Consciousness and Cognition 33 (2015) 386–397

Contents lists available at ScienceDirect

Consciousness and Cognition journal homepage: www.elsevier.com/locate/concog

Reducing the vigilance decrement: The effects of perceptual variability David R. Thomson ⇑, Daniel Smilek, Derek Besner Department of Psychology, University of Waterloo, Canada

a r t i c l e

i n f o

Article history: Received 13 February 2015

Keywords: Sustained attention Vigilance Perceptual variability Resource depletion

a b s t r a c t The longer we are required to monitor for rare but critical events, the accuracy and speed with which we detect such events tend to suffer (the ‘vigilance decrement’) with more difficult tasks yielding larger decrements. Here, we present a striking example of a situation in which increasing the difficulty and complexity of a novel vigilance task actually decreases the vigilance decrement. In a ‘Stable’ condition participants monitored for the same critical target throughout the task, whereas in a ‘Variable’ condition, participants monitored for many possible instantiations of the critical target. Despite the fact that the Variable condition was objectively more difficult, the vigilance decrement was significantly reduced in response times relative to the Stable condition. We discuss these findings in light of ‘overload’ and ‘underload’ theories of the vigilance decrement and suggest that perceptual variability may provide bottom-up support for the maintenance of attentional resource allocation to an external task. Ó 2015 Elsevier Inc. All rights reserved.

1. Introduction Improving the capacity for observers to successfully monitor for rare, but critical, events has long plagued human factors researchers and cognitive psychologists alike. Consider for a moment, the case of a security screener at an airport, who must visually scan thousands of items contained within hundreds of bags in the course of a given day. It is easy to imagine that with such a constant, high-rate of perceptual input, and with dangerous items such as weapons being encountered so rarely, it may become rather difficult to remain vigilant over the course of a typical work shift. Given that the potential costs of failing to detect a deadly weapon concealed within a piece of luggage are significant, understanding the psychology of attention over time-on-task holds not only theoretical interest, but is also of practical importance. The so-called ‘vigilance decrement’ was first demonstrated experimentally by Mackworth (1948), who was tasked with finding ways to mitigate the apparent drop-off in detection accuracy of British naval radar operators as their watch periods progressed. Using a simple task designed to parallel the job of a radar operator, Mackworth found that the ability of observers to respond to rare, but critical target events declined significantly over time. Since then, researchers have delineated the factors that lead to the vigilance decrement, allowing them to develop abbreviated vigilance tasks in which performance decrements can be observed after only a few minutes. Specifically, it has been found that robust decrements will be observed as long as: (1) the event rate is relatively high, (2) the target rate is relatively low, (3) stimuli are presented successively as

⇑ Corresponding author at: Department of Psychology, University of Waterloo, 200 University Avenue West, Waterloo, Ontario N2L 3G1, Canada. Fax: +1 519 746 8631. E-mail address: [email protected] (D.R. Thomson). http://dx.doi.org/10.1016/j.concog.2015.02.010 1053-8100/Ó 2015 Elsevier Inc. All rights reserved.

D.R. Thomson et al. / Consciousness and Cognition 33 (2015) 386–397

387

opposed to simultaneously, and (4) signal salience is moderate to low (Parasuraman, 1979; Parasuraman & Davies, 1977; Parasuraman & Mouloua, 1987). Currently, the leading theoretical account of the vigilance decrement is known as the ‘resource depletion’ account (Grier et al., 2003; Helton & Warm, 2008; Warm & Dember, 1998; Warm, Parasuraman, & Matthews, 2008). This theory holds that observers have a limited pool of information processing resources, and that as a vigil unfolds, those resources become depleted due to mental overload. As a consequence, the attentional focus required to maintain a high level of performance outweighs the available resources and so performance decrements are observed. While some researchers have described time-on-task effects as reflecting more than just resource depletion (e.g. motivation/effort and arousal, in addition to information processing limitations; Sanders, 1983, 1998), current theorizing with respect to the vigilance decrement still centers around resource depletion as the primary factor. For example, Helton and Russell (2012) state that: ‘‘Advocates of resource theory argue that the information processing resources required for vigilance are limited. . .The continuous information processing demands of vigilance tasks deplete the necessary cognitive resources, hence, a decline in performance efficiency over the watch-keeping session’’ (pp. 37–38). Likewise, Grier et al. (2003) posit that: ‘‘. . .performance failures in vigilance tasks result from a decline in information-processing resources rather than from a diminution of arousal’’ (p. 350). A key tenet of the resource depletion account is that vigilance decrements should be more pronounced in more demanding tasks, since the available resources will be depleted at a faster rate. Empirical support for the resource depletion hypothesis stems from a host of studies in which increasing the demands on information processing resources has resulted in larger vigilance decrements. For instance, Smit, Eling, and Coenen (2004) had participants perform an abbreviated vigilance task in which they were required to respond to the rare occurrence of an ‘a’ following an ‘x’ in a sequence of letters. In a ‘high’ demand condition, participants performed the task concurrently with a secondary task in which they had to look for and respond to dark gray squares amongst light gray squares that were presented simultaneously with the letter stimuli for the primary vigilance task. Crucially, the vigilance decrement was larger for the high demand condition. Similarly, Helton and Warm (2008) varied the nominal demands on information processing resources by varying stimulus salience (the contrast of the stimuli with respect to the background). The vigilance decrement was found to be much steeper for the low-salience (harder) condition, relative to the high-salience (easier) condition (see also, Parasuraman et al., 2009 for a similar result). It has also been shown that the addition of a working memory load to a standard vigilance task magnifies performance decrements (Helton & Russell, 2011). And finally, vigilance decrements are more pronounced when observers must simultaneously monitor several displays for multiple critical events relative to when they only have to monitor one or two displays (Grubb, Warm, Dember, & Berch, 1995). Taken together, manipulations of task difficulty consistently result in steeper performance decrements, thus providing strong support for the idea that the vigilance decrement arises because of the depletion of information processing resources over time-on-task. Nevertheless, there is some (albeit indirect) evidence that not all manipulations of task difficulty, or ‘complexity’ result in larger vigilance decrements. For example, it has been shown that when observers monitor for a single critical event in the context of a perceptually complex flight simulation task, the probability of detection does not depend on when the signal is presented during the vigil. In contrast, probability of detection of a single critical event during a perceptually ‘simple’ (i.e. ‘standard’) vigilance task, is much higher when the signal is presented within the first ten minutes of the vigil (Molloy & Parasuraman, 1996). Similarly, in a simulated air traffic control task, the vigilance decrement is abolished (with practice) when observers perform a secondary task that is tied to the task-relevant stimuli (i.e. click on incoming air craft) relative to when they passively monitor for critical events (i.e. possible collisions; Pop, Stearman, Kazi, & Durso, 2012). These findings suggest that the vigilance decrement may not stem from mental overload that depletes attentional resources, but rather from under-stimulation due to the monotony of vigilance tasks (see Manly, Robertson, Galloway, & Hawkins, 1999; Robertson, Manly, Andrade, Baddeley, & Yiend, 1997). It may be the case that whenever manipulations of task complexity/difficulty increase ‘engagement’ with the task-relevant stimuli (and decrease the monotony of the task), vigilance decrements may be alleviated. Importantly, this may be true even when such manipulations result in an objective increase in task difficulty. Indeed, perhaps the main reason why findings such as those reported by Pop et al. (2012) have not been considered overly problematic for the resource depletion hypothesis is that the conditions shown to exhibit the smallest decrements could in fact be the easier conditions (and thus should exhibit smaller decrements by a resource depletion account). What is needed is a manipulation of task complexity that objectively increases the difficulty of the vigilance task, while at the same time increasing observers’ engagement with the task-relevant stimuli. If performance decrements are less pronounced in such a condition, this would be difficult to reconcile with the resource depletion hypothesis. Likewise, if the magnitude of the vigilance decrement is similar across conditions that vary in difficulty, this would, at the very least, demonstrate that task difficulty is not the primary determinant of the vigilance decrement, but simply one of many influences. Again, while several researchers (both recently and historically) have acknowledged that vigilance decrements likely owe to many factors, task difficulty (leading to resource depletion) is largely agreed upon to be the key factor. How then, does one go about increasing both the objective complexity and difficulty of a sustained attention task in a way that will ‘hold’ the attention of the observer for longer periods of time? The answer may lie in research into motivation theory and aesthetics conducted by D.E. Berlyne, which has thus far been overlooked by vigilance researchers. For example, it has been shown that the complexity (i.e. the density of information content) of visual patterns is strongly related to subjective ratings of ‘interestingness’ and ‘pleasingness’ (Berlyne, Ogilvie, & Parham, 1968). Similarly, and perhaps more relevant for the present work, perceptual variability results in ratings of ‘pleasantness’ that remain fairly high over successive

