Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Executive function use during exercise predicts performance on laboratory measures of executive functioning Kozik, Pavel 2021

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Notice for Google Chrome users:
If you are having trouble viewing or searching the PDF with Google Chrome, please download it here instead.

Item Metadata

Download

Media
24-ubc_2021_may_kozik_pavel.pdf [ 2.64MB ]
Metadata
JSON: 24-1.0395601.json
JSON-LD: 24-1.0395601-ld.json
RDF/XML (Pretty): 24-1.0395601-rdf.xml
RDF/JSON: 24-1.0395601-rdf.json
Turtle: 24-1.0395601-turtle.txt
N-Triples: 24-1.0395601-rdf-ntriples.txt
Original Record: 24-1.0395601-source.json
Full Text
24-1.0395601-fulltext.txt
Citation
24-1.0395601.ris

Full Text

EXECUTIVE FUNCTION USE DURING EXERCISE PREDICTS PERFORMANCE ON LABORATORY MEASURES OF EXECUTIVE FUNCTIONING  by Pavel Kozik   B.A. (Hons), University of British Columbia, 2013 M.A., University of British Columbia, 2015   A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF  DOCTOR OF PHILOSOPHY in THE FACULTY OF GRADUATE AND POSTDOCTORAL STUDIES (Psychology)  THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver)   January 2021   © Pavel Kozik, 2021 ii  The following individuals certify that they have read, and recommend to the Faculty of Graduate and Postdoctoral Studies for acceptance, the dissertation entitled: Executive function use during exercise predicts performance on laboratory measures of executive functioning  submitted by Pavel Kozik in partial fulfillment of the requirements for the degree of Doctor of Philosophy in Psychology  Examining Committee: Dr. James Enns, Department of Psychology, UBC Supervisor  Dr. Todd Handy, Department of Psychology, UBC Supervisory Committee Member  Dr. Nicola Hodges, Department of Kinesiology, UBC University Examiner Dr. Ruth Grunau, Department of Pediatrics, UBC University Examiner  Additional Supervisory Committee Members: Dr. Victoria Savalei, Department of Psychology, UBC Supervisory Committee Member        iii  Abstract This dissertation explores the hypothesis that cognitive engagement is an important predictor of the relationship between exercise and executive functioning. Chapter 1 introduces the background claim that exercise benefits executive functioning. This includes reviewing the relationship between exercise and improvements in executive functioning via changes in cerebral blood flow and neuroplasticity. The exercise-executive function relationship is also reviewed via literature on exercise history, duration, intensity, and type. This review concludes by introducing the primary hypothesis of this dissertation, namely, that cognitively-engaging exercise should predict better executive functioning. Chapter 2 tested this hypothesis through an empirical study (N = 145) of undergraduates who self-reported their executive function use during exercise, and then completed executive function tasks (i.e., flanker and backward span). Students reporting engagement in exercise that relied on inhibitory control were found to perform better on a flanker task, and students reporting engagement in exercise that relied on cognitive flexibility performed better on a backward span task. Chapter 3 recruited an independent sample of undergraduates (N = 228) and had them complete different executive function tasks (i.e., stop-signal and trail making B). The main finding was that when students reported engaging in exercise that relied on inhibitory control they had faster stop-signal reaction time and made fewer trail making errors, and when they reported engaging in exercise that relied on cognitive flexibility they had slower stop-signal reaction time and trail making completion time. Chapter 4 recruited a more diverse sample of participants (e.g., older, more males; N = 225) and had them complete the same executive function tasks as chapter 2. The main finding was that correlations now ran in opposite directions. When individuals engaged in exercise that relied on inhibitory control, they performed worse on a flanker task, and when they engaged in exercise that relied on  iv  cognitive flexibility, they performed worse on a backward span task. Chapter 5 summarizes these findings and speculates that cognitively-engaging exercise may predict better or worse executive functioning depending on the underlying motivation and context driving one to exercise, as well as discussing the potential role of leisure activity.                 v  Lay Summary Exercise improves cognition, but this relationship is complex and depends on many factors, including exercise history, duration, intensity, type, cerebral blood flow, and a family of proteins called neurotrophins. This thesis tests the hypothesis that an additional important factor is the extent to which one’s exercise requires cognitive engagement, i.e., inhibitory control over actions, and, flexible thinking. Research in chapters 2 and 3 found that among undergraduate students, exercise self-reported to rely on these cognitive abilities, generally predicted better performance on laboratory measures of those abilities. Chapter 4 recruited a more diverse sample of participants and found the opposite relation; exercise self-reported to rely on various cognitive abilities predicted worse cognitive performance on laboratory measures. These findings suggest that cognitively-engaging exercise may have a positive or negative relationship with cognitive functioning, likely depending on the individual’s motivation and the circumstances surrounding one’s exercise experience.               vi  Preface I prepared the content of this dissertation. Research was completed within the Vision lab at the University of British Columbia where James Enns is the Principal Investigator. Chapters 2 through 5 are based on research that was designed, conducted, and written by me, following approval by the University of British Columbia’s Research Ethics Board: H18-03515.                   vii  Table of Contents Abstract ......................................................................................................................................... iii Lay Summary .................................................................................................................................v Preface ........................................................................................................................................... vi Table of Contents ........................................................................................................................ vii List of Tables ..................................................................................................................................x List of Figures ............................................................................................................................... xi Acknowledgements .................................................................................................................... xiii Chapter 1: Introduction ................................................................................................................1 1.1 Exercise & Cerebral Blood Flow ............................................................................................3 1.2 Exercise & Neurotrophins .......................................................................................................4 1.3 Exercise, Structural Changes in the Brain, & Executive Functioning................................6 1.4 Exercise Qualifiers: History, Intensity, Duration, & Type ..................................................8 1.5 Cognitively-Engaging Exercise .............................................................................................12 1.6 Measuring Cognitive Engagement During Exercise ...........................................................15 1.6.1 Measuring Use of Inhibitory Control During Exercise .............................................. 16 1.6.2 Measuring Use of Cognitive Flexibility During Exercise ........................................... 18 1.7 Cognitively Engaging Exercise & Executive Functioning: A Potential Mechanism .......20 1.8 Overview of Dissertation Studies ..........................................................................................22 Chapter 2: Study 1 .......................................................................................................................25 2.1 Introduction ............................................................................................................................25 2.2 Method ....................................................................................................................................26 2.2.1 Power Analysis ............................................................................................................... 26  viii  2.2.2 Participants ..................................................................................................................... 27 2.2.3 Procedure ........................................................................................................................ 28 2.2.4 Data Analytic Approach ................................................................................................ 34 2.3 Results .....................................................................................................................................41 2.3.1 Exercise Results .............................................................................................................. 41 2.3.1.1 Measurement Item Statistics .................................................................................. 41 2.3.1.2 Confirmatory Factor Analysis & Measurement Model Identification .............. 41 2.3.1.3 Path Analysis & Structural Equation Modelling ................................................. 46 2.3.2 Leisure Results ............................................................................................................... 48 2.3.2.1 Measurement Item Statistics .................................................................................. 48 2.3.2.2 Confirmatory Factor Analysis & Measurement Model Identification .............. 49 2.3.2.3 Path Analysis & Structural Equation Modelling ................................................. 50 2.4 Discussion................................................................................................................................51 Chapter 3: Study 2 .......................................................................................................................70 3.1 Introduction ............................................................................................................................70 3.2 Method ....................................................................................................................................73 3.2.1 Power Analysis ............................................................................................................... 73 3.2.2 Participants ..................................................................................................................... 75 3.2.3 Procedure ........................................................................................................................ 76 3.3 Results .....................................................................................................................................83 3.3.1 Exercise Results .............................................................................................................. 83 3.3.2 Leisure Results ............................................................................................................... 85 3.4 Discussion................................................................................................................................86  ix  Chapter 4: Study 3 .......................................................................................................................98 4.1 Introduction ............................................................................................................................98 4.2 Method ..................................................................................................................................101 4.2.1 Participants ................................................................................................................... 101 4.2.2 Procedure ...................................................................................................................... 103 4.3 Results ...................................................................................................................................105 4.3.1 Exercise Results ............................................................................................................ 105 4.3.2 Leisure Results ............................................................................................................. 106 4.4 Discussion..............................................................................................................................108 Chapter 5: General Discussion .................................................................................................121 5.1 Summary of Key Findings ..................................................................................................121 5.2 Implications of a Positive Exercise & Executive Function Relationship ........................122 5.3 Implications of a Negative Exercise & Executive Function Relationship .......................124 5.4 Implications for Leisure & Executive Functioning...........................................................127 5.5 Implications for Measuring Exercise & Leisure Executive Function Use ......................130 References ...................................................................................................................................133 Appendix: Executive Function Use During Exercise – Measurement Items .......................164          x  List of Tables Table 2.1 Mean and standard deviation values for all exercise measurement items ............ 58 Table 2.2 Confirmatory factor analysis of the 30 exercise measurement items .................... 59 Table 2.3 Confirmatory factor analysis of positively worded exercise items ........................ 60 Table 2.4 Confirmatory factor analysis of full-two exercise factor model ............................ 60 Table 2.5 Confirmatory factor analysis for the final set of exercise items ............................ 61 Table 2.6 Mean and standard deviation values for all leisure measurement items .............. 62 Table 2.7 Confirmatory factor analysis of the 30 leisure measurement items ...................... 63 Table 2.8 Confirmatory factor analysis of positively worded leisure items .......................... 64 Table 2.9 Confirmatory factor analysis of positively worded leisure items .......................... 64 Table 2.11 Confirmatory factor analysis for the final set of leisure items ............................. 65 Table 3.1 Mean and standard deviation values for all exercise measurement items ............ 91 Table 3.2 Mean and standard deviation values for all leisure measurement items .............. 91 Table 4.1 Mean and standard deviation values for all exercise measurement items .......... 116 Table 4.2 Mean and standard deviation values for all leisure measurement items ............ 116            xi  List of Figures Figure 2.1 Frequencies of self-reported primary exercise. ..................................................... 66 Figure 2.2 Frequencies of self-reported primary leisure ......................................................... 66 Figure 2.3 Performance on the flanker task and backward span .......................................... 67 Figure 2.4 CFA of the two-factor exercise model..................................................................... 68 Figure 2.5 Exercise path analysis model ................................................................................... 68 Figure 2.6 Exericse SEM ............................................................................................................ 68 Figure 2.7 CFA of the two-factor leisure model ....................................................................... 69 Figure 2.8 Leisure path analysis model..................................................................................... 69 Figure 2.9 Leisure SEM .............................................................................................................. 69 Figure 3.1 Frequencies of self-reported primary exercise ...................................................... 92 Figure 3.2 Frequencies of self-reported primary leisure ......................................................... 92 Figure 3.3 Present study stop-signal task ................................................................................. 93 Figure 3.4 Stop-signal reaction time illustration ...................................................................... 93 Figure 3.5 Stop-signal reaction time and non-cancelled error rates ...................................... 94 Figure 3.6 Trail making completion time and error rates ...................................................... 95 Figure 3.7 CFA of the two-factor exercise model..................................................................... 96 Figure 3.8 Exercise path analysis model ................................................................................... 96 Figure 3.9 Exercise SEM ............................................................................................................ 96 Figure 3.11 CFA of the two-factor leisure model ..................................................................... 97 Figure 3.12 Leisure path analysis model................................................................................... 97 Figure 3.13 Leisure SEM ............................................................................................................ 97 Figure 4.1 Frequencies of self-reported primary exercise .................................................... 117  xii  Figure 4.2 Frequencies of self-reported primary leisure ....................................................... 117 Figure 4.3 Performance on the flanker task and backward span ........................................ 118 Figure 4.4 CFA of the two-factor exercise model................................................................... 119 Figure 4.5 Exercise path analysis model ................................................................................. 119 Figure 4.6 Exercise SEM .......................................................................................................... 119 Figure 4.7 CFA of the two-factor leisure model ..................................................................... 120 Figure 4.8 Leisure path analysis model................................................................................... 120 Figure 4.9 Leisure SEM ............................................................................................................ 120                     xiii  Acknowledgements I would like to thank my research supervisor Dr. James Enns for his advice and guidance throughout my graduate studies, Dr. Todd Handy and Dr. Victoria Savalei for their contributions and insightful feedback, as well as the Natural Sciences and Engineering Research Council of Canada for funding my research. Lastly, I would like to thank my family and friends who have remained an endless source of support and inspiration.         1  Chapter 1: Introduction A fundamental goal of cognitive psychology is to understand the mental processes used by individuals to interpret, navigate, and experience the external world. Among the various mental processes that cognitive psychologists measure in the pursuit of this goal are a subset titled inhibitory control, cognitive flexibility, and working memory (Diamond, 2013; Miyake et al., 2000; Johann, Könen & Karbach, 2020; Soga et al., 2018). These three processes are collectively referred to as executive functions, and, these functions are critical for various real-world behaviours and activities, including planning, problem-solving, reasoning, and multitasking. The importance of these functions is readily apparent when considering sport performance, as an athlete in one circumstance might be ignoring the jeers of a crowd (inhibitory control), while attempting to identify winning strategies (cognitive flexibility), with consideration to the instructions given by their coach and how much time remains in that game period (working memory).  Inhibitory control refers to the processes by which individuals manage to make appropriate choices, stop inappropriate impulses, and otherwise coordinate their behavior when confronted with conflicting information. Inhibitory control allows one to ignore distracting sounds when completing a work-related task, stop oneself from acting on impulse when receiving unexpected news, and more generally allows for more cautious and measured behaviour when faced with multiple channels of information. Through this executive function one can favor a specific path of action relative to other competing paths. Cognitive flexibility refers to how efficiently one can mentally shift from one mode of thinking to another. Through the mental freedom afforded via cognitive flexibility, one can empathize with another individual’s situation, identify novel thoughts and patterns, and find new  2  solutions to otherwise seemingly irresolvable problems. Without this mental freedom, one might struggle to understand another individual’s viewpoint, and may adopt the same approach to different problems regardless of their context and unique circumstance. Cognitive flexibility allows one to escape a rigid mindset. Working memory indexes how effectively an individual can mentally store and manipulate multiple streams of information. A core aspect of working memory is remaining aware of what is happening in one’s environment while also meaningfully interpreting events as they unfold. Working memory as such is closely implicated in multitasking ability, as when multitasking an individual must manage and complete various activities simultaneously. An individual with higher working memory capacity might for instance find it easier to engage in a meaningful conversation while also completing a work-related task relative to an individual with lower working memory capacity, who instead might prefer to do these activities sequentially rather than concurrently. Researchers commonly study executive functions, not only because of their theoretical importance for understanding cognition, but because they are important for various everyday activities and behaviours. Executive functioning has been implicated in stress regulation (Williams, Suchy, & Rau, 2009), career advancement (Bailey, 2007; Diamond & Ling 2016), substance abuse (Miller et al., 2011; Peeters et al., 2015), academic achievement (Samuels, Tournaki, Blackman & Zilinski, 2016), weight control (Crescioni, et al., 2011; Gettens & Gorin, 2017), criminality (Hancock, Tapscott & Hoaken, 2010; Meijers, Harte, Meynen & Cuijpers, 2017), and reading ability (Johann, Könen & Karbach, 2020). Identifying mechanisms through which executive functioning can be improved and optimized is thus worth pursuing for theoretical insight, as well as for various meaningful real-world outcomes. This dissertation  3  advances the claim that one mechanism through which executive functioning can be improved, is through participation in exercise that is cognitively engaging. The following sections of this dissertation set the groundwork for this hypothesis, beginning first with research that suggests exercise has the potential to improve brain functioning through changes in cerebral blood flow, as well as through neurotrophin-elicited neuroplasticity. 1.1 Exercise & Cerebral Blood Flow   The brain requires a continuous supply of oxygen, and declines of oxygen availability associate with worse cognitive performance (Herold, Wiegel, Scholkmann & Müller, 2018). Oxygen, glucose, and other nutrients are provided to the brain through cerebral blood flow, that if stopped, would lead to irreversible brain damage within minutes (Querido & Sheel, 2007; Joris et al., 2018; Kisler et al., 2017). Exercise has been found to increase oxygenation in parts of the brain, like the prefrontal cortex, that are critical for executive functioning (Jung et al. 2015; Moriarty et al., 2019; Kujach et al., 2018). In one study Yanagisawa and colleagues (2010) had participants complete a Stroop task before and after either an exercise or resting condition. The researchers found that an acute bout of exercise improved inhibitory control (i.e., less Stroop task interference), and that this increase in performance coincided with greater cerebral blood flow within the left dorsolateral prefrontal cortex (DLPFC), an area of the brain that has been implicated for efficient inhibitory control (MacDonald et al., 2000). The relationship between exercise and greater oxygenation within the prefrontal cortex has led some researchers to suggest that this is the mechanism through which exercise improves executive functioning (Kujach et al., 2018). Within this research, Hyodo et al. (2016) measured the cardiorespiratory fitness of older men and recorded their cerebral blood flow while they completed a Stroop task. They found that greater inhibitory control (less Stroop task  4  interference) correlated with greater fitness and greater left-lateralized DLPFC activation (left DLPFC activation minus right DLPFC activation), and that greater left-lateralized DLPFC activation correlated with greater aerobic fitness. A follow-up mediation model revealed that left-lateralized DLPFC activation partially accounted for the relationship between fitness and better inhibitory control, thereby providing support for the hypothesis that exercise improves executive functioning through changes in cerebral hemodynamics. A study completed by Byun et al. (2014), found similar results to Hyodo et al. (2016), and suggested that this underlying mechanism is associated with changes in experienced psychological arousal. Other researchers have suggested that exercise, via noradrenergic and dopaminergic pathways, acts as an arousing stressor that modulates attentional resources (Li, O’Connor, O’Dwyer & Orr, 2017).   In studies of the relationship between exercise and cerebral blood flow, researchers typically study participants who are already exercise adherents, or they introduce an exercise intervention. Alfini et al. (2016) reversed this pattern, asking older adults who exercised regularly to cease this activity for a 10-day period. Resting cerebral blood flow decreased over this period in several brain sites, most notably in bilateral regions of the hippocampus. This finding suggests that the exercise-related changes in executive functioning that occur via cerebral hemodynamics may also be reversible. 1.2 Exercise & Neurotrophins  Neurotrophins are a family of proteins, within which brain-derived neurotrophic factor (BDNF) is perhaps the most studied. BDNF is widely expressed in the central nervous system, and is implicated in the maintenance, regulation, survival, and formation of new neurons (Thoenen, 1995; Gorski et al., 2003; Bathina & Das, 2015). Given these properties, BDNF is central for both neurogenesis and neuroplasticity, through which neuronal remodelling and  5  structural changes of the brain are made possible (Szuhany, Bugatti & Otto, 2015; Calabrese et al., 2014; Castrén & Antila, 2017). It is through these changes that BDNF is thought to elicit richer neural interconnectedness and heightened neural efficiency (Bherer et al., 2015; Erickson & Kramer, 2009; Heisz & Kovacevic, 2016).   Exercise increases BDNF concentration (Liu & Nusslock, 2018; Winter et al., 2007; Knaepen et al., 2010). A meta-analysis by Szuhany, Bugatti and Otto (2015) for instance found that a single episode of exercise increases BDNF, that this increase is greater among those who regularly exercise, and that regular exercise adherence results in a small though notable increase in resting levels of BDNF. This connection between exercise and increased BDNF has been hypothesized to explain how exercise improves cognition and executive functioning (Hötting & Röder, 2013; Erickson, Miller & Roecklein, 2012). An example of this hypothesis is that exercise elicits BDNF, BDNF stimulates hippocampal neurogenesis and long-term potentiation, which improves hippocampus efficiency, leading to improved performance on a variety of memory tasks (Cotman, Berchtold & Christie, 2007; Cotman & Engesser-Cesar, 2002; Erickson, Miller & Roecklein, 2012; Schmidt-Kassow et al., 2012; Liu & Nusslock, 2018). Two neurotrophic factors also found to increase via exercise are insulin-like growth factor (IGF-1) and vascular endothelial growth factor (VEGF; Raichlen & Polk, 2013; Gustafsson et al., 2001; Koziris et al., 1999; Fabel et al., 2003; Heisz et al., 2017). Among their functions, both IGF-1 and VEGF are important for blood vessel survival and growth, as well modulating neuroplasticity and neurogenesis (Jacobo & Kazlauskas, 2015.; Stillman et al, 2020). Within rats for instance, blocking IGF-1 from entering the brain has been found to inhibit exercise-elicited hippocampal neurogenesis (Trejo, Carro & Torres-Alemán, 2001). More generally IGF-1, VEGF, and BDNF are thought to act in concert with one another, as for  6  instance IGF-1 is thought to increase BDNF signalling following exercise (Cotman, Berchtold & Christie, 2007). Different forms of exercise might also differentially elicit changes in BDNF, VEGF and IGF-1 concentration. BDNF for instance has often been found to increase following aerobic exercise, whereas resistance training has been found to elicit elevated IGF-1 expression (Soga, Masaki, Gerber & Ludyga, 2018). Collectively, IGF-1, VEGF, and BDNF thereby all are implicated in structural changes within the brain. 1.3 Exercise, Structural Changes in the Brain, & Executive Functioning  The brain structurally changes in response to exercise (Erickson, Leckie & Weinstein, 2014; Weinstein et al., 2012; Erickson, Miller & Roecklein, 2012). Colcombe et al., (2006) assigned older adults to either an exercise group or a control group for 6 months, with the exercise group exhibiting increases of brain volume by the end of this time period. The largest of these increases was in the prefrontal and temporal cortices, which are critical for efficient executive functioning (Colcombe et al., 2006; Yuan & Raz, 2014; Funahashi & Andreau, 2013). In a related study, older adults who exercised for a one-year time period exhibited a 2% increase of hippocampal volume, whereas a control group over this same time period exhibited a 1.5% volume decrease (Erickson et al., 2011). Relatedly, the hippocampus is central for various types of memory, including working memory, thereby allowing exercise to modify a brain region implicated for this specific executive function (Axmacher et al., 2010; Burgess, Maguire & O'Keefe, 2002; Spellman et al., 2015).  Arguably some of the most convincing evidence that exercise yields better executive functioning through structural changes in the brain can be found in various mediation studies. In these studies, researchers commonly measure cardiorespiratory fitness (e.g., VO2max) and/or physical activity levels (e.g., accelerometry), and then test whether these variables predict greater  7  executive functioning through an indirect effect of brain volumetric mass. Erickson et al. (2009) found that greater cardiorespiratory fitness (VO2max) in older adults predicted both better spatial working memory and greater hippocampal volume. The relationship between greater fitness and greater spatial working memory was also shown to be mediated through greater hippocampal volume. A related study by Weinstein et al. (2012) reported that in older adults, cardiorespiratory fitness (VO2max) predicted lesser Stroop task interference (an index of inhibitory control), as well as increased volume of the dorsolateral prefrontal cortex. The relationship between greater fitness and greater inhibitory control was then found to be mediated by volume of the dorsolateral prefrontal cortex (and more narrowly, volume of the right inferior frontal gyrus and precentral gyrus).  Studies with young adults though seemingly less common too have shown similar results. Bento-Torres et al. (2019) measured cardiorespiratory fitness (VO2max) and daily physical activity levels (via an armband accelerometer) in young adults who went on to complete a Stroop task. The primary outcome of interest was intraindividual variability (IIV) in reaction time on incongruent trials of the Stroop task. This measure was taken as an index of executive control consistency. The researchers found that while fitness and activity levels did not predict reaction time IIV, both of these factors predicted greater thickness of the rostral anterior cingulate cortex, a brain region that is implicated in error monitoring (Polli et al., 2005; Taylor et al., 2006). The researchers then found that despite there not being a direct effect of fitness or activity level on IIV, both were predictive of IIV through an indirect effect of rostral anterior cingulate cortex thickness. Inhibitory control thus appeared to be more consistent among individuals who were more active, with this effect being mediated by the thickness of the rostral anterior cingulate cortex.  8  1.4 Exercise Qualifiers: History, Intensity, Duration, & Type    Exercise is important for executive functioning (Colcombe & Kramer, 2003; Hillman, Erickson & Kramer, 2008), but exercise also radically differs across a variety of factors. Among these factors, those that have been found to qualify the exercise-executive function relationship are exercise history, duration, intensity, and type (Soga et al., 2018; Chang et al., 2012; Roth et al., 2003; Colcombe & Kramer, 2003). This section serves to introduce and review these important exercise qualifiers. Numerous studies suggest that the length of time one has been engaged in exercise is positively associated with executive functioning. This may be because regular exercise increases BDNF, which contributes to neuroplasticity and enduring structural brain changes (Szuhany, Bugatti & Otto, 2015). A review of numerous studies by Etnier et al. (1997) reported that although acute bouts of exercise offer minor cognitive benefit, the benefits of long-term chronic exercise are more pronounced. More recent research suggests that longer-term exercise adherence is particularly beneficial for inhibitory control (Padilla, Pérez, Andres & Parmentier, 2013; Padilla, Pérez & Andrés, 2014). For instance, Pérez, Padilla., Parmentier and Andrés (2014) classified young adults based on whether they had been actively engaged in cardiovascular exercise over the past 10 years or not, and found that active participants had greater inhibitory control (assessed via flanker trials of an attention network test).  Review articles have also compared the effects of acute and chronic exercise on executive functioning. Hsieh et al., (2020) focused on the beneficial effects of acute and chronic high-intensity interval training (HIIT), and found that acute HIIT generally benefited inhibitory control among children, adolescents and adults. Other executive functions were more difficult to discern given inconsistent or limited findings, as for instance, among three studies looking at  9  cognitive flexibility in young adults, two reported a positive effect and one yielded no effect. Chronic HIIT also appeared to be particularly beneficial for children, with positive effects emerging for both inhibitory control and working memory outcomes. Among adults, chronic HIIT was again more difficult to parse, with limited research looking at cognitive flexibility and working memory outcomes. Soga et al. (2018) also recently completed a review comparing the effects of acute and chronic exercise with a focus on resistance training. The authors concluded that both acute and chronic resistance training benefit inhibitory control, with the support for other executive functions being mixed and less strong.  Longer individual bouts of exercise have sometimes been found to be more beneficial for cognition than shorter bouts. In their meta-analysis Colcombe and Kramer (2003) found that exercise lasting more than 30 minutes was more beneficial for cognitive functioning than exercise that was shorter. Some research also suggests that intermediate timespans are better than ones that are overly prolonged. In one such study Chang et al. (2015) had participants complete a Stroop task after a reading condition, or after a cycling condition that lasted 10 minutes, 20 minutes, or 45 minutes. Cycling for 20 minutes was found to yield the greatest performance benefits, as indicated by faster response times and higher accuracy for both congruent and incongruent Stroop trials. A study by Schmidt-Kassow et al. (2012) further found that BDNF concentration appears to peak after 20 minutes of exercise, thereby collectively reinforcing the idea that between 20 to 30 minutes of exercise may be optimal for a single session.  Exercise can be completed at various levels of intensity. Some exercise is relatively light (e.g., brisk walking), whereas other exercise is substantially more demanding or exhaustive (e.g., sprinting). Numerous studies suggest that higher intensity exercise may be more beneficial for executive functioning than lighter intensity exercise. A study by Kamijo et al. (2009) for instance  10  found that the response times of young adults and old adults on a flanker task were quicker following moderate exercise relative to lighter exercise (here assessed by cycling at 50% and 30% of VO2max respectively). More recently Gejl et al. (2018) had adolescents complete a flanker task after resting or completing a five-minute cycling exercise set to 50%, 65% or 80% of their maximal oxygen uptake reserve. Exercising at any intensity resulted in more accurate performance, whereas response times only quickened following the most intense exercise condition (80%). These results also align with research showing that BDNF concentration is greater following more intensive and demanding exercise (Schmidt-Kassow et al., 2012; Winter et al., 2007; Saucedo Marquez et al., 2015). However, some research also suggests that overly intense exercise may be deleterious for executive functioning, and that exercise intensity ought to be carefully matched given an individual’s baseline fitness (Labelle, Bosquet, Mekary & Bherer, 2013; Hüttermann & Memmert, 2014).  