The Use of Eye Movements as an Objective Measure of Mind Wandering

Transcription

1 The Use of Eye Movements as an Objective Measure of Mind Wandering by Sarah Uzzaman A thesis submitted in conformity with the requirements for the degree of Masters of Arts Department of Psychology University of Toronto Copyright by Sarah Uzzaman 2010

2 The Use of Eye Movements as an Objective Measure of Mind Wandering Abstract Sarah Uzzaman Masters of Arts Department of Psychology University of Toronto 2010 Previous research on mind wandering has used subjective verbal reports as a means to acquire the frequency and rate of its occurrence. I proposed the use of eye movements as an objective measure of mind wandering while participants attempted a reading task. Participants were placed in a self-classified probe-caught mind wandering paradigm while their eye movements were recorded. They were randomly probed every 2-3 minutes and were required to indicate whether their mind had been wandering before they were allowed to continue reading. The overall pattern showed that eye movement behaviour was slower and less frequent when participants reported mind wandering episodes, with duration and frequency of within-word regressions reaching levels of statistical significance. It may be that the cognitive processes that normally guide eye movements during reading exert less control during mind wandering episodes. Other implications and limitations are discussed. ii

3 Table of Contents Table of Contents... iii List of Figures... iv List of Appendices... v 1 Introduction Mind Wandering Methodology Direct Methods Indirect Methods Tasks Contexts Associated with Mind Wandering Eye Movements Reading Present Study Methods Participants Apparatus Procedure Results Discussion References Appendix 1: List of Eye Movement Variables Appendix 2: Experiment Instructions iii

4 List of Figures Figure 1: Paired differences in means of the eye movement variables iv

5 Appendix 1: Eye Movement Variables Appendix 2: Experiment Instructions List of Appendices v

6 EYE MOVEMENTS DURING MIND WANDERING EPISODES 1 1 Introduction It is most likely the case that as you begin to read this paper, you are paying at least some attention to the words, to the overall meaning of the passage, and to the manner in which it fits with the larger meaning of the text. However, there is a high probability that sometime during the course of reading this paper your attention may wander elsewhere. Should that occur, your eyes may continue to read the words across the page, the phonetics of the words may still be captured and sounded in your mind, yet semantic processing of the text may not occur. At some point, you will realize that your attention has left the text, and that you were not aware of the contents and meaning of what you were just reading. This phenomenon of zoning out or mind wandering while reading is common in our daily lives. These are situations in which you fail to notice that your immediate goal of task completion and comprehension has been temporarily displaced by another concern (Smallwood & Schooler, 2006). Even though the breadth of literature within this area is small, there is empirical evidence to support the existence of mind wandering as an actual cognitive phenomenon (Smallwood & Schooler, 2006; Schooler, Riechle & Halphern, 2004; Schooler, 2002). However, this evidence is primarily based on subjective self report and indirect methods rather than objective measures. The purpose of this paper is to reaffirm the existence of mind wandering episodes during reading and to explore eye movements as a potential objective measure. 1.1 Mind Wandering Mind wandering involves situations in which individuals fail to notice that their immediate goal of task completion has been temporarily displaced by another concern (Smallwood & Schooler, 2006). In order to move towards an objective measurement of mind wandering, it is important to have a clear understanding of what it is we are trying to measure. A critical characteristic relevant to this understanding is the distinction between experiential consciousness and metaconsciousness (Schooler, 2002). Experiential consciousness refers to the contents of one s

7 EYE MOVEMENTS DURING MIND WANDERING EPISODES 2 consciousness at some point of time. Thus while reading, we experience the text in the sense that it populates our consciousness. Meta-consciousness refers to one s explicit awareness of the contents of consciousness. Thus, if we are aware of experiencing the text, then we are in a metaconscious state. A mind wandering episode while reading is a great example of this dissociation: when an individual experiences a mind wandering episode, they lack the meta-consciousness that their mind have wandered away from the primary task, while still being experientially conscious of whatever topic that has grabbed their attention. At some point, they suddenly become metaconscious that they have zoned out, and they were reading without comprehension. This temporal dissociation occurs when meta-consciousness is directed towards an experience that previously occurred in the absence of explicit awareness (Schooler, 2002). Thus, when individuals zone out while reading, they lack the meta-consciousness that their mind has wandered away from the primary task, while still being experientially conscious of whatever topic that has grabbed their attention (Schooler, Reichle & Halphern, 2004). At some point, they suddenly become meta-conscious that they have zoned out and have been reading without comprehension (Schooler, Reichle & Halphern, 2004). This distinction can be adaptive, for example when one is in deep concentration, or engaged in automatic or intuitive behaviours (Csikszentmihalya, 1990; Schooler, Ohlsoon & Brooks, 1993). However, lack of metaconsciousness can be detrimental in some cases, such as reading text or listening to an instructor Methodology The subjective component of mind wandering episodes has made it difficult to capture and measure them objectively, which has been one of the core bases for the limited research within the domain. To date, there have been two distinct ways of assessing mind wandering in laboratory settings: direct and indirect. Direct methods mainly consist of verbal self report in different formats, whereas indirect methods consist mainly of memory tests and behavioural responses as a result of mind wandering episodes during a given task. Neither of the methods truly provides an objective measure that is void of biases or confounding variables, as both are based on reports provided by the participants.