388

D.R. Thomson et al. / Consciousness and Cognition 33 (2015) 386–397

presentations of sequences of irregular shapes (Berlyne, 1970). It may therefore be the case that by increasing the perceptual variability of critical targets in a vigilance task, one might arrive at a situation in which the novelty of the task persists for longer, thus holding attention and reducing performance decrements. This may occur despite the fact that search for perceptually variable targets may prove to be more difficult overall than search for perceptually stable targets. Specifically, it has been argued that: ‘‘. . .the subjective novelty of a stimulus. . .declines more steeply over a series of consecutive or intermittent presentations of the same stimulus than over a series of heterogeneous stimuli with the same temporal distribution.’’ (Berlyne et al., 1968, p. 422). In other words, all other things being equal, greater perceptual variability should hold the interest of the observer for longer. In the work that follows, we employ a novel vigilance task that allows us to vary the perceptual variability of the critical target stimuli along several dimensions (i.e., color pattern, size, spatial frequency, and brightness). We have (loosely) modeled our vigilance task after the job of the airport security screener described in our earlier example. Specifically, observers must monitor objects as they pass across a ‘window’ and must identify rare ‘targets’. The crucial manipulation is that in one condition, the targets take on only one form (as in a standard vigilance task, hereafter referred to as the Stable condition), and in another take on many potential forms (similar to the real-world task of monitoring for ‘prohibited items’, which take many forms, hereafter referred to as the Variable condition). The experimental stimuli are depicted in Fig. 1. Despite the fact that the Variable condition is expected to place additional demands on information processing resources, and thus increase the difficulty of the task relative to the Stable condition, we hypothesize, that the Variable condition will better hold the attention of the observer over time, resulting in a smaller performance decline. An alternative outcome that is also consistent with our hypothesis is that vigilance decrements will be similar in both conditions – recall that under a resource depletion account, the more difficult condition should yield the steeper performance decline. It may very well be the case that any beneficial effect of perceptual variability of critical targets is equal to the detrimental effects of increasing task difficulty. The latter outcome would suggest that these two influences are equal contributors to vigilance performance, an outcome that is clearly at odds with the resource depletion account as it is currently put forth. Finally, if performance declines are shown to be worse in the more difficult condition (assuming our Variable condition is shown to be objectively more difficult, either through slower detection times, poorer signal sensitivity, or both), this would provide clear support for the resource depletion account of the vigilance decrement, and implicate task difficulty as the primary determinant of sustained attention performance over time. 2. Experiment 1 Experiment 1 had three goals: (1) to verify that our novel vigilance task is effective in producing performance decrements over time, (2) to confirm that our Stable and Variable conditions do in fact differ in terms of difficulty (with the Variable condition being the more difficult), and (3) to assess whether any observed performance decrements differ in terms of accuracy and/or response times for correct detections between the Stable and Variable Target conditions. In addition, one of our broad secondary aims is to begin to make an effort to increase the ecological validity of laboratory vigilance tasks, which

Fig. 1. A depiction of the stimuli presented in Experiments 1 and 2. The non-target stimulus is shown at the top. The critical target stimuli are shown, in which size, brightness, color pattern, or spatial frequency was manipulated with respect to the non-target stimulus.