The environment in which exercise occurs can differ radically from one setting to the next (Highlen & Bennett, 1979; Poulton, 1957). Some exercise occurs within an environment that is relatively stable, consistent, and predictable. Such exercise is often called static (or closed-skill), and includes running, swimming, and cycling. Static exercise is often completed at a pace set by the individual performing that exercise rather than by external constraints imposed by the environment. Other types of exercise occur within an environment that changes and this exercise tends to be less predictable. Such exercise is referred as dynamic (or open-skill), and includes various team sports like soccer, hockey and basketball. Given these environmental conditions, static exercise tends to emphasize rehearsal and mastery of closed-loop motor skills, whereas dynamic exercise places greater emphasis on open-loop motor skills (Arvinen-Barrow et al., 2007; Coelho et al., 2007).   11  The term static should not be taken to imply that such exercise is immobile or without movement. Instead, “static” is descriptive of the predictable and routine pace of such exercise, as for instance a jogger is likely to continue pursuing a comparable pace throughout their routine, and shifts from this pace are likely to be determined by their own internal motives rather than external pressures. By contrast, dynamic exercise may involve similar physical motions but at a fluctuating pace in response to changing circumstances. During soccer, for example, the jogging pace and direction of an athlete will be dependent on other players and the evolving gameplay. Caution should also be used in treating static versus dynamic exercise as a dichotomy. Although this is sometimes methodologically convenient, it ignores the underlying continuum (Arvinen-Barrow et al., 2007; Coelho et al., 2007). An exercise may be predominately dynamic or static, but it may also involve elements of both. For instance, basketball is a typically dynamic activity, but within it the free throw is considered a static event.  In rock climbing, a climber may have past experience with a given route and so finds it predictable, whereas on another route that is completely uncharted, important decisions must be made in real time. A related quality that is correlated with static and dynamic exercise distinction is social interaction. Hockey, soccer, football, volleyball and basketball are all instances of dynamic sports involving interaction with other players on two teams, whereas static exercises like sprinting, swimming, and cycling are often done independently. This however does not mean that static exercise is not influenced by social factors. For instance, perceived competition from an athletic rival may quicken race times on the running track (Kilduff, 2014) as well as cycling circuit (Corbet et al., 2012), and both dynamic and static exercise are susceptible to the presence of observers as skilled tennis players perform better in front of audience (Dube & Tatz, 1991) and maximum bench-press performance increases in the presence of spectating competitors  12  (Rhea, Landers, Alvar & Arent, 2003). The social dimension of many dynamic sports may however be one of the factors that increases their demand on executive functioning over static sports. The distinction between dynamic and static exercise has been important for researchers studying executive functioning. Dynamic exercise for instance appears to prompt less rigid and more flexible distribution of attentional resources (Lum, Enns & Pratt 2002; Castiello & Umiltà, 1992). Wang et al. (2013) also found that tennis players (dynamic) had greater inhibitory control on a stop-signal task (faster action cancellation) relative to swimmers (static) and nonathletes. More recently Chiu, Chen and Muggleton (2017) conducted a study in which athletes from static sports (runners and swimmers) had comparable flanker task accuracy to nonathletes, whereas athletes from dynamic sports (volleyball) were more accurate than nonathletes. Both static and dynamic exercise have also been found to increase BDNF concentration, although, a study by Hung et al. (2018) recently showed that dynamic exercise (badminton) elicited a greater BDNF response than did static exercise (running). A recent review article by Gu et al. (2019) concluded that dynamic exercise appeared to be more effective for improving cognitive functioning relative to static exercise. 1.5 Cognitively-Engaging Exercise Exercise history, duration, intensity, and type, are four important qualifiers within the exercise-executive function relationship. Some researchers have recently speculated that an important, but understudied, additional quality is the extent to which a given exercise is cognitively engaging (Best, 2010; Fabel & Kempermann 2008; Diamond & Ling 2016). Specifically, the cognitive engagement hypothesis states that how much an exercise improves a cognitive ability is dependent on how much that exercise itself relies on and rehearses that  13  cognitive ability. It is this hypothesis that forms the crux of this dissertation and is reviewed within this section. Cognitively-engaging exercise can be defined broadly as any exercise that recruits and challenges a complex cognitive ability like working memory. The hypothesis of this dissertation is that greater participation in engaging exercise will improve that cognitive function to a greater extent than less engaging exercise. An early precursor of this hypothesis was proposed by Tomporowski (1997), who wrote “Older adult’s cognitive performance is affected positively by both physical and mental exercise interventions. It remains to be determined, however, if the two types of interventions produce their effects through totally separate mechanisms or whether there are sufficient commonalities between physical exercise and mental exercise to benefit from common explanatory theories (p. 18)”. This initial framework suggested that both physical and mental exercise benefit cognitive functioning, and, might do so through a related common mechanism.  Over time the cognitive engagement hypothesis has been defined more explicitly. Fabel and Kempermann (2008), studying exercise in relation to hippocampal neurogenesis, wrote "We thus propose that it is not isolated physical activity that is ‘good for the brain’, but physical activity in the context of cognitive challenges (p. 59)."  Best (2010), reviewing exercise and cognition in children, stated, “Cognitively engaging exercise appears to have a stronger effect than non-engaging exercise on children’s executive function (p. 331)”.  Diamond and Ling (2016) also advocated for this hypothesis, stating that exercise lacking cognitive engagement was unlikely to yield cognitive growth. In their words, “Boring exercise is particularly unlikely to yield cognitive benefits (pp.  38-39)”.   14  Although the cognitive engagement hypothesis has been proposed by various researchers, each with their own research specialities, direct evidence for the cognitive engagement hypothesis has so far been scarce. It is not difficult to understand why. For example, how can one determine which exercise is cognitively engaging for an individual and which is not? While using provocative terms such as “boring” or “mindless” is potentially descriptive of some exercise (Diamond & Ling, 2016), these terms do little to help identify which exercises meet that description (Hillman et al., 2019). For example, the same exercise might be considered boring to a newcomer and highly engaging to an expert. Experts might be sensitive to subtle differences in the techniques employed during an exercise (e.g., breathing approach), they may be competing against other highly trained competitors where subtle differences make all the difference (e.g., for placement in competitive rankings), and they may be performing their exercise at very demanding levels of performance (e.g., a marathon). The same exercise that is categorized by one individual as cognitively engaging may also be considered boring by another. An individual might simply not find a given exercise as appealing (regardless of whether it has a cognitive component or not), they may not be performing the activity at a sufficiently high level to find it engaging, or, they may already be so fluent with the underlying cognitive demands of that activity that they now find it uninteresting.  The challenge for researchers wanting to test the cognitive engagement hypothesis is to find a way to do so while taking into account the many idiosyncratic variations that exist between individuals, in the ways an exercise can be performed, and in the environments in which exercise is undertaken. The solution I propose here is a direct one. One can ask adherents of a given exercise whether they personally find it cognitively engaging or not. This approach,  15  centered on individual differences in exercise experience, is worth pursuing for at least four reasons. First, the very idea of whether an exercise is cognitively engaging or not is a phenomenological question. It can only be answered by the individual performing the exercise, and not by an external observer (e.g., the researcher). Second, an individual differences approach explicitly acknowledges that the same exercise can be differentially engaging for different individuals. Third, studying individual differences results in a more precise definition of “cognitive engagement”. Rather than relying on a term that can only be defined vaguely at a group level, questions can be created to target specific underlying cognitive processes. For instance, individuals in the upcoming chapters 2 through 4 were asked the degree to which their exercise experience involved separate facets of executive function, including inhibitory control and cognitive flexibility (e.g., inhibitory control was assessed with the question “When completing my primary exercise, I filter and ignore distracting information”). Fourth, defining cognitive engagement with respect to the theoretical framework of executive functioning (Diamond, 2013; Miyake et al., 2000; Zelazo, 2015) promotes the testing of very specific predictions. For example, in the studies reported in chapters 2 through 4 individuals self-reported the degree to which their primary exercise made demands on inhibitory control. This allowed us to test the specific hypothesis that exercise reliant on inhibitory control would predict better performance on a laboratory measure of inhibitory control (e.g., flanker task), and not performance on a laboratory measure of working memory (e.g., backward span). 1.6 Measuring Cognitive Engagement During Exercise  To test if cognitively-engaging exercise is beneficial for executive functioning requires a tool that can measure executive function use during exercise. To do so individuals across all  16  chapters within this dissertation were asked to report their primary exercise (described in greater detail within the method section of Chapter 2), after which individuals were asked questions intended to measure how much this exercise relied on inhibitory control, cognitive flexibility, and, working memory. In chapter 2 participants answered a total of 30 questions, 10 per executive function, with half of all questions being reverse worded (see Appendix A). The present section outlines the background literature that inspired the creation and wording of these questions. 1.6.1 Measuring Use of Inhibitory Control During Exercise   An important and often first-mentioned aspect of inhibitory control is the ability to selectively attend to some information while filtering and ignoring other information. On describing this aspect of inhibitory control Diamond (2013) wrote “Inhibitory control of attention (interference control at the level of perception) enables us to selectively attend, focusing on what we choose and suppressing attention to other stimuli (p. 137)”. This quality of inhibitory control as a mechanism through which distracting stimuli are filtered and ignored was measured through the measurement item: “filter and ignore distracting information”.1  Inhibitory control is also thought to be involved in various aspects of self-regulation (Tan & Holub, 2011; Graziano, Calkins & Keane, 2010), including self-control (Diamond, 2013; Katzir, Eyal, Meiran & Kessler, 2010), discipline (Diamond 2013; Jacobson, & Matthaeus, 2014) and impulse management (Diamond, 2013). This self-regulatory aspect of inhibitory control has been captured through phrases like “Without inhibitory control we would be at the mercy of impulses, old habits of thought or action (conditioned responses), and/or stimuli in the  1 All measurement items are prompted with the text “When completing my primary exercise I”  17  environment that pull us this way or that (Diamond, 2013; p. 137)”. Three measurement items were written to capture this self-regulatory aspect of inhibitory control: “decide what to do through impulse alone”, “practice self-control and discipline”, and, “act without self restraint” Sometimes despite best efforts to act in a measured approach, a mistake is made through one’s utterances or actions. These mistakes may arise as a result of impulsive behaviour or misreading of a situation. In such circumstances inhibitory control can help one to stop a response or action that has already been initiated (Verbruggen & Logan, 2008). Sometimes this aspect of inhibitory control is more narrowly called action-cancellation to distinguish it from the type of inhibitory control centered on management of perceptual information (van Velzen et al., 2014; Chu et al., 2015). Huster et al. (2020) described this aspect of inhibitory control when writing “Inhibitory control, i.e., the ability to stop or suppress actions, thoughts, or memories, represents a prevalent and popular concept in basic and clinical neuroscience as well as psychology (p. 334)”. Three questions were written to reflect this aspect of inhibitory control: “anticipate making fast or sudden adjustments”, “pause and double check what I am doing”, and “follow every action to completion”.  A common theme concerning the above descriptions of inhibitory control centres on behaving cautiously (Verbruggen et al., 2013). Pardini, Lochman and Wells (2004) for instance wrote “Specifically, there has been an increased interest in the construct of inhibitory control, which has been described as a tendency to be cautious and controlling of one’s personal behavior (p. 507)”. An additional three measurement items were written to capture cautious tendencies thought in alignment with this aspect of inhibitory control: “slow down to avoid making mistakes”, “start and complete actions without thinking” and “care most about speed and performing quickly”.  18  1.6.2 Measuring Use of Cognitive Flexibility During Exercise   Arguably the most prominent aspect of cognitive flexibility is the freedom to mentally shift from one perspective or mindset to another. Zelazo (2015) addressed this aspect of cognitive flexibility when writing “Cognitive flexibility involves thinking about something in multiple ways (p. 57)”. Diamond (2013) similarly suggested that “Cognitive flexibility is the opposite of rigidity (p. 149)”. Without cognitive flexibility for instance one might ruminate the same thoughts repeatedly, and fall into a routine that limits divergent thinking. Five items were written to capture this mental freedom aspect of cognitive flexibility: “have little-to-no flexibility to modify what I do”, “hold the same mindset from start to finish”, “think the same thoughts over and over”, “follow the same routine”, and, “have a plan that I stringently follow”. An important outcome made possible through cognitive flexibility is adaption and innovation. By considering different perspectives, one might find new techniques and strategies through which to achieve their desired outcome. This adaptive aspect of cognitive flexibility was described by Kesler, Lacayo and Jo (2011) in writing “Cognitive flexibility is the ability to generate alternate solutions to problems and fluidly shift between ideas and actions (p. 102)”. Buttelmann and Karbach (2017) expressed a similar sentiment when writing “Cognitive flexibility enables [us] to think divergently, change perspective and adapt to a continuously changing environment (p. 1)”. To include this adaptive aspect of cognitive flexibility, five measurement items were written: “adapt and change how things are done”, “try to identify new techniques or strategies”, “practice creativity”, “rely on a diverse skillset”, and “encounter and solve new problems”. 1.6.3 Measuring Use of Working Memory During Exercise  Working memory involves the ability to simultaneously hold and manage various pieces of mental information. In describing working memory Conway et al. (2007) wrote that “Working  19  memory (WM) is the term that cognitive psychologists use to describe the ability to simultaneously maintain and process goal-relevant information (p. 3)". Similar descriptions are often used when describing the importance of working memory for successful multitasking ability (Bühner et al., 2006; Colom et al., 2010). This aspect of working memory was measured through the three measurement items: “pay attention to many things at the same time”, “multitask”, and “connect and combine different ideas”.   An additional quality of working memory that researchers often emphasize, is remaining continuously updated and aware of one’s environment (Miyake et al., 2000). In discussing’s this facet of working memory, Ecker, Oberauer and Lewandowsky (2014) for instance wrote “The ability to keep working memory content up to date is vital for a number of higher cognitive functions such as navigation and reasoning, but it is also crucial for the effective operation of working memory itself (p. 1)”. This ability to remain aware of what is happening was measured through five measurement items: “monitor what is happening on a second-to-second basis”, “disengage from what is happening around me”, “understand everything that is happening even when absent-minded”, “avoid time-keeping” and “do not care about the order in which things happen”.  A practical utility and implication of continuously being aware of one’s environment is the ability to make inferences and predictions about what is likely to happen next. Goldman-Rakic (1992) for instance when writing on working memory stated that “It enables humans to plan for the future and to string together thoughts and ideas… (p. 111)”. Various studies similarly suggest that higher working memory capacity is associated with more time spent in future orientated thought (Baird, Smallwood & Schooler, 2011). To include this future orientated  20  aspect of working memory one additional measurement question was included: “make predictions about what will happen next”. 1.7 Cognitively Engaging Exercise & Executive Functioning: A Potential Mechanism  The primary focus of this dissertation is to test whether cognitively-engaging exercise predicts performance on laboratory measures of executive functioning. But even if this hypothesis is supported in the data, it would not by itself explain the underlying mechanism through which cognitively-engaging exercise sharpens executive functioning. This section of the manuscript considers possible mechanisms for how cognitively-engaging exercise may improve executive functioning.  As described earlier, exercise may benefit executive functioning through changes in cerebral blood flow and neuroplasticity (see 1.1 Exercise & Cerebral Blood Flow, 1.2 Exercise & Neurotrophins). With this in mind, Hötting and Röder (2013) suggested that an additional property of exercise is preparing the brain for anticipated cognitive stimulation. Exercise may provide the initial starting point for neuronal remodelling, which cognitively-engaging exercise may enhance even further. Raichlen and Alexander (2017) expressed a similar sentiment in their review, noting that while exercise is important for BDNF-initiated neuronal birth and synaptic connections, the maintenance of these neural changes is dependent on the cognitive challenges experienced during exercise. This hypothesis is also supported in animal research, as newly generated neurons within the rat hippocampus are more likely to survive following training on tasks that themselves rely on hippocampal functioning (Gould, Beylin, Tanapat, Reeves, & Shors, 1999).   Exercise that elicits neuroplasticity can be expected to selectively enhance an executive function by recruiting and making use of that executive function. Exercise reliant on working  21  memory, for instance, might selectively enhance working memory by accentuating the effects of hippocampal neurogenesis, allowing for more efficient working memory through these neural changes. In doing so, the structural changes elicited via neuroplasticity may be selectively acted upon and favored dependent on executive function use during exercise. Notably this speculative mechanism posits a contrast between exercise activities and other leisure activities without exercise.  First, non-exercise leisure activities are not expected to improve executive functioning through either cerebral blood flow or BDNF-initiated neuroplasticity. This is not to say that certain types of leisure activities cannot benefit executive functioning, but the claim is that they could not do so through exercise-related mechanisms. Second, because non-exercise leisure does not elicit BDNF-initiated neuroplasticity, cognitively engaging leisure activities cannot selectively enhance the effects of neuroplasticity.  These two reasons are consistent with the hypothesis that exercise that is cognitively engaging (but not non-exercise leisure) will be correlated with better performance on laboratory measures of executive functioning. The three studies in this dissertation test this hypothesis by considering two types of analyses, one focused on exercise, and a second focused on non-exercise leisure. This entails asking individuals about their primary exercise, as well as their primary non-exercise leisure, and measuring activity qualifiers (history, type, intensity and duration) as well as executive functioning use during these activities (section 1.6). Participants will then also perform laboratory measures of executive functioning (e.g., flanker task) to assess if either exercise, or leisure, is predictive of greater task performance. A further benefit is that by including both exercise and leisure analyses, an additional possibility can be tested for. Namely, it may be that individuals with high executive functioning in general seek out activities (exercise and leisure) that are highly demanding on executive functioning. If this is true, the results of both  22  exercise and leisure analyses ought to be similar, whereas if only exercise is found to predict greater executive functioning (and not leisure), this would suggest something distinct about the relationship between exercise and executive functioning. 1.8 Overview of Dissertation Studies  Chapter 2 in this dissertation sought to accomplish three goals. First, undergraduate students reported their primary exercise and then completed 30 items designed to measure executive function use during this exercise. These items were then analyzed through a set of confirmatory factor analyses until a final measurement model was identified. Two, a path analysis was completed in which exercise qualifiers were used to predict performance on laboratory measures of executive functioning. Specifically exercise history, duration, intensity, and type were used as predictors of performance on the flanker task (inhibitory control) and backward span (working memory). Three, armed with subjective reports of executive function use during exercise, exercise qualifiers, and laboratory measures of executive functioning, a structural equation model was built that combined both prior sets of analyses (path analysis and confirmatory factor analysis). In doing so, this structural equation model tested whether individual differences in executive function use during exercise predicted performance on laboratory measures of executive functioning.  This analytic approach followed a deliberately sequential process meant (1) to provide a novel methodology through which executive function use during exercise can be measured, (2) to create a model of data that is comparable to those from past research where common exercise qualifiers predicted executive functioning, and then (3) to hybridize points 1 and 2, thereby creating a model that combines both novel and familiar research methods. This same analytic process was then followed in chapters 3 and 4.   23  Chapter 3 expanded on the initial investigation made in chapter 2 by recruiting a new set of undergraduate students. These students similarly identified their primary exercise and self-reported whether this exercise was reliant on executive functioning (e.g., inhibitory control). This allowed for a replication of the final measurement model identified in chapter 2. Chapter 3 also adopted conceptually different laboratory measures of executive functioning through the stop-signal task (inhibitory control) and trail making B (hybrid measure of cognitive flexibility and working memory) in order to test the generality of the exercise-executive function link found in chapter 2.  Chapter 4 is an extension of the methodology of chapter 2 to a different participant population. Here, we tested a sample of participants recruited via Amazon’s Mechanical Turk. This allowed the findings of chapter 2 to be contrasted with a participant sample that differed on various demographic properties (e.g., gender and age) as well as their choice of regular exercise activities (e.g., approximately 25% of these participants reported walking as their primary exercise). In all three of these chapters (2 through 4) a parallel investigation was also undertaken that assessed leisure activity. Participants reported their primary leisure activity (that did not involve exercise), answered questions pertaining to leisure history, duration and intensity, and completed an identical set of items meant to measure executive function use during leisure. This permitted a complimentary comparison between the predictive relationships involving leisure versus exercise on laboratory tests of executive functioning. For example, if exercise that made heavier demands on inhibitory control also predicted better flanker task performance, but leisure activity that made heavier demands on inhibitory control did not, then this would support the  24  interpretation that the effects of cognitively engaging exercise are unique in predicting better executive functioning.                        25  Chapter 2: Study 1 2.1 Introduction  Chapter 2 tested the hypothesis that cognitively-engaging exercise is predictive of better executive functioning (Best, 2010; Fabel & Kempermann 2008; Diamond & Ling 2016). To do so, participants reported their primary exercise and completed 30 questions meant to measure executive function use during exercise, and then completed the flanker task (inhibitory control) and a backward span task (working memory). The 30 questions were subjected to a set of confirmatory factor analyses, with the intent of finding a good-fitting model of executive function use during exercise. Because the exact structure of this model was not yet known, a more general hypothesis was formulated regarding this model and performance on laboratory measures of executive functioning. This hypothesis stated that exercise reliant on executive functioning ought to predict better performance on the flanker task and backward span. There was, however, a more focused expected outcome as well, because the use of an executive function during exercise should provide benefits that are specific to a given laboratory measure of executive functioning. This meant for instance, that reports of greater inhibitory control during exercise were expected to predict better performance on the flanker task, and not performance on the backward span.  A flanker task and backward span were used as measures of inhibitory control and working memory, respectively. These two tasks were chosen because of their frequent use within the executive function literature, as well as because of their common use within the exercise and sport literatures (Eriksen & Eriksen, 1974; Rosen & Engle, 1997; Guiney & Machado, 2013; Chiu, Chen & Muggleton, 2017; Kluding, Tseng & Billinger, 2011; Kamijo et al., 2009; Wylie et al., 2018).  26  An identical set of analyses were also completed in which participants reported their primary (non-exercise) leisure activity. This additional set of analyses were included to have a reference condition against which anticipated exercise effects could be compared. For instance, if exercise reported to be more reliant on inhibitory control was found to predict better flanker task performance, and the same was not true concerning leisure activity, this would imply that exercise is unique. Alternatively, if both exercise and leisure analyses yielded the same results, this would suggest that an underlying mechanism likely accounts for both sets of analyses (e.g., a propensity for certain individuals to seek out executively demanding activities in general). Concerning the results of the leisure models more narrowly, no specific hypotheses were anticipated a priori, rather, the more direct purpose of these models was to contextualize exercise results. 2.2 Method 2.2.1 Power Analysis  To determine an appropriate sample size a power analysis was conducted given an anticipated measurement model. This model consisted of 3 latent factors (inhibitory control, cognitive flexibility, and working memory), each allowed to freely correlate, and each indicated by 10 items. Following the procedures outlined in MacCallum, Browne and Sugawara (1996), a sample size of 50 participants resulted in an estimated power of .86, given α = .05, 𝜀𝑜 = 0.07, 𝜀𝑎 = .10 and df = 402. Values for 𝜀𝑜 and 𝜀𝑎 were selected to be more liberal, given the relatively new nature of this research. Estimated power however had to be held in tension with anticipated modifications to the original model in the event of further exploratory analysis, as well as model non-convergence, potential Heywood cases, and other concerns common to newly tested constructs deviating away from an ideal solution. Furthermore, a well-fitting model identified  27  through confirmatory factor analysis (CFA) while critical, is only the initial first step toward the goal of creating a structural equation model (SEM) in which executive function use during exercise is used to predict performance on laboratory measures of executive functioning. It was also difficult to anticipate a priori what proportion of the recruited participants will endorse regularly engaging in exercise (some might for instance sign up for the study but only endorse participating in non-exercise leisure). For all these reasons, a sample size of 50 was thought likely insufficient, and instead a sample size of 200 participants was targeted for this initial study, with the intent to perform more informed power analyses in follow-up studies. A sample size of 200 is a commonly suggested starting value when discussing adequate power for newly proposed SEM models (Boomsma, 1985; Holbert & Stephenson, 2002).  2.2.2 Participants   All participants were recruited through the University of British Columbia human subject pool, following review and approval of the research plan by the Behavioral Research Board of the University of British Columbia (H18-03515). A total of 196 participants completed the study. All participants completed the study within an hour and were provided 1 course credit. Following data filtering steps, the sample size for participants reporting a primary leisure was 178, and 145 for participants reporting a primary exercise.  Participants who reported exercising were on average 20.43 years old (SD = 2.02), the majority were women (73.10%) and most identified as East Asian (45.52%), followed by European/Caucasian (23.45%), Indian-South (16.55%), Other (6.70%), African (2.76%), Native American (2.07%) and Latin American (.69%). Figure 2.1 shows the frequency with which various types of exercise were reported, with exercises considered to be static versus dynamic indicated with shading. The most frequently reported exercises were running (12.41%) and  28  weight training (12.41%), followed by gym (11.03%). Participants reported an average exercise history of 4.36 (SD = 1.06) on a 5-point Likert scale, which generally suggested a starting date of more than 7 months ago. The average exercise frequency was 3.18 (SD = 1.30) on a 5-point Likert scale, which corresponded to roughly 60 to 90 minutes of exercise per week. The average exercise intensity was 3.26 (SD = .71) on a 4-point Likert scale, thereby suggesting moderate exercise intensity.  Participants who reported a primary leisure activity (most of whom had also reported exercising) were on average 20.44 years old (SD = 1.99), the majority were women (73.03%) and most identified as East Asian (49.44%), followed by European/Caucasian (20.79%), Indian-South (17.98%), Other (6.74%), African (2.25%), Native American (2.25%) and Latin American (.56%). Figure 2.2 shows the frequency with which various leisure activities were reported, with activities considered to be passive versus active indicated with shading. The most frequently reported leisure activity was viewing (e.g., television; 25.28%), followed by gaming (11.80%), and reading (11.24%). Participants reported an average leisure history of 4.70 (SD = .79) on a 5-point Likert scale, which generally suggested a starting date of more than 9 months ago. The average leisure frequency was 3.61 (SD = 1.36) on a 5-point Likert scale, which fell between response options 60 to 90 minutes, and 90 to 120 minutes per week. The average leisure intensity was 2.07 (SD = .77) on a 4-point Likert scale, thereby suggesting low leisure intensity. 2.2.3 Procedure   On arriving to the laboratory, participants were given a consent form to read and sign. Participants were seated in a small testing room and were told they would be completing a series of computerized questionnaires, before being tested on two cognitive tasks. Participants were encouraged to ask the study experimenter for clarification at any point in the session.  29   The first question participants answered was “Do you complete or take part in any form of exercise?”, with the option of answering “Yes” or “No”. If participants answered “Yes”, they completed a series of exercise and leisure questions, with the order of the question set (exercise or leisure) randomly determined. If participants answered “No”, they only answered a series of leisure questions.  Exercise questions began with the prompt, “People engage in many forms of exercise, including jogging, swimming, cycling, weight training, soccer, basketball, volleyball, rock-climbing, golf, yoga and various martial arts. Although there are many types of exercise, try to think of the one that you consider most central and primary for you. In the box below, type the name of your primary exercise. Only type the name of one exercise. If more than one exercise comes to mind, type the one you have done the most frequently in the past six weeks.” This prompt is similar to one used by Chekroud et al. (2018) in which participants were asked to report their primary exercise, and were provided exercise examples as well as a 1-month timeframe. After participants freely typed and submitted an answer, the next screen asked a series of Likert scale questions meant to measure executive functioning use during exercise.  All Likert scale questions were prompted with, “When completing my primary exercise I…” which was followed by 30 questions organized into 3 sections (see Appendix A). These 3 sections consisted of 10 questions each, meant to measure inhibitory control (e.g., “filter and ignore distracting information”), cognitive flexibility (e.g., “try to identify new techniques or strategies”), and working memory (e.g., “pay attention to many things at the same time”). Each question was answered by selecting a value ranging from 1 through 5, representing the terms “Never”, “Rarely”, “Sometimes”, “Often” and “Always”. Half of all questions were positively  30  worded and half were negatively worded, with wording type alternating after each question. Negatively worded items were reverse scored for all analyses.   Following these 30 exercise questions, a set of qualifier questions measured exercise history, duration, and intensity (Chang et al., 2012; Roth et al., 2003; Colcombe & Kramer, 2003; Soga et al., 2018). History was measured by asking participants when they first began their primary exercise, with response options being “less than 1 month ago”, “1 to 3 months ago”, “4 to 6 months ago”, “7 to 9 months ago”, and “more than 9 months ago”. Both duration and intensity were next measured through self-report questions comparable to those used in other studies (e.g., Roth et al. 2003; Chekroud et al., 2018). Duration was measured by asking participants over the past 6 weeks how much time in a typical week was spent actively engaging in their primary exercise, with response options being “less than 30 minutes”, “30 to 60 minutes”, “60 to 90 minutes”, “90 to 120 minutes”, and “120 minutes or more”. Intensity was measured asking participants to rate the intensity of their primary exercise with response options being “none”, “low intensity”, “moderate intensity”, and “high intensity”.  