8 EYE MOVEMENTS DURING MIND WANDERING EPISODES Direct Methods Thought sampling, assessing the inner experience of an individual as they complete a task in a controlled experimental setting, has been the most common method in capturing mind wandering episodes in participants. For example, during an experimental task, participants would be asked to state what is going through their mind, what they were thinking of, when questioned. Ecologically valid methods of thought sampling also exist where participants are asked to evaluate their subjective experiences during the course of the day using electronic devices, such as pagers or smartphones (Smallwood & Schooler, 2006; Shiffman, 2000). For example, Shiffman (2000) provided participants with pagers and randomly probed participates throughout the day. When probed, the participants were asked to describe their internal experiences and thoughts. The most common thought sampling technique is assessing the number of taskunrelated-thoughts experienced by an individual, which refers to any thoughts directed away from the current situation or task, such as a daydream, the mental reliving of a prior event or even thoughts concerning future events or experiences. In empirical investigations, taskunrelated-thoughts are gathered through two main techniques: (a) probe-caught mind wandering, in which individuals are interrupted during the performance of a task and asked to report their experiences, or (b) self-caught mind wandering, where participants are asked to monitor their awareness for off-task episodes and report them each time they occur (Smallwood & Schooler, 2006). Probe-caught mind wandering can be further categorized into self-classified or experimenterclassified probe-caught mind wandering. Participants in the self-classified conditions are initially trained to recognize mind wandering episodes. Thus when probed during a given task, these trained participants could accurately report the occurrence of mind wandering episode. Typically, these probes are framed as simple yes/no judgment calls. In experimenter-classified conditions, participants are required to state their content of experience (i.e., what is passing through their minds) when probed, and these verbal statements are recorded. Subsequently, the experimenter codes these experiences as mind wandering episodes accordingly.

9 EYE MOVEMENTS DURING MIND WANDERING EPISODES 4 Both of these probe-caught methods of identifying mind wandering episodes have their advantages and disadvantages. Self-classified probe-caught mind wandering conditions are less disruptive to the current task as participants are only required to make a yes/no judgment, while in experimenter-classified probe-caught mind wandering conditions, participants are required to explain and elaborate on their prior thoughts. Participants in the experimenter-classified condition are not informed regarding the true nature of the experiment, and are only told about the occurrence of the probes, thus providing an unbiased outlook. However, these participants may have some reservations as to truly provide the contents of their experience, thus leading to misrepresentation of the data. In addition, participants themselves are better apt at classifying their own mind wandering episodes than the experimenter. Unlike what occurs when the probe-caught mind wandering technique is used, when the selfcaught mind wandering technique is utilized, participants must be aware of the not only of what they are thinking about, but also what they are supposed to be thinking about. Specifically, when participants indicate a mind wandering episode via the self-caught paradigm, there are two distinct activities occurring: firstly, the participants must undergo a mind wandering episode, and secondly they must become meta-conscious of the mind wandering episode in order to report it. Thus, there is an extra step needed when the self-caught mind-wandering technique is used which is not required in the probe-caught mind-wandering technique. Self-caught mind wandering conditions are less disruptive to the participant as only a yes/no judgment call is required, although being explicitly informed of the probes and definition of mind wandering may have an effect on the frequency of occurrence of mind wandering episodes and rate of reporting them. The potential effects of monitoring one s own thought content has been a cause of concern when using the self-caught mind wandering technique, as well as self-classified probe-caught mind wandering methodologies. For example, if you are told not to think about a purple elephant, chances are that you are thinking of a purple elephant. Thus when informed about a topic or a