D.R. Thomson et al. / Consciousness and Cognition 33 (2015) 386–397

389

have been criticized by human factors researchers for their lack of correspondence with the real world situations in which humans must monitor largely automated systems (Hancock, 2013). 2.1. Method 2.1.1. Participants One hundred and nineteen undergraduates from the University of Waterloo (44 male, 75 female, with a mean age of 19.8 years) participated in exchange for course credit. 2.1.2. Apparatus The experimental program was written in Python (www.python.org) and presented using PsychoPy software (Peirce, 2007). The experiment was run on a Mac Mini computer with a 2.40 GHz processor connected to a 24 in. Phillips 244E LCD monitor, placed at an approximate viewing distance of 57 cm. 2.1.3. Stimuli and procedure Participants were randomly assigned to the Variable Target or Stable Target conditions. The Stable Target condition began with the following instructions: ‘‘In this task, you will see items moving across the screen from right to left, one at a time. Most of the items are ’non-targets’, and require you to do nothing. In order to learn what the non-targets look like, press the space bar to view several trials.’’ Participants then viewed 15 non-target trials. Following this, participants were told that: ‘‘Some of the items you will see are ‘targets’ and require you to press the space bar as quickly as you can each time one appears. In order to learn what the target item looks like, press the space bar to view several trials.’’ Participants then viewed 15 target-trials. The target item for a given participant in the Stable Target condition was randomly selected from one of eight possible targets. Finally, participants in the Stable Target condition were told that: ‘‘From now on, targets and non-targets will be mixed together. But, whenever you see a target item, press the space bar. Your goal is to identify as many target items as possible as quickly as possible’’. Participants in the Variable Target condition were given the following task instructions: ‘‘In this task, you will see items moving across the screen from right to left, one at a time. Most of the items are identical, and require you to do nothing. However, sometimes the item will appear different in some way. Whenever you see an item that looks different from the standard item, press the space bar as quickly as possible. In order to learn what the standard item looks like, press the space bar to view several trials.’’ Participants then viewed 15 non-target trials. Following this, they were told: ‘‘From now on, some of the items will appear different from the standard item you just saw. Whenever you see an item that is different from the standard item, press the space bar. Your goal is to identify as many target items as possible, as quickly as possible.’’ All participants were then given the opportunity to ask any questions of the experimenter after which, the experimental session began. Speed and accuracy were equally stressed. The experiment consisted of a total of 1, 176 trials, divided into six Watch Periods, each of which consisted of 196 trials (180 non-targets and 16 targets, resulting in a target frequency of 8.16%). An illustration of the non-target, as well as the target items is presented in Fig. 1. For the Stable Target group, one of the eight possible targets was randomly selected and presented 16 times within each Watch Period. In contrast, for the Variable Target group, each of the eight target items was presented exactly twice within each Watch Period in an order that was determined randomly. Except for the ‘small’ and ‘large’ target items, all stimuli subtended a visual angle of 2°. The ‘small target’ subtended a visual angle of 1.5°, and the ‘large target’ subtended a visual angle of 2.5°. Stimuli were presented within a square window that subtended 14° of visual angle. This window also acted as a colored background mask that was intended to increase the difficulty of target identification. This background was created by randomly selecting RGB color values within each 0.1° square section of the background window. On each trial, the stimuli first appeared at the right-hand side of the background window and then moved across the screen in a straight line until it had completely traversed the background window. This took 1150 ms, which was the available response window for a given trial. The position of a given stimulus on the y-axis within the background window was randomly chosen on each trial from one of seven equally-spaced locations. The entire experimental session lasted approximately 25 min. 2.2. Results Hits and false alarms, as well as response times for hits were collected for each participant in each of the six Watch Periods. These descriptive statistics are displayed in Table 1. 2.2.1. Response times Response times (RTs) for trials on which targets were correctly responded to (hits) were submitted to a mixed analysis of variance (ANOVA) that treated Target Group (Stable, Variable) as a between-subjects factor and Watch Period (1, 2, 3, 4, 5, 6) as a within-subject factor. These data are displayed in Fig. 2a. There was a significant main effect of Target Group, F(1, 117) = 13.56, p < .001, gp2 = .10, with RTs being faster in the Stable group (M = 582 ms, SE = 10 ms) than in the Variable group (M = 631 ms, SE = 9 ms). There was also a main effect of Watch Period, F(5, 585) = 30.81, p < .001, gp2 = .21, with RTs increasing as the task progressed. Importantly, there was a significant Target Group by Watch Period interaction, F(5, 585) = 5.93,

390

D.R. Thomson et al. / Consciousness and Cognition 33 (2015) 386–397

Table 1 Mean percentage of hits (H) and false alarms (FA), as well as mean response times (RT) in milliseconds, for hits, for each Watch Period in Experiment 1. Standard errors are in parentheses. Watch period 1 Stable Target condition H 96.7(1.2) FA 2.5(0.6) RT 533(13) Variable Target condition H 95.8(0.7) FA 0.9(0.1) RT 615(9)

Response Time (ms)

(A)

2

3

4

5

6

96.9(0.9) 0.6(0.1) 558(14)

96.0(1.4) 0.5(0.1) 580(11)

93.2(1.7) 0.4(0.1) 594(11)

91.4(2.4) 0.5(0.1) 610(12)

90.1(2.4) 0.5(0.1) 616(14)

95.6(0.7) 0.8(0.1) 617(8)

93.5(0.9) 0.6(0.1) 631(7)

95.4(0.8) 0.5(0.2) 633(9)

92.2(1.3) 0.5(0.1) 647(9)

90.9(1.7) 0.8(0.1) 643(11)

700 650 600 550

stable Linear (stable)

500

variable Linear (variable)

450 1

2

3

4

5

6

Watch Period

Mean A' Scores

(B)

1 0.99 0.98 0.97

stable Linear (stable)

0.96

variable Linear (variable)

0.95 1

2

3

4

5

6

Watch Period Fig. 2. Response times for correct target detections (A) and mean A’ scores (B) as a function of Watch Period in Experiment 1 are depicted for both the Stable (open circles) and Variable (open squares) target groups. Dashed and solid lines represent the best linear fit to the Stable and Variable conditions respectively. Error bars represent one standard error of the mean.

p < .001, gp2 = .05. An analysis of the slopes representing the best fitting line for the Stable and Variable groups respectively, revealed that the interaction is driven by a significantly larger increase in RT over Watch Period for the Stable relative to the Variable condition, t(116) = 3.56, p = .001, d = .65. In order to confirm that the increase in RT in the Stable Target condition was not driven by those subjects who received particular targets, an ANOVA was conducted that included Watch Period as a within-subject factor and Target Identity as a between-subjects factor. Importantly, there was no Target Identity by Watch Period interaction (F < 1), indicating that the effects of time on RT did not differ as a function of Target Identity. Finally, independent sample t-tests comparing RT in the Stable and Variable Target groups at each watch period reveals RT to be significantly slower at Watch Periods 1, 2, 3, 4, and 5, t(117) = 5.36, p < .001, d = .98, t(117) = 3.70, p < .001, d = .67, t(117) = 4.06, p < .001, d = .72, t(117) = 2.77, p = .007, d = .51, and t(117) = 2.40, p = .018, d = .43, respectively. RTs were not significantly different from one another in the final Watch Period, t(117) = 1.54, p = .13. Given that the vigilance decrement in RT was smaller in the Variable condition relative to the Stable condition, and given that overall, RT was slower in the Variable condition, we wished to address the notion that perhaps RT did not slow over Watch Period in the Variable condition as much as in the Stable condition because there was ‘less room’ for it to slow. Put differently, vigilance decrements tend to be more pronounced earlier in the task and often ‘level off’. Consequently, RTs in the Variable condition may have simply started off closer to this ‘leveling off’ point. We do not believe this is likely given that mean RTs in the Variable group were not especially high (631 ms) – recall that each trial lasted for 1150 ms and so there is nothing procedural that would impose an upper limit on RTs this fast. However, we sought to support this contention