The primary exercise participants listed was then coded as either static or dynamic. Whether an exercise was coded as static or dynamic was determined by common guidelines as well as past precedent (Arvinen-Barrow et al., 2007; Coelho et al., 2007; Chiu, Chen, & Muggleton, 2017; Di Corrado, Guarnera, & Quartiroli, 2014; Ericsson & Smith, 1991; Highlen & Bennett, 1979; Lum, Enns & Pratt, 2002; McLeod, 1985; Starkes & Ericsson, 2003; Vaeyens, Lenoir, Williams & Philippaerts, 2008; Wang et al., 2013).  If participants reported not exercising, or were randomly assigned to complete leisure questions first, they instead began with the prompt: “People engage in many different leisure activities that are not related to exercise. These include drawing, writing,  31  painting, reading, cooking, travelling, puzzle solving, carpentry, photography, theatre, politics, playing instruments, listening to music, and many others. Although there are many types of leisure activity try to think of the one that you consider most central and primary for you. In the box below, type the name of your primary (non-exercise related) leisure activity. Only type the name of one leisure activity. If more than one leisure activity comes to mind, type the one you have done the most frequently in the past six weeks.” After answering this question, participants completed a complimentary set of 30 questions measuring executive function use during leisure activity, as well as questions meant to measure leisure history, duration and intensity. These leisure questions mirrored the exercise set of questions.  To be consistent with how exercise type was categorized, leisure activities were similarly dichotomized. Various studies have provided guidelines for categorizing leisure (e.g., Richards, Hardy & Wadsworth, 2003; Wang, Xu & Pei, 2012), here however we adopted a classification structure similar to Dardis, Soberon-Ferrer & Patro (1994) where leisure activity is differentiated based on whether that activity requires active involvement (e.g., cooking) or if instead the leisure activity can be completed without interaction (e.g., watching television). Leisure activities were thus categorized as being either active or passive.   Cognitive testing began after all exercise and/or leisure questions were answered. Participants completed the flanker task and backward span, with task order being randomized. The flanker task used in this study is similar to many others that have been used in attention and sport research (e.g., Wylie et al., 2018). Participants were informed they would be responding to the direction of a central arrow presented on the computer screen. If the arrow pointed left participants were told to press “Z” on an adjacent keyboard, and if the arrow pointed right they were told to press “/”. Participants were also informed that the central arrow would be flanked by  32  adjacent arrows that could be either congruent (e.g., > > > > >) or incongruent (e.g., < < > < <) with the central arrow. Participants were instructed to respond as quickly and accurately as possible to the orientation of the central arrow and to ignore all other arrows. The start of a trial was signified by a blank screen (200ms), followed by a fixation cross (500ms), and another blank screen (200ms) after which the central arrow with adjacent flankers appeared. Participants completed 10 practise trials which included feedback (i.e., correct or incorrect), after which followed 200 test trials. Whether a given trial was congruent or incongruent was randomly determined, with the constraint that half of all trials were congruent and half were incongruent (i.e., 100 trials each).   Inhibitory control on the flanker task was assessed via accuracy and reaction time differences between congruent and incongruent trials (Chen et al., 2014; O’Leary et al., 2011; Raine et al., 2018; Sandroff, Benedict & Motl, 2015; Wylie et al., 2018). Specifically, accuracy for congruent trials was subtracted by accuracy for incongruent trials, and reaction time for incongruent trials was subtracted by reaction time for congruent trials. For both metrics larger values are interpreted to mean worse inhibitory control. By this interpretation, if incongruent flankers are effectively inhibited, they should be no more interfering for task performance than congruent flankers. Figure 3.3 summarizes average flanker task performance among exercise and leisure participants. Both groups demonstrated the expected flanker effect where incongruent trials are completed more slowly and less accurately than congruent trials.  The backward span task used in this study was modelled after Bourrier, Berman and Enns (2018). On starting this task, participants were informed that they would see a series of digits presented one at a time on the computer screen. Each digit appeared on screen for 1 second, and, after all digits had appeared, a dialog box would become available and participants were asked to  33  type the digits they saw in backward order. Participants were given examples of how to complete a trial, for example participants viewed the digits 517 and were then informed the correct answer to this trial was 715. Participants also completed 2 practise trials that included feedback (i.e., correct or incorrect), after which followed 14 test trials. The first two test trials had participants view 3 digits presented one at a time before being asked to report them in backward order. After every two trials the number of displayed digits increased by 1 (4 digits, 5 digits, 6 digits, etc.) until participants viewed 9 digits in total. Performance was calculated such that for every correct digit reported a capacity score of 1 was assigned. For instance, if the presented digits had been “215” and the response was “542”, the capacity score would be 2 for correct placement of the digits “5” and “2”. Once all trials were completed, an average capacity score per subject was calculated and represented working memory performance. Figure 2.3 summarizes backward span performance among exercise and leisure participants, and depicts the expected trend of high recollection for trials with few digits (3 and 4), followed by recollection of around 4 or so digits on harder trials (5 through 9; Cowan, 2010).  Performance on the flanker task was screened to include reaction times between 300ms and 1500ms (Chen, Zhao, Fan & Chen, 2018; van Leeuwen et al., 2007; White, Ratcliff, & Starns, 2011; White, Brown, & Ratcliff, 2012), amounting to fewer than 3% of all trials being excluded from analysis. Participants who had near chance levels of performance on either the flanker task or backward span were also excluded as follows: Four subjects had approximately 50% accuracy on the flanker task and were excluded from analysis for falling below a cut-off value of 51%. Seven subjects had approximately 11% accuracy on the backward span and were excluded from analysis for falling below a cut-off of 14%. An additional four subjects were excluded from analysis for listing an identical primary exercise as their leisure (e.g., soccer for  34  both). Participants who had been involved with their primary exercise and/or leisure for less than one month were also excluded, for two reasons. First, it only became apparent during exit-interviews and debriefing that in some cases students had only very recently started their primary exercise and/or leisure (e.g., start of that semester, or as part of university orientation) and therein had limited experience with this activity. Second, various reviews and studies (e.g., Etnier et al., 1997; Pérez et al., 2014) report that chronic rather than acute exercise is most likely to reveal an impact on cognitive functioning. Participants thus had to have a history of at least 1 month with their reported exercise and/or leisure to qualify for analysis (Smith et al., 2010). These considerations resulted in a final sample size of 178 subjects who reported a primary leisure activity and 145 who reported a primary exercise.2  Cognitive tasks were created using the Matrix Laboratory (MatLab), along with the Psychophysis Toolbox (Brainard, 1997; Kleiner, Brainard & Pelli, 2007). On completion of both cognitive tasks, participants provided basic demographic information on their age, gender, and ethnicity.  2.2.4 Data Analytic Approach   All three studies in this thesis follow a similar data analytic approach. First, executive function use during exercise was modelled through confirmatory factor analyses (CFA). Second, path analyses tested if exercise qualifiers (history, duration, intensity, and type) predicting performance on laboratory measures of executive function. Third, this path analysis and the prior CFA were then combined to form a structural equation model (SEM) that tested if performance on laboratory tasks of executive function was predicted by individual differences in executive  2 Once we established these exclusion criteria for chapter 2, they were applied for chapters 3 and 4.  35  function use during exercise. This same process was then mirrored and applied to leisure data. The remainder of this section describes this process in greater detail specific to chapter 3, although chapters 4 and 5 also follow a similar approach. Confirmatory factor analysis began by modelling the 30 items measuring executive function use during exercise. Inhibitory control, cognitive flexibility and working memory were each latent constructs in an initial CFA model. Each latent construct was indicated by 10 items and these three constructs were allowed to correlate freely with one another. This initial model, as well as all subsequent confirmatory factor analysis models, were first evaluated using a chi-square (𝜒2) test. A non-significant result on this test is interpreted to mean the proposed model should be retained rather than rejected. One caution is that the chi-square test tends to be sensitive to relatively minor model misspecification given a large enough sample size (Bentler & Bonett, 1980; MacCallum, Browne, & Sugawara, 1996; for a further discussion on the limitations of the chi-square test see Steiger, 2007). Because of this, alternative fit indices are also reported based on guidelines provided by Jackson, Gillaspy and Purc-Stephenson (2009) and Kline (2016). These alternative fit indices include the root mean square error of approximation (RMSEA), the comparative fit index (CFI), the standardized root mean square residual (SRMR) and the largest standardized absolute residual (LSAR). In adopting these fit measures, no single index was alone decisive in characterizing model fit and all were considered (and are reported) collectively for consideration. These fit indices are discussed briefly in the paragraphs that follow, though for greater detail readers are directed to Kline (2016), Savalei and Bentler (2006), and Vriens (2006). When performing a chi-square test, the sample covariance matrix (S) serves as an estimate of the population covariance matrix (Σ). The population covariance matrix is  36  hypothesized to have a certain structure wherein some parameters are thought to be functions of other parameters. This hypothesized structure is specified by the researcher to create their proposed model, which is then applied to S. The estimate of the population covariance matrix through a proposed model is called ?̂?, and is obtained through a fitting function (F) that aims to minimize the discrepancy between S and ?̂?. If both S and ?̂? are estimates of the same entity, their discrepancy will be zero. In all analyses within this manuscript maximum likelihood estimation was the applied fitting function (i.e., 𝐹𝑀𝐿).3 Parameter estimates from this proposed model are collectively are referred to as θ and the null hypothesis is that Σ = Σ(θ), meaning that the population covariance matrix is a function of the estimated parameters captured by θ. This hypothesis is tested through the statistic 𝑇, the formula for which is 𝑇 = (𝑁 −  1)𝐹𝑀𝐿, wherein N reflects the number of observed cases (typically the sample size). When assumptions are met 𝑇 approximates a chi-square distribution with degrees of freedom equal to 𝑝∗(𝑝+1)2−𝑞, wherein p is equal to number of variables defining S, and q is the number of estimated parameters (i.e., number of items compromising θ). Retention of the null-hypothesis (a non-significant chi-square) suggests that the observed data do not sufficiently deviate enough from the proposed model to merit model rejection. A non-significant chi-square suggests one ought to retain their proposed model as there is insufficient evidence to reject it.  The chi-square test is often referred to as a measure of exact-fit, such that the proposed model is thought to be exactly correct in the population and only sampling error is responsible for model imprecision (Bentler & Bonett, 1980; MacCallum, Browne, & Sugawara, 1996). By contrast, RMSEA is a measure of approximate-fit wherein distance from exact-fit is quantified  3 Full information maximum likelihood was used in one instance for study 1 because there were 3 missing data points for leisure models.  37  through a non-centrality parameter and fit is evaluated on the basis of this distance.4 Calculation of RMSEA [1] relies on subtracting the chi-square of the proposed model (𝜒𝑃𝑟𝑜𝑝𝑜𝑠𝑒𝑑 𝑀𝑜𝑑𝑒𝑙2 ) by its degrees of freedom (𝑑𝑓𝑃𝑟𝑜𝑝𝑜𝑠𝑒𝑑 𝑀𝑜𝑑𝑒𝑙).  The result is then divided by (𝑁 − 1) ∗ 𝑑𝑓𝑃𝑟𝑜𝑝𝑜𝑠𝑒𝑑 𝑀𝑜𝑑𝑒𝑙, where 𝑁 represents the number of model cases (this typically being the study sample size). The square-root of this value is then used as a measure of model misfit, with larger values denoting greater misfit (and under exact-fit RMSEA equals 0). Values falling below .06, .08. and .10 are then often interpreted to mean one’s model respectively has great, good, and mediocre fit (Browne & Cudeck, 1992; MacCallum, Browne & Sugawara, 1996). Researchers have also recommended that because the distributional properties of RMSEA are known, confidence intervals ought to be constructed to provide estimate uncertainty and imprecision, this recommendation is adopted in this thesis for all instances in which RMSEA is reported. 𝑅𝑀𝑆𝐸𝐴 =  √𝜒2 − 𝑑𝑓𝑃𝑟𝑜𝑝𝑜𝑠𝑒𝑑 𝑀𝑜𝑑𝑒𝑙(𝑁 − 1)𝑑𝑓𝑃𝑟𝑜𝑝𝑜𝑠𝑒𝑑 𝑀𝑜𝑑𝑒𝑙 [1] The comparative fit index (CFI), as suggested by its name, is a is a measure of relative fit where the proposed model is compared to a poorly fitting model. Typically, as is the case here, the poorly fitting model is one in which all covariances are set to zero and only individual variances are estimated [2]. The chi-square for this poorly fitting model (𝜒𝑁𝑢𝑙𝑙 𝑀𝑜𝑑𝑒𝑙2 ), and its degrees of freedom are obtained (𝑑𝑓𝑁𝑢𝑙𝑙 𝑀𝑜𝑑𝑒𝑙) are then used in the calculation of the CFI [3]. Specifically, the CFI is a ratio of the proposed model chi-square minus the proposed model degrees of freedom, over the null model chi-square minus the null model degrees of freedom. One minus this ratio is the CFI and is interpreted such that values closer to 1 represent greater  4 Browne and Cudeck (1992) have also described the RMSEA as “a measure of the discrepancy per degree of freedom for the model"  38  model fit. A value above .90 is typically taken to mean mediocre-to-acceptable fit, with values above .95 signifying good fit (Hu & Bentler, 1999). 𝑁𝑢𝑙𝑙 𝑀𝑜𝑑𝑒𝑙 =(  𝜎1     0 𝜎2    0 0 𝜎3   0 0 0 𝜎…  0 0 0 0 𝜎𝑘)  [2] 𝐶𝐹𝐼 = 1 − (𝜒𝑃𝑟𝑜𝑝𝑜𝑠𝑒𝑑 𝑀𝑜𝑑𝑒𝑙2 − 𝑑𝑓𝑃𝑟𝑜𝑝𝑜𝑠𝑒𝑑 𝑀𝑜𝑑𝑒𝑙)(𝜒𝑁𝑢𝑙𝑙 𝑀𝑜𝑑𝑒𝑙2 − 𝑑𝑓𝑁𝑢𝑙𝑙 𝑀𝑜𝑑𝑒𝑙)[3]  SRMR is a summary characterization of the standardized residual matrix. The formula for the SRMR is shown in equation [4], and at its core is the observed data correlation matrix (𝐶𝑀𝑂𝑏𝑠𝑒𝑟𝑣𝑒𝑑) subtracted by by the model implied correlation matrix (𝐶𝑀𝐼𝑚𝑝𝑙𝑖𝑒𝑑). Deviations resulting through this subtraction reflect the degree of residual between what the data are and what the model predicts ought to be. Redundant values are removed, since the matrix is symmetrical along the diagonal, and the remaining values are then squared, summed, and divided by  𝑝 ∗ (𝑝 + 1)/ 2, where p is the number of model variables. The square-root of this value is then interpreted as a summary statistic of the residual matrix. Smaller SRMR values indicate less discrepency between what the model predicts and what the observed data are, with values below .08 commonly interpreted as signifying a well-fitting model (Hu & Bentler, 1999).  𝑆𝑅𝑀𝑅 = √(∑(𝐶𝑀𝑂𝑏𝑠𝑒𝑟𝑣𝑒𝑑 − 𝐶𝑀𝐼𝑚𝑝𝑙𝑖𝑒𝑑)2 𝑝 ∗(𝑝 + 1)2  ) [4] One limitation of the thus far listed fit indices is an emphasis of global-model fit over local-model fit. SRMR, for example, provides a summary metric of the residual matrix and in doing so evaluates global-fit. While this is meaningful information, this summary metric may  39  conceal large residuals indicatative of local-model misfit (i.e. the model implied correlation matrix may predict a value to be substantially larger in magnitude than otherwise found in the observed data correlation matrix). For this reason the LSAR is reported [5], and is the largest standardized absolute value within the residual matrix (that is, the single largest absolute difference between the observed correlation matrix and model implied correlation matrix). Although no common cut-offs exist, researchers have recently recommended that this value be reported as an indicator of local misfit (Kline, 2016). 𝐿𝑆𝐴𝑅 = 𝑚𝑎𝑥|𝐶𝑀𝑂𝑏𝑠𝑒𝑟𝑣𝑒𝑑 − 𝐶𝑀𝐼𝑚𝑝𝑙𝑖𝑒𝑑| [5] Lastly, for all CFA models coefficient omega (𝜔) is reported as a measure of construct reliability. Coefficient omega is used instead of Cronbach’s alpha because Cronbach’s alpha assumes measurement of a unidimensional construct whereas coefficient omega is applicable to confirmatory factor analysis models in which are multiple factors (Hayes & Coutts, 2020; Watkins, 2017). Coefficient omega is a ratio of reliable variance over reliable variance plus error variance [6]. To calculate reliable variance the covariance matrix of the model factor loadings (𝛬) is multiplied by the covariance matrix of factors (𝛷), and this product is then multiplied by the transposed covariance matrix of factor loadings (𝛬′). The denominator of this ratio includes this same term plus the covariance matrix of residuals (𝛹). Coefficient omega is interpreted as the proportion of variance in all indicators accounted for by all model factors. Common cut-off values for Cronbach’s alpha vary from .50 to .90, reflecting poor to excellent reliability (Crutzen & Peters, 2017) and here a similar gradient is adopted for interpreting coefficient omega.  𝜔 = ∑(𝛬𝛷𝛬′)∑(𝛬𝛷𝛬′) + ∑𝛹[6]  40  Path analysis in each chapter consisted of testing if exercise qualifiers in history, duration, intensity and type (dynamic or static) predicted performance on laboratory measures of executive functioning. In chapter 2, these exercise qualifiers each predicted inhibitory control as measured by the flanker task, and working memory as measured by the backward span. A structural equation model (SEM) was then built by combining this path analysis with the prior exercise CFA model. The primary purpose of this SEM was to test if exercise perceived to be reliant on executive functioning was predictive of performance on laboratory measures of executive functioning. Fit indices for structural equation models are reported, although are less focused on than were for CFA models (i.e., the measurement portion of the structural equation model). As noted in chapter 1, this analytic approach follows a deliberately sequential process meant to (1) provide a novel methodology through which executive function use during exercise is measured, (2) to create a model that is comparable to past research wherein exercise predicts executive functioning (3) to hybridize points 1 and 2, thereby creating a model that combines both novel and familiar research methods, and tests if executive function use during exercise predicts performance on laboratory measures of executive functioning. This same analytic process followed in chapters 3 and 4 of this dissertation.  A complimentary set of analyses (CFA, path analysis and SEM) were also conducted using leisure data, such to allow for a comparison between leisure and exercise models. In doing so, whether exercise uniquely predicts executive functioning is considered. All models were built using the statistical language R, through the use of packages lavaan (Rosseel, 2012) and semTools (Jorgensen, et al., 2018).   41  2.3 Results 2.3.1 Exercise Results 2.3.1.1 Measurement Item Statistics  Descriptive statistics for all exercise measurement items are summarized in Table 2.1. Most items had a mean of about 3, with the lowest being 2.23 (item 6) and highest being 3.73 (item 18). Items generally had comparable variance, with each item having a standard deviation of approximately 1. All item categories were endorsed for every item (i.e., all items had a minimum value of 1 and maximum value of 5). A mean composite score was calculated for inhibitory control, cognitive flexibility, and working memory. Items 1 through 10 represented inhibitory control, 11 through 20 cognitive flexibility, and 21 through 30 working memory. Inhibitory control had a mean of 3.18 (SD = .35), cognitive flexibility had a mean of 3.09 (SD = .48), and working memory had a mean of 3.00 (SD = .38). Greater inhibitory control correlated with greater cognitive flexibility, r(143) = .36 (95CI [.21, .50]) p < .001, but did not correlate with working memory, r(143) = -.09 (95CI [-.25, .07]) p = .265, greater working memory however did correlate with greater cognitive flexibility r(143) = .34 (95CI [.19, .48]) p < .001. This initial evidence suggests that cognitive flexibility is related to both inhibitory control and working memory, whereas working memory is only related to cognitive flexibility.  2.3.1.2 Confirmatory Factor Analysis & Measurement Model Identification  A set of confirmatory factor analyses were completed next, with the intent of identifying an optimally fitting model of executive function use during exercise. An optimal model here was defined as one that had a CFI above .95, an SRMR below .08. and a RMSEA below .06. For every analysis the chi-square is also reported for which a non-significant result is favorable. However, as this is typically not an attainable result, optimal fit was primarily defined through the aforementioned alternative fit indices. As the measurement model identified through these  42  analyses would subsequently be used in future studies (chapters 3 and 4), meeting these indices of optimal fit was seen as an important preparatory step in anticipation of these subsequent studies.  An initial confirmatory factor analysis was conducted on the 30 items measuring executive function use during exercise. Inhibitory control, cognitive flexibility, and working memory were latent constructs in this model, were allowed to freely correlate with one another, and were each indicated by 10 items. This model hereafter will be referred to as the original three-factor model, and is summarized in Table 2.2. The original three-factor model posed a multitude of measurement problems. Among these problems was high heterogeneity and inconsistency for individual factor loadings. Positively worded items tended to load positively and significantly (p < .05), whereas negatively worded items loaded sometimes positively, sometimes negatively, and sometimes non-significantly. Examining the correlations among latent constructs revealed that inhibitory control correlated with cognitive flexibility (r = .85), as well as working memory (r = .82), and that cognitive flexibility correlated particularly highly with working memory (r = .96). This prominent correlation between cognitive flexibility and working memory may be suggestive of poor discriminant validity, as for instance, Kline (2016), Kenny (2016), and Brown (2015) all write that correlations above .85 for latent constructs are symptomatic of such an issue. Lastly, the overall fit indices associated with this model were indicative of poor fit, 𝜒2(402) = 978.80, p < .001, RMSEA = .10 (90CI[.09, .11]), CFI = .50, SRMR = .12. Among the problems present in the original three-factor model, reverse worded items seemed the most salient. There is a large body of literature suggesting that use of reverse worded items may elicit statistical noise, and that these items often are unreliable, are confusing, are  43  prone to producing method-based factors, and may be more burdensome or demanding for participants to interpret by virtue of their phrasing (DiStefano & Motl, 2006; Woods, 2006; Zhang, Noor & Savalei, 2016). The general pattern of loading inconsistency found in the present results was thought attributable to these negative qualities of reverse worded items.  The first approach in creating a better fitting and more meaningful model was thus removal of these items. The resulting model is hereafter referred to as the positive three-factor model, and is summarized in Table 2.3.  The positive three-factor model as a whole retained poor fit, 𝜒2(87) = 218.06, p < .001, RMSEA = .10 (90CI[.09, .12]), SRMR = .11, though did see a slight increase denoted most prominently by the CFI (.50 to .78). All factor loadings were also now significant, with the exception of inhibition item 9. Latent constructs continued to significantly and highly correlate with one another. Inhibitory control correlated with cognitive flexibility (r = .82) as well as working memory (r = .60). Notably however, the correlation between cognitive flexibility and working memory was now at an impossibly high value of 1.02. This result was interpreted to mean that these two constructs lacked discriminant validity and were no longer statistically distinguishable. This was seen as the most pressing and problematic aspect of the positive three-factor model that had to be remedied. One suggestion offered by Brown (2015) and Kenny (2016) is to combine highly correlated factors (greater than .85) into a single latent factor. Prior to executing this suggestion, it is worth considering if the underlying items of these two factors may indeed be indicating a related construct. A read through the measurement items would suggest this is tenable scenario. For instance, working memory item 25 states “During my primary exercise I connect and combine different ideas” and here both working memory and cognitive flexibility may be  44  required. Cognitive flexibility may underlie the initial identification of multiple ideas that working memory then combines and evaluates in creating an overall coherent interpretation. Brown (2015) notes that ideally fit will not greatly deviate between two models where the only difference is a conjunction of two factors into one, thereby suggesting an expectation of little difference in model fit. The resulting model that combined cognitive flexibility and working memory into a single construct is hereafter referred to as the full two-factor model, and is summarized in Table 2.4. The fit indices of the full two-factor model were highly comparable to the positive three-factor model and as such retained poor fit, 𝜒2(89) = 225.50, p < .001, RMSEA = .10 (90CI[.09, .12]), CFI = .77, SRMR = .10. This is in line with both the expectation set by Brown (2015) as well as the rational that a more parsimonious model typically does not fit better. Among indicator items, again, only item 9 had a non-significant factor loading. A significant correlation (r = .79) remained between inhibitory control and the hybrid factor representing cognitive flexibility and working memory. Poor fit indices were now seen as the most pressing problem defining the current measurement model, and a series of steps were next completed to identify a better fitting model.  A common data-driven approach to improve model fit concerns use of modification indices. Through a succession of these indices researchers add additional paths to their model (e.g., correlating measurement errors or allowing for cross-loadings) that progressively contribute to improved fit. However, while this approach may help produce a model that better describes and accommodates a current dataset, this approach has also often been found to yield models that do not replicate and fail cross-validation efforts (Hermida, 2015; MacCallum, Roznowski & Necowitz, 1992). As the current study sought to establish a common set of  45  measurement items that may then be carried forward into future studies, a more parsimonious and efficient approach was thought to be one that produced a smaller subset of items that more concisely measured the constructs of interest (Hair et al., 2009). This would also potentially aid scale use such that future studies would not be required to add additional paths (e.g., correlate residuals), but rather, the theorized optimal measurement model would be captured solely through the inherent scale items.  Various guidelines and suggestions have been put forward for item selection and removal geared toward improving model fit (Brown, 2015; Pituch & Stevens, 2015). Hair et al. (2009) for instance suggested that items with non-significant factor loadings ought to be dropped, and that items with weak or low factor loadings ought to at least be considered for removal. What constitutes a low factor loading somewhat varies across researchers, Hair et al. (2009) used .50 as a benchmark, whereas Briggs and MacCallum (2003) defined weak loadings as those falling below .45. Some researchers have suggested lower values still, within the range of .30 through .40, although these values were also framed in the context of exploratory factor analysis (Brown, 2015; Pituch & Stevens, 2015). Within the present study, the more conservative .50 benchmark was adopted and used to screen for items that ought to be considered for removal (Hair et al., 2009). Specifically, all items falling below this threshold value were first identified, the lowest loading item was then dropped, and this process repeated. This iterative process was applied rather than dropping all items below .50 simultaneously, because the removal of a single item may result in fluctuation of other item loadings. Items were no longer dropped once fit indices suggested strong model fit, denoted by a CFI above .95, an SRMR below .08. and a RMSEA below .06. An anticipatory requirement was also placed such that this process would be stopped and be re-evaluated if fewer than 3 items per  46  latent construct were suggested, as removal of items beyond that point would have to be weighed in tension with potential convergence problems in future studies. Inhibition item 9 was the first item removed (non-significant, and with a low factor loading of .19), followed by hybrid items 27 (significant and with a low-loading of .19), 21 (significant and with a low-loading of .30), 29 (significant and with a low-loading of .32), and lastly inhibition item 5 (significant and with a borderline loading of .48). After removal of items 9, 27 and 21, only the SRMR denoted optimal fit trough a value below .08 (𝜒2(53) = 86.04, p < .001, CFI = .93, SRMR = .07, RMSEA = .07). Further removal of item 29 brought most fit indices directly to their threshold value of optimal fit (𝜒2(43) = 65.00, p = .02, CFI = .95, SRMR = .06, RMSEA = .06), and finally removal of item 5 resulted in a model that surpassed these thresholds (𝜒2 = 44.83, p = .10, CFI = .97, SRMR = .06, RMSEA = .05). Having achieved this hard-stop criteria, with all fit indices now denoting optimal fit (in addition to a non-significant chi-square) no further items were removed as the model was thought to be within a range of fit that realistically could not be improved. This model, hereafter referred to as the two-factor model, is summarized in Table 2.5 (illustrated in Figure 2.4) and consisted of 3 items measuring inhibitory control, and 7 items measuring cognitive flexibility. The title of cognitive flexibility was retained for this latter construct, as 5 of the 7 items had originally been intended to measure this executive function. 2.3.1.3 Path Analysis & Structural Equation Modelling   A path analysis next tested whether inhibitory control and working memory, as measured by performance on the flanker task and backward span, could be predicted by exercise qualifiers in history, duration, intensity, and type. Exercise qualifiers were allowed to freely correlate with one another, and performance on the flanker task freely correlated with performance on the  47  backward span. This resulted in a saturated model, as shown in Figure 2.5, and for this reason no fit indices are reported. None of the exercise qualifiers were found to predict either inhibitory control or working memory. Correlations among the predictor variables revealed that dynamic exercise was longer in duration and history, and greater exercise intensity was also correlated with exercise of longer duration and history. Correlations among the executive functions were not significant.  Having thus far been unsuccessful in predicting performance on laboratory measures of inhibitory control and working memory with exercise qualifiers (history, duration, intensity, and type), a structural equation model was built that combined this path analysis with the previously identified two-factor exercise model. This SEM model is shown in Figure 2.6 and tested if perceived executive function use during exercise predicted executive function as measured by the flanker task and backward span, while controlling for common exercise qualifiers in exercise history, duration, intensity and type. Exercise perceived to be reliant on inhibitory control predicted greater inhibitory control on the flanker task (i.e., a smaller accuracy difference between congruent and incongruent trials), β = -.33, B = -3.30 (95.CI = [-6.46, -.13]), p = .0415, and exercise perceived to be reliant on cognitive flexibility predicted greater working memory capacity on the backward span, β = .29, B = .57 (95.CI = [.02, 1.12]), p = .043. The overall model yielded a significant chi-square, 𝜒2(98) = 163.50, p < .001, with alternative fit indices being, RMSEA = .07 (90CI[.05, .09]), CFI = .87, SRMR = .12, LSAR = .49.   5 All p values for parameter estimates within this manuscript are based on unstandardized values (B and not β). This is because p- errors for standardized estimates are obtained through a delta method, which has been cautioned to potentially exacerbate issues related to non-linearity, kurtosis, skewness, etc., and may bias inferential decisions. Historically, concern between using covariances and correlation matrices interchangeably for inferential decisions has been discussed elsewhere (Cudeck, 1989; MacCallum & Austin, 2000).  48  Given that multivariate normality may be an untenable assumption to hold when completing a structural equation model, a Satorra-Bentler (1994) correction was applied here, and for every SEM model reported. Significance of parameter estimates remained largely the same following this correction. Exercise perceived to be reliant on inhibitory control continued to predict a smaller accuracy difference between congruent and incongruent trials on the flanker task, β = -.33, B = -3.30 (95.CI = [-6.13, -.46]), p = .023, and exercise perceived to be reliant on cognitive flexibility continued to predict greater working memory capacity on the backward span, β = .29, B = .57 (95.CI = [.05, 1.09]), p = .033. The one notable change was that exercise perceived to be reliant on cognitive flexibility now predicted poorer inhibitory control (a larger accuracy difference between congruent and incongruent trials on the flanker task), β = .26, B = 2.69 (95.CI = [.21, 5.16]), p = .033.  2.3.2 Leisure Results 2.3.2.1 Measurement Item Statistics  Descriptive statistics for all 30 leisure measurement items are summarized in Table 2.6. Most items had a mean of about 3, with the lowest being 2.58 (item 28) and highest being 3.90 (item 2). Items generally had comparable variance, with each item having a standard deviation of approximately 1. All item categories were endorsed for every item (i.e., all items had a minimum value of 1 and maximum value of 5). A mean composite score was calculated for inhibitory control, cognitive flexibility, and working memory. Items 1 through 10 represented inhibitory control, 11 through 20 cognitive flexibility, and 21 through 30 working memory. Inhibitory control had a mean of 3.14 (SD = .39), cognitive flexibility had a mean of 3.26 (SD = .51), and working memory had a mean of 3.11 (SD = .35). Greater inhibitory control correlated with greater cognitive flexibility, r(175) = .48 (95CI [.36, .59]) p < .001, but did not correlate with  49  working memory, r(175) = .07 (95CI [-.08, .21]) p = .361, and working memory was also not found to correlate with cognitive flexibility r(176) = -.03 (95CI [-.18, .11]) p < .001. 