10 EYE MOVEMENTS DURING MIND WANDERING EPISODES 5 process, we become highly cognizant of it, which can have implications when inquiring into the frequency of that particular process. In this case, by informing participants of mind wandering episodes, we run the risk of elevating their frequency. Studies have tried to resolve this concern by using behaviours that cannot occur during mind wandering episodes. For example, encoding for the primary task is difficult if your mind is wandering on an internal topic. Thus by correlating the number of mind wandering episodes captured via self-caught and probe-caught techniques during retrieval of task content, one can determine if being aware of one s own thought process has an effect on the frequency of mind wandering episodes and if either of the methods is better apt at capturing it (Smallwood, Baracaia, Lowe & Obonsawin, 2003; Smallwood, Davies, et al., 2004). Several studies have found that both the techniques, self caught and probe caught, are relatively equal at detecting mind wandering episodes (r = 0.60, Smallwood, Baracaia, et al., 2003; r = 0.50, Smallwood, Davies, et al., 2004; r = 0.50, Smallwood, O Connor, et al., 2004; Smallwood & Schooler, 2006) Although both are subjective measures, when paired together the probe-caught and self-caught mind wandering techniques can provide useful information regarding the process by which mind wandering episodes arise and the rate at which participants become meta-conscious of them. Probe-caught mind wandering condition allows one to get a baseline frequency of the occurrence of mind wandering as it does not require participant s awareness, while self-caught mind wandering technique provides the frequency at which one becomes meta-conscious of mind wandering episodes. One of the experiments conducted by Schooler, Reichle and Halpern (2004) asked participants to read passages while both probe-caught and self-caught methods were implemented. Results indicated that participants were not meta-conscious of their mind wandering 13% of the time, as the probe caught them experiencing a mind wandering episode before they caught themselves. Manipulations, such as intoxication or fatigue, can allow one to observe the differences within probe-caught and self-caught mind wandering techniques. As one would predict, alcohol intoxication or nicotine deprivation caused a decrease in self-caught mind wandering episodes as it would be difficult to reach a state of meta-conscious when intoxicated or craving nicotine (Sayette, Kirchner, Reichle, & Schooler, 2006; Sayette, Schooler & Reichle, 2009). The differences in frequencies of mind wandering episodes between self-caught and probe-caught techniques can allow us to gain insight into the workings of meta-consciousness.

11 EYE MOVEMENTS DURING MIND WANDERING EPISODES 6 Mind wandering studies have mostly relied on subjective self reports of participants. However, verbal self reports cannot be taken on face value, regardless of how earnest the participant may be. Although it has been shown, by way of correlation, that knowledge and awareness of mind wandering episodes does not drastically alter a secondary task response (memory), the use of self reports as evidence is a cause of concern when conducting empirical investigations and validating the existence and occurrence of a cognitive phenomenon Indirect Methods The premise for indirect methods is as follows: when mind wandering occurs, attention and resources are focused internally and as a result, task performance suffers. Models of processing propose that the amount of mind wandering experienced during any given task is inversely proportional to the amount of resources deployed towards the primary task (Smallwood, Baracaia, Lowe & Obonsawin, 2003). Using the same model, it can also be stated that mind wandering would be more likely to occur when a task can be performed in a relatively automatic manner, such as reading or driving, and less likely to occur when the task is more demanding, such as mental rotations or n-back tests. Depending on the primary task, performance can be measured in a number of different ways. For example one could assess memory of the task content, comprehension of the text, error detection within the text, and reaction times and response inhibition during a go/no-go task. Memory of the primary task is one of the most common forms of indirect methods for capturing mind wandering episodes. Intuitively, as the mind wanders away from the primary task, one would be less likely to recall information presented. For example, during a word recall paradigm, where a list of words is being displayed for you to remember at the later time, you experience a mind wandering episode. Your memory for the words presented will be lower than if you had not experienced the mind wandering episode. Several studies have used this measure paired with

12 EYE MOVEMENTS DURING MIND WANDERING EPISODES 7 direct methods as evidence of mind wandering (Sayette, Kirchner, Reichle, & Schooler, 2006; Smallwood, Baracaia, et al., 2003; see review Smallwood & Schooler 2006). Jacoby s (1991) distinction between recollection and familiarity have also been paired with comprehensive reading and mind wandering, respectively. Recollection for an item is when the participant knows with a high degree of confidence that they ve seen the item previously, while familiarity arises when a previously viewed item affects their behaviour without conscious knowledge of it. As mind wandering occurs when attention shifts away from the external environment, then subsequent retrieval of information may occur via familiarity rather than recollection. Initial research has shown results consistent with this hypothesis. By fashioning manipulations that increase the frequency of mind wandering, such as decreasing interest or increasing the length of the primary task, studies show recollection measures for the primary task to be reduced, while familiarity measures of the same primary task remain stable (Giambra & Grodsky, 1989). In addition, higher frequency of mind wandering is associated with higher false alarm rates in immediate recognition (Smallwood, Baracaia, et al., 2003). Based on the same premise as memory tests, comprehension tests go a bit further by trying to observe the level and depth of processing taking place regarding the primary task. Participants not only have to remember the content of the task, but also must be able to think logically and coherently about the contents. Smallwood, McSpadden and Schooler (2008) investigated the ability of participants to identify the killer based on clues provided in a murder story. Participants were placed in a self-classified probe-caught paradigm, and were subsequently given a comprehension test. Participants who indicated higher frequency of mind wandering episodes performed poorly on the comprehension tests, and were less likely to correctly identify the killer in the story. Although these methods are not as subjective as verbal self reports, they still can be influenced by many factors, such as motivation, interest and difficulty. Performance measures may be able to distinguish mind wandering episodes better than selfreport and/or memory procedures. Tasks that need continuous monitoring and responding can allow us to observe behavioural performance online. Smallwood, McSpadden and Schooler (2007) attempted to use reaction times and response inhibition during a simple go/no-go task to