D.R. Thomson et al. / Consciousness and Cognition 33 (2015) 386–397

391

with additional analyses. For example, one might expect that if RTs increased over time toward some asymptotic level, that RT variability might decrease. Consequently, we submitted mean RT variability per subject per Watch Period to a mixed analysis of variance that treated Target Group (Stable, Variable) as a between-subject factor and Watch Period (1, 2, 3, 4, 5, 6) as a within-subject factor. While there was indeed a main effect of Watch Period, F(5, 585) = 2.99, p = .011, this was due to an increase in RT variability over time (11 ms in Watch Period 1, 15 ms in Watch Period 6). There was no main effect of Target Group, nor was there a Target Group by Watch Period interaction (F’s < 1). 2.2.2. Accuracy Identical analyses were carried out on mean detection accuracy. The percentage of hits and false alarms for each participant at each Watch Period were used to compute A0 (as per Macmillan & Creelman, 2005), which is a more appropriate measure of sensitivity than d0 when there are hit rates of 100% and/or false alarm rates of zero. A0 scores were submitted to an ANOVA that treated Target Group (Variable, Stable) as a between-subjects factor and Watch Period (1, 2, 3, 4, 5, 6) as a within-subject factor. These data are displayed in Fig. 2b. There was no main effect of Target Group, F < 1, with A0 scores being similar in the Stable group (M = .982, SE = .003 ms) and in the Variable group (M = .983, SE = .003). There was a main effect of Watch Period, F(5, 585) = 8.09, p < .001, gp2 = .07, with A0 scores decreasing as the task progressed. Unlike in the pattern of RTs, there was no Target Group by Watch Period interaction, F < 1. Overall, the pattern of A0 scores across Watch Period is not indicative of a speed-accuracy tradeoff interpretation of the pattern of RTs. Put differently, the smaller decrement in RT observed in the Variable condition relative to the Stable condition, is not accompanied by a larger decrement in sensitivity. 2.3. Discussion One of the goals of Experiment 1 was to assess whether the vigilance decrement would be observed in a novel task designed to better approximate the challenges of real world human operators. The vigilance decrement was observed in sensitivity for critical target events as Watch Period progressed for both the Stable and Variable Target conditions. In addition, there were pronounced increases in the response times for correct target detections across Watch Period for both conditions. We have therefore established the utility of the task devised here in measuring vigilance decrements. In addition, we wished to examine the vigilance decrement across two conditions that differ in objective measures of task difficulty. While overall sensitivity for critical target events was similar in the two conditions, response times for correct detections were significantly slower in the Variable condition relative to the Stable condition, indicating that the Variable condition was objectively the more difficult of the two. The primary goal of Experiment 1 was to compare performance decrements between the Stable and Variable Target conditions. Given that our Variable condition was shown to be the more difficult, the resource depletion theory clearly predicts that this condition should exhibit the more pronounced performance decrement. The results of Experiment 1 are not in line with this prediction. First, the vigilance decrement was similar across the Stable and Variable conditions when measured in terms of sensitivity (A0 ), second, the vigilance decrement was actually smaller in the Variable condition when measured in terms of RT. We next seek to replicate and extend these findings in a follow-up experiment since the task employed in Experiment 1 was a novel one, and the critical result somewhat counter-intuitive. 3. Experiment 2 The primary purpose of Experiment 2 was to replicate the effects of target variability on the vigilance decrement observed in Experiment 1. As a secondary goal, we also wished to examine subjective reports of attentional focus over the course of the vigil, based on research into the phenomenon of ‘mind wandering’, which posits that once attention is withdrawn from an external task (a phenomenon known as ‘perceptual decoupling’; Schooler et al., 2011; Smallwood, Beach, Schooler, & Handy, 2008), the result is a re-direction of attention toward self-generated task-unrelated thought (i.e., mind wandering; Smallwood & Schooler, 2006). We therefore included a question at the end of each Watch Period, asking observers to rate the extent to which their attention was focused on task-unrelated thought during the immediately preceding Watch Period. It has been shown previously that reports of mind wandering tend to increase over the course of a standard vigilance task (Cunningham, Scerbo, & Freeman, 2000; McVay & Kane, 2012). This finding is consistent with the idea that the vigilance decrement owes to the withdrawal of attention from the external task due to monotony (or ‘boredom’), and inconsistent with the idea that the vigilance decrement stems from the depletion of information processing resources (which would result in a decrease in task-unrelated thought, since the attentional resources required for task-unrelated thought are argued to become reduced over time – Thomson, Besner, & Smilek, 2015). We therefore took the opportunity to assess the nature of task-unrelated thought over time in the present task. 3.1. Method 3.1.1. Participants Eighty-six undergraduates from the University of Waterloo (26 male, 60 female, mean age of 20.1 years) participated in exchange for course credit.