2.3.2.2 Confirmatory Factor Analysis & Measurement Model Identification  Leisure data were next analyzed through the same series of models as completed in the exercise section such to make these two analyses maximally comparable. To do so leisure data were fitted to the original three-factor model, the positive three-factor model, the full two-factor model, and the two-factor model.6 The original three-factor model (Table 2.7) resulted in a significant chi-square, 𝜒2(402) = 971.85, p < .001, and showed poor fit under alternative indices, RMSEA = .09 (90CI[.08, .10]), CFI = .67, SRMR = .10. Correlations among the latent constructs revealed that inhibitory control correlated with cognitive flexibility (r = .88), as well as working memory (r = .48), and that cognitive flexibility correlated with working memory (r = .64). All negative worded items were next removed, and the resulting positive three-factor model (Table 2.8) yielded a significant chi-square, 𝜒2(87) = 216.65, p < .001, and showed generally poor fit on alternative indices, RMSEA = .09 (90CI[.08, .11]), CFI = .87, SRMR = .07. Correlations among the latent constructs revealed that inhibitory control correlated with cognitive flexibility (r = .93), as well as working memory (r = .55), and that cognitive flexibility correlated with working memory (r = .57). Cognitive flexibility and working memory were next combined into a single latent construct, and the resulting full two-factor model (Table 2.9) was found to yield a significant chi-square, 𝜒2(89) = 239.43, p < .001, with alternative indices suggesting comparably poor fit to the positive three-factor model, RMSEA = .10 (90CI[.08, .11]), CFI = .85, SRMR = .08. The correlation between inhibitory control and cognitive  6 There was missing data for one participant who omitted two questions about executive function use during leisure, and for another participant who did not report leisure intensity. Rather than excluding these participants from path analysis and SEM, full information maximum likelihood estimation was used.  50  flexibility was also notably high with a value of .93. Finally, the two-factor model in which three items indicated inhibitory control and seven items indicated cognitive flexibility (summarized in Table 2.11 and illustrated in Figure 2.7) yielded a significant chi-square, 𝜒2(34) = 74.16, p < .001, with alternative indices suggesting good fit, RMSEA = .08 (90CI[.06, .11]), CFI = .95, SRMR = .05. Within this model the correlation between inhibitory control and cognitive flexibility was .90.  2.3.2.3 Path Analysis & Structural Equation Modelling  A path analysis model (shown in Figure 2.8) next tested whether leisure qualifiers in history, duration, intensity, and type (active or passive) were predictive of executive functioning as measured by the flanker task and backward span. Overall, leisure qualifiers were not predictive of executive functioning. The only exception was that leisure of greater intensity predicted greater working memory capacity on the backward span task, β = .16, B = .21 (95.CI = [.01, .41]), p = .035. Among leisure qualifiers, active leisure correlated with more intense leisure. Executive function measures were not found to correlate with one another. Having been largely unsuccessful in predicting performance on laboratory measures of executive functioning with leisure qualifiers, a structural equation model was built that combined this path analysis with the previously identified two-factor leisure model. No variables were predictive of executive function within this SEM, as shown in Figure 2.9. For instance, leisure that is reliant on inhibitory control did not predict an accuracy difference between congruent and incongruent flanker trials, β = .17, B = 1.12 (95.CI = [-3.66, 5.91]), p = .645, and leisure reliant on cognitive flexibility did not predict backward span performance, β = .77, B = 1.00, (95.CI = [-.15, 2.14]), p = .089. For the purposes of the present study, these leisure data were sufficient to demonstrate that the findings for exercise were unique. It was clearly the case that participants  51  self-reports of executive function use during leisure were not predictive of laboratory measures of executive function via the flanker task and backward span. The overall model yielded a significant chi-square, 𝜒2(98) = 257.30, p < .001, with alternative fit indices being, RMSEA = .10 (90CI[.08, .11]), CFI = .82, SRMR = .13, LSAR = .51. Applying a Satorra-Bentler (1994) correction for non-normality did not yield meaningful differences. 2.4 Discussion   Self-reported executive function use during exercise was predictive of better executive functioning as measured by the flanker task and backward span. When participants reported that their primary exercise required greater inhibitory control, they also tended to perform more efficiently on the flanker task as indicated by a smaller accuracy difference between congruent and incongruent trials. When these same participants reported that their primary exercise required high levels of cognitive flexibility, they also tended to have higher estimates of working memory capacity on the backward span. This pair of findings suggests that the positive relationship between exercise and executive function is not uniform across participants or types of exercise, but rather tends to hold for participants who engage in exercise activities that make demands on inhibitory control and cognitive flexibility. Conversely, it implies that participants who do engage in exercise activities not reliant on executive functioning tend to perform more poorly on laboratory tasks of executive functioning. These results were interpreted as support for the cognitive-engagement hypothesis, because exercise reliant on executive functioning was found to predict better executive functioning. These observed correlations between the perceived cognitive demands of one’s exercise and the ability to perform well on laboratory tasks designed to index these cognitive demands represents the greatest novelty of this chapter. In attempting to interpret the underlying  52  mechanism of this result, section 1.7 provided one plausible interpretation. Namely, exercise has been found to increase concentration of neurotrophins like BDNF (Szhuhany, Bugatti & Ottom 2015). This increase in neurotrophins then prompts the brain into a state of general neuroplasticity and neurogenesis from which structural changes can occur. These structural changes help the brain adapt and improve neural efficiency, leading to improved executive functioning (Erickson et al., 2009; Weinstein et al., 2012; Bento-Torres et al., 2019). The present study suggests that this mechanism may be more narrowly fine-tuned via the cognitive engagement inherent in one’s exercise. Within rats for instance, survival of new neurons within the hippocampus is enhanced following completion of behavioural tasks that themselves rely on the hippocampus (Gould, Beylin, Tanapat, Reeves, & Shors, 1999). Neurogenesis and neuroplasticity elicited via exercise might be selectively enhanced and acted upon through the cognitive demands of one’s exercise (Raichlen & Alexander, 2017; Hötting & Röder 2013). The present results are worth considering in light of recent work by Heisz et al. (2017). In this study participants viewed items (e.g., an apple) and later categorized items as new (e.g., a mailbox), old (e.g., an apple), or similar (e.g., an apple but not one that was previously seen) on a subsequent memory test. Participants then completed a second memory test after a 6-week exercise (cycling), exercise and cognitive training (cycling and memory task), or control condition. Both exercise groups went on to perform better on similar trials than did the control group. This result suggests a general benefit of exercise for cognitive functioning. Participants who showed the greatest fitness improvements, also tended to those who exhibited higher BDNF and IGF-1 concentration, replicating past work connecting exercise with increased neurotrophins. Critically however, among participants who made the greatest fitness improvements, performance on similar trials was greater in the combined group over the exercise  53  only group. These results suggest that greater cognitive improvement may occur through a combination of exercise and cognitive training, relative to exercise alone. The present study results are analogous, with the exception that instead of assigning individuals to conditions of varying cognitive engagement, they came from measuring the self-reported cognitive engagement of voluntarily completed exercise.  So far, I have interpreted the results of the present study as implying that cognitively-engaging exercise has a positive influence on executive functioning. But it must be noted that the results could also reflect that people with better executive function simply seek out exercise activities that capitalize on those strengths. Correlational studies of exercise and cognition generally face similar limitations of interpretation (Wang et al., 2013; Jacobson & Matthaeus, 2014; Sakamoto et al., 2018). This is where the leisure data are helpful in pointing to causal directions in the data. There was no evidence in the present study that reports of executive function use during leisure predicted performance on the flanker task or backward span. Leisure qualifiers (history, duration, intensity and type) were also not predictive. These absent correlations suggest that the exercise-dependent correlations we have observed are not simply reflecting a broader tendency for people with certain executive function abilities to seek out activities that are generally demanding on executive function, but rather, that there is something distinct about exercise and executive function. Why might cognitively-engaging exercise, and not leisure activities, predict better performance on laboratory measures of executive functioning? One distinction between exercise and non-exercise leisure, is that exercise uniquely activates the BDNF pathway through which structural changes in the brain may occur. Hence, even if both exercise and leisure are comparable in their reliance on executive functioning, exercise may have an advantage through  54  its direct contributions to neuroplasticity and neurogenesis. Exercise is also beneficial for executive functioning through changes in cerebral blood flow, including research showing that exercise elicits greater activation within the left dorsolateral prefrontal cortex, an area of the brain critical for inhibitory control (Yanagisawa et al., 2010; Hyodo et al., 2016; Byun et al., 2014; MacDonald et al., 2000). Another interpretation worth considering is that exercise and leisure represent different types of cognitive challenge. Note that the most frequently reported leisure activity in the present study was viewing (e.g., watching television), which is not usually very demanding of executive functioning. In contrast, a central aspect of many exercise activities is the pursuit of continual growth, skill refinement, and challenge. This drive is reflected through desires to sprint faster, jog longer, shoot more accurately, lift heavier sets, communicate more effectively with teammates, better conceal intended actions from opposition, and advance to higher tiers of competition. If the hypothesis that cognitively-engaging exercise benefits executive functioning holds true, then, exercise also offers a continual gradient of difficulty through which continued executive function maturation and adaption is made possible. This is not to say leisure that cannot be cognitively engaging, but, that exercise may collectively offer greater opportunity to rehearse executive functioning. Exercise history, duration, intensity and type were not found to predict executive functioning in the present study. One potential reason why these relationships were not seen here, when they have been found in others, is that relatively long periods of exercise may be required for these effects to emerge (Etnier et al., 1997). For example, Pérez et al. (2014) reported that physically active participants had greater inhibitory control (indicated by a smaller reaction time difference between incongruent and congruent trials on a flanker task) than did  55  non-active participants. In this study, active participants were defined as those who had been performing aerobic exercise for at least 10 years, for an average of 6 or more hours per week distributed across at least three days a week. Such results may suggest that exercise history matters for executive functioning, but perhaps more as a lifestyle quality rather than as an activity one performs periodically for only brief periods of time. Although it is also worth noting there are numerous studies finding acute bouts of exercise to benefit executive functioning (O’Leary et al., 2011; Chang et., 2015. Chen et al., 2014). The present results add possible insight to the findings of previous studies in which the relationship between exercise qualifiers and executive functioning are reported to be nonexistent or inconclusive (Barella, Etnier & Chang, 2010; Quaney et al., 2009; Loprinzi & Kane, 2015; Marmeleira, Godinho & Fernandes, 2009). For example, Ho, Gooderham and Handy (2018) asked university students to report their physical activity over the past week before completing a flanker task, and did not find significant correlations between physical activity, regardless of intensity, and flanker task performance. If the relationship between exercise and executive functioning is dependent on individual differences, as is suggested by the present data, then it may be that group studies not finding an effect are capturing circumstances where the measured exercise was not sufficiently cognitively-engaging for the group as a whole. All of the present study results are premised on the two-factor measurement model of executive function use during exercise. Within this measurement model there were two latent factors: inhibitory control and cognitive flexibility. It is worth stressing that this model was identified through a data-driven approach, and hence a few broader implications are worth considering. First, reverse worded items were found to be unhelpful, and thought likely to have elicited statistical noise more than useful measurement (DiStefano & Motl, 2006; Woods, 2006;  56  Zhang, Noor & Savalei, 2016). Following this, the majority of removed items from the original 30 were those intended to measure working memory. This suggests that how working memory had been measured was largely ineffective. The two working memory items that remained, on review, may have had stronger association with cognitive flexibility and the potential to load onto this construct. Item 25 (“connect and combine different ideas”) could be expected to rely on working memory as one must concurrently hold multiple ideas such to connect and combine them, whereas the original identification of these ideas may require viewing a problem from varying angles and perspectives, and thereby positing cognitive flexibility. Similarly, item 23 (“Make predictions about what will happen next”) may rely on cognitive flexibility such to identify what future scenarios may be possible, while working memory than judges and evaluates these possibilities given on-going awareness of current events.  Among the excluded working memory items, ambiguity likely played a role in their weaker loadings and contribution to poor model fit. For instance, question 21 asked individuals if they pay attention to “many” things at the same time when exercising. What constitutes “many” however is left vague and up to the reader rather than providing a consistent baseline from which participants may judge the question. Rephrasing this question may help alleviate this problem (e.g., “When doing my primary exercise, I pay attention to more than one thing at a time”). A separate account for why working memory was difficult to measure, may be that working memory does not loan itself well to self-reporting. The ability to accurately report one’s use of executive functioning may itself be correlated with executive functioning ability, thereby leading to some of the noise thought captured by the measurement model. The inhibitory control items that remained within the final measurement model, in hindsight, appear more concrete in their description relative to those that were dropped. Items 1  57  and 7 for instance asked individuals if a specific event tends to occur when they exercise (item 1: “slow down to avoid making mistakes”; item 7: “pause and double check what I am doing”). This presumably allows individuals to make a simpler judgement and response for these behaviours. Item 3 although a bit more abstract does describe (in a way that is fairly common within the executive function literature) how individuals are likely to resolve competing streams of perceptual information (item 3: “filter and ignore distracting information”). These questions can be contrasted to those that were dropped. Item 5 asks individuals if during their primary exercise they practice “self-control” and “discipline”. The use of these terms leaves open various interpretations as to what either may actually mean. Discipline for instance might be taken to mean slowing down, ensuring accuracy, not being distracted, dedication to exercise adherence, or a more general sense of task completion. Without greater context, these terms are ambiguous. A similar stance could be made for what constitutes “sudden” within question item 9.           58  Measurement Item Mean SD 1. slow down to avoid making mistakes 2.88 0.97 2. care most about speed and performing quickly (R) 3.28 1.05 3. filter and ignore distracting information 3.50 0.91 4. decide what to do through impulse alone (R) 3.26 1.01 5. practice self-control and discipline 3.74 0.93 6. follow every action to completion (R) 2.23 0.84 7. pause and double check what I am doing 2.96 1.08 8. start and complete actions without thinking (R) 3.10 0.98 9. anticipate making fast or sudden adjustments 3.20 1.09 10. act without self restraint (R) 3.62 0.91 11. adapt and change how things are done 3.50 0.88 12. have a plan that I stringently follow (R) 2.74 0.96 13. try to identify new techniques or strategies 3.52 0.97 14. follow the same routine (R) 2.64 0.91 15. practice creativity 2.83 1.11 16. hold the same mindset from start to finish (R) 2.79 0.94 17. encounter and solve new problems 3.06 1.08 18. have little-to-no flexibility to modify what I do (R) 3.73 0.97 19. rely on a diverse skillset 3.11 1.11 20. think the same thoughts over and over (R) 3.01 0.93 21. pay attention to many things at the same time 3.12 1.09 22. understand everything that is happening even when absent-minded (R) 2.68 0.90 23. make predictions about what will happen next 3.52 0.97 24. avoid time-keeping (R) 3.00 1.12 25. connect and combine different ideas 3.17 1.07 26. do not care about the order in which things happen (R) 3.09 1.01 27. multitask 2.84 1.22 28. disengage from what is happening around me (R) 2.84 0.98 29. monitor what is happening on a second-to-second basis 2.96 1.05 30. focus all my attention entirely on one thing before moving onto the next (R) 2.78 1.04 Table 2.1. Mean and standard deviation values for all exercise measurement items. Items were rated on a 5-point scale ranging from “Never” to “Always”. Reverse worded items are denoted with an R. Each item was prefaced with the prompt “When engaging in my primary exercise I”. Inhibitory control, cognitive flexibility and working memory were intended to be measured through items 1-10, 21-20, and 21-30 respectively.       59  Measurement Item I. C. W. 1. slow down to avoid making mistakes .38a   2. care most about speed and performing quickly (R) -.20   3. filter and ignore distracting information .59   4. decide what to do through impulse alone (R) -.11   5. practice self-control and discipline .54   6. follow every action to completion (R) -.63   7. pause and double check what I am doing .50   8. start and complete actions without thinking (R) .05   9. anticipate making fast or sudden adjustments .30   10. act without self restraint (R) -.08   11. adapt and change how things are done  .65a  12. have a plan that I stringently follow (R)  -.28  13. try to identify new techniques or strategies  .66  14. follow the same routine (R)  .29  15. practice creativity  .67  16. hold the same mindset from start to finish (R)  -.19  17. encounter and solve new problems  .69  18. have little-to-no flexibility to modify what I do (R)  .29  19. rely on a diverse skillset  .70  20. think the same thoughts over and over (R)  -.04  21. pay attention to many things at the same time   .33a 22. understand everything that is happening even when absent-minded (R)   -.28 23. make predictions about what will happen next   .60 24. avoid time-keeping (R)   -.36 25. connect and combine different ideas   .72 26. do not care about the order in which things happen (R)   -.08 27. multitask   .17 28. disengage from what is happening around me (R)   -.04 29. monitor what is happening on a second-to-second basis   .36 30. focus all my attention entirely on one thing before moving onto the next (R)   -.34 𝜒2                      RMSEA                    CFI                SRMR               LSAR               𝜔             978.80**     .10 (90CI[.09, .11])             .50                   .12                     .42                 .58 Table 2.2. Confirmatory factor analysis of the 30 exercise measurement items. Bolded items denote statistical significance (p < .05). The factors are titled such that I = Inhibition, C = Cognitive Flexibility, and, W = Working memory. a Indicator item       60  Measurement Item I. C. W. 1. slow down to avoid making mistakes .52a   3. filter and ignore distracting information .54   5. practice self-control and discipline .50   7. pause and double check what I am doing .60   9. anticipate making fast or sudden adjustments .11   11. adapt and change how things are done  .66a  13. try to identify new techniques or strategies  .64  15. practice creativity  .67  17. encounter and solve new problems  .70  19. rely on a diverse skillset  .70  21. pay attention to many things at the same time   .35a 23. make predictions about what will happen next   .59 25. connect and combine different ideas   .71 27. multitask   .21 29. monitor what is happening on a second-to-second basis   .34 𝜒2                      RMSEA                    CFI                SRMR               LSAR               𝜔             218.06**     .10 (90CI[.09, .12])             .78                   .11                     .41                 .84 Table 2.3. Confirmatory factor analysis of positively worded exercise items. Bolded items denote statistical significance (p < .05). The factors are titled such that I = Inhibition, C = Cognitive Flexibility, and, W = Working memory. a Indicator item  Measurement Item I. C/W. 1. slow down to avoid making mistakes .49a  3. filter and ignore distracting information .56  5. practice self-control and discipline .49  7. pause and double check what I am doing .57  9. anticipate making fast or sudden adjustments .19  11. adapt and change how things are done  .65a 13. try to identify new techniques or strategies  .64 15. practice creativity  .68 17. encounter and solve new problems  .71 19. rely on a diverse skillset  .71 21. pay attention to many things at the same time  .31 23. make predictions about what will happen next  .57 25. connect and combine different ideas  .72 27. multitask  .19 29. monitor what is happening on a second-to-second basis  .34 𝜒2                      RMSEA                    CFI                SRMR               LSAR               𝜔       225.50**     .10 (90CI[.09, .12])             .77                   .10                     .39                 .84 Table 2.4. Confirmatory factor analysis of full-two exercise factor model in which cognitive flexibility and working memory have been collapsed into a single latent construct. Bolded items denote statistical significance (p < .05). The factors are titled such that I = Inhibition, C = Cognitive Flexibility, and, W = Working memory. a Indicator item   61  Measurement Item I. C. 1. slow down to avoid making mistakes .60a  3. filter and ignore distracting information .47  7. pause and double check what I am doing .70  11. adapt and change how things are done  .63a 13. try to identify new techniques or strategies  .64 15. practice creativity  .72 17. encounter and solve new problems  .70 19. rely on a diverse skillset  .70 23. make predictions about what will happen next  .54 25. connect and combine different ideas  .74 𝜒2                      RMSEA                    CFI                SRMR               LSAR               𝜔       44.83           .05 (90CI[.00, .08])             .97                   .06                     .13                 .84 Table 2.5. Confirmatory factor analysis for the final set of exercise items. This model is referred to as the two-factor model. Bolded items denote statistical significance (p < .05). The factors are titled such that I = Inhibition, and C = Cognitive Flexibility. a Indicator item                     62  Measurement Item Mean SD 1. slow down to avoid making mistakes 2.89 1.15 2. care most about speed and performing quickly (R) 3.90 0.96 3. filter and ignore distracting information 3.34 0.94 4. decide what to do through impulse alone (R) 3.13 1.00 5. practice self-control and discipline 3.08 0.99 6. follow every action to completion (R) 2.66 1.07 7. pause and double check what I am doing 3.03 1.20 8. start and complete actions without thinking (R) 3.09 1.06 9. anticipate making fast or sudden adjustments 2.93 1.15 10. act without self restraint (R) 3.38 1.03 11. adapt and change how things are done 3.35 1.05 12. have a plan that I stringently follow (R) 3.22 1.07 13. try to identify new techniques or strategies 3.26 1.25 14. follow the same routine (R) 3.01 0.95 15. practice creativity 3.38 1.30 16. hold the same mindset from start to finish (R) 3.05 1.00 17. encounter and solve new problems 3.21 1.24 18. have little-to-no flexibility to modify what I do (R) 3.81 0.98 19. rely on a diverse skillset 3.02 1.18 20. think the same thoughts over and over (R) 3.28 1.00 21. pay attention to many things at the same time 3.44 1.00 22. understand everything that is happening even when absent-minded (R) 2.69 0.93 23. make predictions about what will happen next 3.72 0.96 24. avoid time-keeping (R) 2.76 1.09 25. connect and combine different ideas 3.67 0.94 26. do not care about the order in which things happen (R) 3.02 1.09 27. multitask 3.50 1.12 28. disengage from what is happening around me (R) 2.58 0.95 29. monitor what is happening on a second-to-second basis 2.83 1.11 30. focus all my attention entirely on one thing before moving onto the next (R) 2.85 1.02 Table 2.6. Mean and standard deviation values for all leisure measurement items. Items were rated on a 5-point scale ranging from “Never” to “Always”. Reverse worded items are denoted with an R. Each item was prefaced with the prompt “When engaging in my primary leisure I”. Inhibitory control, cognitive flexibility and working memory were intended to be measured through items 1-10, 21-20, and 21-30 respectively.        63  Measurement Item I. C. W. 1. slow down to avoid making mistakes .72a   2. care most about speed and performing quickly (R) -.25   3. filter and ignore distracting information .34   4. decide what to do through impulse alone (R) .07   5. practice self-control and discipline .70   6. follow every action to completion (R) -.72   7. pause and double check what I am doing .70   8. start and complete actions without thinking (R) .13   9. anticipate making fast or sudden adjustments .45   10. act without self restraint (R) .04   11. adapt and change how things are done  .77a  12. have a plan that I stringently follow (R)  -.59  13. try to identify new techniques or strategies  .85  14. follow the same routine (R)  -.02  15. practice creativity  .75  16. hold the same mindset from start to finish (R)  -.25  17. encounter and solve new problems  .79  18. have little-to-no flexibility to modify what I do (R)  -.04  19. rely on a diverse skillset  .84  20. think the same thoughts over and over (R)  -.12  21. pay attention to many things at the same time   .37a 22. understand everything that is happening even when absent-minded (R)   -.56 23. make predictions about what will happen next   .40 24. avoid time-keeping (R)   -.18 25. connect and combine different ideas   .62 26. do not care about the order in which things happen (R)   -.30 27. multitask   .12 28. disengage from what is happening around me (R)   -.19 29. monitor what is happening on a second-to-second basis   .47 30. focus all my attention entirely on one thing before moving onto the next (R)   -.46 𝜒2                      RMSEA                    CFI                SRMR               LSAR               𝜔             971.85**     .09 (90CI[.08, .10])             .67                   .10                     .41                 .61 Table 2.7. Confirmatory factor analysis of the 30 leisure measurement items. Bolded items denote statistical significance (p < .05). The factors are titled such that I = Inhibition, C = Cognitive Flexibility, and, W = Working memory a Indicator item       64  Measurement Item I. C. W. 1. slow down to avoid making mistakes .71a   3. filter and ignore distracting information .34   5. practice self-control and discipline .67   7. pause and double check what I am doing .65   9. anticipate making fast or sudden adjustments .48   11. adapt and change how things are done  .77a  13. try to identify new techniques or strategies  .85  15. practice creativity  .76  17. encounter and solve new problems  .78  19. rely on a diverse skillset  .84  21. pay attention to many things at the same time   .42a 23. make predictions about what will happen next   .41 25. connect and combine different ideas   .62 27. multitask   .15 29. monitor what is happening on a second-to-second basis   .45 𝜒2                      RMSEA                    CFI                SRMR               LSAR               𝜔             216.65**     .09 (90CI[.08, .11])             .87                   .07                     .39                 .89 Table 2.8. Confirmatory factor analysis of positively worded leisure items. Bolded items denote statistical significance (p < .05). The factors are titled such that I = Inhibition, C = Cognitive Flexibility, and, W = Working memory a Indicator item  Measurement Item I. C/W. 1. slow down to avoid making mistakes .71a  3. filter and ignore distracting information .34  5. practice self-control and discipline .67  7. pause and double check what I am doing .65  9. anticipate making fast or sudden adjustments .48  11. adapt and change how things are done  .77a 13. try to identify new techniques or strategies  .85 15. practice creativity  .76 17. encounter and solve new problems  .78 19. rely on a diverse skillset  .84 21. pay attention to many things at the same time  .21 23. make predictions about what will happen next  .15 25. connect and combine different ideas  .44 27. multitask  -.03 29. monitor what is happening on a second-to-second basis  .27 𝜒2                      RMSEA                    CFI                SRMR               LSAR               𝜔       239.43**     .10 (90CI[.08, .11])             .85                   .08                     .46                 .87 Table 2.9. Confirmatory factor analysis of positively worded leisure items after cognitive flexibility and working memory items were collapsed into a single construct. Bolded items denote statistical significance (p < .05). The factors are titled such that I = Inhibition, C = Cognitive Flexibility, and, W = Working memory a Indicator item   65  Measurement Item I. C. 1. slow down to avoid making mistakes .74a  3. filter and ignore distracting information .31  7. pause and double check what I am doing .67  11. adapt and change how things are done  .76a 13. try to identify new techniques or strategies  .86 15. practice creativity  .77 17. encounter and solve new problems  .77 19. rely on a diverse skillset  .84 23. make predictions about what will happen next  .13 25. connect and combine different ideas  .45 𝜒2                      RMSEA                    CFI                SRMR               LSAR               𝜔       74.16**       .08 (90CI[.06, .11])             .95                   .05                     .24                 .89 Table 2.11. Confirmatory factor analysis for the final set of leisure items. This model is hereafter referred to as the two-factor model. Bolded items denote statistical significance (p < .05). The factors are titled such that I = Inhibition, and C = Cognitive Flexibility  a Indicator item       66  Figure 2.1 Frequencies of self-reported primary exercise. Figure 2.2 Frequencies of self-reported primary leisure 67      Figure 2.3 Performance on the flanker task (left, accuracy; middle reaction time) and backward span (right, capacity) among exercise (top half) and leisure (bottom half) participants. 68   Figure 2.4 CFA of the two-factor exercise model.  Figure 2.5 Exercise path analysis model.  Figure 2.6 Exercise SEM.  69   Figure 2.7 CFA of the two-factor leisure model.  Figure 2.8 Leisure path analysis model.  Figure 2.9 Leisure SEM.  70  Chapter 3: Study 2 3.1 Introduction  The main finding of chapter 2 was that participants who characterized their primary exercise as relying on executive functioning also tended to perform more efficiently on laboratory tasks of executive functioning via the flanker task and backward span. Chapter 3 sought to build on these results in three ways. First, the two-factor structure of executive function use during exercise that was identified in chapter 2, was tested in a solely confirmatory framework in chapter 3 (i.e., novel model re-specification was not undertaken). This was done by providing individuals the 15 positively worded items from the original 30 measurement items, and testing the two-factor model identified in chapter 2. Although two-factor model requires only 10 of these 15 items, in the event that the two-factor model was found to fit poorly, the remaining items were recorded for potential insight and comparison purposes. Second, trail making test B was used as hybrid measure of cognitive flexibility and working memory, instead of using the backward span. This was an attempt to align the laboratory measurement tool with the hybrid cognitive flexibility-working memory latent factor identified in chapter 2. Third, the stop-signal task was used as a measure of inhibitory control, instead of the flanker task, in an effort to capture the motor and action requirements inherent in many sport activities and exercises. The next section offers a brief overview of these two executive function tasks such to introduce and contextualize these measures. The stop-signal task measures a form of inhibitory control that is sometimes called action cancellation. During this task participants are seated in front of a computer and complete trials where a target appears and must be then identified (e.g., reporting whether a circle appears on the left or right side of a computer screen). On a minority of these trials a second stimulus (e.g., a  71  central circle) appears after the target and indicates that all responses must be immediately stopped. Experiments typically vary the temporal lag between the “Go” and “Stop” stimuli until a threshold level of stopping accuracy is achieved (e.g., 50%). This threshold temporal lag is then used (and potentially slightly shifted) to estimate a response time measure of inhibitory control called the stop-signal reaction time. Among studies using this task in sport and exercise research are two by Padilla et al. (2013, 2014) where participants who regularly exercised exhibited greater inhibitory control (i.e., faster stop-signal reaction time) on a strategic version stop-signal task when compared to passive participants. Wang et al. (2013) also found that the stop-signal task differentiated athletes based on environment type. Dynamic sport athletes (i.e., tennis players) showed greater inhibitory control on the stop-signal task than did static sport athletes (i.e., swimming) and non-athletes. A set of studies by Verburg et al. (2014, 2016) further found that the stop-signal task differentiated inhibitory control among expert and novice soccer players, whereas the flanker task did not.  The second executive function task used in chapter 3 is called trail making test B, and is a commonly used measure of cognitive flexibility and working memory (Reitan, 1958; Hobert et al., 2011; Crowe 1998; Sánchez-Cubillo et al., 2009). To complete this task participants are given a sheet of paper on which are dispersed numbers and letters. Participants then use a pen to connect these numbers and letters in alternating and ascending order (i.e., 1-A-2-B-3-C-etc.), and time to complete this task is measured. If a mistake is made (e.g., drawing a line to C instead of 3), an experimenter informs the participant of this error, and a correction must be applied (e.g., returning to point B and restarting). Errors made during trail making are often thought to be captured through this penalization process that results in slower task completion time. Trail making B completion time has been found to differentiate athletes of greater skill from those of  72  lesser skill (Han et al., 2011), has been found to improve following a session of exercise (Murray & Russoniello, 2012; Harveson et al., 2016), and is sensitive to varying exercise intensities (Tierney et al., 2010).   Initial research suggested that error frequency is not an important metric to record for trail making tests. Klusman, Cripe and Dodrill (1989) for instance found that while trail making completion time discriminated control subjects from those who had experienced a closed head-injury, total number of errors did not. Clinical studies since however have found that trail making errors are predictive of frontal brain-lesions (Stuss et al., 2001) and various mental health illnesses and conditions, including schizophrenia (Mahurin et al., 2006) as well as dementia (Ashendorf, 2008; Rasmusson et al., 1998). Amongst this research Amieva et al. (1998) found that individuals with dementia of the Alzheimer Type (DAT) required more time to complete trail making B, and, made more errors compared to control participants. The authors suggested that committing an error on this task may indicate a lapse of inhibitory control, as individuals must alternate between numbers and letters (1-A-2-B-3-C) rather than on an overlearned tendency to connect numbers and letters in serial order (1-2-3-4 and A-B-C-D).  