13 EYE MOVEMENTS DURING MIND WANDERING EPISODES 8 explore the relationship of the frequency of mind wandering with the performance measures. Shorter reaction times and failures to correctly withhold a response were assumed to be indicative of attention being directed away from the primary task. Using a self-classified probecaught technique, mind wandering was classified in two categories: zoning out (classic mind wandering) and tuning out (intentionally mind wandering, where you are aware that your thoughts have drifted away from the task, yet you continue to read). Observed results displayed a pattern consistent with expectations, where shorter reaction times and poor response inhibition were associated with zoning out, yet not with tuning out. The authors claimed that being meta-conscious of mind wandering was the key distinction between performance levels, rather than mind wandering itself. In an attempt to show the importance of meta-consciousness, Smallwood, et al. (2007) may have misrepresented the data due to their definition of zoning out and tuning out. The distinction between zoning out and tuning out was the awareness that accompanied tuning out. However, the definition of tuning out is a bit inconsistent how can one continue a task while one is aware that their thoughts have drifted to something else? There are two potential renditions for tune outs: either (a) the task is cognitively non-demanding where the participant can allow their mind to wander while still continuing the task, thus undermining the mind wandering concept; or (b) the participant is not really thinking of something unrelated, but rather is thinking of something that is related to the task on hand. For example, in a memory paradigm, participants may try to link the items to a central theme or schema, to allow for deeper processing for subsequent retrieval, thus not really thinking of unrelated ideas. Although the directionality of causation was not revealed in the study, it did introduce performance measures as a potential method to tap into the mind wandering episodes Tasks Associated with Mind Wandering Many mind wandering studies have utilized less demanding tasks wherein the occurrence of mind wandering episodes does not necessarily impact performance to a large extent. Nonetheless

14 EYE MOVEMENTS DURING MIND WANDERING EPISODES 9 these studies have provided researchers the opportunity to explore the nature and cause of mind wandering in a variety of domains. For example, signal detection studies have shown that mind wandering episodes are more likely to occur when participants are stressed, highly aroused, or when experimenter and participant are of the opposite sex (Antrobus, Coleman & Singer, 1967; Singer, 1988). Go/no-go tasks have shown that low target probability increases mind wandering episodes as it increases in the interval between task-critical events which makes the task less cognitively straining (Giambra & Grodsky, 1989; Smallwood, McSpadden & Schooler, 2007). Memory tasks have shown that encoding is rather superficial during mind wandering episodes, and that there is a positive correlation between the reports of task-unrelated-thoughts and false alarm rates during recollection and recognition tests (Smallwood, Baracaia, Lowe & Obonsawin, 2003). These findings allow us to get a better picture of the phenomenon of mind wandering in various domains. However, the limitation of the aforementioned studies is that they require relatively shallow processing, and thus may be artificial in their production of mind wandering episodes. When questioned, most people describe mind wandering episodes as most frequently occurring as they are listening to or reading text. Arguably one of the most complex cognitive activities humans regularly partake in, reading requires individuals to engage in deep and parallel processing in which words are identified, encoded and parsed to create a meaningful propositional statement (Smallwood, McSpadden & Smallwood, 2007; Riechle, Warren & McConnel, 2009). In the domain of reading, two main methodologies have been used to capture mind wandering episodes: error detection (detection of errors embedded within the text) and comprehension tests (Schooler, 2002). Studies utilizing error detection indicate that good readers are better than poor readers in detecting inconsistencies within the text, although even good readers show a surprisingly high capacity to miss textual inconsistencies, and reading skill was determined by a questionnaire prior to experiment (Glenberg, Wilkison & Epstien, 1982; Smallwood, McSpadden, & Smallwood, 2008). Studies with comprehension monitoring also indicate higher performance for good readers than poor readers. The task of silently reading provides a good conceptual framework to observe mind wandering episodes.