392

D.R. Thomson et al. / Consciousness and Cognition 33 (2015) 386–397

3.1.2. Apparatus All apparatus were identical to that of Experiment 1. 3.1.3. Stimuli and procedure Stimuli and procedure were identical to that of Experiment 1, with the following exceptions. At the conclusion of each Watch Period, participants were asked to make a subjective judgment regarding the relative amounts of ‘on-task’ and ‘off-task’ thought they engaged in during the most recent Watch Period (referred to as thought probes). Participants were given instructions regarding the thought probes prior to observing the practice trials for the vigilance task. These instructions read as follows: ‘‘At various intervals throughout this task, you will be asked to reflect on the focus of your attention. Specifically, during these ‘thought probes’ you will be asked to rate the extent to which your thoughts were focused on things unrelated to the task at hand since the last thought probe.’’ Each thought probe read as follows: ‘‘Please indicate the extent to which your thoughts were focused on things other than the task since the last thought probe.’’ Participants made a judgment on a seven-point scale by pressing the corresponding number on the keyboard. The rating scale was anchored with a rating of ‘1’ being ‘‘almost always focused ON task’’, and a rating of ‘7’ being ‘‘almost always focused OFF task.’’ Although many mind wandering researchers have begun to use thought probes, in which several responses are possible, including ‘on task’, ‘off task’, ‘task related’, ‘other’ (e.g. McVay & Kane, 2009, 2012), we wanted to keep our thought probes simple and to keep the deliberation period of the participants as short as possible so as not to overly disrupt the vigilance task itself. 3.2. Results Percent hits and false alarms, response times for hits, and degree of ‘off-task’ thought were collected for each participant in each of the six Watch Periods. These descriptive statistics are displayed in Table 2. 3.2.1. Response times Response times (RTs) for trials on which targets were correctly responded to (hits) were submitted to a mixed analysis of variance (ANOVA) that treated Target Group (Stable, Variable) as a between-subjects factor and Watch Period (1, 2, 3, 4, 5, 6) as a within-subject factor. These data are displayed in Fig. 3a. There was a significant main effect of Target Group, F(1, 84) = 11.37, p = .001, gp2 = .12, with RTs being faster in the Stable group (M = 593 ms, SE = 12 ms) than in the Variable group (M = 645 ms, SE = 9 ms). There was also a main effect of Watch Period, F(5, 420) = 21.77, p < .001, gp2 = .21, with RTs increasing as the task progressed. Importantly, there was a significant Target Group by Watch Period interaction, F(5, 420) = 4.45, p = .001, gp2 = .05. An analysis of the slopes representing the best fitting line for the stable and variable groups respectively, revealed that the interaction is driven by a significantly larger linear increase in RT over Watch Period for the Stable relative to the Variable condition, t(84) = 2.66, p = .009, d = .57. In order to confirm that the increase in RT in the Stable condition was not driven by those subjects who received particular targets, an ANOVA was conducted that included Watch Period as a within-subject factor and Target Identity as a between-subjects factor. Importantly, there was no Target Identity by Watch Period interaction (F < 1), indicating that the effects of time on RT did not differ as a function of Target Identity. Finally, independent sample t-tests comparing RT in the Stable and Variable Target groups at each Watch Period reveals RT to be significantly slower at Watch Periods 1, 2, 3, and 6, t(84) = 4.90, p < .001, d = 1.06, t(84) = 3.50, p = .001, d = .75, t(84) = 2.99, p = .004, d = .65, and t(84) = 2.51, p = .014, d = .54, respectively. RTs were not significantly different from one another in Watch Period 4 or 5, t(84) = 1.95, p = .055, and t(84) = 1.48, p = .142, respectively. As with Experiment 1, we sought to examine whether RTs in the Variable condition started out closer to some functional ‘asymptote’ thus limiting the amount of RT slowing that was possible. Again, we do not believe this is likely given that mean RTs in the Variable group were not especially high (645 ms) and participants could respond at any point during the 1150 ms duration of a given trial. We posit that one might expect that if RTs increased over time toward some asymptotic level, that RT variability might well decrease. Consequently, we submitted mean RT variability per subject per Watch Period to a mixed

Table 2 Mean percentage of hits (H) and false alarms (FA), as well as mean response times (RT) in milliseconds, for hits, and mean level of reported task-unrelated thought (TUT) for each Watch Period in Experiment 2. Standard errors are in parentheses. Watch period 1 Stable Target condition H 97.7(0.5) FA 1.7(0.4) RT 538(12) TUT 4.02(0.27) Variable Target condition H 94.9(1.3) FA 1.9(2.7) RT 624(13) TUT 3.69(0.26)

2

3

4

5

6

98.0(0.6) 1.1(0.5) 572(14) 4.14(0.26)

97.3(0.9) 0.7(0.3) 594(14) 4.20(0.25)

95.7(1.0) 0.5(0.1) 612(14) 4.27(0.27)

93.9(1.3) 0.7(0.3) 623(14) 4.55(0.25)

94.5(1.1) 0.7(0.2) 621(14) 4.45(0.28)

93.6(1.4) 1.0(0.1) 633(10) 3.74(0.23)

93.5(1.5) 0.8(0.1) 647(10) 4.05(0.24)

93.3(1.5) 0.5(0.1) 646(11) 4.38(0.25)

90.9(1.9) 0.6(0.1) 648(9) 4.79(0.27)

92.4(1.7) 0.4(0.1) 671(14) 4.50(0.28)

393

D.R. Thomson et al. / Consciousness and Cognition 33 (2015) 386–397

Response Time (ms)

(A)

700 650 600 550

stable Linear (stable)

500

variable Linear (variable)

450 1

2

3

4

5

6

Watch Period

Mean A' Scores

(B)

1 0.995 0.99 0.985 0.98 0.975 0.97 0.965 0.96

1

stable

variable

Linear (stable)

Linear (variable)

2

3

4

5

6

(C)

Level of Task-Unreltaed Thought

Watch Period 7 6 5 4 3 2

stable

variable

Linear (stable)

Linear (variable)

1 1

2

3

4

5

6

Watch Period Fig. 3. Response times for correct target detections (A), mean A’ scores (B), and mean reports of task-unrelated thought (C) as a function of Watch Period in Experiment 2 are depicted for both the Stable (open circles) and Variable (open squares) target groups. Dashed and solid lines represent the best linear fit to the Stable and Variable conditions respectively. Error bars represent one standard error of the mean.

analysis of variance that treated Target Group (Stable, Variable) as a between-subject factor and Watch Period (1, 2, 3, 4, 5, 6) as a within-subject factor. While there was a main effect of Watch Period, F(5, 420) = 4.15, p = .001, this was again due to an increase in RT variability over time (12 ms in Watch Period 1, 17 ms in Watch Period 6). There was no main effect of Target Group, nor was there a Target Group by Watch Period interaction (F’s < 1). 3.2.2. Accuracy Identical analyses were carried out on accuracy as on mean response time. The percentage of hits and false alarms for each participant at each Watch Period were used to compute A’ as in Experiment 1. A’ scores were submitted to an ANOVA that treated Target Group (Variable, Stable) as a between-subjects factor and Watch Period (1, 2, 3, 4, 5, 6) as a within-subject factor. These data are displayed in Fig. 3b. There was a main effect of Target Group, F(1, 84) = 4.54, p < .001, gp2 = .05, with A’ scores being higher in the Stable group (M = .988, SE = .002) than in the Variable group (M = .980, SE = .003). There was a main effect of Watch Period, F(5, 420) = 2.74, p = .019, gp2 = .03, with A’ scores decreasing as the task progressed. Unlike in the pattern of RTs, there was no Target Group by Watch Period interaction, F < 1. Overall, the pattern of A’ scores across Watch Periods is not indicative of a speed-accuracy tradeoff interpretation of the pattern of RTs. Again, the smaller decrement in RT observed in the Variable condition relative to the Stable condition, is not accompanied by a larger decrement in sensitivity. 3.2.3. Thought probes Thought probe responses were submitted to a mixed analysis of variance that treated Target Group (Stable, Variable) as a between-subjects factor and Watch Period (1, 2, 3, 4, 5, 6) as a within-subject factor. These data are displayed in Fig. 3c. There was no main effect of Target Group, with mean reports of task-unrelated thought (TUT) being similar in the Stable (M = 4.27,