An additional reason why trail making errors may traditionally have received less focus than completion time, is because of their somewhat subjective nature. When for instance does an experimenter decide that an error has been made – can a subject for instance dart in the wrong direction, or must they fully connect an ongoing line to the wrong target. Similarly, this error identification process relies on an experimenter (or multiple experimenters) remaining vigilant, having an unobstructed view of task progression, consistently giving the same feedback instructions, and holding the same standard as to what does and does not constitute an error across participants. All of this is possible, but might entail one reason why the exact number of  73  errors is often not emphasized when the trail making test is used. The present study works around many of these issues by creating a computerized version of trail making test B (as opposed to the traditional pen-and-paper set up) wherein participants use a mouse to click from one target (e.g., 3) to the next (e.g., C). In doing so, all errors were automatically recorded, were defined by the same criteria (i.e., any click 20 pixels away or more from the current correct target), and feedback was given such that a connecting line from one target to the next only appeared following a correct click. In doing so, the present study measured both trail making completion time and number of errors committed by participants. The general goals of chapter 3 are similar to those to chapter 2. First, I determined whether the two-factor model of executive function use during exercise (identified in chapter 2) yielded good fit. Second, I tested whether the exercise qualifiers of history, duration, intensity, and type, were predictive of performance on a laboratory measures of executive functioning. Third, I combined goals 1 and 2, by creating a structural equation model in which exercise reliant on executive functioning serves to predict performance on laboratory measures of executive functioning. The central hypotheses concerning this SEM model were that greater inhibitory control during exercise was expected to predict faster stop-signal reaction time and fewer trail making errors, and exercise higher in cognitive flexibility was expected to predict faster trail making completion time. Just as in chapter 2, a parallel investigation was completed by having individuals self-report on their leisure activities, and for these leisure models, no a priori hypotheses were generated. 3.2 Method 3.2.1 Power Analysis   74  Having identified a tenable two-factor model measuring executive function use during exercise within chapter 2 a power analysis was conducted to estimate an appropriate CFA sample size for chapter 3. Given that a sizeable portion of recruited participants in chapter 2 indicated they did not engage in any exercise, or had only engaged in exercise and/or leisure for less than 1 month, caution have to be placed over estimating power through a single sample size quantity. To address this a minimal and maximal sample size was determined in order to establish acceptable lower and upper bounds for power. Following the procedures of MacCallum et al. (1996), a sample size of 200 resulted in an estimated power of .69, given α = .05, 𝜀𝑜 = 0.07, 𝜀𝑎 = .10 and df = 34. Increasing this sample size to 300 resulted in an estimated power of .86. Given these values, a sample size of approximately 300 participants was recruited, and all subsequent analysis were considered to have adequate power if they were based on between 200 and 300 cases.  Chapter 2 also identified the SEM architecture through which executive functioning outcomes were predicted. This SEM architecture was expected to be used again in chapter 3, and so, was used to estimate power for all individual model parameters via simulation. Using a sample size of 250 (midway between the minimal and maximal bounds of acceptable sample sizes) suggested that all regression paths predicting executive function outcomes would have at least approximately .70 power, and coverage falling between 94-95%.7  7 Simulations were run using the R package simsem (Pornprasertmanit, Miller, & Schoemann, 2015). The population model was specified such that all factor loadings had a standardized value of .65 and correlations among latent variables were .70. Latent constructs (i.e. inhibitory control and cognitive flexibility) were specified to predict executive functioning via a standardized beta value of .25, with the remaining predictors (i.e. exercise/leisure type, history, duration and intensity) having a value of .15. All correlations among exercise (and leisure) variables in history, duration, intensity and type were set to .30. The correlation between trail making completion time and stop-signal reaction time was set to .40, and the correlation between trail making completion time and trail making errors was set to .70. This larger correlation was estimated given that studies often suggest errors made during trial making are encapsulated through a longer completion time thereby partially equating the two metrics. A total of 20,000 replications were completed, all of which converged, and none of which identified Heywood cases, correlations greater than 1, or other commonly encountered identification issues.  75  3.2.2 Participants   All participants were recruited through the University of British Columbia human subject pool, following review and approval of the research plan by the University Behavioral Research Board (H18-03515). A total of 299 participants completed the study, and among these 243 reported exercising. All participants completed the study within an hour and were provided 1 course credit. As in chapter 2, participants with a history less of than 1 month for either their primary exercise or leisure, were excluded from analysis. Among subjects who reported exercising, 228 reported a history of more than one month with their primary exercise, and 291 reported a history of more than one month with their primary leisure.  Participants who exercised were on average 20.40 years old (SD = 3.12), the majority were women (77.63%) and most identified as East Asian (45.18%), followed by European/Caucasian (26.32%), Indian-South (10.96%), Other (10.09%), Middle Eastern (2.63%), Latin American (2.19%), Native American (1.32%), and African (.88%). The most frequently reported exercise was weight training (16.67%), followed by running (9.65%) and jogging (8.33%), see Figure 3.1 for a further summary and classification of exercise types (static or dynamic). Participants reported an average exercise history of 4.25 (SD = 1.19) on a 5-point Likert scale, which generally suggested a starting date of more than 7 months ago. The average exercise frequency was 3.14 (SD = 1.28) on a 5-point Likert scale, which corresponded to roughly 60 to 90 minutes of exercise per week. The average exercise intensity was 3.20 (SD = .66) on a 4-point Likert scale, thereby suggesting moderate exercise intensity.  Participants who reported a primary leisure (most of whom had also reported exercising) were on average 20.39 years old (SD = 2.97), the majority were women (79.38%), and most identified as East Asian (50.52%), followed by European/Caucasian (21.99%), Indian-South  76  (11.34%), Other (9.62%), Middle Eastern (2.41%), Latin American (2.06%), Native American (1.03%), and African (.69%)8. The most frequently reported leisure activity was viewing (e.g., television; 20.96%), followed by listening to music (15.81%) and reading (12.37%), see Figure 3.2 for a further summary and classification of leisure activities (active or passive). Participants reported an average leisure history of 4.75 (SD = .76) on a 5-point Likert scale, which generally suggested a starting date of more than 9 months ago. The average leisure frequency was 3.42 (SD = 1.34) on a 5-point Likert scale, which fell between response options 60 to 90 minutes, and 90 to 120 minutes per week. The average leisure intensity was 1.99 (SD = .82) on a 4-point Likert scale, thereby suggesting low leisure intensity. 3.2.3 Procedure   Chapter 3 had an identical procedure to chapter 2 with two exceptions. One, participants now completed half as many measurement items about executive function use during exercise and/or leisure, because all negatively worded items were removed (DiStefano & Motl, 2006; Woods, 2006). The two-factor measurement model, as found in chapter 2, was then directly evaluated on the basis of various fit indices, these being the chi-square significance test, RMSEA, CFI and SRMR. Just as in chapter 2, the LSAR and coefficient omega are also reported as indicators of local misfit and construct reliability. The second exception is that cognitive testing now consisted of completing the stop-signal task and trail making test B, with task order again being randomized across participants.  The stop-signal task measures inhibitory control by requiring cancellation of an already initiated action (Logan & Cowan, 1984). In this task participants responded to the appearance of  8 Six participants selected not to provide their age, and one participant selected not to provide their ethnicity, these values are omitted among the exercise and leisure sample descriptive statistics.  77  a stimulus, and sometimes shortly after this a second stimulus appeared thereby requiring immediate cessation of all responses. On these “Stop” trials, if no response is made the participant has successfully demonstrated inhibitory control, whereas if a response is made, this suggests inhibitory control was insufficient for response cessation. Various recommended guidelines were adopted when creating the present study stop-signal task (Matzke, Verbruggen & Logan, 2018; Verbruggen et al., 2019) including the use of a standard two-choice discrimination task, instructions encouraging quick responding, an initial set of “Go” trials without “Stop” signals, use of a salient stop-signal, the presence of stop-signals on a minority of trials, feedback to discourage intentional slowing, and the use of multiple stop-signal delays to calculate an average stop-signal reaction time. Outside of these more general recommendations, the present stop-signal task was modelled after Wang et al., (2013), Muggleton et al. (2010) and Hsu et al. (2011). A more detailed description of the stop-signal task is provided below along with the accompanying visual reference in Figure 3.3.  Participants first completed a set of 50 “Baseline” trials in which they initiated and completed a “Go” response absent of “Stop” signals. To complete this task participants responded to the appearance of a white circle on a black computer screen. If the circle appeared on the left side of the screen participants were instructed to press “Z”, and if the circle appeared on the right side of the screen participants were required to press “/”. Participants were instructed to respond as quickly and accurately as possible. After making a response a fixation cross would appear for 500ms followed by a blank screen for 200ms, after which the next target appeared. On completion of these 50 “Baseline” trials, a screen appeared informing participants what their average reaction time and accuracy had been. This set of trials introduced participants to the task,  78  normalized fast responding, and allowed for an estimate of average reaction time absent of stop-signals.  Participants next completed a “Calibration” block, in which were 24 “Go” trials and 8 “Stop” trials. Participants were informed that on a minority of these trials after the appearance of the target (again a left or right white circle) a central white circle would appear. The appearance of this central circle meant that all responses had to be immediately halted, and that any keyboard response at this point would be incorrect. Two seconds after the appearance of this stop-signal, or if participants made a keyboard response, the next trial would begin. Participants were encouraged to perform as quickly and as accurately as possible, even though on a minority of trials they would be required to immediately stop all responses. During the first block of “Calibration” trials the time interval between the “Go” signal and “Stop” signal, called the stop-signal delay (SSD), was set to 170ms. This means that if participants were completing a “Stop” trial, the stop-signal would appear 170ms after the target.  To discourage intentional slowing that might occur in anticipation of a stop-signal, performance on “Go” trials (trials without a stop-signal), had to be faster than a prespecified threshold value. This threshold value was the average reaction time that a participant achieved during the “Baseline” trials, plus three standard deviations. Performing slower than this threshold prompted a message to appear that requested participants perform the task as quickly and accurately as possible. This message remained on screen for 1750ms, after which the next trial would automatically begin.   Participants were provided feedback after each block of “Calibration” trials. This feedback included average reaction time and accuracy, as well as a message stating the next set of trials would be easier (or harder). This message remained on screen for 25 seconds. Whether  79  the next block of trials was made easier or harder depended on how successfully participants had inhibited their “Go” response given a “Stop” signal in the prior block. If the probability of making a “Go” response given a “Stop” signal, i.e., 𝑝(𝐺𝑜 | 𝑆𝑡𝑜𝑝), was greater than 67.5% than the next block of trials was made easier via lessening the stop-signal delay by 40ms (e.g., 170ms minus 40ms). Alternatively, if the probability of making a “Go” response given a “Stop” signal was less than 32.5%, the next block of trials was made harder by elongating the stop-signal delay by 40ms (e.g., 170ms plus 40ms). If the probability of making a “Go” response given a “Stop” signal fell between the values of 67.5% and 32.5%, the same stop-signal delay was used in the subsequent block of trials. The probability of making a “Go” response given a “Stop” signal has elsewhere been called the noncancelled error rate, a term that hereafter is used (Wang et al., 2013).  By shifting the stop-signal delay in increments of 40ms, a noncancelled error rate of approximately .50 was sought for each subject. The underlying rational is that the noncancelled error rate will be 0 when the stop signal occurs early enough and 1 when the stop signal occurs late enough, and so the median noncancelled error rate (.50) can be attained through shifting of the stop-signal delay. The stop-signal delay that results in a noncancelled error rate of approximately .50 is called the critical stop-signal delay. The “Calibration” phase ended when participants achieved a noncancelled error rate of approximately .50 (between 67.5% and 32.5%) for two consecutive blocks and in doing so their critical stop-signal delay was attained. On average participants completed 5.58 “Calibration” blocks (SD = 3.14) before meeting this requirement and completing this phase. The average critical stop-signal delay for the final two blocks during this phase was 109.93ms (SD = 66.92). This means that on average, participants  80  needed a stop-signal delay of 109.93ms such to achieve a noncancelled error rate of approximately .50.  The final phase of the stop-signal task is the “Test” phase and consisted of 108 “Go” trials and 36 “Stop” trials. The “Stop” trials were subdivided such that 12 were at the critical stop-signal delay (obtained for each participate in the prior “Calibration” phase), 12 were at this critical delay minus 40ms, and 12 were at this critical delay plus 40ms. On completion of these trials a stop-signal reaction time can be calculated by obtaining the noncancelled error rate for a given delay (critical, plus, and minus), identifying the associated percentile within the “Go” reaction time distribution, and subtracting from this reaction time the associated stop-signal delay. For instance, if a participant completed trials with a critical stop-signal delay of 100ms, and attained a noncancelled error rate of .55, this suggests that 55% of the time their “Stop” reaction time was too slow to halt their “Go” reaction time. To calculate the stop-signal reaction time, the 55th percentile of the “Go” reaction time distribution is first identified from the 108 “Go” trials this participant completed, and this reaction time is then subtracted by the associated stop-signal delay (100ms). Continuing this example, let us say that 400ms corresponds to the 55th percentile within the “Go” reaction time distribution for this participant, the stop-signal reaction time would thus be 400ms minus the critical stop-signal delay of 100ms, resulting in a stop-signal reaction time of 300ms. This approach can more generally be defined as 𝑆𝑆𝑅𝑇 =𝑛𝑡ℎ𝐺𝑜 𝑅𝑒𝑎𝑐𝑡𝑖𝑜𝑛 𝑇𝑖𝑚𝑒 − 𝑆𝑡𝑜𝑝 𝑆𝑖𝑔𝑛𝑎𝑙 𝐷𝑒𝑙𝑎𝑦. Because multiple stop-signal delays were used during the “Test” phase (critical, minus 40 and plus 40), 𝑆𝑆𝑅𝑇𝐴𝑣𝑒𝑟𝑎𝑔𝑒 is calculated by taking the  81  mean of 𝑆𝑆𝑅𝑇𝐶𝑟𝑖𝑡𝑖𝑐𝑎𝑙, 𝑆𝑆𝑅𝑇𝑀𝑖𝑛𝑢𝑠 and 𝑆𝑆𝑅𝑇𝑃𝑙𝑢𝑠.9 See Figure 3.4 for a visual illustration of  𝑆𝑆𝑅𝑇𝐴𝑣𝑒𝑟𝑎𝑔𝑒.  Figure 3.5 summarizes the mean reaction times for exercise and leisure participants during “Baseline”, “Calibration” and “Test” trials. As expected, reaction times were fastest during “Baseline” trials (in which a stop-signal never appeared), followed by trials in which a stop-signal appeared but participants still responded, and slowest on trials where a stop-signal could appear but did not. Similarly, and as found in other studies, (e.g., Wang et al., 2013) the noncancelled error rate during the “Test” trials was found to increase as the stop-signal delay was elongated.  In addition to the stop-signal task participants completed trail making test B (Reitan, 1958). Trail making test B is a commonly used pen-and-paper neuropsychological task that measures cognitive flexibility and working memory (Hobert et al., 2011; Crowe 1998; Sánchez-Cubillo et al., 2009). In the present study a computerized version of this task was constructed and employed. On starting this task participants were informed they would be connecting spatially dispersed numbers and letters in alternating order (e.g., 1-A-2-B-3-C) as they appeared on a computer screen. Each trial began with the appearance of the number “1”, and immediately after clicking on this number, the remaining numbers and letters appeared on the computer screen. This was done to ensure that completion time of the task was coupled with initial exposure to all test stimuli. Participants were then required to move their mouse and click on the next target (e.g., after clicking on number 1, the next target is the letter A). After a correct click, a line  9 During the calibration phase, 32 subjects were only able to attain a noncancelled error rate between 32.5% and 67.5% when the critical delay was between 0 - 10ms.  These subjects went on to have an identical stop-signal delay (i.e. 0) underlying 𝑆𝑆𝑅𝑇𝐶𝑟𝑖𝑡𝑖𝑐𝑎𝑙  and 𝑆𝑆𝑅𝑇𝑀𝑖𝑛𝑢𝑠. 18 subjects also had noncancelled error rates of 100% for a given delay (e.g. 𝑆𝑆𝑅𝑇𝑃𝑙𝑢𝑠), in these cases the mean of the remaining delays was used to calculate 𝑆𝑆𝑅𝑇𝐴𝑣𝑒𝑟𝑎𝑔𝑒, e.g., (𝑆𝑆𝑅𝑇𝐶𝑟𝑖𝑡𝑖𝑐𝑎𝑙 + 𝑆𝑆𝑅𝑇𝑀𝑖𝑛𝑢𝑠)/ 2  82  would connect the prior target (1) with the current target (A), and participants then began searching for next target (2). All numbers and letters were written in monospace font Consolas (font size 40), and participants had to click within a 20-pixel radius of the current target to qualify as a correct click. After being provided written instructions participants completed one practise trial where the final target letter was “D”.    Participants completed a total of 5 trials of trail making B. Target and letter locations for these 5 test trials were randomly determined with the restraint that a current or upcoming target was never obscured (e.g., crossed by a line and/or another letter or number). All participants completed the same 5 test trials presented in the same order. On each of these trials the final target letter was “M”, that when clicked on would prompt the next trial with the appearance of the number “1”. Completion time for these 5 trials was averaged to calculate mean trail making completion time. The total number of errors made (defined as any click outside a 20-pixel radius from the current target) was also recorded and summed across trials to calculate an error rate.   Average trail making completion time and errors made across the 5 trials is summarized in Figure 3.6 for exercise and leisure participants. Past studies have shown considerable variation in trail making completion time, in part likely stemming from methodological constraints (e.g., an experimenter having to detect and inform participants when they committed a mistake). Generally, however average completion time for the current study (approximately 60 seconds per trail) was comparable to completion times reported elsewhere (e.g., Woods, Wyma, Herron & Yund, 2015). Though it should also be noted that comparing completion times across studies might be limited as studies might use a different maximum of letters/numbers (e.g., a task where the final target is L versus one where the final target is M), and in the case of a computerized version, different screen resolutions.    83   All cognitive tasks were created using the Matrix Laboratory (MatLab), and specifically the Psychophysis Toolbox (Brainard, 1997; Kleiner, Brainard & Pelli, 2007). On completion of both cognitive tasks, participants provided basic demographic information on their age, gender, and ethnicity. 3.3 Results 3.3.1 Exercise Results  A confirmatory factor analysis was completed that mimicked the final measurement model identified in chapter 2. Within this model were two latent constructs titled inhibitory control (indicated with 3 items) and cognitive flexibility (indicated by 7). This model (Figure 3.7) yielded a significant chi-square, 𝜒2(34) = 112.01, p < .001, showed mediocre-to-acceptable-fit, RMSEA = .10 (90CI[.08, .12]), CFI = .91, SRMR = .07, LSAR = .20, and retained high reliability (𝜔 = .89). Mean and standard deviation values for individual items are summarized in Table 3.1. Path analysis next tested whether executive functioning, as measured by performance on the stop-signal task and trail making B, could be predicted by commonly measured exercise qualifiers in history, duration, intensity, and type. Exercise qualifiers were allowed to freely correlate with one another, and performance on the stop-signal task correlated freely with performance on trail making B (completion time and errors). Because this model was saturated no fit indices are reported (Figure 3.8). Exercise qualifiers were largely unpredictive of executive functioning, with one exception being that longer exercise history predicted slower trail making completion time, β = .19, B = 2.25 (95.CI = [.70, 3.79]), p = .004. Correlations among the predictor variables showed that dynamic exercise correlated with longer exercise duration and intensity, longer exercise  84  history correlated with greater exercise duration and intensity, and longer exercise duration also correlated with greater intensity. Correlations among executive functions revealed that faster stop-signal reaction time correlated with faster trail making completion time and fewer trail making errors. A correlation was also found such that more errors made during trail making associated with slower test completion time. The magnitude of this correlation (r = .46) suggested that while these two constructs demonstrated considerable association, they were also distinguishable.  Having been largely unsuccessful in predicting performance on laboratory measures of executive function with exercise qualifiers, a structural equation model was built that combined this path analysis with the prior two-factor exercise model. The resulting model is depicted in Figure 3.9. Exercise perceived to be reliant on inhibitory control was found to predict faster stop-signal reaction time, β = -.26, B = -18.12 (95.CI = [-33.70, -.2.55]), p = .023 and fewer trail making errors, β = -.27, B = -11.57 (95.CI = [-21.44, -1.70]), p = .022, but did not predict faster trail making completion time, β = -.21, B = -4.55 (95.CI = [-9.14, .05]), p = .052. Conversely, exercise perceived to be reliant on cognitive flexibility predicted slower stop-signal reaction time, β = .20, B = 13.63 (95.CI = [.39, 26.87]), p = .044, and slower trail making completion time, β = .29, B = 6.05 (95.CI = [2.07, 10.04]), p = .003, but was not predictive of trail making errors, β = .16, B = 6.90 (95.CI = [-.1.46, 15.25]), p = .106. Participating in dynamic exercise was also predictive of faster stop-signal reaction time, β = -.15, B = -17.72 (95.CI = [-33.96, -1.49]), p = .032, and faster trail making completion time, β = -.16, B = -5.90 (95.CI = [-10.75, -1.06]), p = .017. Longer exercise history continued to predict slower trail making completion time, β = .16, B = 1.94 (95.CI = [.42, 3.46]), p = .013. The overall model yielded a significant chi-square, 𝜒2(98) = 304.68, p < .001, with alternative fit indices being, RMSEA = .10  85  (90CI[.08, .11]), CFI = .82, SRMR = .13, LSAR = .48. Applying a Satorra-Bentler (1994) correction for non-normality did not greatly change significance of parameter estimates, except exercise perceived to be reliant on inhibitory control was no longer predictive of trail making errors, β = -.27, B = -11.57 (95.CI = [-24.31, 1.16]), p = .075. 3.3.2 Leisure Results  Leisure data were given the same data analytic approach as the exercise data. The two-factor leisure model (Figure 3.11) yielded a significant chi-square, 𝜒2(34) = 110.22, p < .001, showed acceptable-to-good fit on alternative fit indices, RMSEA = .09 (90CI[.07, .11]), CFI = .95, SRMR = .05, LSAR = .28, and retained high reliability (𝜔 = .91). Mean and standard deviation values for individual items are summarized in Table 3.2. A path analysis next tested whether leisure qualifiers in history, duration, intensity and type were predictive of executive functioning as measured by the stop-signal task and trail making B (Figure 3.12). No variables in this model were predictive of executive functioning. Correlations among leisure qualifiers showed that active leisure correlated with shorter leisure history and duration, as well as greater leisure intensity. Correlations among executive functions showed that faster stop-signal reaction time correlated with faster trail making completion time and fewer trail making errors, and that slower trail making completion time correlated with more trail making errors. Having thus far been unsuccessful in predicting performance on laboratory measures of executive functioning with leisure qualifiers, a structural equation model was built that combined this path analysis with the prior two-factor leisure model. No variables in this model (Figure 3.13) were found to predict executive functioning. Among the paths with the largest standardized  86  beta coefficients, leisure perceived to be reliant on inhibitory control did not predict trail making completion time, β = -.21, B = -2.78 (95.CI = [-6.27, .71]), p = .119, or trail making errors, β = -.15, B = -4.04 (95.CI = [-10.99, 2.90]), p = .254, and leisure perceived to be reliant on cognitive flexibility did not predict trail making completion time, β = .12, B = 1.66 (95.CI = [-1.80, 5.11]), p = .348. The overall model had a significant chi-square, 𝜒2(98) = 402.39, p < .001, with alternative fit indices being, RMSEA = .10 (90CI[.09, .11]), CFI = .83, SRMR = .16, LSAR = .63. Applying a Satorra-Bentler (1994) correction for non-normality did not greatly change these results. 3.4 Discussion  The main finding of chapter 3 was that self-reports of executive function use during exercise were predictive of executive functioning, as measured in the lab by the stop-signal task and trail making B. As hypothesized, participants who self-reported that their exercise was reliant on inhibitory control tended to have faster stop-signal reaction times and made fewer trail making errors. Exercise reliant on inhibitory control thus appeared to quicken one’s ability to cancel an already initiated action, and reduced impulsivity to committing erroneous actions. These results were interpreted as support for the cognitive-engagement hypothesis, because exercise reported to rely on executive functioning was found to benefit executive functioning. However, not all results supported this hypothesis, and some actually ran in the opposite direction. Notably, participants who self-reported that their exercise was reliant on cognitive flexibility tended to have slower stop-signal reaction time and slower trail making completion time. These participants were less effective in cancelling an already initiated action, and required more time to complete a visual search task reliant on alternating between number and letter sets.   87  Why might exercise perceived to be reliant on greater cognitive flexibility predict worse executive functioning? In addressing this, it is worth considering that cognitive flexibility reflects a diverse subset of cognitive skills (Ionescu, 2012), some of which might be beneficial for task performance and others deleterious. Cognitive flexibility is in part defined by the ability to efficiently shift from one mindset or pattern of thinking to another. This aspect of cognitive flexibility could be expected to increase the speed of trail making performance by allowing one to rapidly alternate between number-sets (1-2-3) and letter sets (A-B-C) when searching for the currently appropriate target (1-A-2-B-3-C). However, cognitive flexibility also represents whether one practises innovation and creativity. These other aspects of cognitive flexibility might be expected to slow trail making completion time as an individual may examine the dispersion of numbers and letters on the computer screen and attempt to identify a hidden pattern or underlying structure. Depending on which aspect of cognitive flexibility had been refined through exercise, faster or slower trail making completion time might be expected. This is admittedly only speculation at this point.  Whether an individual exercises within a structured or unstructured environment may underlie beneficial changes in cognitive flexibility (Fardilha & Allen, 2020). Greco, Memmert and Morales (2010) assigned youth basketball players to a traditional, highly structured training condition (e.g., isolated activities to develop dribbling, passing and throwing skills), or to a condition that emphasized unstructured, play-orientated situations (e.g., 3 players competing against 4). Participants then were tested on various game situations (e.g., attackers were required to pass the ball among themselves without allowing defenders to take possession) and their performance was rated by judges. Judges rated participants on their tactical intelligence (reliant on convergent thinking) as well as their tactical creativity (reliant on divergent thinking). Higher  88  ratings for both these outcomes were obtained by participants in the unstructured condition, suggesting that such exercise may be better suited to promoting creativity and innovativeness. The relationship between exercise and improved inhibitory control may also be more consistent than the relationship between exercise and other executive functions. Two recent review articles found that HIIT (Hsieh et al., 2020) and resistance training (Soga et al., 2018) were both more consistently linked to improved inhibitory control than cognitive flexibility and working memory. In one study, Liu-Ambrose et al. (2010) compared older adult women assigned to either weekly resistance training or to balance and tone training. After 12-months, participants in both groups showed comparable cognitive flexibility (assessed via trail making) and working memory (assessed via digit span), but participants assigned to resistance training showed greater inhibitory control than those that completed balance and tone training (via lesser Stroop task interference). When considering exercise qualifiers (history, duration, intensity, and type), participation in dynamic exercise was found to predict faster stop-signal reaction time and trail making completion time. This replicates the findings of Wang et al. (2013) who found that athletes from dynamic sports (tennis) had faster stop-signal reaction time than athletes from static sports (swimmers) and nonathletes. Dynamic sport is characterized by an environment that is fast-paced and requires reacting to ongoing gameplay, thereby providing opportunities to refine action cancellation ability. Action cancellation in static sports by contrast is likely less important of a skill to master and refine, as these sports are characterized by an environment that is relatively stable and consistent. This granted however, it is important to note these are generalizations that while often applicable, leave open the potential for individual differences as found in the present study.  89  Longer exercise history was found to predict slower trail making completion time. This finding is in opposition to studies showing that longer exercise history is beneficial for executive functioning (Pérez et al., 2014; Padilla et al., 2013; Padilla et al., 2014). One potential interpretation is that exercise, if completed for a long period of time, may become stale and repetitive. Exercise that is undertaken without challenging one’s ability to problem-solve or without inviting any change in perspective, might reinforce a more rigid and routine mindset. A separate account may again suggest that inhibitory control is more sensitive to improvement following exercise relative to other executive functions (Hsieh et al., 2020; Soga et al. 2018; Liu-Ambrose et al., 2010). Chapter 3 confirmed two key findings first reported in chapter 2. First, the relationship between exercise and executive function was not found to be uniform, but rather it varied as a function of individual differences in perceived exercise experience. Second, leisure qualifiers (history, duration, intensity and type) and whether one’s leisure was perceived to be reliant on executive functioning (inhibitory control or cognitive flexibility) again did not predict performance on laboratory measures of executive functioning. These results, taken collectively, suggest that exercise (and not leisure) is specifically important for executive functioning at the level of individual differences.  As in chapter 2, the results of chapter 3 are grounded within the measurement model of executive function use during exercise. A data-driven approach was completed to identify this initial model in chapter 2. Chapter 3 then tested this model with a new set of data, and it is therefore worthwhile to consider where these two models diverged. When considering inhibitory control items, standardized loadings across both models were highly similar (chapter 2: .60, .47, and .70, chapter 3: .61, .44, and .75). Items measuring cognitive flexibility varied more, but were  90  also largely comparable (chapter 2: .63, .64, .72, .70, .70, .54, and .74, chapter 3: .69, .72, .71, .83, .79, .59, and .71). On a slightly grander level, inhibitory control and cognitive flexibility items yielded average standardized loadings of .59 and .67 in chapter 2, which were comparable to the values of .60 and .72 of chapter 3. These average values suggest that while there is some measurement fluctuation between the two models, the differences are within a small range. The two-factor exercise model identified in chapter 2 did however include a larger correlation between the latent constructs inhibitory control and cognitive flexibility (.68) relative to the two-factor model of the present chapter (.59). This likely in part contributed to the better fit of the model found in chapter 2.   Concerning fit indices more directly, the two-factor exercise model identified in chapter 2 had undeniably better fit than the two-factor exercise model of chapter 3. A degradation of fit may be expected to some extent, as the two-factor model identified in chapter 2 was deliberately made through a data-driven approach specific to that dataset, whereas the model in chapter 3 served as a confirmatory test of this novel measurement model. Fit indices did not point to especially poor model fit within chapter 3, but, they were also not supportive of especially strong model fit either, leaving open an interpretation that might best be described as promising or tentatively mediocre. It is important also to note that although the two-factor model was designed with exercise in mind, the two-factor leisure model had comparable, if not better, fit than the two-factor exercise model. Taken collectively, these results point to a measurement model for which further research is needed.    91  Measurement Item Mean SD 1. slow down to avoid making mistakes 3.01 1.07 2. filter and ignore distracting information 3.75 1.01 3. pause and double check what I am doing 3.23 1.13 4. adapt and change how things are done 3.54 0.99 5. try to identify new techniques or strategies 3.68 1.09 6. practice creativity 2.90 1.18 7. encounter and solve new problems 2.98 1.20 8. rely on a diverse skillset 3.15 1.26 9. make predictions about what will happen next 3.34 1.20 10. connect and combine different ideas 3.12 1.21 Table 3.1. Mean and standard deviation values for all exercise measurement items. Items were rated on a 5-point scale ranging from “Never” to “Always”. Each item was prefaced with the prompt “When engaging in my primary exercise I”. Inhibitory control was measured via items 1 through 3, and cognitive flexibility items 4 through 10.  Measurement Item Mean SD 1. slow down to avoid making mistakes 2.70 1.22 2. filter and ignore distracting information 3.49 0.96 3. pause and double check what I am doing 2.92 1.23 4. adapt and change how things are done 3.04 1.20 5. try to identify new techniques or strategies 3.10 1.30 6. practice creativity 3.26 1.34 7. encounter and solve new problems 2.85 1.30 8. rely on a diverse skillset 2.74 1.35 9. make predictions about what will happen next 3.32 1.14 10. connect and combine different ideas 3.53 1.02 Table 3.2. Mean and standard deviation values for all leisure measurement items. Items were rated on a 5-point scale ranging from “Never” to “Always”. Each item was prefaced with the prompt “When engaging in my primary exercise I”. Inhibitory control was measured via items 1 through 3, and cognitive flexibility items 4 through 10.    