15 EYE MOVEMENTS DURING MIND WANDERING EPISODES 10 Using a self-caught mind wandering paradigm, Giambra and Grodsky (1989) conducted one of the first studies where participants reported mind wandering episodes while reading passages. Participants were trained in recognizing two types of mind wandering episodes: intentional mind wandering defined as deliberately thinking about something unrelated to the text and unintentional mind wandering defined as unintentionally thinking about something unrelated to the text. Passages varied with respect to participant s interest and level of difficulty. Consistent with prior research, good readers regularly reported mind wandering episodes, although less than poor readers. In addition, level of interest with the text was a determining factor in the frequency of mind wandering episodes, where low interest resulted in higher reports; whereas frequency of mind wandering reports did not differ depending on the level of text difficulty. Grodsky and Giambra (1990) found a positive correlation between mind wandering episodes during reading and signal detection studies, demonstrating a level of consistency among distinct contexts. However, the factors that determined mind wandering frequency varied across experimental situation: easy blocks rendered higher rates of mind wandering episodes compared to difficult blocks in signal detection, while conceptually easy text did not increase mind wandering episodes in reading tasks (Grodsky & Giambra, 1990). The studies described in the previous paragraph show that reading is susceptible to mind wandering episodes similar to signal detection tasks, yet has different factors associated with its frequency. Although these experiments provided a gateway into the domain of reading for studying mind wandering episodes, strong conclusions cannot be drawn due to two key limitations. Firstly, the relationship between frequency of mind wandering episodes and reading performance was not explored. Results from signal detection studies suggest that mind wandering episodes would negatively impact performance, however this can t be said conclusively nor could severity of impact be predicted. Secondly, both the studies focused on the participant s intend in engaging in mind wandering rather than their awareness of the mind wandering episode. Although potentially related, there is a distinct difference between the intention to do something and the awareness that one is doing something. For example, you don t intend to walk into a door, but are definitely aware of doing so when you hit your head. It is this awareness, or lack thereof, that is the critical difference in being meta-conscious of one s experience.

16 EYE MOVEMENTS DURING MIND WANDERING EPISODES 11 Schooler, Reichle and Halphern (2004) conducted a study in which participants read a rather dull text and reported mind wandering episodes via self-classified probe-caught techniques, selfcaught techniques, or both. Mind wandering episodes were labelled as zoning out, defined as situations when participants had no idea what they had just read, thus lacking awareness. After reporting a mind wandering episode, participants in the self-caught group were asked Were you aware that you were zoned out while you were zoning out? Lastly, a forced choice comprehension test and recognition test were given to determine the relationship between mind wandering and performance. Results from the self-caught group indicated that 67% of the participants who reported a mind wandering episode were not aware that they had zoned out while they were zoning out. In addition, the probe was successful in catching mind wandering episodes in participants before they realized it themselves. As briefly mentioned previously, 13% of participants were zoned out without being aware of it. Although performance on the comprehension test was not significantly related to the number of mind wandering episodes that occurred, performance on the recognition test was positively correlated with the number of mind wandering episodes within the probe-caught group. This study demonstrates that participants, at least initially, lack awareness of mind wandering episodes, and that mind wandering episodes are associated with poorer overall comprehension of the text. Reading for comprehension is a semantically meaningful task that involves deep processing, yet has been shown to be positively correlated with signal detection tasks in the number of mind wandering episodes experienced. Reading tasks can be manipulated in terms of interest, conceptual and textual difficulty, and is a task that is ecologically valid. We read on a daily basis, so using reading tasks in experiments creates a truly ecologically valid task. This study will use a reading task coupled with a self-classified probe caught mind wandering paradigm to obtain a subjective account of mind wandering episodes. We propose to use eye movements as an objective measure of mind wandering episodes while participants are partaking in the reading task. Eye movement literature, and its links to reading, will briefly be reviewed.

17 EYE MOVEMENTS DURING MIND WANDERING EPISODES Eye Movements When a sixteenth century French anatomist and medical scientist, Du Laurens (1596), referred to eyes as windows of the mind, his view reflected both an ancient and modern observation (Van Gompel, Fischer, Murray & Hill, 2007). Our perception of the world is heavily influenced by our sense of sight, which cannot be consciously controlled at the lower level of oculomotor behaviour. This lack of conscious control provides a great biological and reflexive marker through which behaviour can be objectively measured. Thus, eye movements can reveal much about the workings of the brain and mind. Unlike the smooth trajectory we feel when scanning the horizon, eyes movements are actually made up of short bursts of stationary visual display, fixations, interspersed with rapid, ballistic eye movements, saccades. During fixations, the eye remains relatively motionless and visual information processing is achieved, while saccades serve the function of transporting the eye from one fixation to the next, with no information extracted in its duration. Although people do have the ability to control the general location of where they are looking, the ability to consciously control or manipulate fixations and saccades is not physically possible (Rayner, 1998). In recent years, the study of eye movements for investigating aspects of the mind and brain has exploded across a wide variety of topic areas Reading Perhaps one of the most complex cognitive activity humans regularly engage in, reading is a demanding task that requires coordination of many different stages of information processing, including attention, visual processing, word identification, semantics, and integrative ability to create a meaningful prose. Reading provides a domain in which basic mechanisms of visual processing can be studied in a highly controlled visual environment during a meaningful and ecologically valid task. Technological advancements during the last few decades have made it possible to precisely measure eye movements during a given task. The logic of capturing and studying eye movements is predicated on the simple assumption that the location of what the reader is looking at reflects, to some degree, the processes that are ongoing in his or her mind at