394

D.R. Thomson et al. / Consciousness and Cognition 33 (2015) 386–397

SE = .21) and Variable (M = 4.20, SE = .20) groups. There was however a main effect of Watch Period, F(5, 420) = 6.45, p < .001, gp2 = .07, with reports of task-unrelated thought increasing in a linear manner as the task progressed, F(1, 84) = 13.37, p < .001, gp2 = .14. Finally, there was no significant Target Group by Watch Period interaction, indicating that the increase in reports of task-unrelated thought across Watch Period was similar for the Stable and Variable groups. We next wanted to assess whether increases in TUT across Watch Period predicted changes in RT and/or A0 across Watch Period in either of the Stable or Variable conditions. To that end, slopes representing TUT, RT, and A’ were entered into a Pearson correlation matrix. None of these relations reached significance however, indicating that changes in TUT over time-on-task were unrelated to changes in performance over that same time period. 3.3. Discussion The primary goal of Experiment 2 was to replicate the finding that the vigilance decrement observed in Experiment 1 was not larger for the more demanding condition. In fact, increases in RT over Watch Period were more pronounced in the easier Stable condition relative to the more challenging Variable condition. This pattern was again observed in Experiment 2. This is particularly noteworthy since performance was overall slower and less accurate in the Variable condition, thus objectively verifying that it was the more difficult condition. According to the resource depletion account, the more demanding (difficult) condition should have exhibited the larger decrement. Thus the results of the present experiment are once again at odds with the depletion account, which posits task difficulty as the primary determinant of the magnitude of the vigilance decrement. A secondary goal of Experiment 2 was to determine the contribution (if any) of task-unrelated thought (mind wandering) to the vigilance decrement. First, subjective reports indicated a linear increase in pre-occupations with task-unrelated thought as a function of time-on-task (consistent with prior work – Cunningham et al., 2000; McVay & Kane, 2012; Thomson, Seli, Besner, & Smilek, 2014). Second, the extent of this increase did not differ between the Stable and Variable conditions, and correlational analyses indicated that increases in the amount of reported off-task thought did not predict increases in response times for correct detections of critical events, or accuracy scores across Watch Period. 4. General discussion The purpose of the present work was to assess the counter-intuitive hypothesis that increasing the perceptual variability of a target in the context of a vigilance task might actually alleviate performance decrements over time that are typically observed. Importantly, it was hypothesized that increasing the perceptual variability of the task-relevant stimuli would also increase the difficulty of the task. As a result, we were also afforded the opportunity to test the resource depletion account of vigilance performance (and indeed, of sustained attention in general), which explicitly states that performance decrements should be steeper in more difficult tasks (Grubb et al., 1995; Helton & Russell, 2011; Smit et al., 2004; Warm & Dember, 1998). In fact, it has even been stated that: ‘‘. . .it is not task duration per se, but rather resource demands that determine vigilance performance’’ (Smit et al., 2004, p. 41). In both of the experiments reported here, perceptually variable targets were responded to more slowly than perceptually stable targets (and less accurately, in Experiment 2), objectively verifying the greater relative difficulty of the Variable condition with respect to the Stable condition. Crucially however, the vigilance decrement in RT was less pronounced when observers monitored for variable targets relative to when they monitored for stable targets, and the decrement in sensitivity was equivalent between the two conditions. The results of the work reported here stand in contrast to a large body of research in support of the resource depletion account of the vigilance decrement. Specifically, relative increases in task difficulty in prior work have been shown to magnify performance decrements over time-on-task (e.g., Helton & Russell, 2011; Helton & Russell, 2013; Smit et al., 2004). In addition, prior attempts to increase task engagement have failed to alleviate the decrement (Helton & Russell, 2012; Ossowski, Malinen, & Helton, 2011). The experiments reported here however, differ from those of prior work in at least two important ways: (1) Whereas prior work has primarily manipulated task demands by inserting additional tasks that draw attention away from the critical events, our manipulation of perceptual variability was tied directly to the task-relevant stimuli. (2) Prior work assessing the difficulty of vigilance tasks has primarily relied upon retrospective self-reports of perceived workload and stress, whereas in the present experiments, we use task performance as an objective measure of task difficulty. Thus, by manipulating variability with respect to the task-relevant stimuli, we have provided the first demonstration of an objectively difficult vigilance task that yields a smaller decrement than an objectively easier version of that same task. We contend that the present findings cast doubt on the notion that the vigilance decrement primarily owes to task difficulty (as is argued by depletion theorists), and instead are more consistent with the idea that decrements derive from an increasing state of under-stimulation as the task progresses. Indeed, the results of Experiment 2 may even be taken to suggest that observers cope with this increasing lack of engagement by pursuing self-generated, task-unrelated thoughts (which again suggests that resources do not become depleted over time-on-task, since it is typically assumed that information processing resources are needed to engage in off-task thought – Smallwood, 2010; Smallwood & Schooler, 2006). Future work is necessary however, in order to more directly assess the link between mind wandering and the vigilance decrement. For example, it has been shown that mind wandering can be further broken down into ‘task-related’ thought (TRT, e.g. thinking