92   Figure 3.1 Frequencies of self-reported primary exercise.   Figure 3.2 Frequencies of self-reported primary leisure.   93   Figure 3.3 The stop-signal task used in chapter 3. In phase 1 participants completed 50 “Go” trials, without the appearance of a “Stop” signal. In phase 2 participants completed blocks of 24 “Go” trials and 8 “Stop” trials. The first block of trials had a stop-signal delay of 170ms, and, this value was adjusted by 40 milliseconds until the noncancelled error rate fell between 32.5% and 62.5% for two consecutive blocks. The stop-delay at which this occurred is the critical delay. In phase 3 participants completed 108 “Go” trials and 36 “Stop” trials. “Stop” trials were distributed such that 12 occurred given the critical delay as calculated in phase 2, 12 were the critical delay minus 40 milliseconds and 12 were the critical delay plus 40 milliseconds.   Figure 3.4 The stop-signal reaction time calculated for three different stop-signal delays. 𝑆𝑆𝑅𝑇𝐴𝑣𝑒𝑟𝑎𝑔𝑒 is then taken as the mean of 𝑆𝑆𝑅𝑇𝑃𝑙𝑢𝑠, 𝑆𝑆𝑅𝑇𝐶𝑟𝑖𝑡𝑐𝑎𝑙, and 𝑆𝑆𝑅𝑇𝑀𝑖𝑛𝑢𝑠.   94     Figure 3.5 Stop-signal task reaction times (left) and non-cancelled error rates (right) among exercise (top) and leisure (bottom) participants. Go reaction times, as well as Go | No Stop reaction times were calculated only using trials where the target was correctly identified. Only the final two blocks of calibration trials were used in calculating mean values.    95    Figure 3.6 Trail making completion time (left) and error rates (right) among exercise (top) and leisure (bottom) participants across 5 trails of trail making B.     96   Figure 3.7 CFA of the two-factor exercise model.  Figure 3.8 Exercise path analysis model.  Figure 3.9 Exercise SEM model.  97   Figure 3.11 CFA of the two-factor leisure model.  Figure 3.12 Leisure path analysis model.  Figure 3.13 Leisure SEM model.  98  Chapter 4: Study 3 4.1 Introduction  Chapters 2 and 3 found that performance on laboratory measures of executive functioning were predicted by self-reported executive function use during exercise. The participants in both of these studies were university undergraduate students, which represents a relatively homogeneous sample in terms of age, gender, stage in life, family resources, and educational achievement, among other variables. University students tend to be younger, more educated, and more resourceful than the broader population (Henrich, Heine & Norenzayan, 2010) and reside within an environment meant to emphasize education, learning, mental stimulation and cognitive growth (Campanholo et al, 2014; Hamdan & Hamdan, 2009; Tombaugh, 2004; Wu et al., 2013). University students hence might reflect a population that is likely to exhibit already peak levels of executive functioning from which exercise might only offer modest benefit.   Chapter 4 sought to expand the diversity of the participant population by recruiting individuals through Amazon’s Mechanical Turk. Mechanical Turk is an online platform that individuals register with voluntarily in order to complete human intelligence tasks (HITS) for financial compensation. Notably, use of online platforms like Mechanical Turk has become increasingly popular amongst researchers seeking to recruit samples that represent greater diversity than the typical undergraduate student population (Sheehan, 2018). Mechanical Turk represents a dynamic population of individuals, because some individuals are regularly joining while others are leaving the platform. As a result, Mechanical Turk demographics are likely to exhibit shifts over time and differ across studies employing various inclusion and exclusion criteria. Studies that have set out to capture the demographics of  99  Mechanical Turk report that participants are largely from the United States, and among United States participants, most are female, are older than 30, commonly have at least a bachelors degree, and often report a lower income than the general population (Barger et al., 2011; Ipeirotis, 2010; Ross, Zaldivar, Irani, & Tomlinson, 2010). Mechanical Turk users from the United States although more diverse than a college sample, have also been reported to be younger and more educated than the broader United States population (Sheehan, 2018; Stewart, Chandler & Paolacci, 2017).   Numerous studies have assessed the replicability, reliability, and overall quality of data obtained through Mechanical Turk in comparison to traditional university subject pools. Crump, McDonnell and Gureckis (2013) for instance had Mechanical Turk participants complete a series of common cognitive tasks (e.g., flanker task, Stroop task, attentional blink) and found that the standard effects definitive of these tasks, as found in laboratory studies, tended to replicate. Hauser and Schwarz (2016) also reported that Mechanical Turk participants perform comparably, or better, on attention checks than do university undergraduates. Further research has suggested that Mechanical Turk participants exhibit high test-retest reliability (Holden, Dennie, & Hicks, 2013), and when compared to university students yield comparable psychometric properties on individual difference measures (Stewart, Chandler & Paolacci, 2017; Behrend et al., 2011). Although there exist numerous studies reporting favorably on Mechanical Turk, others have noted various limitations concerning the platform. For instance, Mechanical Turk participants been found to score higher on social desirability than university students (Behrend et al., 2011), and results by Chmielewski and Kucker (2020) have more recently suggested that data quality appears to be declining (e.g., lower reliability). One strategy adopted in the current study comes from Peer, Vosgerau and Acquisti (2014) who found that participants  100  with an approval rating greater 95% were more likely to pass attention check questions and had higher reliability on various scales than those with lower approval ratings.  Executive functioning in chapter 4 was measured through the flanker task and backward span. The rationale behind this was three-fold. First, these tasks would permit a direct replication of the measures used in chapter 2 through a more diverse participant sample. Second, chapter 2 yielded clearer results than chapter 3 with regard to the relationship between exercise and executive function. Third, the flanker task and backward span lend themselves more readily to remote online testing than does trail making B and the stop-signal task. For example, the flanker and backward span both present visual materials to the center of a screen, and thereby are less sensitive to the diversity of the display conditions for remotely-tested participants. Questionnaires and cognitive tasks in chapter 4 were created through JavaScript and specifically the jsPsych library (de Leeuw, 2015; de Leeuw & Motz, 2016). Research has suggested that cognitive tasks run online and through JavaScript are comparable to more traditional in lab studies using software like E-Prime and MatLab, although reaction times might be slightly delayed within the range of 25ms (de Leeuw & Motz, 2015; Hilbig, 2016; Pinet et al., 2017; Crump et al., 2013).  The guiding hypothesis in chapter 4 remained the same as in the previous two chapters.  The main goal was to see whether, under a broader demographic of study participants, perceived executive functioning use during exercise predicted better performance on laboratory measures of executive functioning. The more specific hypothesis was to replicate chapter 2, such that exercise perceived to be reliant on inhibitory control predicted more efficient inhibitory control on the flanker task, and that exercise perceived to be reliant on cognitive flexibility predicted better working memory capacity on the backward span. No relationship was hypothesized to  101  underlie leisure activity, though in light of the leisure results obtained in chapters 2 and 3, it would seem reasonable to again expect null leisure findings. 4.2 Method 4.2.1 Participants  Following the power analysis results from chapter 3, a similar sample size of 300 participants was sought for the present study. All participants were recruited through Amazon’s Mechanical Turk, following review and approval of the research plan by the Behavioral Research Board of the University of British Columbia (H18-03515). Participants self-enrolled in this study by selecting the accompanying human intelligence task (HIT) listed on the Mechanical Turk online interface. To view this study participants were required to have an average approval rating greater than 95% (Peer, Vosgerau & Acquisti, 2014) and had to reside within Canada or the United States of America. Eligible and interested participants were briefly informed that the study involved exercise and leisure habits and their relation to cognition, and that the study required a desktop or laptop for completion (i.e., not tablet, mobile phone, etc.). Participants prior to beginning the study were required to voluntarily sign a consent form. A total of 297 participants completed the study, although four subjects were found to have completed the study twice and their data (for both the first and second completion) was excluded from analysis thereby reducing the total sample size from 297 to 289. Among these 289 participants, 264 reported exercising. Following data filtering steps, the final sample size for participants reporting a primary leisure was 255, and 225 for participants reporting a primary exercise. Average completion time for the study was approximately 30 minutes and all subjects were compensated with $3.00.  102  Participants who reported exercising were on average 37.24 years old (SD = 11.18), and the majority were men (64%), followed by women (35.56%), with the remainder of participants selecting not to specify their gender (or indicated that their gender was not captured in the available response categories). Most participants were college or university graduates (48%), followed by some college or university experience (24%), completion of high school (13.33%), a master degree or doctoral degree (12%), some graduate or professional school experience (2.22%), and some high school experience (.44%). Most participants reported being European or Caucasian (75.11%), followed by African (7.11%), Latin-American (7.11%), East Asian (4.89%), Native American (2.22%), Other (2.22%) and Indian-South (1.33%). The most frequently reported exercise was walking (22.67%), followed by weight training (18.67%) and running (11.11%), see Figure 4.1 for a further summary and classification of exercise types (static or dynamic). Participants reported an average exercise history of 4.52 (SD = .95) on a 5-point Likert scale, which generally suggested a starting date of more than 7 months ago. The average exercise frequency was 2.95 (SD = 1.20) on a 5-point Likert scale, which suggested 60 to 90 minutes of exercise per week. The average exercise intensity was 2.98 (SD = .64) on a 4-point Likert scale, thereby suggesting moderate exercise intensity.  Participants who reported a primary leisure (most of whom had also reported exercising) were on average 37.08 years old (SD = 11.13), the majority were men (61.57%), followed by women (37.65%), with the remainder of participants selecting not to specify their gender (or indicated that their gender was not captured in the available response categories). Most participants were college or university graduates (47.45%), followed by some college or university experience (25.10%), completion of high school (14.12%), a master degree or doctoral degree (10.20%), some graduate or professional school experience (2.75%), and some high  103  school experience (.39%). Most participants reported being European or Caucasian (76.08%), followed by African (7.06%), Latin-American (5.88%), East Asian (4.71%), Other (2.75%), Native American (1.96%) and Indian-South (1.57%). The most frequently reported leisure was gaming (22.75%), followed by reading (18.04%) and viewing (e.g., television; 12.55%), see Figure 4.2 for a further summary and classification of leisure activities (active or passive). Participants reported an average leisure history of 4.82 (SD = .57) on a 5-point Likert scale, which generally suggested a starting date of more than 9 months. The average leisure frequency was 3.75 (SD = 1.20) on a 5-point Likert scale, which fell between response options 60 to 90 minutes, and 90 to 120 minutes per week. The average leisure intensity was 1.98 (SD = .78) on a 4-point Likert scale, thereby suggesting low leisure intensity. 4.2.2 Procedure   The procedure for the chapter 4 was identical to chapters 2 and 3 with the following exceptions. (1) Participants completed the study online rather than within a laboratory, (2) all tasks were created using JavaScript and the jsPsych library (de Leeuw, 2015; de Leeuw & Motz, 2016) rather than MatLab and the Psychophysics Toolbox (Brainard, 1997; Kleiner, Brainard & Pelli, 2007), (3) in the present chapter negatively worded measurement items were excluded, as they had been in chapter 3, (4) during cognitive testing additional examples and practise trials were provided to ensure participants had further feedback and instruction given the absence of an experimenter, (5) participants were instructed and requested to only rely on their memory when completing the backward span (and not to write anything down or use any external tool), (6) on study completion participants were provided a dialogue box should they seek to provide optional feedback, and (7) on study completion participants were provided a randomized code to then enter into the Mechanical Turk interface signifying that they have completed the study.   104  Figure 4.3 summarizes average flanker task performance for exercise and leisure participants. Both groups demonstrated the expected flanker effect where incongruent trials are completed more slowly and less accurately than congruent trials. Relative to chapter 2 reaction times were slightly slower, though within the expected range of a 25ms or so delay (de Leeuw & Motz, 2015). Figure 4.3 also summarizes backward span performance among exercise and leisure participants, and depicts the expected trend of high recollection for trials with few digits (3 and 4), followed by recollection of around 4 or so digits on harder trials (5 through 9; Cowan, 2010). The filtering process used in chapter 2 was adopted here, with one important exception. Just as in chapter 2, trials on the flanker task were excluded if responses were shorter than 300ms or longer than 1500ms (Chen, Zhao, Fan & Chen, 2018; van Leeuwen et al., 2007; White, Ratcliff, & Starns, 2011; White, Brown, & Ratcliff, 2012). Within chapter 2 this disqualified less than 3% of all trials. The majority of Mechanical Turk data too fell within his range (e.g., most subjects had 180 or more valid trials of the total 200). However, a small number of participants following this step had 20 or fewer valid trials, including some with fewer than 5. A decision was made such that participants were included for analysis only if they had 50 or more valid trials (i.e., 25% of trials) on the grounds that (1) approximately 25 trials are needed to reasonably estimate a mean response time in a condition (congruent, incongruent), and (2) shortly after this cut-off the number of valid trials per subject rapidly accelerated to comparable quantities as in chapter 2. This filtering process resulted in 22 subjects being excluded from analysis. With regard to the other exclusion criteria, 7 participants were removed for having flanker task accuracy below 51% and 2 were removed for having backward span accuracy below 14%. No participant reported the same activity for exercise and leisure, although one participant reported  105  “nothing” as their primary leisure and accordingly was excluded from leisure analysis. This resulted in a final sample of 255 participants reporting a history of more than 1 month with their primary leisure, and 225 participants reporting a history of more than 1 month with their primary exercise. 4.3 Results 4.3.1 Exercise Results  A confirmatory factor analysis was completed that mimicked the final measurement model identified in chapter 2. Within this model were two latent constructs titled inhibitory control (indicated with 3 items) and cognitive flexibility (indicated by 7 items). This two-factor model (Figure 4.4) yielded a significant chi-square, 𝜒2(34) = 70.65, p < .001, showed good fit under alternative indices, RMSEA = .07 (90CI[.05, .09]), CFI = .96, SRMR = .05, LSAR = .12, and retained high reliability (𝜔 = .90). Mean and standard deviation values for individual items are summarized in Table 4.1. Path analysis next tested whether inhibitory control and working memory, as measured by performance on the flanker task and backward span, could be predicted by commonly measured exercise qualifiers in history, duration, intensity, and type (dynamic or static). Exercise qualifiers were freely allowed to correlate with one another, and performance on the flanker task freely correlated with performance on the backward span. Because this model is saturated no fit indices are reported (Figure 4.5). Exercise qualifiers were not found to predict executive functioning. Correlations among the predictor variables revealed that dynamic exercise had greater intensity, and longer exercise history correlated with longer exercise duration. Among executive function outcomes, a smaller reaction time difference between incongruent and  106  congruent trials on the flanker task correlated with a smaller accuracy difference between congruent and incongruent trials, as well as greater backward span capacity.  Having thus far been unsuccessful in predicting performance on laboratory measures of executive functioning through exercise qualifiers, a structural equation model was built that combined this path analysis with the two-factor exercise model. This model is shown in Figure 4.6. Exercise perceived to be reliant on inhibitory control predicted poorer inhibitory control on the flanker task (i.e., a larger accuracy difference between congruent and incongruent trials), β = .31, B = 4.73 (95.CI = [1.18, 8.28]), p = .009, and exercise perceived to be reliant on cognitive flexibility predicted worse working memory capacity on the backward span, β = -.22, B = -.41 (95.CI = [-.81, -.00]), p = .047. No other variables were predictive of executive functioning. The overall model yielded a significant chi-square, 𝜒2(98) = 205.66, p < .001, with alternative fit indices being, RMSEA = .07 (90CI[.06, .08]), CFI = .91, SRMR = .08, LSAR = .40. Applying a Satorra-Bentler (1994) correction for non-normality did not meaningfully change these results.  4.3.2 Leisure Results  Leisure data were given the same data analytic approach as the exercise data. The two-factor leisure model (Figure 4.7) yielded a significant chi-square, 𝜒2(34) = 87.23, p < .001, showed good fit under alternative fit indices, RMSEA = .08 (90CI[.06, .10]), CFI = .96, SRMR = .04, LSAR = .17, and retained high reliability (𝜔 = .91). Mean and standard deviation values for individual items are summarized in Table 4.2. A path analysis next tested whether leisure qualifiers in history, duration, intensity and type (active or passive) were predictive of performance on laboratory measures of inhibitory control (flanker task) and cognitive flexibility (backward span). The results of this model are  107  shown in Figure 4.8. Longer history with leisure predicted greater working memory capacity on the backward span, β = .14, B = .28 (95.CI = [.03, .53]), p = .027, as well as greater inhibitory control on the flanker task as indicated by a smaller accuracy difference between congruent and incongruent trials, β = -.24, B = -5.28 (95.CI = [-8.01, -2.54]), p < .001, and a smaller reaction time difference between incongruent and congruent trials, β = -.20, B = -11.57 (95.CI = [-18.72, -4.42]), p = .002. Leisure rated as being more intense also predicted worse inhibitory control on the flanker task (i.e., a larger accuracy difference between congruent and incongruent trials), β = .16, B = 2.59 (95.CI = [.61, 4.57]), p = .010. Correlations among the predictor variables showed that active leisure tended to be more intense, and that longer leisure history correlated with greater leisure duration and lesser intensity. A final structural equation model was built that combined this path analysis with the two-factor leisure model (Figure 4.9). Just as in the path analysis model, longer history with leisure predicted greater working memory capacity on the backward span, β = .15, B = .29 (95.CI = [.04, .54]), p = .023, as well as greater inhibitory control via a smaller accuracy difference between congruent and incongruent flanker trials, β = -.23, B = -5.10 (95.CI = [-7.82, -2.39]), p < .001, and a smaller reaction time difference between incongruent and congruent flanker trials, β = -.20, B = -11.23 (95.CI = [-18.36, -4.10]), p = .002. Greater intensity leisure also predicted worse inhibitory control via a larger accuracy difference between congruent and incongruent flanker trials, β = .13, B = 2.11 (95.CI = [.15, 4.07]), p = .035. Latent constructs were unpredictive of executive functioning, as for instance leisure perceived to be reliant on cognitive flexibility did not predict backward span capacity, β = -.02, B = -.03 (95.CI = [-.32, .26]), p = .820, and leisure perceived to be reliant on inhibitory control did not predict an accuracy difference between congruent and incongruent trials on the flanker task, β = .13, B = 1.81 (95.CI = [-1.12, 4.74]), p =  108  .226. The overall model had a significant chi-square, 𝜒2(98) = 293.34, p < .001, with alternative fit indices being, RMSEA = .09 (90CI[.08, .10]), CFI = .87, SRMR = .12, LSAR = .45. Following a Satorra-Bentler (1994) correction, longer history with leisure no longer predicted working memory capacity, β = .15, B = .29 (95.CI = [-.02, .60]), p = .070, but continued to significantly predict greater inhibitory control via a smaller accuracy difference between congruent and incongruent flanker trials, β = -.23, B = -5.10 (95.CI = [-10.14, -.06]), p = .047, as well as a smaller reaction time difference between incongruent and congruent trials, β = -.20, B = -11.23 (95.CI = [-19.79, -2.67]), p = .010. Leisure perceived to be reliant on inhibitory control was also no longer predictive of an accuracy difference on the flanker task, β = .13, B = 2.11 (95.CI = [-.05, 4.27]), p = .056. 4.4 Discussion  The main finding of chapter 4 was that self-reported executive function use during exercise was predictive of less efficient performance on laboratory measures of executive functioning. Less efficient here refers to the findings that participants who reported greater executive function use during exercise also exhibited worse inhibitory control on the flanker task, and displayed lower working memory capacity on the backward span. These are the primary exercise results of chapter 4, and notably, they run counter to both the cognitive- engagement hypothesis as well as the results of chapter 2, where exercise reported to rely on executive function predicted more efficient flanker task and backward span performance. Hence any discussion of chapter 4 results must be directly contrasted with the results of chapter 2.   There are numerous important methodological differences between chapters 2 and 4, with the most pronounced being the nature of the participants. Chapter 2 was based on a sample of university undergraduate students, which is typical of most studies in the background literature.  109  Participants in chapter 4 by contrast were recruited via Amazon’s Mechanical Turk in order to obtain a more diverse and less homogenous sample relative to university undergraduates. A comparison of these two samples indicated that participants who reported exercising in chapter 4 were nearly twice the age of participants in chapter 2 (means of 37.08 and 20.43 years, respectively), and the variability in their ages was substantially greater (standard deviations of 11.13 and 2.02, respectively). Most participants in chapter 4 also reported being men, Caucasian, and had college or university degrees. In contrast, participants in chapter 2 were mostly women, identified primarily as East Asian, and all were undergraduate students enrolled in at least one psychology course. Participants were also distributed geographically in different ways. Chapter 4 participants were distributed across locations in the United States and Canada, whereas chapter 2 participants were all geographically centered near the University of British Columbia, Canada. Differences in study compensation were also such that chapter 4 participants were compensated with $3.00 and chapter 2 participants 1 course credit. Of the many demographic differences between these two samples, two in particular are worth considering: gender and age. Studies on exercise and executive functioning do not typically break down their results by sex. To help remedy this, Colcombe and Kramer (2003) reported a meta-analysis in which they coded studies based on whether most of the sample was male or female. Samples in which more than half of participants were female tended to show a more pronounced effect of exercise on cognition than did samples where most participants were male. Numerous studies also suggest that men tend to have faster reaction times than do women, but, that this difference dissipates when comparing men and women who regularly exercise (Lum, Enns & Pratt, 2002; Silverman, 2006; Alves et al., 2013). Hence women might potentially be expected to show a larger beneficial effect of exercise on executive functioning.  110   Running counter to the above interpretation are studies finding that neurotrophins differ between men and women in response to exercise. Schmidt-Kassow et al. (2012) found that while men and women showed comparable BDNF concentration during baseline and exercise-recovery phases, men as a group showed a higher BDNF peak in response to exercise. In their review Szuhany, Bugatti and Otto (2015) also reported that studies with more women tend to show smaller BDNF changes. Differences between men and women regarding BDNF are further complicated as some research suggests men show greater neurotrophin fluctuation than do women. A study by Piccinni et al. (2008) for instance found that among men, plasma BDNF concentration is highest in the morning and steadily decreases throughout the day, whereas among women BDNF concentration was more consistent and varied less throughout the day. This may suggest that there are different periods of time for men and women during which cognitively-engaging exercise might be more beneficial or expected to exert a positive effect. The second demographic characteristic that distinguished the samples in chapters 2 and 4 was age. Older age typically is associated with numerous forms of cognitive decline. This decline is often readily apparent by 50 years of age, though it has also been reported as early as the late 20s and the early 30s in otherwise healthy samples of educated adults (Salthouse, 2009). Consider Hommel, Li and Li (2004), who tested participants aged 6 through 88 years on visual search tasks (both easy feature search and difficult conjunction search), and found that peak performance was consistently observed between 23 and 33 years of age. More broadly numerous studies suggest that older adults perform worse on various executive function measures including trail making (Kennedy, 1981; Hamdan & Hamdan 2009; Titova et al., 2016; Tombaugh, 2004), stop-signal (Bedard et al., 2002) and Stroop tasks (West & Alain, 2000). Along with age related cognitive decline, are studies showing that BDNF concentration declines in older age, and that  111  age-related cognitive decline itself is associated with reduced BDNF (Li et al., 2009; Buchman et al., 2016; Laske, et al., 2011; Erickson, Miller & Roecklein, 2012) These findings pertaining to age suggest that exercise is particularly important for maintaining cognitive functioning among older adults (Guiney & Machado, 2013; Erickson & Kramer, 2009; Huang et al., 2014; Ludyga et al., 2016; Mekari et al., 2019). Numerous studies support this claim, including those showing that exercise among older adults quickens trail making completion time (Benedict, et al., 2013; Barnes et al., 2008), as well as flanker task reaction time (Kamijo et al., 2009), and reduces the broader risk of cognitive impairment (Bherer, Erickson & Liu-Ambrose, 2013). A related study by Bullock, Mizzi, Kovacevic and Heisz (2018) found that older adults performed worse on the difficult trials of a memory task relative to younger adults. However, only among older adults was greater aerobic fitness found to predict better performance on these trials. Importantly, exercise among older adults heightens BDNF concentration, thereby allowing for structural changes in the brain to occur, which may help to promote more efficient executive functioning among a population more likely to face cognitive decline (Thoenen, 1995; Gorski et al., 2003; Bathina & Das, 2015; Heisz et al., 2017; Szhuhany, Bugatti & Ottom 2015; Erickson et al., 2011; Weinstein et al., 2012; Erickson, Leckie & Weinstein, 2014) All the above research can be taken to imply, that within a sample of older participants, cognitively-engaging exercise ought to be especially beneficial for executive functioning.  Yet the opposite result was found. To resolve this paradox, one starting point is to consider the underlying motivation and circumstances under which an individual may engage in exercise. Nearly 25% of participants in chapter 4 reported walking as their primary form of exercise. By contrast, running and weight training were each tied in chapter 2 for the most  112  frequently reported primary exercise, and collectively accounted for about 25% of all reported exercise. This at first glance suggests that while chapter 4 participants were more diverse in their demographics, they may have been more homogenous in their exercise activities. Arguably more important, however, is questioning what personal experiences might underlie these reports. An individual who reports using high levels of executive control when walking may be someone who is facing physical challenges with this form of locomotion, with their motivation to walk, or they may be engaged in mind-absorbing rumination. Understanding to what extent one’s primary exercise is a negative or stressful experience may help determine whether cognitively-engaging exercise is predictive of improved, or deleterious, executive functioning.  Stress reduces BDNF, and with this may limit neuroplasticity as well as beneficial structural remodelling of the brain (Dell'Osso et al., 2009; Pittenger & Duman, 2008; Smith, Makino, Kvetnansky & Post, 1995). Both acute and chronic stress exposure have deleterious effects including dendritic atrophy and apoptotic cell dell within the hippocampus (Vyas et al., 2016). The prefrontal cortex, which is vital for executive functioning, is among the brain regions thought to be the most sensitive to the adverse effects of stress (Arnsten, 2009), and following a stressful event cognitive flexibility, (Alexander et al., 2007), working memory (Luethi, Meier, & Sandi, 2009), and inhibitory control (Roos et al., 2017) have all been found to decline. Within rats BDNF also increases less under forced exercise, relative to voluntarily initiated exercise, and, forced exercise elicits greater corticosterone concentration (Ke et al., 2011; Uysal et al., 2015).  Given the above research, it is important to understand why one is engaging in a given exercise and what the underlying motivation may be (Duncan, Hall, Wilson & Jenny 2010). An individual might be exerting high levels of executive functioning when exercising because they  113  are attempting to refine their skills and to advance toward more elite competition, they may do so because they value the goal progression afforded through exercise, because they enjoy the challenge this activity offers, or because they appreciate the opportunistic escape exercise can provide. Under these circumstances, cognitively-engaging exercise is ideally situated to advance improvements in executive functioning. In contrast to these scenarios, exercise that is completed out of necessity or outside of personal interest may contribute to a more stressful experience. An individual might engage in exercise because it is part of their commute to work, because of physician recommendations, in order to upkeep appearance, or because of other complex and interacting factors (e.g., guilt, shame, etc.). These scenarios suggest that one’s exercise may be cognitively engaging, not because one is enjoying the activity, but because the activity is undesirable, requires effort to forcefully complete, and cannot be otherwise avoided. Such exercise might be expected to provide less benefit, and potentially even a negative impact, on executive functioning. Outside of these theoretical considerations, the online nature of the study may have contributed to these aberrant results. When research is conducted outside of a laboratory, various factors may bias participants performance and divert their attention from the task at hand. For instance, individuals who report high executive functioning use during exercise may also be those who are more likely to multitask while completing the study (e.g., by simultaneously performing other studies or other non-study related tasks). If the online participants were multitasking in this way, performance could be expected to decline, or to be more variable and sporadic on the flanker task and backward span. Participants in study 3 however still exhibited the expected effects associated with these executive function tasks, including better performance  114  on congruent versus incongruent flanker trials, and more digits being correctly recalled for shorter backward span lengths. Leisure results within chapter 4 were also largely similar to those of chapters 2 and 3, with two noteworthy exceptions. The first, and more prominent, is that longer history of leisure predicted better performance on every executive function outcome. Specifically, longer history of leisure was associated with greater inhibitory control on the flanker task (both accuracy and reaction time differences) and predicted larger working memory capacity on the backward span. This finding suggests that having a consistent long-term hobby is important, whatever that hobby might be, remaining perseverant and developing expertise is critical. The second, and more minor exception, was that leisure rated as being more intense was also found to predict poorer inhibitory control, as marked by a greater accuracy difference between congruent and incongruent trials on the flanker task. This latter finding may suggest that low to moderately intense leisure activity is ideal. As was true for chapters 2 and 3, the underlying results of chapter 4 are grounded within the measurement model of executive function use during exercise. The two-factor exercise model identified in chapter 4 had comparable qualities to the two-factor exercise model reported in chapter 2. When considering inhibitory control items, both studies yielded similar standardized loadings (chapter 2: .60, .47, and .70, chapter 4: .73, .43, and .84). A similar statement can also be made for cognitive flexibility loadings (chapter 2: .63, .64, .72, .70, .70, .54, and .74, chapter 4: .47, .76, .79, .80, .81, .62, and .80). When considering the average loading for inhibitory control and cognitive flexibility items, values found in chapter 2 (.59 and .67) were comparable to those found in present chapter 4 (.67 and .72). Both models also had highly comparable correlations between inhibitory control and cognitive flexibility latent  115  constructs (.68 and .70). The model reported in chapter 2 however did have better fit than the model reported in chapter 4, but, this difference was not as large as that between chapters 2 and 3. This difference acknowledged, it is also the case that the two-factor exercise model of chapter 4 generally yielded good fit through a CFI of .96, an SRMR of .05, and RMSEA of .07.                  116  Measurement Item Mean SD 1. slow down to avoid making mistakes 2.93 1.10 2. filter and ignore distracting information 3.50 1.06 3. pause and double check what I am doing 3.05 1.19 4. adapt and change how things are done 2.59 1.34 5. try to identify new techniques or strategies 3.21 1.16 6. practice creativity 2.67 1.19 7. encounter and solve new problems 2.64 1.06 8. rely on a diverse skillset 2.61 1.15 9. make predictions about what will happen next 2.91 1.11 10. connect and combine different ideas 2.79 1.12 Table 4.1. Mean and standard deviation values for all exercise measurement items. Items were rated on a 5-point scale ranging from “Never” to “Always”. Each item was prefaced with the prompt “When engaging in my primary exercise I”. Inhibitory control was measured via items 1 through 3, and cognitive flexibility items 4 through 10.  Measurement Item Mean SD 1. slow down to avoid making mistakes 2.97 1.12 2. filter and ignore distracting information 3.67 0.96 3. pause and double check what I am doing 3.09 1.16 4. adapt and change how things are done 3.32 1.09 5. try to identify new techniques or strategies 3.35 1.28 6. practice creativity 3.39 1.21 7. encounter and solve new problems 3.24 1.19 8. rely on a diverse skillset 3.17 1.22 9. make predictions about what will happen next 3.58 0.95 10. connect and combine different ideas 3.55 0.98 Table 4.2. Mean and standard deviation values for all leisure measurement items. Items were rated on a 5-point scale ranging from “Never” to “Always”. Each item was prefaced with the prompt “When engaging in my primary exercise I”. Inhibitory control was measured via items 1 through 3, and cognitive flexibility items 4 through 10.        117  Figure 4.1 Frequencies of self-reported primary exercise.  Figure 4.2 Frequencies of self-reported primary leisure.   