18 EYE MOVEMENTS DURING MIND WANDERING EPISODES 13 that time. Silent reading creates a distinct pattern when compared to reading out loud, or silently reading while listening to a voice reading the same text. These eye tracking methodologies have provided the means for researchers to inquire into the cognitive processes that lead to and occur during reading. Eye tracking has allowed us to gain accurate measurements of the different variables of eye movements during silent reading. Fixations last approximately ms and the mean saccade size is about 7-9 letters (Rayner, 1998). Gaze-contingent moving window studies have demonstrated that readers extract information from a limited area of text, ranging from 3-4 letter spaces to the left of fixation and approximately letter spaces to the right of fixation (Rayner, Pollatsek, Drieghe, Slattery & Reichle, 2007). Although reading on the basis of only parafoveal or peripheral information is difficult to impossible, the limited region that readers extract information from is greater than the specific word being fixated upon (Rayner, Ashby, Pollatsek & Reichle, 2004). In addition, foveal processing of each word is not necessary as many words are not fixated upon at all. Specifically, content words are fixated upon about 85% of the time they occur while function words are fixated upon only about 35% of the times they occur. This is assumed to reflect the facts that function words tend to be shorter and their relationship to semantics of the sentence straightforward (Rayner, 1998; Rayner, Pollatsek, Drieghe, et al., 2007). There is a strong relationship between the probability of fixating on a word and its respective length as the length increases, the probability of word fixation increases as well, where 2-3 letter words are fixated upon about 25% of the time, words 8 letters or longer are almost always fixated upon, and often more than once (Rayner, 1998). In addition, lower frequency words are fixated on a higher rate than high frequency words (Pollatsek, Juhaz, Reichle & Rayner, 2008; Rayner, Ashby, et al., 2004). Readers also tend to make backward saccades called regressions, right-to-left movements to previously read lines, that tend to occur about 10-15% of the time and most readers are surprisingly unaware of them (Rayner, 1998; Reichle, Pollatsek, Fisher, et al., 1998). Many are small and go only a few letters back, usually moving to an earlier position within the same word, although some regressions are longer, going back several words or even an earlier sentence. They also occur when the reader makes too long of a saccade and thus needs to make a short regression back for reading efficiently. Although the occurrence of regressions at low levels is normal, an exceptionally high frequency of within-

19 EYE MOVEMENTS DURING MIND WANDERING EPISODES 14 word regressions is indicative of problems processing the current fixated word or poor reading ability (Reichle, Pollatsek, Fisher, et al., 1998). These are few of the basic phenomena that occur while silently reading that have been discovered due to eye tracking technologies. The development of computational models of eye movements has seen great growth and attention as eye tracking methodologies advance. Eye movement models allow researchers to observe and manipulate the variables that create the distinct eye movement patterns that occur while reading. Two broad areas have surfaced regarding the factors that influence eye movement. Oculomotor models hypothesize that eye movement are controlled mainly by low level oculomotor and visuomotor processes, while cognitive models are based on the assumption that ongoing cognitive processes, such as lexical processes, steer eye movements during reading (Nuthmann & Engbert, 2009). Regardless of the type of model, continuous reading involves two concurrent streams of processes: the targeting and timing of saccades and fixations to provide adequate spatio-temporal conditions for extracting information, and the processing of written language where the acquisition of orthographically coded linguistic information feeds into the construction of a cognitive text representation (Radach, Reilly & Inhoff, 2007). Thus, every model reflects the combination of these two processes, where a higher emphasis is given to either the oculomotor or cognitive aspects that lead to the ability to read. Currently there are two comprehensive computational models of visuomotor and cognitive control in reading: E-Z reader and SWIFT model. The major distinction between these models lies in assumptions concerning the allocation of visual attention. E-Z reader assumes sequential attention shifts (SAS) where the main mechanism guiding eye movements is the serial allocation of visual attention from one word to the next (Nuthmann & Engbert, 2009; Reichle, Pollatsek, Fisher, et al., 1998). Thus in E-Z reader, word processing occurs in a strictly serial fashion, while saccadic programming can occur in parallel (Nuthmann & Engbert, 2009; Reichle, Pollatsek, Fisher, et al., 1998). Alternatively, the SWIFT model is based on the guidance by attentional gradient (GAG) premise, where attention is distributed continuously as a gradient so several words are processed in parallel (Engbert, Longtin & Kliegl, 2002). Both models have been able