D.R. Thomson et al. / Consciousness and Cognition 33 (2015) 386–397

395

about one’s performance on the task) and ‘task-unrelated’ thought (TUT, e.g. thinking about personal concerns or worries; e.g., McVay & Kane, 2012). While the design of the current study allowed only a small number of thought probes (and thus prevented us from assessing the nature of reported mind wandering in a more detailed manner), future work might employ vigilance tasks that are more amenable to a qualitative distinction between the various types of mind wandering and their relation to vigilance performance (but see Carter, Russell, & Helton, 2013, who attempted to distinguish TUT from TRT in a retrospective manner). Novel theoretical accounts of sustained attention have recently been forwarded that may offer greater explanatory power than the resource depletion theory in accounting for the results of the present study. It could be the case, for instance, that some additional process, such as executive control, is what falters over time in vigilance tasks, resulting in both performance declines as well as increases in task-unrelated thought (Thomson, Besner, & Smilek, 2013). Specifically, Thomson et al. (2015) have recently forwarded a ‘resource control’ account of sustained attention in which executive control serves to maintain the distribution of attentional resources among competing goals. Over time, a greater exertion of executive control is required to maintain attentional resources on the task-relevant stimuli (due to increasing monotony and boredom associated with the task) and to prevent attentional resources from becoming overly preoccupied with self-generated task-unrelated thought. When executive control is relaxed to the point that the amount of attentional resources devoted to the external task is insufficient to maintain a high level of attentional focus, performance decrements are observed. In the context of the present work, increased perceptual variability may serve to decrease task monotony, thus making it easier to control the allotment of attentional resources to the vigilance task for longer periods of time. A somewhat related theoretical account of vigilant attention has also been forwarded, in which it is argued that maintaining the focus of attention on an unrewarding task causes a depletion of self-control. As a consequence, active goal maintenance becomes increasingly difficult with the passage of time (Langer, Willmes, Chatterjee, Eickhoff, & Sturm, 2010; see also, Langer & Eickhoff, 2013). Again, the variability manipulation employed in the present work may serve to increase the subjective appeal of the vigilance task, and thus, maintaining attention to the task may not require such a high degree of self-regulation. The fact that the difference in performance over time between the Variable and Stable Target conditions emerged in response times, but not accuracy measures deserves further comment. In our effort to create a vigilance task that would more closely approximate a real-world situation, we had our targets traverse across the computer screen at a rate of one per 1.15 s (recall that our inspiration for this task was abstracted from the case of the airport security screener). Compared to vigilance tasks used in prior work, this is a relatively long period of time, since target presentation times typically vary from 50 ms (Helton & Russell, 2012) to 200 ms (Caggiano & Parasuraman, 2004; Grier et al., 2003). However, in prior work, the target stimuli are perceptually very simple, such as line segments, dots, or letters. In the present study, we created stimuli that are perceptually more complex than in prior tasks (even in the Stable condition), and in which there was spatial uncertainty with regards to where stimuli would appear on the screen. For this reason, we extended the available ‘dwell’ time of the targets compared with prior studies. In addition, we again wanted to approximate real-world conditions in which fixated objects are not immediately ‘masked’ as in prior vigilance studies. Thus, the present paradigm may simply be better suited for observing differences in decrements in response times, but not sensitivity. Nonetheless, it is important to note that even though sensitivity decrements were not larger for the Stable relative to the Variable condition here, they were not smaller either. Recall, that under a resource depletion account, more difficult tasks should yield larger decrements. Thus, even a failure to find a difference between sensitivity decrements between two conditions that objectively differ in task difficulty, is largely inconsistent with the depletion account, and at the very least, indicates that vigilance decrements owe to much more than the difficulty of the monitoring task. It is not our contention that increasing task difficulty will alleviate the vigilance decrement in a strictly linear manner. That is, increasing task variability will have a positive effect on one’s ability to maintain attention to the task, but only to a point. Eventually, the task will become sufficiently difficult that even when observers are fully attentive, performance declines will likely be significant. In other words, an observer’s motivation to maintain attention to the task at hand will likely wane to the extent that performing the task becomes untenable, or overly difficult. Indeed, this idea has been formalized as the ‘inverted-U’ hypothesis, which states that there is an optimal level of arousal in vigilance tasks that promotes the best performance over time. Being too bored, or too busy, are equally detrimental to performance, and thus the relation between arousal and vigilance performance is best described as an inverted ‘U’ shaped function (Wiener, Curry, & Faustina, 1984). Finally, we wish to make explicit that the desirable effects of perceptual variability seen in the experiments reported here are specific to the change in performance over time. That is, even though the vigilance decrement in response times is smaller in the Variable condition, performance is still worse overall than in the Stable condition. Thus, while increasing perceptual variability may indeed help maintain the coupling of attentional resources to the external task over time, building in additional task-related complexity into everyday tasks such as air traffic control and security screening is likely not advisable. 4.1. Conclusions The work reported here provides clear evidence for the counterintuitive finding that increasing the difficulty of a task can actually help performance to remain more stable over time. The leading theoretical account of the vigilance decrement (the ‘resource depletion’ account) explicitly posits that the primary determinant of vigilance performance over time is task difficulty, with more difficult tasks producing larger performance decrements. While there are a few examples in the literature of beneficial effects of task complexity on vigilance performance, the present experiments are the first to show that objective