118    Figure 4.3 Performance on the flanker task (left, accuracy; middle reaction time) and backward span (right, capacity) among exercise (top half) and leisure (bottom half) participants  119   Figure 4.4 CFA of the two-factor exercise model.  Figure 4.5 Exercise path analysis.   Figure 4.6 Exercise SEM.  120   Figure 4.7 CFA of the two-factor leisure model.  Figure 4.8 Leisure path analysis model.  Figure 4.9 Leisure SEM.  121  Chapter 5: General Discussion 5.1 Summary of Key Findings   This thesis is comprised of three independent studies as described in chapters 2 through 4. Each study tested the cognitive engagement hypothesis, namely, that an important factor in the positive link between exercise and executive functioning is the extent to which participants find their exercise activities to be cognitively engaging. This hypothesis was tested by examining whether the degree to which executive functions are reported to be used during regular exercise is a significant predictor of performance on laboratory measures of executive functioning. Each of the studies also included a parallel investigation that tested the same hypothesis with respect to an individual’s non-exercise leisure activity. This feature of the design allowed us to distinguish between an exercise-specific link to executive function and a general-engagement hypothesis. A comparison of the findings in these three studies leads to four broad conclusions. First, individual differences in executive function use during exercise are predictive of performance on laboratory measures of executive functioning. Second, exercise qualifiers that are commonly studied in the exercise literature (e.g., history, duration, intensity, and type of exercise), were largely not predictive of performance on executive function tasks. Third, self-reported executive function use during leisure activities was never predictive of performance on executive function tasks. Lastly, performance on laboratory measures of executive functioning was generally not predicted by leisure history, duration, intensity or type.  It is important to note that these general findings were observed across different sample demographics and different laboratory measures of executive functioning. Specifically, chapters  122  2 and 3 tested undergraduate students from the university of British Columbia, and chapter 4 tested participants from Amazon’s Mechanical Turk. Chapters 2 and 4 both used the flanker task and backward span, whereas chapter 3 used trail making B and the stop-signal task. Yet despite all these differences in sample demographics and study measures, results were consistent such that the perceived executive demands of exercise predicted performance on executive function tasks. What was not consistent, however, was the direction of these relationships, which is discussed in the following sections below. 5.2 Implications of a Positive Exercise & Executive Function Relationship  The results in chapters 2 and 3 offer the most convincing evidence in support of the cognitive-engagement hypothesis. Chapter 2 found that among undergraduate students, exercise that made greater demands on inhibitory control predicted a smaller accuracy difference between congruent and incongruent trials on a flanker task. Chapter 3 found this same exercise-specific relationship in a different group of undergraduate students using different measures of inhibitory control. Here exercise that made demands on inhibitory control predicted faster stop-signal reaction time and fewer trail making errors. Notably exercise was found to benefit inhibitory control across three distinct measurements of this construct, that are sometimes are referred to as interference control (flanker task), action cancellation (stop-signal task), and impulsivity (trail making B). A natural question arises from these findings: how does cognitively-engaging exercise benefit executive functioning? A substantial body of research suggests that exercise elicits an increase in BDNF (Szuhany, Bugatti & Otto, 2015; Heisz et al. 2017). BDNF aids in the maintenance, regulation, survival, and formation of new neurons (Thoenen, 1995; Gorski et al., 2003; Bathina & Das, 2015), and these properties collectively allow for neuroplasticity and  123  neurogenesis of the brain. Exercise as such prompts a state in which structural changes to the brain can be made, thereby resulting in improved neuronal efficiency (Szuhany, Bugatti & Otto, 2015; Calabrese et al., 2014; Castrén & Antila, 2017; Bherer et al., 2015; Erickson & Kramer, 2009; Heisz & Kovacevic, 2016). In addition to this, BDNF has also been thought to help prep the brain for anticipated cognitive stimulation (Hötting & Röder, 2013), and it is here that cognitively-engaging exercise is thought to exert an influence.  Research involving rats has found that when new neurons form within the hippocampus, their survival is improved through completion of behavioral tasks that are themselves reliant on hippocampal functioning (Gould, Beylin, Tanapat, Reeves, & Shors, 1999). This suggests a potential cumulative effect, wherein exercise may initiate neurogenesis and neuroplasticity, that is then further refined and selectively enhanced through the cognitive demands of one’s exercise. Cognitive engagement as such may accentuate neuroplasticity initiated through exercise within specific and behaviourally relevant brain regions. Raichlen and Alexander (2017) had provided a similar account suggesting that while BDNF may prompt neurogenesis and richer neural interconnectedness, maintaining these changes is likely dependent on the cognitive stimulation afforded through one’s exercise. Growing research similarly suggests that the survival and integration of new neurons is dependent on learning and adaption, which here is posited to be achieved through cognitively-engaging exercise (Olariu et al., 2005). A clear implication of this hypothesized mechanism is that it ought to be testable using brain imaging. For instance, various mediation models have found that exercise and cardiorespiratory fitness predict improved cognition through an indirect effect of an applicable brain site. One such study was completed by Erickson et al., (2009) in which cardiorespiratory fitness (VO2max) of older adults was found to predict better spatial working memory, as well as  124  greater hippocampal volume. The relationship between greater fitness and greater spatial working memory was then found to be mediated through greater hippocampal volume. Similar mediation models have been successfully used by Weinstein et al. (2012) as well as by Bento-Torres et al. (2019). A review article by Erickson, Miller and Roecklein (2012) has also argued that memory functioning improves through exercise via BDNF-elicited neuroplasticity of the hippocampus. Cognitively-engaging exercise is thought to operate within this model such that exercise-elicited neuroplasticity is selectively enhanced and maintained at specific brain regions via the cognitive demands of one’s exercise.   5.3 Implications of a Negative Exercise & Executive Function Relationship   Chapter 4 found that among participants recruited through Amazon’s Mechanical Turk, self-reports of exercise that made greater demands on executive functioning predicted poorer performance on laboratory measures of executive functioning. Specifically, exercise reported to rely on inhibitory control predicted less efficient flanker task performance, and exercise reported to rely on cognitive flexibility predicted smaller working memory capacity on the backward span. These results directly oppose those of chapter 2, where self-reports of cognitive engagement during exercise among undergraduate students predicted more efficient flanker task performance, and larger working memory capacity. In an attempt to interpret these different results, chapter 4 considered demographic differences between these two samples. Particular focus was given to the relatively older age of participants in chapter 4 (37.08 years) relative to participants in chapter 2 (20.43 years), and to the larger proportion of women in chapter 2 versus men in chapter 4.  Neither the factors of age nor gender helped to elucidate the findings of chapter 4. Older age for instance is associated with cognitive decline, lower neurotrophin concentration, and  125  greater neural degradation (Li et al., 2009; Buchman et al., 2016; Laske, et al., 2011; Erickson, Miller & Roecklein, 2012; Drag & Bieliauskas, 2010; Salat et al., 2009; Fjell & Walhovd, 2010). For these reasons, cognitively-engaging exercise in this sample could be expected to have an even greater beneficial effect than was found in chapter 2. Instead, chapter 4 found that exercise reported to rely on executive functioning predicted poorer executive functioning. The study described within chapter 4 was completed online which may have contributed to these aberrant results. Relative to participants who completed the study within a lab, participants in chapter 4 may have been more likely to multitask while completing the study (e.g., by simultaneously performing other studies or other non-study related tasks). If the online participants were multitasking in this way, performance could be expected to decline, or to be more variable and sporadic on the flanker task and backward span. Participants in study 3 however still exhibited the expected effects associated with these tasks, including better performance on congruent versus incongruent flanker trials, and more digits being correctly recalled for shorter backward span lengths. A full understanding of the results of chapter 4 may require a richer assessment of why individuals find a particular exercise to be cognitively engaging. Individuals may participate in a given exercise because it is part of their identity (e.g., I am a runner, I am a hockey player), because they value the health benefits (e.g., cardiovascular fitness), because of external pressures (e.g., advice from a physician), because it is part of their routine or habit (e.g., morning run), because it is readily available, and/or because of a complex set of other interacting factors (e.g., enjoyment, guilt, etc.). These motivating factors each imply a different relationship between exercise and executive functioning, and may help qualify whether cognitively engaging exercise might predict better, or worse, executive functioning.   126  An individual exerting high levels of executive functioning when exercising may be doing so because they are facing talented opposition, because they are attempting to master a new skillset, because they inherently enjoy the afforded challenge, because they find the activity to be destressing, or due to a host of other factors. Under these circumstances, cognitively engaging exercise is situated within an ideal position from which executive functioning might improve. Conversely if one’s exercise is overwhelming, stressful, and promotes negative affect, then a beneficial impact on executive functioning may be less tenable or pronounced.  Research suggests that stress and depression both implicate reduced BDNF (Erickson, Miller & Roecklein, 2012; Dell'Osso et al., 2009; Pittenger & Duman, 2008; Smith, Makino, Kvetnansky & Post, 1995; Duman & Monteggia, 2006; Schmidt, & Duman, 2010). Exercise that elicits or associates with stress, rumination, or other such negative states, may then lead to a supressed BDNF response and consequently limited potential for neuroplasticity and neurogenesis. Certain brain sites, like the prefrontal cortex, are also particularly sensitive to stress (Arnsten, 2009), which can worsen executive functioning (Gohie et al., 2009; Berman et al., 2011; Roos et al., 2017; Johnco, Wuthrich & Rapee, 2015; Lee & Goto, 2015; Mika et al., 2012), including impaired cognitive flexibility, (Alexander et al., 2007), working memory (Luethi, Meier, & Sandi, 2009), and inhibitory control (Roos et al., 2017). Within rats BDNF has also been found to increase less under forced exercise conditions, relative to voluntarily initiated exercise, and, forced exercise elicits greater corticosterone concentration (Ke et al., 2011; Uysal et al., 2015).  An individual who exercises for cardiovascular benefits and personal enjoyment likely has greater control over the exercise they decide to voluntarily pursue. This may include following an exercise they are interested in (e.g., swimming versus jogging), whether this  127  exercise is social or not (e.g., individual or team orientated), what days of the week they can commit (e.g., weekends), what duration is manageable given their schedule (e.g., minutes per session), what level of competition or intensity is appropriate (e.g., beginner versus intermediate), whether they should pursue external help (e.g., coaching), and when for a given week they can stop this activity given a change in circumstance (e.g., physical injury or sickness). Should this individual also lose interest or no longer enjoy a particular exercise, they may freely quit or pursue a different form of exercise. All of these qualities allow for a more positive and enjoyable exercise experience, from which the cognitive demands of this exercise may help optimize executive functioning. Conversely, exercise that is completed out of necessity or outside of personal interest, will often involve less agency and control over the factors listed above. An individual for instance might engage in exercise because it is part of their commute to work. This might suggest that only certain types of exercise are applicable (e.g., cycling or jogging), predetermined by what days are relevant (e.g., work days), without the option of social interaction (e.g., commuting to independently), for a duration determined by external factors (e.g., distance to work), and often regardless of whether one is feeling unwell or unfit. Other scenarios may include exercise that is completed on the basis of physician recommendations, or for upkeeping appearance. This is not to downplay the potential benefits of such exercise, given the large amount of research supporting its benefits. Rather, the claim here is that some exercise scenarios contribute more to the beneficial effects for executive functioning than other scenarios. 5.4 Implications for Leisure & Executive Functioning  Across chapters 2, 3 and 4, performance on laboratory measures of executive functioning was not predicted by perceived executive function use during leisure. Failing to find an  128  association between leisure and executive functioning makes the association between exercise and executive functioning unique. It is worth considering here what factors make exercise distinct from non-exercise leisure that could account for these results. Three possible answers are reviewed here, although these do not exhaust the possibilities. One, exercise has been found to elicit changes in cerebral blood flow that associate with greater executive functioning. In one study, Yanagisawa et al. (2010) found that an acute bout of exercise improved inhibitory control (i.e., less Stroop task interference), and that this increase in performance coincided with greater cerebral blood flow within the left dorsolateral prefrontal cortex (DLPFC). Note that the left dorsolateral prefrontal cortex has been implicated in more efficient inhibitory control (MacDonald et al., 2000; Kujach et al., 2018; Hyodo et al. (2016; Byun et al. 2014). Leisure activity that does not have exercise components would therefore not be expected to elicit and benefit from this mechanism. Two, exercise results in an increase of neurotrophic factors like BDNF, VEGF and IGF-1, which allow for structural changes within the brain and improved neural efficiency (Raichlen & Polk, 2013; Gustafsson et al., 2001; Koziris et al., 1999; Fabel et al., 2003; Szuhany, Bugatti & Otto, 2015; Heisz et al. 2017). As one example, Erickson, Miller and Roecklein (2012), suggest that exercise leads to the production of BDNF, which leads to increased hippocampal volume, which aids memory functioning. By eliciting this physiological cascade exercise further discerns itself from non-exercise leisure. Three, exercise offers a continual gradient through which executive functioning can be progressively challenged and refined. A tennis player for instance can strive to better read their opponent, anticipate what moves are likely to come next, and better respond to various deceptive plays. Runners similarly can strive to sprint faster, better manage their breathing, extend their  129  track duration, adopt a more consistent running regime, and seek to improve their overall performance and form. Leisure too offers many avenues of progress and growth, however, for many leisure activities, such progress and growth at least on face-value appears less tangible or evident (e.g., viewing media for entertainment). This suggests that the demands of exercise, relative to certain leisure activities, may offer a continual gradient through which executive functioning can be continuously challenged and mastered. These potential cognitive demands may then selectively enhance and act upon the more general benefits of exercise-elicited neuroplasticity. This interpretation should not be taken to mean that leisure is unimportant for executive functioning. Leisure might for instance provide an important contribution through other mechanisms that are not directly implicated with exercise. If exercise is an effortful activity that draws on executive functioning, leisure might be an activity that allows for recovery and restoration of depleted resources. To illustrate, consider a study by Berman, Jonides and Kaplan (2008) where participants completed a backward span task and then an additional exhausting memory task. Participants were then assigned to a condition where they walked within a nearby park (nature) or within a downtown district (industrial), before returning to the lab and completing another session of the backward span task. Participants who had traversed through nature exhibited greater improvement on this task relative to participants who had traversed through the city. Furthermore, a second study found that merely viewing photos of nature, versus industrial landscapes, resulted in greater task improvement on both the backward span, as well as the inhibitory control portion (i.e., flanker task) of the attention network test. Hence the act of simply viewing nature was associated with a restorative effect through which executive  130  functioning benefitted. Through such restorative mechanism’s leisure might exert an important influence on executive functioning. 5.5 Implications for Measuring Exercise & Leisure Executive Function Use   All three studies in this dissertation share a common backbone in the measurement model that was used. This measurement model was initially identified in chapter 2 through a data-driven approach, and was subsequently applied to independent data collected in chapters 3 and 4. Within this model, executive functioning consists of two latent constructs, inhibitory control, indicated by 3 items, and cognitive flexibility, indicated by 7 items. One limitation of this model is the absence of a well-defined working memory construct. This may be because the working memory measurement items were not sufficiently targeted to this construct, or because working memory and cognitive flexibility are too overlapped during exercise. Although all measurement items were inspired by how executive functions are commonly discussed within the scientific literature (sections 1.6.1, 1.6.2, and 1.6.3), this approach does not guarantee a well-fitting measurement model. A related concern is that poor working memory may itself impair the ability to report on one’s own working memory capacity, making this construct difficult to measure using self-reports.  When the two-factor exercise model was applied to a new set of undergraduate students in chapter 3, it yielded acceptable though weaker fit relative to chapter 2. When the same model in chapter 4 was applied to a set of participants recruited through Amazon’s Mechanical Turk, the overall model fit was acceptable-to-good, though it was still weaker relative to chapter 2.  Taken collectively, these results suggest that the two-factor model has fit ranging from mediocre to acceptable. Reliability, measured through coefficient omega, also never fell below .80 (note that Cronbach’s alpha was not used to measure reliability because it assumes a unidimensional  131  construct; Hayes & Coutts, 2020; Watkins, 2017). These results imply that this measurement model is worth further exploration, while at the same time remaining open to modifications and revision.  One simple consideration is whether negatively worded items, when re-written to be positively worded, may help identify the underlying executive function constructs. Another consideration pertains to standardizing the wording of all questions more strictly (e.g., replacing the word “many” with “more than one”). A separate basis for evaluating the two-factor exercise model concerns the conceptual and theoretical meaning of the underlying items. The latent construct representing inhibitory control for instance consisted of three indicator items. These three items could be seen as measuring different facets of inhibitory control. Slowing down to avoid making mistakes (item 1) is a common strategy that participants employ when completing executive measures like the stop-signal task. Filtering and ignoring distracting information (item 3) is descriptive of what must be done for successful completion of the flanker task. Lastly, pausing to double check what one is doing (item 7) may help prevent one from making an impulse based mistake, as may be captured by trail making errors. Given these considerations, new items, and item revision, might do well to a adopt more task-descriptive based approach rather than one that is more generic to the overall cognitive construct under study. It is worth noting that the used two-factor measurement model was intended to capture differences in executive function use during exercise, though nonetheless tended to have fit indices that were quite comparable when it was employed to reports of cognitive engagement during leisure activity. This may suggest that executive function use during exercise, as well as during leisure, can be comparably measured through the same set of measurement items.   132  Finally, it is worth considering the research applicability of two-factor exercise model in light of the collective results of the three studies reported here. The place to begin might well be to heed the sage advice that while all models are wrong, some are useful (Box, 1979). The present findings clearly show that the two-factor exercise model was able to predict performance on tasks of executive functioning. It has therefore satisfied the criterion of usefulness, and in doing so further poses under what context and motivation might exercise be expected to maximally benefit executive functioning.               133  References Alexander, J. K., Hillier, A., Smith, R. M., Tivarus, M. E., & Beversdorf, D. Q. (2007). Beta-adrenergic modulation of cognitive flexibility during stress. Journal of cognitive neuroscience, 19(3), 468-478. Alfini, A. J., Weiss, L. R., Leitner, B. P., Smith, T. J., Hagberg, J. M., & Smith, J. C. (2016). Hippocampal and cerebral blood flow after exercise cessation in master athletes. Frontiers in aging neuroscience, 8, 184. Alves, H., Voss, M., Boot, W. R., Deslandes, A., Cossich, V., Inacio Salles, J., & Kramer, A. F. (2013). Perceptual-cognitive expertise in elite volleyball players. Frontiers in psychology, 4, 36. Amieva, H., Lafont, S., Auriacombe, S., Rainville, C., Orgogozo, J. M., Dartigues, J. F., & Fabrigoule, C. (1998). Analysis of error types in the Trail Making Test evidences an inhibitory deficit in dementia of the Alzheimer type. Journal of clinical and experimental neuropsychology, 20(2), 280-285. Arnsten, A. F. (2009). Stress signalling pathways that impair prefrontal cortex structure and function. Nature reviews neuroscience, 10(6), 410-422. Arvinen-Barrow, M., Weigand, D. A., Thomas, S., Hemmings, B., & Walley, M. (2007). Elite and novice athletes' imagery use in open and closed sports. Journal of Applied Sport Psychology, 19(1), 93-104. Ashendorf, L., Jefferson, A. L., O'Connor, M. K., Chaisson, C., Green, R. C., & Stern, R. A. (2008). Trail Making Test errors in normal aging, mild cognitive impairment, and dementia. Archives of Clinical Neuropsychology, 23(2), 129-137.  134  Axmacher, N., Henseler, M. M., Jensen, O., Weinreich, I., Elger, C. E., & Fell, J. (2010). Cross-frequency coupling supports multi-item working memory in the human hippocampus. Proceedings of the National Academy of Sciences, 107(7), 3228-3233. Bailey, C. E. (2007). Cognitive accuracy and intelligent executive function in the brain and in business. Annals of the New York Academy of Sciences, 1118(1), 122-141. Baird, B., Smallwood, J., & Schooler, J. W. (2011). Back to the future: Autobiographical planning and the functionality of mind-wandering. Consciousness and cognition, 20(4), 1604-1611. Barella, L. A., Etnier, J. L., & Chang, Y. K. (2010). The immediate and delayed effects of an acute bout of exercise on cognitive performance of healthy older adults. Journal of Aging and Physical Activity, 18(1), 87-98. Barger, P., Behrend, T. S., Sharek, D. J., & Sinar, E. F. (2011). IO and the crowd: Frequently asked questions about using Mechanical Turk for research. The Industrial-Organizational Psychologist, 49(2), 11-17. Barnes, D. E., Blackwell, T., Stone, K. L., Goldman, S. E., Hillier, T., & Yaffe, K. (2008) Cognition in older women: the importance of daytime movement. Journal of the American Geriatrics Society, 56(9), 1658-1664. Bathina, S., & Das, U. N. (2015). Brain-derived neurotrophic factor and its clinical implications. Archives of medical science: AMS, 11(6), 1164. Berman, M. G., Kross, E., Krpan, K. M., Askren, M. K., Burson, A., Deldin, P. J., ... & Jonides, J. (2012). Interacting with nature improves cognition and affect for individuals with depression. Journal of affective disorders, 140(3), 300-305  135  Bedard, A. C., Nichols, S., Barbosa, J. A., Schachar, R., Logan, G. D., & Tannock, R. (2002). The development of selective inhibitory control across the life span. Developmental neuropsychology, 21(1), 93-111. Behrend, T. S., Sharek, D. J., Meade, A. W., & Wiebe, E. N. (2011). The viability of crowdsourcing for survey research. Behavior research methods, 43(3), 800. Benedict, C., Brooks, S. J., Kullberg, J., Nordenskjöld, R., Burgos, J., Le Grevès, M., ... & Lind, L. (2013). Association between physical activity and brain health in older adults. Neurobiology of aging, 34(1), 83-90. Bentler, P. M., & Bonett, D. G. (1980). Significance tests and goodness of fit in the analysis of covariance structures. Psychological bulletin, 88(3), 588. Bento-Torres, J., Bento-Torres, N. V. O., Stillman, C. M., Grove Jr, G. A., Huang, H., Uyar, F., ... & Erickson, K. I. (2019). Associations between cardiorespiratory fitness, physical activity, intraindividual variability in behavior, and cingulate cortex in younger adults. Journal of sport and health science, 8(4), 315-324. Berman, M. G., Jonides, J., & Kaplan, S. (2008). The cognitive benefits of interacting with nature. Psychological science, 19(12), 1207-1212. Best, J. R. (2010). Effects of physical activity on children’s executive function: Contributions of experimental research on aerobic exercise. Developmental Review, 30(4), 331-351. Bherer, L. (2015). Cognitive plasticity in older adults: effects of cognitive training and physical exercise. Annals of the New York Academy of Sciences, 1337(1), 1-6. Bherer, L., Erickson, K. I., & Liu-Ambrose, T. (2013). A review of the effects of physical activity and exercise on cognitive and brain functions in older adults. Journal of aging research, 2013.  136  Boomsma, A. (1985). Nonconvergence, improper solutions, and starting values in LISREL maximum likelihood estimation. Psychometrika, 50(2), 229-242. Bourrier, S. C., Berman, M. G., & Enns, J. T. (2018). Cognitive strategies and natural environments interact in influencing executive function. Frontiers in psychology, 9, 1248 Box, G. E. (1979). All models are wrong, but some are useful. Robustness in Statistics, 202, 549. Brainard, D. H. (1997). The psychophysics toolbox. Spatial vision, 10(4), 433-436. Briggs, N. E., & MacCallum, R. C. (2003). Recovery of weak common factors by maximum likelihood and ordinary least squares estimation. Multivariate Behavioral Research, 38(1), 25-56. Brown, T. A. (2015). Confirmatory factor analysis for applied research. Guilford publications. Browne, M. W., & Cudeck, R. (1992). Alternative ways of assessing model fit. Sociological methods & research, 21(2), 230-258. Buchman, A. S., Yu, L., Boyle, P. A., Schneider, J. A., De Jager, P. L., & Bennett, D. A. (2016). Higher brain BDNF gene expression is associated with slower cognitive decline in older adults. Neurology, 86(8), 735-741. Bühner, M., König, C. J., Pick, M., & Krumm, S. (2006). Working memory dimensions as differential predictors of the speed and error aspect of multitasking performance. Human Performance, 19(3), 253-275. Bullock, A. M., Mizzi, A. L., Kovacevic, A., & Heisz, J. J. (2018). The association of aging and aerobic fitness with memory. Frontiers in aging neuroscience, 10, 63. Burgess, N., Maguire, E. A., & O'Keefe, J. (2002). The human hippocampus and spatial and episodic memory. Neuron, 35(4), 625-641.  137  Buttelmann, F., & Karbach, J. (2017). Development and plasticity of cognitive flexibility in early and middle childhood. Frontiers in psychology, 8, 1040. Byun, K., Hyodo, K., Suwabe, K., Ochi, G., Sakairi, Y., Kato, M., ... & Soya, H. (2014). Positive effect of acute mild exercise on executive function via arousal-related prefrontal activations: an fNIRS study. Neuroimage, 98, 336-345. Calabrese, F., Rossetti, A. C., Racagni, G., Gass, P., Riva, M. A., & Molteni, R. (2014). Brain-derived neurotrophic factor: a bridge between inflammation and neuroplasticity. Frontiers in cellular neuroscience, 8, 430. Campanholo, K. R., Romão, M. A., Machado, M. D. A. R., Serrao, V. T., Coutinho, D. G. C., Benute, G. R. G., & Lucia, M. C. S. D. (2014). Performance of an adult Brazilian sample on the Trail Making Test and Stroop Test. Dementia & neuropsychologia, 8(1), 26-31 Castiello, U., & Umiltà, C. (1992). Orienting of attention in volleyball players. International Journal of Sport Psychology. 23, 301–310.  Castrén, E., & Antila, H. (2017). Neuronal plasticity and neurotrophic factors in drug responses. Molecular psychiatry, 22(8), 1085-1095. Chang, Y. K., Labban, J. D., Gapin, J. I., & Etnier, J. L. (2012). The effects of acute exercise on cognitive performance: a meta-analysis. Brain research, 1453, 87-101. Chang, Y. K., Pesce, C., Chiang, Y. T., Kuo, C. Y., & Fong, D. Y. (2015). Antecedent acute cycling exercise affects attention control: an ERP study using attention network test. Frontiers in human neuroscience, 9, 156 Chekroud, S. R., Gueorguieva, R., Zheutlin, A. B., Paulus, M., Krumholz, H. M., Krystal, J. H., & Chekroud, A. M. (2018). Association between physical exercise and mental health in 1· 2  138  million individuals in the USA between 2011 and 2015: a cross-sectional study. The Lancet Psychiatry, 5(9), 739-746. Chen, Z., Zhao, X., Fan, J., & Chen, A. (2018). Functional cerebral asymmetry analyses reveal how the control system implements its flexibility. Human brain mapping, 39(12), 4678-4688. Chen, A. G., Yan, J., Yin, H. C., Pan, C. Y., & Chang, Y. K. (2014). Effects of acute aerobic exercise on multiple aspects of executive function in preadolescent children. Psychology of Sport and Exercise, 15(6), 627-636 Chiu, C. N., Chen, C. Y., & Muggleton, N. G. (2017). Sport, time pressure, and cognitive performance. In Progress in brain research (Vol. 234, pp. 85-99). Elsevier.  Chmielewski, M., & Kucker, S. C. (2020). An MTurk crisis? Shifts in data quality and the impact on study results. Social Psychological and Personality Science, 11(4), 464-473.   Chu, C. H., Alderman, B. L., Wei, G. X., & Chang, Y. K. (2015). Effects of acute aerobic exercise on motor response inhibition: An ERP study using the stop-signal task. Journal of Sport and Health Science, 4(1), 73-81. Coelho, R. W., De Campos, W., Silva, S. G. D., Okazaki, F. H. A., & Keller, B. (2007). Imagery intervention in open and closed tennis motor skill performance. Perceptual and motor skills, 105(2), 458-468. Colcombe, S. J., Erickson, K. I., Scalf, P. E., Kim, J. S., Prakash, R., McAuley, E., ... & Kramer, A. F. (2006). Aerobic exercise training increases brain volume in aging humans. The Journals of Gerontology Series A: Biological Sciences and Medical Sciences, 61(11), 1166-1170  139  Colcombe, S., & Kramer, A. F. (2003). Fitness effects on the cognitive function of older adults: a meta-analytic study. Psychological science, 14(2), 125-130. Colom, R., Martínez-Molina, A., Shih, P. C., & Santacreu, J. (2010). Intelligence, working memory, and multitasking performance. Intelligence, 38(6), 543-551. Conway, A. R. A., Jarrold, C., Kane, M. J., Miyake, A.,& Towse, J. N. (Eds.). (2007). Variation in working memory. New York: Oxford University Press Corbett, J., Barwood, M. J., Ouzounoglou, A., Thelwell, R., & Dicks, M. (2012). Influence of competition on performance and pacing during cycling exercise. Medicine & Science in Sports & Exercise, 44(3), 509-515. Corso, H. V., Cromley, J. G., Sperb, T., & Salles, J. F. (2016). Modeling the relationship among reading comprehension, intelligence, socioeconomic status, and neuropsychological functions: The mediating role of executive functions. Psychology & Neuroscience, 9(1),   32 Cotman, C. W., Berchtold, N. C., & Christie, L. A. (2007). Exercise builds brain health: key roles of growth factor cascades and inflammation. Trends in neurosciences, 30(9), 464-472. Cotman, C. W., & Engesser-Cesar, C. (2002). Exercise enhances and protects brain function. Exercise and sport sciences reviews, 30(2), 75-79. Cowan, N. (2010). The magical mystery four: How is working memory capacity limited, and why? Current directions in psychological science, 19(1), 51-57. Crowe, S. F. (1998). The differential contribution of mental tracking, cognitive flexibility, visual search, and motor speed to performance on parts A and B of the Trail Making Test. Journal of clinical psychology, 54(5), 585-591.  140  Crump, M. J., McDonnell, J. V., & Gureckis, T. M. (2013). Evaluating Amazon's Mechanical Turk as a tool for experimental behavioral research. PloS one, 8(3). Cudeck, R. (1989). Analysis of correlation matrices using covariance structure models. Psychological Bulletin, 105(2), 317. Crutzen, R., & Peters, G. J. Y. (2017). Scale quality: alpha is an inadequate estimate and factor-analytic evidence is needed first of all. Health Psychology Review, 11(3), 242-247. Drag, L. L., & Bieliauskas, L. A. (2010). Contemporary review 2009: cognitive aging. Journal of geriatric psychiatry and neurology, 23(2), 75-93. Dardis, R., Soberon-Ferrer, H., & Patro, D. (1994). Analysis of leisure expenditures in the United States. Journal of Leisure Research, 26(4), 309-321. De Leeuw, J. R., & Motz, B. A. (2016). Psychophysics in a Web browser? Comparing response times collected with JavaScript and Psychophysics Toolbox in a visual search task. Behavior Research Methods, 48(1), 1-12. De Leeuw, J. R. (2015). jsPsych: A JavaScript library for creating behavioral experiments in a Web browser. Behavior research methods, 47(1), 1-12. Dell'Osso, L., Carmassi, C., Del Debbio, A., Dell'Osso, M. C., Bianchi, C., da Pozzo, E., ... & Piccinni, A. (2009). Brain-derived neurotrophic factor plasma levels in patients suffering from post-traumatic stress disorder. Progress in Neuro-Psychopharmacology and Biological Psychiatry, 33(5), 899-902. Diamond, A. (2013). Executive functions. Annual review of psychology, 64, 135-168. Diamond, A., & Ling, D. S. (2016). Conclusions about interventions, programs, and approaches for improving executive functions that appear justified and those that, despite much hype, do not. Developmental cognitive neuroscience, 18, 34-48.  141  Di Corrado, D., Guarnera, M., & Quartiroli, A. (2014). Vividness and transformation of mental images in karate and ballet. Perceptual and motor skills, 119(3), 764-773. DiStefano, C., & Motl, R. W. (2006). Further investigating method effects associated with negatively worded items on self-report surveys. Structural Equation Modeling, 13(3), 440-464 Dube, S. K., & Tatz, S. J. (1991). Audience effects in tennis performance. Perceptual and motor skills, 73(3), 844-846. Duman, R. S., & Monteggia, L. M. (2006). A neurotrophic model for stress-related mood disorders. Biological psychiatry, 59(12), 1116-1127. Duncan, L. R., Hall, C. R., Wilson, P. M., & Jenny, O. (2010). Exercise motivation: a cross-sectional analysis examining its relationships with frequency, intensity, and duration of exercise. International Journal of Behavioral Nutrition and Physical Activity, 7(1), 7. Ecker, U. K., Oberauer, K., & Lewandowsky, S. (2014). Working memory updating involves item-specific removal. Journal of Memory and Language, 74, 1-15. Erickson, K. I., & Kramer, A. F. (2009). Aerobic exercise effects on cognitive and neural plasticity in older adults. British journal of sports medicine, 43(1), 22-24. Erickson, K. I., Leckie, R. L., & Weinstein, A. M. (2014). Physical activity, fitness, and gray matter volume. Neurobiology of aging, 35, S20-S28. Erickson, K. I., Miller, D. L., & Roecklein, K. A. (2012). The aging hippocampus: interactions between exercise, depression, and BDNF. The Neuroscientist, 18(1), 82-97. Erickson, K. I., Prakash, R. S., Voss, M. W., Chaddock, L., Hu, L., Morris, K. S., ... & Kramer, A. F. (2009). Aerobic fitness is associated with hippocampal volume in elderly humans. Hippocampus, 19(10), 1030-1039.  