20 EYE MOVEMENTS DURING MIND WANDERING EPISODES 15 to account for and predict basic and complex empirical phenomena in eye movements while silently reading. Even though the precise factors that influence eye movements while reading have not been completely understood, the ability to compare eye movement behaviour between and within tasks can provide a reliable and valid method to dissociate phenomena. By comparing eye movement patterns, Rayner and Fisher (1996) were able to show that lexical processes (e.g., reading normal sentences versus scanning z-strings in random or block format) created a cognitive demand that was reflective in the eye movement behaviour, in terms of word skipping, fixation duration and refixation (within-word regressions). Likewise, by analyzing and comparing eye movement behaviours while participants read for comprehension versus reading during mind wandering episodes, eye movements may provide a valid and objective experimental method to distinguish the two conscious states. Although Schooler, Reichle and Halphern (2004) empirically demonstrated the existence of mind wandering episodes during reading tasks, their evidence was based on self report, and thus may not be entirely free of bias. With the ability to monitor and track eye movements during reading, one can truly differentiate between mind wandering episodes during reading and comprehensive reading episodes without solely relying on subjective verbal reports from the participants. 1.3 Present Study The aim of the paper is to investigate whether eye movement behaviour differs when one is reading for comprehension versus when one is experiencing a mind wandering episode. This study will use a reading task coupled with a self-classified probe-caught mind wandering paradigm to obtain a subjective account of mind wandering episodes. I propose to use eye movements as an objective measure of mind wandering episodes while participants are reading. By using both an objective and a subjective method of measurement, I can determine if an association is present between either, or both of the measures with mind wandering episodes. I

21 EYE MOVEMENTS DURING MIND WANDERING EPISODES 16 expect eye movement behaviour 5-seconds preceding the probe-caught mind wandering episodes to be different than eye movement behaviour 5-seconds preceding probe-caught comprehension reading. I chose a 5-second time period to be consistent with previous research (Smilek, Carriere, & Cheyne, 2010). It is possible that the cognitive process that normally guides eye movements during reading exerts less control during mind wandering episodes. I believe that the cognitive factors do have some influence on oculomotor behaviour and, thus changes in one s cognitive orientation towards the primary task (i.e., reading for meaning versus wandering) will be reflected in eye movements. Essentially, I think that when one s mind is wandering, the eye movements will primarily reflect very low-level text scanning processes, rather than some combination of these processes and cognitive processes. Hence eye movements should become simpler during mind wandering instances. The experimental paradigm is based on the study by Schooler, Reichle and Halphern (2004) with the added feature of eye tracking. Prior to the start of the experiment, participants will be explicitly informed regarding the definition of mind wandering episodes, and that they will be probed intermittently regarding their mind state. Participants will be reading the first four chapters of War and Peace by Tolstoy on a computer screen while their eye movements are being tracked and recorded. Specifically, I will be analyzing nine variables that I think will be indicative of mind wandering episodes (see Appendix 1 for list and definition of the eye movement variables). Some are basic eye movement characteristics, such as average fixation duration, fixation count, saccade count and run count, which will be measured to observe any differences during mind wandering episodes, but also to correlate with other eye tracker results for reliability purposes. Additionally, other variables will be collected that I believe may be linked to mind wandering and inattentive states. For example, blink count will also be analyzed as higher number of blinks reduce the processing of external stimuli by decreasing visual input opportunities, and thus may be indicative of decreasing visual processing (Rayner, 1998; Smilek, Carriere, & Cheyne, 2010). Higher frequencies of regressions have been shown to be linked to lexical processing difficulties and lack of attention to the text, thus frequency and duration of within-word regressions will be recorded and analyzed (Reichle, Pollatsek, Fisher, et al., 1998). In addition, average fixation pupil size has been shown to be reflective of cognitive processing, where larger diameter is indicative of higher cognitive effort (Kahneman, 1997). I will analyze

22 EYE MOVEMENTS DURING MIND WANDERING EPISODES 17 any differences between these variables within the 5-second window preceding probes of mind wandering episodes versus probes of comprehensive reading episodes. 2 Methods 2.1 Participants A total of 33 undergraduate students from the University of Toronto Scarborough participated in this study on a volunteer basis for course credit. All participants gave informed consent. In accordance with ethical guidelines, participants were explicitly informed about the workings of the experiment, the apparatus used (Eye Tracker) and the aim of the study. Lastly, the participants were informed that they could leave at any time during the experiment without any repercussions if they felt uncomfortable. 2.2 Apparatus A Pentium III Dell computer with a 26 SVGA color monitor was used to carry out the experiment. The novel excerpt was displayed on this monitor, with the monitor rotated to the portrait position such that it resembled a large sheet of paper. Eye movements were recorded by an EyeLink 1000 tracker (SR Research Ltd.) with a 250 Hz temporal resolution and a 0.2 spatial resolution. This system uses infrared video based tracking technology to measure the pupil center and the size of the pupil of both eyes. The eye tracker was connected to another host computer which was responsible for recording the output of eye tracker. The spatial location of the right pupil while viewing the excerpts could be viewed through this computer.