396

D.R. Thomson et al. / Consciousness and Cognition 33 (2015) 386–397

increases in task difficulty need not result in larger vigilance decrements, and in some cases, can even yield smaller decrements. Moreover, we have employed a novel vigilance task intended to more closely approximate the challenges faced by human operators in the real-world (admittedly, the paradigm employed here makes only a modest step forward in this regard, but nonetheless makes progress toward addressing the problem of vigilance decrements representing iatrogenic artifacts – see Hancock, 2013 for a discussion of this idea). Importantly, we have shown that when it comes to the monitoring of automated systems (as in the example of the airport security screener with which we began this article), performance declines over time owe to more than simply task difficulty. Acknowledgments This research was funded by NSERC Discovery Grants awarded to DS and DB. We would like to thank Kirill Zaitsev, Jenny Wan, and Talia Hashmani for assistance with data collection. In addition, we thank James Allen Cheyne for insightful theoretical discussions during the development of this research. Additionally, we thank William S. Helton for constructive comments on an earlier version of this manuscript. References Berlyne, D. E. (1970). Novelty, complexity, and hedonic value. Perception & Psychophysics, 8, 279–286. Berlyne, D. E., Ogilvie, J. C., & Parham, L. C. (1968). The dimensionality of visual complexity, interestingness, and pleasingness. Canadian Journal of Psychology, 22, 376–387. Caggiano, D. M., & Parasuraman, R. (2004). The role of memory representation in the vigilance decrement. Psychonomic Bulletin & Review, 11, 932–937. Carter, L., Russell, P. N., & Helton, W. S. (2013). Target predictability, sustained attention, and response inhibition. Brain and Cognition, 82, 35–42. Cunningham, S., Scerbo, M. W., & Freeman, F. G. (2000). The electrocortical correlates of daydreaming during vigilance tasks. Journal of Mental Imagery, 24, 61–72. Grier, R. A., Warm, J. S., Dember, W. N., Matthews, G., Galinsky, T. L., Szalma, J. L., et al (2003). The vigilance decrement reflects limitations in effortful attention, not mindlessness. Human Factors: The Journal of the Human Factors and Ergonomics Society, 45, 349–359. Grubb, P. L., Warm, J. S., Dember, W. N., & Berch, D. B. (1995). Effects of multiple-signal discrimination on vigilance performance and perceived workload. Proceedings of the Human Factors and Ergonomics Society Annual Meeting, 39, 1360–1364. Hancock, P. A. (2013). In search of vigilance: The problem of iatrogenically created psychological phenomena. American Psychologist, 68, 97–109. Helton, W. S., & Russell, P. N. (2011). Working memory load and the vigilance decrement. Experimental Brain Research, 212, 429–437. Helton, W. S., & Russell, P. N. (2012). Brief mental breaks and content-free cues may not keep you focused. Experimental Brain Research, 219, 37–46. Helton, W. S., & Russell, P. N. (2013). Visuospatial and verbal working memory load: Effects on visuospatial vigilance. Experimental Brain Research, 224, 429–436. Helton, W. S., & Warm, J. S. (2008). Signal salience and the mindlessness theory of vigilance. Acta Psychologica, 129, 18–25. Langer, R., & Eickhoff, S. B. (2013). Sustaining attention to simple tasks: A meta-analytic review of the neural mechanisms of vigilant attention. Psychological Bulletin, 139, 870–900. Langer, R., Willmes, K., Chatterjee, A., Eickhoff, S. B., & Sturm, W. (2010). Energetic effects of stimulus intensity on prolonged simple reaction-time performance. Psychological Research, 74, 499–512. Mackworth, N. H. (1948). The breakdown of vigilance during prolonged visual search. Quarterly Journal of Experimental Psychology, 1, 6–21. Macmillan, N. A., & Creelman, C. D. (2005). Detection Theory: A User’s Guide Lawrence Erlbaum Associates. Mahwah, NJ. Manly, T., Robertson, I. H., Galloway, M., & Hawkins, K. (1999). The absent mind: Further investigations of sustained attention to response. Neuropsychologia, 37, 661–670. McVay, J. C., & Kane, M. J. (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. McVay, J. C., & Kane, M. J. (2012). 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. Molloy, R., & Parasuraman, R. (1996). Monitoring an automated system for a single failure: Vigilance and task complexity effects. Human Factors: The Journal of the Human Factors and Ergonomics Society, 38, 311–322. Ossowski, U., Malinen, S., & Helton, W. S. (2011). The effects of emotional stimuli on target detection: Indirect and direct resource costs. Consciousness and Cognition, 20, 1649–1658. Parasuraman, R. (1979). Memory load and event rate control sensitivity decrements in sustained attention. Science, 205, 924–927. Parasuraman, R., & Davies, D. R. (1977). A taxonomic analysis of vigilance performance. In R. Mackie (Ed.), Vigilance (pp. 559–574). US: Springer. Parasuraman, R., de Visser, E., Clarke, E., McGarry, W. R., Hussey, E., Shaw, T., et al (2009). Detecting threat-related intentional actions of others: Effects of image quality, response mode, and target cuing on vigilance. Journal of Experimental Psychology: Applied, 15, 275–290. Parasuraman, R., & Mouloua, M. (1987). Interaction of signal discriminability and task type in vigilance decrement. Perception & Psychophysics, 41, 17–22. Peirce, J. W. (2007). PsychoPy—psychophysics software in Python. Journal of Neuroscience Methods, 162, 8–13. Pop, V. L., Stearman, E. J., Kazi, S., & Durso, F. T. (2012). Using engagement to negate vigilance decrements in the NextGen environment. International Journal of Human-Computer Interaction, 28, 99–106. Robertson, I. H., Manly, T., Andrade, J., Baddeley, B. T., & Yiend, J. (1997). Oops!’: Performance correlates of everyday attentional failures in traumatic brain injured and normal subjects. Neuropsychologia, 35, 747–758. Sanders, A. F. (1983). Towards a model of stress and human performance. Acta Psychologica, 53, 61–97. Sanders, A. F. (1998). Elements of human performance: Reaction processes and attention in human skill. New Jersey: Erlbaum. Schooler, J. W., Smallwood, J., Christoff, K., Handy, T. C., Reichle, E. D., & Sayette, M. A. (2011). Meta-awareness, perceptual decoupling and the wandering mind. Trends in Cognitive Sciences, 15, 319–326. Smallwood, J. (2010). Why the global availability of mind wandering necessitates resource competition: Reply to McVay and Kane (2010). Psychological Bulletin, 136, 202–207. Smallwood, J., Beach, E., Schooler, J. W., & Handy, T. C. (2008). Going AWOL in the brain: Mind wandering reduces cortical analysis of external events. Journal of Cognitive Neuroscience, 20, 458–469. Smallwood, J., & Schooler, J. W. (2006). The restless mind. Psychological Bulletin, 132, 946. Smit, A. S., Eling, P. A., & Coenen, A. M. (2004). Mental effort causes vigilance decrease due to resource depletion. Acta Psychologica, 115, 35–42. Thomson, D. R., Besner, D., & Smilek, D. (2013). In pursuit of off-task thought: Mind wandering-performance trade-offs while reading aloud and color naming. Frontiers in Psychology, 4, 360. Thomson, D. R., Besner, D., & Smilek, D. (2015). A resource control account of sustained attention: Evidence from mind wandering and vigilance paradigms. Perspectives on Psychological Science, 10, 82–96.

D.R. Thomson et al. / Consciousness and Cognition 33 (2015) 386–397

397

Thomson, D. R., Seli, P., Besner, D., & Smilek, D. (2014). On the link between mind wandering and task performance over time. Consciousness and Cognition, 27, 14–26. Warm, J. S., & Dember, W. N. (1998). Tests of vigilance taxonomy. In R. R. Hoffman, M. F. Sherrick, & J. S. Warm (Eds.), Viewing psychology as a whole: The integrative science of William N. Dember (pp. 79–112). Washington, DC: American Psychological Association. Warm, J. S., Parasuraman, R., & Matthews, G. (2008). Vigilance requires hard mental work and is stressful. Human Factors: The Journal of the Human Factors and Ergonomics Society, 50, 433–441. Wiener, E. L., Curry, R. E., & Faustina, M. L. (1984). Vigilance and task load: In search of the inverted U. Human Factors: The Journal of the Human Factors and Ergonomics Society, 26, 215–222.