder, J. R. (2006). Bahrami, B., Carmel, D., Walsh, V., Rees, G., & Lavie, N. (2008). Spatial Interrupting the ‘‘stream of consciousness”: An fMRI investigation. attention can modulate unconscious orientation processing. Neuroimage, 29, 1185–1191. Perception, 37(10), 1520–1528. Rees, G., Frith, C. D., & Lavie, N. (1997). Modulating irrelevant motion Bahrami, B., Lavie, N., & Rees, G. (2007). Attentional load modulates perception by varying attentional load in an unrelated task. Science, responses of human primary visual cortex to invisible stimuli. Current 278, 1616–1619. Biology, 17(6), 509–513. Schooler, J. W., Reichle, E. D., & Halpern, D. V. (2004). Zoning out while Brand-D’Abrescia, M., & Lavie, N. (2007). Distractor effects during reading: Evidence for dissociations between experience and processing of words under load. Psychonomic Bulletin and Review, metaconsciousness. In D. T. Levin (Ed.), Thinking and seeing: Visual 14(6), 1153–1157. metacognition in adults and children (pp. 203–226). Cambridge, MA: Cartwright-Finch, U., & Lavie, N. (2006). The role of perceptual load in MIT Press. Inattentional Blindness. Cognition, 102(3), 321–340. Schwartz, S., Vuilleumier, P., Hutton, C., Maravita, A., Dolan, R. J., & Driver, Christoff, K., Ream, J. M., & Gabrieli, J. D. E. (2004). Cognitive and neural J. (2005). Attentional load and sensory competition in human vision: basis of spontaneous thought processes. Cortex, 40, 623–630. Modulation of fMRI responses by load at fixation during task- Clark, D. A., & O’Conner, K. (2005). Thinking is believing: Ego-dystonic irrelevant stimulation in the peripheral visual field. Cerebral Cortex, intrusive thoughts in obsessive-compulsive disorder. In D. A. Clark 15, 770–786. (Ed.), Intrusive thoughts in clinical disorders. New York: The Guilford Segal, S. J., & Fusella, V. (1970). Influence of imaged pictures and sounds Press. on detection of visual and auditory signals. Journal of Experimental Dalton, P., Lavie, N., & Spence, C. (2008). The role of working memory in Psychology, 83, 458–464. tactile selective attention. Quarterly Journal of Experimental Shaw, G. A., & Giambra, L. (1993). Task-unrelated thoughts of college- Psychology, in press [Epub ahead of print]. students diagnosed as hyperactive in childhood. Developmental De Fockert, J. W., Rees, G., Frith, C. D., & Lavie, N. (2001). The role of Neuropsychology, 9, 17–30. working memory in visual selective attention. Science, 291(5509), Smallwood, J. M., Baracaia, S. F., Lowe, M., & Obonsawin, M. (2003). Task 1803–1806. unrelated thought whilst encoding information. Consciousness and Falsetti, S. A., Monnier, J., & Resnick, H. S. (2005). Intrusive thoughts in Cognition, 12, 452–484. posttraumatic stress disorder. In D. A. Clark (Ed.), Intrusive thoughts in Smallwood, J. M., Davies, J. B., Heim, D., Finnigan, F., Sudberry, M., clinical disorders. New York: The Guilford Press. O’Connor, R., et al. (2004). Subjective experience and the attentional Forster, S., & Lavie, N. (2007). High perceptual load makes everybody lapse: Task engagement and disengagement during sustained equal – Eliminating individual differences in distractibility with load. attention. Consciousness and Cognition, 13, 657–690. Psychological Science, 18, 377–381. Smallwood, J., Fishman, D. J., & Schooler, J. W. (2007). Counting the cost of Forster, S., & Lavie, N. (2008a). Attentional capture by entirely irrelevant an absent mind: Mind wandering as an underrecognized influence on distractors. Visual Cognition, 16, 200–214. educational performance. Psychonomic Bulletin and Review, 14, Forster, S., & Lavie, N. (2008b). Failures to ignore entirely irrelevant 230–236. distractors: The role of load. Journal of Experimental Psychology: Smallwood, J., McSpadden, M., & Schooler, J. W. (2007). The lights are on Applied, 14(1), 73–83. but no one’s home: Meta-awareness and the decoupling of attention Giambra, L. M. (1995). A laboratory method for investigating influences when the mind wanders. Psychonomic Bulletin and Review, 14, on switching attention to task-unrelated imagery and thought. 527–533. Consciousness and Cognition, 4, 1–21. Smallwood, J., O’Connor, R. C., Sudberry, M. V., Haskell, C., & Ballantyne, Jonides, J., & Yantis, S. (1988). Uniqueness of abrupt visual onset in C. (2004). The consequences of encoding information on the capturing attention. Perception and Psychophysics, 43, 346–354. maintenance of internally generated images and thoughts: The Kahneman, D., & Treisman, A. (1984). Changing views of attention and role of meaning complexes. Consciousness and Cognition, 13, automaticity. In R. Parasuraman, D. R. Davies & J. Beatty (Eds.), 789–820. Variants of attention (pp. 29–61). New York: Academic Press. Smallwood, J., & Schooler, J. W. (2006). The restless mind. Psychological Lavie, N. (1995). Perceptual load as a necessary condition for selective Bulletin, 132, 946–958. attention. Journal of Experimental Psychology-Human Perception and Teasdale, J. D., Dritschel, B. H., Taylor, M. J., Proctor, L., Lloyd, C. A., Performance, 21, 451–468. Nimmo-Smith, I., et al. (1995). Stimulus-independent thought S. Forster, N. Lavie / Cognition 111 (2009) 345–355 355

depends on central executive resources. Memory and Cognition, 23, Treisman, A., & Gelade, G. (1980). A feature integration theory of 551–559. attention. Cognitive Psychology, 12, 97–136. Teasdale, J. D., Proctor, L., Lloyd, C. A., & Baddeley, A. D. (1993). Working- Yantis, S., & Jonides, J. (1984). Abrupt visual onsets and selective attention memory and stimulus-independent thought – Effects of memory load – Evidence from visual-search. Journal of Experimental Psychology- and presentation rate. The European Journal of Cognitive Psychology, 5, Human Perception and Performance, 10, 601–621. 417–433. Yi, D. J., Woodman, G. F., Widders, D., Marois, R., & Chun, M. M. Treisman, A. (1988). Features and objects: The Fourteenth Bartlett (2004). Neural fate of ignored stimuli: Dissociable effects of Memorial Lecture. Quarterly Journal of Experimental Psychology, perceptual and working memory load. Nature Neuroscience, 7, 40A(2), 201–237. 992–996.