23 EYE MOVEMENTS DURING MIND WANDERING EPISODES Procedure The experiment was conducted in a laboratory setting. Prior to the start of the experiment, participants were explicitly informed that they would be reading an excerpt from War and Peace by Tolstoy, and would be probed periodically regarding their reading comprehension of the text. Participants were told the aim of the experiment was to study the pattern of eye-movements during comprehension and mindless reading. Given that the low-level details of eye movements cannot be consciously altered or controlled, there was no threat in revealing the aim of the study. Mindless reading was explicitly defined as reading without text comprehension, or thinking about anything other than the text on hand, and several examples were provided to ensure participants understood the concept. Participants were instructed to read the text attentively and at a normal pace. Participants were then lead into the eye-tracker room, where the seat and head mount was adjusted to their height. The eye-tracker was then calibrated to the participant s pupils, and a series of validation tests were conducted to make sure the eye-tracker was correctly synced with the participant s pupils. Only participants with validation results of Good, Very Good or Excellent were allowed to continue to the experimental stage. Participants who did not receive adequate validation results were re-calibrated until adequate results were achieved or did not continue the experiment (these participants did received their course credit). Once validation results were satisfactory, the computer program reiterated the aim and instructions of the study, and participants were given the opportunity to ask any further questions while the experimenter was still in the room before the start of the experiment (see Appendix 2 for instructions). The materials used in the experiment consisted of 16 pages of text, with an average of 300 words per page. There was no time limit, and participants moved between the pages by looking at a big arrow bottom right hand corner, which triggered the program to display the next page. Randomly every 2-3 minutes, a probe would appear on top of the text asking ARE YOU ZONED OUT AT

24 EYE MOVEMENTS DURING MIND WANDERING EPISODES 19 THIS PARTICULAR MOMENT? Y/N? Participants would have to answer to continue the experiment. Following the completion of all the pages, participants underwent the validation tests once more to ensure proper recording, and then briefly interviewed (to ensure no problems were encountered) and subsequently debriefed. 3 Results As previously stated, participants were placed in a self-classified probe-caught mind wandering paradigm while their eye movements were recorded as they read through 16 pages of text. They were randomly probed every 2-3 minutes and were required to answer to continue reading. There were two sets of data collected in this experiment: the subjective mind wandering reports and the objective eye movement behaviours. Based on the subjective mind wandering reports of the participants, the objective eye movement behaviours were categorized into mind wandering or reading conditions, thus making the eye movement behaviours the dependant variable. Specifically, there were nine eye movement variables collected and analyzed (see Appendix 1 for list and definition of each variable). Data from 11 participants were excluded from the analysis due to any of the following situations arising: if calibration and validation of the eye tracker was unsatisfactory; if it was discovered that the participant had not read all the pages completely, or if the probe occurred within the first 3 seconds of a new page of text, as a new page would refresh attention. These precautions were implemented to ensure the quality, validity and reliability of the data was upheld. As participants were randomly probed every 2-3 minutes, the number of probes administered relied upon the duration of the experiment and the speed at which participants read. On average participants received 10 probes in total, of which mind wandering was reported on 49% of them. Consistent with previous research, these results show that mind wandering was observable during a controlled laboratory reading experiment (Giambra & Grodsky, 1989; Schooler, Reichle & Halphern, 2004). However as previously mentioned, these episodes of mind wandering were

25 EYE MOVEMENTS DURING MIND WANDERING EPISODES 20 subjective verbal reports that cannot be taken on face value. Eye movement behaviour was proposed as an objective measure, which would be void of bias and conscious control. Withinsubjects analysis was conducted for the eye movement data, which was categorized based on the answers to the probes as either reading or wandering. Critically, analysis was conducted for the 5 seconds time interval preceding the probe for reading and wandering conditions within each participant. I conducted paired t tests, which calculated the difference between the eye movement variables across reports of mind wandering and reports of reading, and analyzed the list of differences based on the assumption that the differences in the entire population followed a normal distribution. Nine pairs of eye movement variables were analyzed, which displayed a spectrum of sensitivity to mind wandering. The overall pattern of the data suggested that eye movement behaviour was less complex (i.e., slower and/or less frequent) when participants reported mind wandering episodes. The first three eye movement variables were sensitive to the distinction between mind wandering and reading episodes. Visited interest-area, pertaining to the frequency of fixations within an interest area, was lower during the 5 second time period prior to the probe of mind wandering episodes count (t (21) = 1.874, p = 0.075). In addition, duration of within-word regressions during mind wandering episodes (t (21) = , p = 0.061) was comparatively less during reports of reading, as were the frequency of within-word regressions (t (21) = 1.929, p = 0.067). The next set of variables demonstrated a weaker sensitivity to the overall trend in spite of being statistically nonsignificant as well. Number of fixations, saccades and run were all lower during the 5 second period when the mind wandering was reported compared to reading, with the paired t value being no greater than and p-value no less than Fixation duration, blink count and average fixation pupil size were all consistent with this overall pattern, despite being the least sensitive measures, where the paired t value was no greater than and p-value no less than Thus, all nine variables suggested that eye movement behaviours were stagnant and less frequent during periods of mind wandering compared to periods of reading reports. The paired t tests analysis provided an escalating scale of sensitivity of the eye movement variables that alluded to the distinction between mind wandering and reading conditions, where