142  Erickson, K. I., Voss, M. W., Prakash, R. S., Basak, C., Szabo, A., Chaddock, L., ... & Wojcicki, T. R. (2011). Exercise training increases size of hippocampus and improves memory. Proceedings of the National Academy of Sciences, 108(7), 3017-3022. Ericsson, K. A., & Smith, J. (Eds.). (1991). Toward a general theory of expertise: Prospects and limits. Cambridge University Press. Eriksen, B. A., & Eriksen, C. W. (1974). Effects of noise letters upon the identification of a target letter in a nonsearch task. Perception & psychophysics, 16(1), 143-149. Etnier, J. L., Salazar, W., Landers, D. M., Petruzzello, S. J., Han, M., & Nowell, P. (1997). The  influence of physical fitness and exercise upon cognitive functioning: A metaanalysis. Journal of sport and Exercise Psychology, 19(3), 249-277.  performance: attentional control theory. Emotion, 7(2), 336. Fabel, K., Fabel, K., Tam, B., Kaufer, D., Baiker, A., Simmons, N., ... & Palmer, T. D. (2003). VEGF is necessary for exercise‐induced adult hippocampal neurogenesis. European Journal of Neuroscience, 18(10), 2803-2812. Fabel, K., & Kempermann, G. (2008). Physical activity and the regulation of neurogenesis in the adult and aging brain. Neuromolecular medicine, 10(2), 59-66. Fardilha, F. D. S., & Allen, J. B. (2020). Defining, assessing, and developing creativity in sport: a systematic narrative review. International Review of Sport and Exercise Psychology, 13(1), 104-127. Fjell, A. M., & Walhovd, K. B. (2010). Structural brain changes in aging: courses, causes and cognitive consequences. Rev Neurosci, 21(3), 187-221 Funahashi, S., & Andreau, J. M. (2013). Prefrontal cortex and neural mechanisms of executive function. Journal of Physiology-Paris, 107(6), 471-482.  143  Gejl, A. K., Bugge, A., Ernst, M. T., Tarp, J., Hillman, C. H., Have, M., ... & Andersen, L. B. (2018). The acute effects of short bouts of exercise on inhibitory control in adolescents. Mental Health and Physical Activity, 15, 34-39.  Gettens, K. M., & Gorin, A. A. (2017). Executive function in weight loss and weight loss maintenance: a conceptual review and novel neuropsychological model of weight control. Journal of Behavioral Medicine, 40(5), 687-701. Gohier, B., Ferracci, L., Surguladze, S. A., Lawrence, E., El Hage, W., Kefi, M. Z., ... & Le Gall, D. (2009). Cognitive inhibition and working memory in unipolar depression. Journal of affective disorders, 116(1-2), 100-105. Goldman-Rakic, P. S. (1992). Working memory and the mind. Scientific American, 267(3), 110-117. Gorski, J. A., Zeiler, S. R., Tamowski, S., & Jones, K. R. (2003). Brain-derived neurotrophic factor is required for the maintenance of cortical dendrites. Journal of  Neuroscience, 23(17), 6856-6865. Gould, E., Beylin, A., Tanapat, P., Reeves, A., & Shors, T. J. (1999). Learning enhances adult neurogenesis in the hippocampal formation. Nature neuroscience, 2(3), 260-265 Graziano, P. A., Calkins, S. D., & Keane, S. P. (2010). Toddler self-regulation skills predict risk for pediatric obesity. International journal of obesity, 34(4), 633-641. Greco, P., Memmert, D., & Morales, J. C. (2010). The effect of deliberate play on tactical performance in basketball. Perceptual and motor skills, 110(3), 849-856. Gu, Q., Zou, L., Loprinzi, P. D., Quan, M., & Huang, T. (2019). Effects of Open Versus Closed Skill Exercise on Cognitive Function: A Systematic Review. Frontiers in psychology, 10, 1707.  144  Guiney, H., & Machado, L. (2013). Benefits of regular aerobic exercise for executive functioning in healthy populations. Psychonomic bulletin & review, 20(1), 73-86. Gustafsson, T., Bodin, K., Sylven, C., Gordon, A., Tyni‐Lenne, R., & Jansson, E. (2001). Increased expression of VEGF following exercise training in patients with heart failure. European journal of clinical investigation, 31(4), 362-366. Hair, J. F., Black, W. C., Babin, B. J., & Anderson, R. E. (2009). Multivariate Data Analysis 7th  Edition Pearson Prentice Hall. Hamdan, A. C., & Hamdan, E. M. L. (2009). Effects of age and education level on the Trail Making Test in a healthy Brazilian sample. Psychology & Neuroscience, 2(2), 199-203. Harveson, A. T., Hannon, J. C., Brusseau, T. A., Podlog, L., Papadopoulos, C., Durrant, L. H., ... & Kang, K. D. (2016). Acute effects of 30 minutes resistance and aerobic exercise on cognition in a high school sample. Research quarterly for exercise and sport, 87(2), 214-220. Hauser, D. J., & Schwarz, N. (2016). Attentive Turkers: MTurk participants perform better on online attention checks than do subject pool participants. Behavior research methods, 48(1), 400-407. Hayes, A. F., & Coutts, J. J. (2020). Use Omega Rather than Cronbach’s Alpha for Estimating Reliability. But…. Communication Methods and Measures, 14(1), 1-24. Heisz, J. J., Clark, I. B., Bonin, K., Paolucci, E. M., Michalski, B., Becker, S., & Fahnestock, M. (2017). The effects of physical exercise and cognitive training on memory and neurotrophic factors. Journal of cognitive neuroscience, 29(11), 1895-1907. Heisz, J. J., & Kovacevic, A. (2016). Exercise impacts age-related changes in cognitive function and neural complexity. Kinesiology Review, 5(1), 30-38.  145  Henrich, J., Heine, S. J., & Norenzayan, A. (2010). Most people are not WEIRD. Nature, 466(7302), 29-29 Hermida, R. (2015). The problem of allowing correlated errors in structural equation modeling:  concerns and considerations. Computational Methods in Social Sciences, 3(1), 5-17. Herold, F., Wiegel, P., Scholkmann, F., & Müller, N. G. (2018). Applications of functional near-infrared spectroscopy (fNIRS) neuroimaging in exercise–cognition science: a systematic, methodology-focused review. Journal of clinical medicine, 7(12), 466. Highlen, P. S., & Bennett, B. B. (1979). Psychological characteristics of successful and nonsuccessful elite wrestlers: An exploratory study. Journal of Sport and Exercise Psychology, 1(2), 123-137. Hilbig, B. E. (2016). Reaction time effects in lab-versus Web-based research: Experimental evidence. Behavior Research Methods, 48(4), 1718-1724. Hillman, C. H., Erickson, K. I., & Kramer, A. F. (2008). Be smart, exercise your heart: exercise effects on brain and cognition. Nature reviews neuroscience, 9(1), 58-65. Hillman, C. H., McAuley, E., Erickson, K. I., Liu-Ambrose, T., & Kramer, A. F. (2019). On mindful and mindless physical activity and executive function: A response to Diamond and Ling (2016). Developmental cognitive neuroscience, 37. Ho, S., Gooderham, G. K., & Handy, T. C. (2018). Self-reported free-living physical activity and executive control in young adults. PloS one, 13(12). Hobert, M. A., Niebler, R., Meyer, S. I., Brockmann, K., Becker, C., Huber, H., ... & Maetzler, W. (2011). Poor trail making test performance is directly associated with altered dual task prioritization in the elderly–baseline results from the TREND study. PLoS One, 6(11)  146  Holbert, R. L., & Stephenson, M. T. (2002). Structural equation modeling in the communication sciences, 1995–2000. Human Communication Research, 28(4), 531-551. Holden, C. J., Dennie, T., & Hicks, A. D. (2013). Assessing the reliability of the M5-120 on Amazon’s Mechanical Turk. Computers in Human Behavior, 29(4), 1749-1754. Hommel, B., Li, K. Z., & Li, S. C. (2004). Visual search across the life span. Developmental psychology, 40(4), 545-558. Hötting, K., & Röder, B. (2013). Beneficial effects of physical exercise on neuroplasticity and cognition. Neuroscience & Biobehavioral Reviews, 37(9), 2243-2257. Hsieh, S. S., Chueh, T. Y., Huang, C. J., Kao, S. C., Hillman, C. H., Chang, Y. K., & Hung, T. M. (2020). Systematic review of the acute and chronic effects of high-intensity interval training on executive function across the lifespan. Journal of Sports Sciences, 1-13. Hu, L., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling: A Multidisciplinary Journal, 6, 1-55. doi:10.1080/10705519909540118 Huang, C. J., Lin, P. C., Hung, C. L., Chang, Y. K., & Hung, T. M. (2014). Type of physical exercise and inhibitory function in older adults: an event-related potential study. Psychology of Sport and Exercise, 15(2), 205-211. Hull, R., Martin, R. C., Beier, M. E., Lane, D., & Hamilton, A. C. (2008). Executive function in older adults: a structural equation modeling approach. Neuropsychology, 22(4), 508. Hung, C. L., Tseng, J. W., Chao, H. H., Hung, T. M., & Wang, H. S. (2018). Effect of acute exercise mode on serum brain-derived neurotrophic factor (BDNF) and task switching performance. Journal of clinical medicine, 7(10), 301.  147  Huster, R. J., Messel, M. S., Thunberg, C., & Raud, L. (2020). The P300 as marker of inhibitory control–fact or fiction?. Cortex. Hüttermann, S., & Memmert, D. (2014). Does the inverted-U function disappear in expert athletes? An analysis of the attentional behavior under physical exercise of athletes and non-athletes. Physiology & behavior, 131, 87-92. Hyodo, K., Dan, I., Kyutoku, Y., Suwabe, K., Byun, K., Ochi, G., ... & Soya, H. (2016). The association between aerobic fitness and cognitive function in older men mediated by frontal lateralization. Neuroimage, 125, 291-300. Ionescu, T. (2012). Exploring the nature of cognitive flexibility. New ideas in psychology, 30(2), 190-200. Ipeirotis, P. (2010). Demographics of mechanical turk. Retrieved from http://hdl.handle.net/  2451/29585. Jackson, D. L., Gillaspy, J. A., & Purc-Stephenson, R. (2009). Reporting practices in confirmatory factor analysis: An overview and some recommendations. Psychological Methods, 14, 6-23. doi:10.1037/a0014694 Jacobo, S. M. P., & Kazlauskas, A. (2015). Insulin-like growth factor 1 (IGF-1) stabilizes nascent blood vessels. Journal of Biological Chemistry, 290(10), 6349-6360. Jacobson, J., & Matthaeus, L. (2014). Athletics and executive functioning: How athletic participation and sport type correlate with cognitive performance. Psychology of Sport and Exercise, 15(5), 521-527. Johann, V., Könen, T., & Karbach, J. (2020). The unique contribution of working memory, inhibition, cognitive flexibility, and intelligence to reading comprehension and reading speed. Child Neuropsychology, 26(3), 324-344.  148  Johnco, C., Wuthrich, V. M., & Rapee, R. M. (2015). The impact of late‐life anxiety and depression on cognitive flexibility and cognitive restructuring skill acquisition. Depression and anxiety, 32(10), 754-762. Jorgensen, T. D., Pornprasertmanit, S., Schoemann, A. M., Rosseel, Y., Miller, P., Quick, C., & Garnier-Villarreal, M. (2018). semTools: Useful tools for structural equation modeling. R package version 0.5-1.  Joris, P. J., Mensink, R. P., Adam, T. C., & Liu, T. T. (2018). Cerebral Blood Flow Measurements in Adults: A Review on the Effects of Dietary Factors and Exercise. Nutrients, 10(5) Joyce, J., Graydon, J., McMorris, T., & Davranche, K. (2009). The time course effect of moderate intensity exercise on response execution and response inhibition. Brain and cognition, 71(1), 14-19. Jung, R., Moser, M., Baucsek, S., Dern, S., & Schneider, S. (2015). Activation patterns of different brain areas during incremental exercise measured by near-infrared spectroscopy. Experimental brain research, 233(4), 1175-1180. Kamijo, K., Hayashi, Y., Sakai, T., Yahiro, T., Tanaka, K., & Nishihira, Y. (2009). Acute effects of aerobic exercise on cognitive function in older adults. Journals of Gerontology: Series B, 64(3), 356-363. Katzir, M., Eyal, T., Meiran, N., & Kessler, Y. (2010). Imagined positive emotions and inhibitory control: The differentiated effect of pride versus happiness. Journal of Experimental Psychology: Learning, Memory, and Cognition, 36(5), 1314.  149  Ke, Z., Yip, S. P., Li, L., Zheng, X. X., & Tong, K. Y. (2011). The effects of voluntary, involuntary, and forced exercises on brain-derived neurotrophic factor and motor function recovery: a rat brain ischemia model. PloS one, 6(2), e16643. Kennedy, K. J. (1981). Age effects on trail making test performance. Perceptual and Motor Skills, 52(2), 671-675. Kenny, D. A. (2016). Multiple latent variable models: Confirmatory factor analysis. Retrieved  from http://davidakenny.net/cm/mfactor.htm Kesler, S. R., Lacayo, N. J., & Jo, B. (2011). A pilot study of an online cognitive rehabilitation program for executive function skills in children with cancer-related brain injury. Brain Injury, 25(1), 101-112. Kilduff, G. J. (2014). Driven to win: Rivalry, motivation, and performance. Social Psychological and Personality Science, 5(8), 944-952. Kisler, K., Nelson, A. R., Montagne, A., & Zlokovic, B. V. (2017). Cerebral blood flow regulation and neurovascular dysfunction in Alzheimer disease. Nature Reviews Neuroscience, 18(7), 419. Kleiner, M., Brainard, D. H., & Pelli, D. G. (2007). What’s new in Psychtoolbox-3? Perception, 36, ECVP Abstract Supplement Kline, R.B. (2016). Principles and Practice of Structural Equation Modeling (4th ed.). New York: Guildford Kluding, P. M., Tseng, B. Y., & Billinger, S. A. (2011). Exercise and executive function in individuals with chronic stroke: a pilot study. Journal of neurologic physical therapy: JNPT, 35(1), 11  150  Klusman, L. E., Cripe, L. I., & Dodrill, C. B. (1989). Analysis of Errors on the Trail Making Test. Perceptual and Motor Skills, 68(3_suppl), 1199–1204. Knaepen, K., Goekint, M., Heyman, E. M., & Meeusen, R. (2010). Neuroplasticity—exercise-induced response of peripheral brain-derived neurotrophic factor. Sports medicine, 40(9), 765-801. Koziris, L. P., Hickson, R. C., Chatterton Jr, R. T., Groseth, R. T., Christie, J. M., Goldflies, D. G., & Unterman, T. G. (1999). Serum levels of total and free IGF-I and IGFBP-3 are increased and maintained in long-term training. Journal of Applied Physiology, 86(4), 1436-1442 Krenn, B., Finkenzeller, T., Würth, S., & Amesberger, G. (2018). Sport type determines differences in executive functions in elite athletes. Psychology of Sport and Exercise, 38, 72-79. Kujach, S., Byun, K., Hyodo, K., Suwabe, K., Fukuie, T., Laskowski, R., ... & Soya, H. (2018). A transferable high-intensity intermittent exercise improves executive performance in association with dorsolateral prefrontal activation in young adults. Neuroimage, 169, 117-125. Labelle, V., Bosquet, L., Mekary, S., & Bherer, L. (2013). Decline in executive control during acute bouts of exercise as a function of exercise intensity and fitness level. Brain and cognition, 81(1), 10-17. Laske, C., Stellos, K., Hoffmann, N., Stransky, E., Straten, G., Eschweiler, G. W., & Leyhe, T. (2011). Higher BDNF serum levels predict slower cognitive decline in Alzheimer's disease patients. International Journal of Neuropsychopharmacology, 14(3), 399-404.  151  Lee, Y. A., & Goto, Y. (2015). Chronic stress effects on working memory: Association with prefrontal cortical tyrosine hydroxylase. Behavioural brain research, 286, 122-127. Li, J. W., O’Connor, H., O’Dwyer, N., & Orr, R. (2017). The effect of acute and chronic exercise on cognitive function and academic performance in adolescents: a systematic review. Journal of science and medicine in sport, 20(9), 841-848 Li, G., Peskind, E. R., Millard, S. P., Chi, P., Sokal, I., Yu, C. E., ... & Montine, T. J. (2009). Cerebrospinal fluid concentration of brain-derived neurotrophic factor and cognitive function in non-demented subjects. PloS one, 4(5), e5424. Liu-Ambrose, T., Nagamatsu, L. S., Graf, P., Beattie, B. L., Ashe, M. C., & Handy, T. C. (2010). Resistance training and executive functions: a 12-month randomized controlled trial. Archives of internal medicine, 170(2), 170-178 Liu, P. Z., & Nusslock, R. (2018). Exercise-mediated neurogenesis in the hippocampus via BDNF. Frontiers in neuroscience, 12, 52. Logan, G. D., & Cowan, W. B. (1984). On the ability to inhibit thought and action: A theory of an act of control. Psychological review, 91(3), 295. Loprinzi, P. D., & Kane, C. J. (2015). Exercise and cognitive function: A randomized controlled trial examining acute exercise and free-living physical activity and sedentary effects. Mayo Clinic Proceedings, 90(4), 450–460. Ludyga, S., Gerber, M., Brand, S., Holsboer‐Trachsler, E., & Pühse, U. (2016). Acute effects of moderate aerobic exercise on specific aspects of executive function in different age and fitness groups: A meta‐analysis. Psychophysiology, 53(11), 1611-1626.  152  Luethi, M., Meier, B., & Sandi, C. (2009). Stress effects on working memory, explicit memory, and implicit memory for neutral and emotional stimuli in healthy men. Frontiers in behavioral neuroscience, 2, 5. Lum, J., Enns, J. T., & Pratt, J. (2002). Visual orienting in college athletes: Explorations of athlete type and gender. Research Quarterly for exercise and sport, 73(2), 156-167. MacCallum, R. C., & Austin, J. T. (2000). Applications of structural equation modeling in psychological research. Annual review of psychology, 51(1), 201-226 MacCallum, R. C., Browne, M. W., & Sugawara, H. M. (1996). Power analysis and determination of sample size for covariance structure modeling. Psychological Methods, 1, 130-149. doi:10.1037//1082-989X.1.2.130 MacCallum, R. C., Roznowski, M., & Necowitz, L. B. (1992). Model modifications in  covariance structure analysis: the problem of capitalization on chance. Psychological  bulletin, 111(3), 490. MacDonald, A. W., Cohen, J. D., Stenger, V. A., & Carter, C. S. (2000). Dissociating the role of the dorsolateral prefrontal and anterior cingulate cortex in cognitive control. Science, 288(5472), 1835-1838. Mahurin, R. K., Velligan, D. I., Hazleton, B., Mark Davis, J., Eckert, S., & Miller, A. L. (2006). Trail making test errors and executive function in schizophrenia and depression. The Clinical Neuropsychologist, 20(2), 271-288. Marmeleira, J. F., Godinho, M. B., & Fernandes, O. M. (2009). The effects of an exercise program on several abilities associated with driving performance in older adults. Accident Analysis & Prevention, 41, 90-97.  153  Marquez, C. M. S., Vanaudenaerde, B., Troosters, T., & Wenderoth, N. (2015). High intensity interval training evokes larger serum BDNF levels compared to intense continuous exercise. Journal of Applied Physiology, jap-00126. Matzke, D., Verbruggen, F., & Logan, G. D. (2018). The Stop‐Signal Paradigm. Stevens' Handbook of Experimental Psychology and Cognitive Neuroscience, 5, 1-45. McCabe, D. P., Roediger III, H. L., McDaniel, M. A., Balota, D. A., & Hambrick, D. Z. (2010). The relationship between working memory capacity and executive functioning: evidence for a common executive attention construct. Neuropsychology, 24(2), 222. McLeod, B. (1985). Field dependence as a factor in sports with preponderance of open or closed skills. Perceptual and motor skills, 60(2), 369-370. Meijers, J., Harte, J. M., Meynen, G., & Cuijpers, P. (2017). Differences in executive functioning between violent and non-violent offenders. Psychol. Med, 47(10), 1784-1793. Mekari, S., Dupuy, O., Martins, R., Evans, K., Kimmerly, D. S., Fraser, S., & Neyedli, H. F. (2019). The effects of cardiorespiratory fitness on executive function and prefrontal oxygenation in older adults. Geroscience, 41(5), 681-690. Mika, A., Mazur, G. J., Hoffman, A. N., Talboom, J. S., Bimonte-Nelson, H. A., Sanabria, F., & Conrad, C. D. (2012). Chronic stress impairs prefrontal cortex-dependent response inhibition and spatial working memory. Behavioral neuroscience, 126(5), 605. Miller, H. V., Barnes, J. C., & Beaver, K. M. (2011). Self-control and health outcomes in a nationally representative sample. American journal of health behavior, 35(1), 15-27. Miyake, A., & Friedman, N. P. (2012). The nature and organization of individual differences in executive functions: Four general conclusions. Current directions in psychological science, 21(1), 8-14.  154  Miyake, A., Friedman, N. P., Emerson, M. J., Witzki, A. H., Howerter, A., & Wager, T. D. (2000). The unity and diversity of executive functions and their contributions to complex “frontal lobe” tasks: A latent variable analysis. Cognitive psychology, 41(1), 49-100 Moriarty, T., Bourbeau, K., Bellovary, B., & Zuhl, M. N. (2019). Exercise intensity influences prefrontal cortex oxygenation during cognitive testing. Behavioral Sciences, 9(8), 83. Murray, N. P., & Russoniello, C. (2012). Acute physical activity on cognitive function: a heart rate variability examination. Applied psychophysiology and biofeedback, 37(4), 219-227.  Olariu, A., Cleaver, K. M., Shore, L. E., Brewer, M. D., & Cameron, H. A. (2005). A natural form of learning can increase and decrease the survival of new neurons in the dentate gyrus. Hippocampus, 15(6), 750-762. O’Leary, K. C., Pontifex, M. B., Scudder, M. R., Brown, M. L., & Hillman, C. H. (2011). The effects of single bouts of aerobic exercise, exergaming, and videogame play on cognitive control. Clinical Neurophysiology, 122(8), 1518-1525. Padilla, C., Pérez, L., Andres, P., & Parmentier, F. B. (2013). Exercise improves cognitive control: evidence from the stop signal task. Applied Cognitive Psychology, 27(4), 505-511.  Padilla, C., Pérez, L., & Andrés, P. (2014). Chronic exercise keeps working memory and inhibitory capacities fit. Frontiers in behavioral neuroscience, 8, 49. Pardini, D., Lochman, J., & Wells, K. (2004). Negative emotions and alcohol use initiation in high-risk boys: The moderating effect of good inhibitory control. Journal of abnormal child psychology, 32(5), 505-518. Peer, E., Vosgerau, J., & Acquisti, A. (2014). Reputation as a sufficient condition for data quality on Amazon Mechanical Turk. Behavior research methods, 46(4), 1023-1031.  155  Peeters, M., Janssen, T., Monshouwer, K., Boendermaker, W., Pronk, T., Wiers, R., & Vollebergh, W. (2015). Weaknesses in executive functioning predict the initiating of adolescents’ alcohol use. Developmental cognitive neuroscience, 16, 139-146. Pérez, L., Padilla, C., Parmentier, F. B., & Andrés, P. (2014). The effects of chronic exercise on attentional networks. PloS one, 9(7). e101478. Piccinni, A., Marazziti, D., Del Debbio, A., Bianchi, C., Roncaglia, I., Mannari, C., ... & Dell'Osso, L. (2008). Diurnal variation of plasma brain‐derived neurotrophic factor (BDNF) in humans: An analysis of sex differences. Chronobiology international, 25(5), 819-826. Pinet, S., Zielinski, C., Mathôt, S., Dufau, S., Alario, F. X., & Longcamp, M. (2017). Measuring sequences of keystrokes with jsPsych: Reliability of response times and interkeystroke intervals. Behavior research methods, 49(3), 1163-1176. Pittenger, C., & Duman, R. S. (2008). Stress, depression, and neuroplasticity: a convergence of mechanisms. Neuropsychopharmacology, 33(1), 88-109. Pituch, K. A., & Stevens, J. P. (2015). Applied multivariate statistics for the social sciences:  Analyses with  SAS and IBM’s SPSS. Routledge. Polli, F. E., Barton, J. J., Cain, M. S., Thakkar, K. N., Rauch, S. L., & Manoach, D. S. (2005). Rostral and dorsal anterior cingulate cortex make dissociable contributions during antisaccade error commission. Proceedings of the National Academy of Sciences, 102(43), 15700-15705. Pornprasertmanit, S., Miller, P., & Schoemann, A. (2015). simsem: SIMulated structural equation modeling. R package version 0.5, URL http://CRAN. R-project. org/package= simsem. Poulton, E. C. (1957). On prediction in skilled movements. Psychological bulletin, 54(6), 467.  156  Quaney, B. M., Boyd, L. A., McDowd, J. M., Zahner, L. H., He, J., Mayo, M. S., & Macko, R. F. (2009). Aerobic exercise improves cognition and motor function poststroke. Neurorehabilitation and neural repair, 23(9), 879-885. Querido, J. S., & Sheel, A. W. (2007). Regulation of cerebral blood flow during exercise. Sports Medicine, 37(9), 765-782. Raichlen, D. A., & Alexander, G. E. (2017). Adaptive capacity: an evolutionary neuroscience model linking exercise, cognition, and brain health. Trends in neurosciences, 40(7), 408-421. Raichlen, D. A., & Polk, J. D. (2013). Linking brains and brawn: exercise and the evolution of human neurobiology. Proceedings of the Royal Society B: Biological Sciences, 280(1750), 20122250. Raine, L. B., Kao, S. C., Pindus, D., Westfall, D. R., Shigeta, T. T., Logan, N., ... & Khan, N. A. (2018). A large-scale reanalysis of childhood fitness and inhibitory control. Journal of cognitive enhancement, 2(2), 170-192. Rasmusson, X. D., Zonderman, A. B., Kawas, C., & Resnick, S. M. (1998). Effects of age and dementia on the Trail Making Test. The Clinical Neuropsychologist, 12(2), 169-178. Reitan, R. M. (1958). Validity of the Trail Making Test as an indicator of organic brain damage. Perceptual and motor skills, 8(3), 271-276. Rhea, M. R., Landers, D. M., Alvar, B. A., & Arent, S. M. (2003). The effects of competition and the presence of an audience on weight lifting performance. Journal of strength and conditioning research, 17(2), 303-306. Richards, M., Hardy, R., & Wadsworth, M. E. (2003). Does active leisure protect cognition? Evidence from a national birth cohort. Social science & medicine, 56(4), 785-792  157  Roos, L. E., Knight, E. L., Beauchamp, K. G., Berkman, E. T., Faraday, K., Hyslop, K., & Fisher, P. A. (2017). Acute stress impairs inhibitory control based on individual differences in parasympathetic nervous system activity. Biological psychology, 125, 58-63. Rosen, V. M., & Engle, R. W. (1997). Forward and backward serial recall. Intelligence, 25(1), 37-47. Ross, J., Irani, L., Silberman, M. S., Zaldivar, A., & Tomlinson, B. (2010). Who are the crowdworkers? Shifting demographics in Mechanical Turk. In CHI'10 extended abstracts on Human factors in computing systems Rosseel, Y. (2012). Lavaan: An R package for structural equation modeling. Journal of Statistical Software, 48, 1-36. Roth, D. L., Goode, K. T., Clay, O. J., & Ball, K. K. (2003). Association of physical activity and visual attention in older adults. Journal of aging and health, 15(3), 534-547. Sakamoto, S., Takeuchi, H., Ihara, N., Ligao, B., & Suzukawa, K. (2018). Possible requirement of executive functions for high performance in soccer. PloS one, 13(8). Salthouse, T. A. (2009). When does age-related cognitive decline begin? Neurobiology of aging, 30(4), 507-514. Samuels, W. E., Tournaki, N., Blackman, S., & Zilinski, C. (2016). Executive functioning predicts academic achievement in middle school: A four-year longitudinal study. The Journal of Educational Research, 109(5), 478-490. Sánchez-Cubillo, I. 1., Perianez, J. A., Adrover-Roig, D., Rodriguez-Sanchez, J. M., Rios-Lago, M., Tirapu, J. E. E. A., & Barcelo, F. (2009). Construct validity of the Trail Making Test: role of task-switching, working memory, inhibition/interference control, and visuomotor abilities. Journal of the International Neuropsychological Society, 15(3), 438-450.  158  Sandroff, B. M., Benedict, R. H., & Motl, R. W. (2015). Nonsignificant associations between measures of inhibitory control and walking while thinking in persons with multiple sclerosis. Archives of Physical Medicine and Rehabilitation, 96(8), 1518-1524. Savalei, V., & Bentler, P. M. (2006). Structural equation modeling. In R. Grover & M. Vriens (Eds.), The handbook of marketing research: Uses, misuses, and future advances. Thousand Oaks, CA: Sage. Schmidt, H. D., & Duman, R. S. (2010). Peripheral BDNF produces antidepressant-like effects in cellular and behavioral models. Neuropsychopharmacology, 35(12), 2378-2391. Schmidt-Kassow, M., Schädle, S., Otterbein, S., Thiel, C., Doehring, A., Lötsch, J., & Kaiser, J. (2012). Kinetics of serum brain-derived neurotrophic factor following low-intensity versus high-intensity exercise in men and women. Neuroreport, 23(15), 889-893. Sheehan, K. B. (2018). Crowdsourcing research: data collection with Amazon’s Mechanical  Turk. Communication Monographs, 85(1), 140-156. Silverman, I. W. (2006). Sex differences in simple visual reaction time: A historical meta-analysis. Sex roles, 54(1-2), 57-68. Sleiman, S. F., Henry, J., Al-Haddad, R., El Hayek, L., Haidar, E. A., Stringer, T., ... & Ninan, I.  (2016). Exercise promotes the expression of brain derived neurotrophic factor (BDNF) through the action of the ketone body β-hydroxybutyrate. Elife, 5, e15092. Smith, P. J., Blumenthal, J. A., Hoffman, B. M., Cooper, H., Strauman, T. A., Welsh-Bohmer, K., ... & Sherwood, A. (2010). Aerobic exercise and neurocognitive performance: a meta-analytic review of randomized controlled trials. Psychosomatic medicine, 72(3), 239.  159  Smith, M. A., Makino, S., Kvetnansky, R., & Post, R. M. (1995). Stress and glucocorticoids affect the expression of brain-derived neurotrophic factor and neurotrophin-3 mRNAs in the hippocampus. Journal of Neuroscience, 15(3), 1768-1777. Soga, K., Masaki, H., Gerber, M., & Ludyga, S. (2018). Acute and long-term effects of resistance training on executive function. Journal of Cognitive Enhancement, 2(2), 200-207. Spellman, T., Rigotti, M., Ahmari, S. E., Fusi, S., Gogos, J. A., & Gordon, J. A. (2015). Hippocampal–prefrontal input supports spatial encoding in working memory. Nature, 522(7556), 309-314. Starkes, J. L., & Ericsson, K. A. (2003). Expert performance in sports: Advances in research on sport expertise. Human Kinetics. Stillman, C. M., Esteban-Cornejo, I., Brown, B., Bender, C. M., & Erickson, K. I. (2020). Effects of exercise on brain and cognition across age groups and health states. Trends in Neurosciences. Steiger, J. H. (2007). Understanding the limitations of global fit assessment in structural equation modeling. Personality and Individual differences, 42(5), 893-898 Stewart, N., Chandler, J., & Paolacci, G. (2017). Crowdsourcing samples in cognitive science. Trends in cognitive sciences, 21(10), 736-748. Stuss, D. T., Bisschop, S. M., Alexander, M. P., Levine, B., Katz, D., & Izukawa, D. (2001). The Trail Making Test: a study in focal lesion patients. Psychological assessment, 13(2), 230 Szuhany, K. L., Bugatti, M., & Otto, M. W. (2015). A meta-analytic review of the effects of exercise on brain-derived neurotrophic factor. Journal of psychiatric research, 60, 56-64.  160  Tan, C. C., & Holub, S. C. (2011). Children’s self-regulation in eating: Associations with inhibitory control and parents’ feeding behavior. Journal of pediatric psychology, 36(3), 340-345. Taylor, S. F., Martis, B., Fitzgerald, K. D., Welsh, R. C., Abelson, J. L., Liberzon, I., ... & Gehring, W. J. (2006). Medial frontal cortex activity and loss-related responses to errors. Journal of Neuroscience, 26(15), 4063-4070. Thoenen, H. (1995). Neurotrophins and neuronal plasticity. Science, 270(5236), 593-598. Tierney, M. C., Moineddin, R., Morra, A., Manson, J., & Blake, J. (2010). Intensity of recreational physical activity throughout life and later life cognitive functioning in women. Journal of Alzheimer's Disease, 22(4), 1331-1338. Titova, O. E., Lindberg, E., Elmståhl, S., Lind, L., Schiöth, H. B., & Benedict, C. (2016). Association between shift work history and performance on the trail making test in middle-aged and elderly humans: the EpiHealth study. Neurobiology of aging, 45, 23-29. Tombaugh, T. N. (2004). Trail Making Test A and B: normative data stratified by age and education. Archives of clinical neuropsychology, 19(2), 203-214. Tomporowski, P. D. (1997). The Effects of Physical and Mental Training on the Mental Abilities of Older Adults. Journal of Aging & Physical Activity, 5(1). Trejo, J. L., Carro, E., & Torres-Alemán, I. (2001). Circulating insulin-like growth factor I mediates exercise-induced increases in the number of new neurons in the adult hippocampus. Journal of Neuroscience, 21(5), 1628-1634. Uysal, N., Kiray, M., Sisman, A. R., Camsari, U. M., Gencoglu, C., Baykara, B., ... & Aksu, I. (2015). Effects of voluntary and involuntary exercise on cognitive functions, and VEGF and BDNF levels in adolescent rats. Biotechnic & Histochemistry, 90(1), 55-68.  161  Vaeyens, R., Lenoir, M., Williams, A. M., & Philippaerts, R. M. (2008). Talent identification and development programmes in sport. Sports medicine, 38(9), 703-714. Van Leeuwen, M., van den Berg, S. M., Hoekstra, R. A., & Boomsma, D. I. (2007). Endophenotypes for intelligence in children and adolescents. Intelligence, 35(4), 369-380. Van Velzen, L. S., Vriend, C., de Wit, S. J., & van den Heuvel, O. A. (2014). Response inhibition and interference control in obsessive–compulsive spectrum disorders. Frontiers in human neuroscience, 8, 419. Verbruggen, F., & Logan, G. D. (2008). Response inhibition in the stop-signal paradigm. Trends in cognitive sciences, 12(11), 418-424. Verbruggen, F., Adams, R. C., van‘t Wout, F., Stevens, T., McLaren, I. P., & Chambers, C. D. (2013). Are the effects of response inhibition on gambling long-lasting?. PloS one, 8(7), e70155. Verbruggen, F., Aron, A. R., Band, G. P., Beste, C., Bissett, P. G., Brockett, A. T., ... & Colzato, L. S. (2019). A consensus guide to capturing the ability to inhibit actions and impulsive behaviors in the stop-signal task. Elife, 8, e46323. Vriens, R. G. M. (2006). The handbook of marketing research: uses, misuses, and future advances. Sage. Vyas, S., Rodrigues, A. J., Silva, J. M., Tronche, F., Almeida, O. F., Sousa, N., & Sotiropoulos, I. (2016). Chronic stress and glucocorticoids: from neuronal plasticity to neurodegeneration. Neural plasticity. Wang, C. H., Chang, C. C., Liang, Y. M., Shih, C. M., Chiu, W. S., Tseng, P., ... & Juan, C. H. (2013). Open vs. closed skill sports and the modulation of inhibitory control. PloS one, 8(2).   162  Wang, H. X., Xu, W., & Pei, J. J. (2012). Leisure activities, cognition and dementia. Biochimica et Biophysica Acta (BBA)-Molecular Basis of Disease, 1822(3), 482-49 Watkins, M. W. (2017). The reliability of multidimensional neuropsychological measures: from alpha to omega. The Clinical Neuropsychologist, 31(6–7), 1113–1126. Weinstein, A. M., Voss, M. W., Prakash, R. S., Chaddock, L., Szabo, A., White, S. M., ... & Erickson, K. I. (2012). The association between aerobic fitness and executive function is mediated by prefrontal cortex volume. Brain, behavior, and immunity, 26(5), 811-819. West, R., & Alain, C. (2000). Age-related decline in inhibitory control contributes to the increased Stroop effect observed in older adults. Psychophysiology, 37(2), 179-189. White, C. N., Brown, S., & Ratcliff, R. (2012). A test of Bayesian observer models of processing in the Eriksen flanker task. Journal of Experimental Psychology: Human Perception and Performance, 38(2), 489.  White, C. N., Ratcliff, R., & Starns, J. J. (2011). Diffusion models of the flanker task: Discrete versus gradual attentional selection. Cognitive psychology, 63(4), 210-238 Winter, B., Breitenstein, C., Mooren, F. C., Voelker, K., Fobker, M., Lechtermann, A., ... & Knecht, S. (2007). High impact running improves learning. Neurobiology of learning and memory, 87(4), 597-609. Williams, P. G., Suchy, Y., & Rau, H. K. (2009). Individual differences in executive functioning: Implications for stress regulation. Annals of Behavioral Medicine, 37(2), 126-140. Woods, C. M. (2006). Careless responding to reverse-worded items: Implications for confirmatory factor analysis. Journal of Psychopathology and Behavioral Assessment, 28(3), 186  163  Woods, D. L., Wyma, J. M., Herron, T. J., & Yund, E. W. (2015). The effects of aging, malingering, and traumatic brain injury on computerized trail-making test performance. PloS one, 10(6). Wu, Y. H., de Rotrou, J., Vidal, J. S., Jeandel, C., Rigaud, A. S., Kesse-Guyot, E., & Hanon, O. (2013). The KT cancellation test in the older adults: Normative data and construct validity. Aging, Neuropsychology, and Cognition, 20(4), 429-442. Wylie, S. A., Bashore, T. R., Van Wouwe, N. C., Mason, E. J., John, K. D., Neimat, J. S., & Ally, B. A. (2018). Exposing an “Intangible” cognitive skill among collegiate football players: enhanced interference control. Frontiers in psychology, 9. Yanagisawa, H., Dan, I., Tsuzuki, D., Kato, M., Okamoto, M., Kyutoku, Y., & Soya, H. (2010). Acute moderate exercise elicits increased dorsolateral prefrontal activation and improves cognitive performance with Stroop test. Neuroimage, 50(4), 1702-1710. Yuan, P., & Raz, N. (2014). Prefrontal cortex and executive functions in healthy adults: a meta-analysis of structural neuroimaging studies. Neuroscience & Biobehavioral Reviews, 42, 180-192. Zelazo, P. D. (2015). Executive function: Reflection, iterative reprocessing, complexity, and the developing brain. Developmental Review, 38, 55-68. Zhang, X., Noor, R., & Savalei, V. (2016). Examining the effect of reverse worded items on the  factor  structure of the need for cognition scale. PloS one, 11(6), e0157795.      164  Appendix: Executive Function Use During Exercise – Measurement Items Inhibitory control When doing my primary exercise I... 1. slow down to avoid making mistakes 2. care most about speed and performing quickly (R) 3. filter and ignore distracting information 4. decide what to do through impulse alone (R) 5. practice self-control and discipline 6. follow every action to completion (R) 7. pause and double check what I am doing 8. start and complete actions without thinking (R) 9. anticipate making fast or sudden adjustments 10. act without self restraint (R)  Cognitive flexibility When doing my primary exercise I... 1. adapt and change how things are done 2. have a plan that I stringently follow (R) 3. try to identify new techniques or strategies 4. follow the same routine (R) 5. practice creativity 6. hold the same mindset from start to finish (R) 7. encounter and solve new problems 8. have little-to-no flexibility to modify what I do (R) 9. rely on a diverse skillset 10. think the same thoughts over and over (R)  Working Memory When doing my primary exercise I... 1. pay attention to many things at the same time 2. understand everything that is happening even when absent-minded (R) 3. make predictions about what will happen next 4. avoid time-keeping (R) 5. connect and combine different ideas 6. do not care about the order in which things happen (R) 7. multitask 8. disengage from what is happening around me (R) 9. monitor what is happening on a second-to-second basis 10. focus all my attention entirely on one thing before moving onto the next (R)  Two-factor model: Inhibitory items 1, 3, and 7. Cognitive flexibility items 1, 3, 5, 7, and 9. Working memory items 3 and 5. 

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            data-media="{[{embed.selectedMedia}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
https://iiif.library.ubc.ca/presentation/dsp.24.1-0395601/manifest

Comment

Related Items