<%BANNER%>

Understanding the Transition from Normal Cognitive Aging to Mild Cognitive Impairment: Comparing the Intraindividual Var...

University of Florida Institutional Repository

PAGE 1

UNDERSTANDING THE TRANSITION FROM NORMAL COGNITIVE AGING TO MILD COGNITIVE IMPAIRMENT: COMPARING THE INTRAINDI VIDUAL VARIABILITY IN COGNITIVE FUNCTION By KARIN J. M. MCCOY A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2004

PAGE 2

Copyright 2004 by Karin J. M. McCoy

PAGE 3

This dissertation is dedicated to my fianc, Joseph Barker. His love, generosity, and goodness of spirit supported me through a ll the hurdles of graduate school.

PAGE 4

ACKNOWLEDGMENTS I would like to express my deepest appreciation to my dissertation chair, Michael Marsiske. He has patiently and thoughtfully provided prodding and direction throughout the inception, production, and presentation of this work, as well as giving graciously and generously of his time and intellect in his role as a mentor. Special recognition goes to Russell Bauer, since, in addition to his role as my cochair, he has been my mentor in neuropsychology throughout my graduate career. In conjunction with my chair and cochair, my committee members, Lise Abrams, Dawn Bowers, and Duane Dede, have provided me with guidance, support, and friendship. Each has contributed greatly to my development as a scholar. I would like to thank the FITMIND team, Sarah Cook, Adrienne Aiken, Amber Domenech, and Jaclyn Pittman, for their tireless assistance with data collection and entry. The FITMIND participants deserve special thanks for their commitment to this project. After all the phone calls and hours of assessment, most returned to our laboratory to provide feedback on their experiences. My work, and the FITMIND project itself, was supported by the University of Florida Provost Fellowship in Aging Research and the America Psychological Association Division 20/ Retirement Research Foundation Research Proposal Award. Finally, my parents, Guy and Maja McCoy, deserve all my gratitude, love, and appreciation for their unwavering support on this journey. iv

PAGE 5

TABLE OF CONTENTS Page ACKNOWLEDGMENTS .................................................................................................iv LIST OF TABLES ...........................................................................................................viii LIST OF FIGURES .............................................................................................................x ABSTRACT .......................................................................................................................xi CHAPTER 1 INTRODUCTION...........................................................................................................1 Cognitive Aging............................................................................................................3 Normal Cognitive Aging.......................................................................................3 Mild Cognitive Impairment...................................................................................5 Intraindividual Variability............................................................................................7 Definition of Intraindividual Variability...............................................................7 Variability in Biological Systems..........................................................................8 Variability in Psychological Domains...................................................................9 Variability in Cognitive Functioning of Older Adults........................................10 Variability and Cognitively Impaired Populations..............................................13 Variability and Learning Over Trials..................................................................15 Unresolved Issues Motivating the Current Study.......................................................18 Variability in Cognitive Impairment...................................................................18 Variability in Learning........................................................................................18 2 STATEMENT OF THE PROBLEM.............................................................................19 Intraindividual Variability in Memory and Other Cognitive Domains......................20 Aim One..............................................................................................................20 Hypothesis One...................................................................................................20 Aim Two..............................................................................................................20 Hypothesis Two...................................................................................................21 Intraindividual Variability in Learning.......................................................................21 Aim Three............................................................................................................21 Hypothesis Three.................................................................................................21 v

PAGE 6

3 METHODS....................................................................................................................23 Study Design...............................................................................................................23 Participants.................................................................................................................23 Participant Recruitment.......................................................................................24 Inclusion/Exclusion Characteristics....................................................................25 All participants.............................................................................................25 MCI participants...........................................................................................26 Measures.....................................................................................................................27 Phase 1: Telephone Screening.............................................................................27 Phase 2: Neuropsychological Intake Assessment................................................28 Phase 3: Daily Cognitive Assessment Battery (DCAB)......................................30 Procedure....................................................................................................................37 Overview of Study Phases...................................................................................37 Rationale for the research partner administration protocol..........................39 Compliance monitoring................................................................................40 Group Membership Assignment: Consensus Conference...................................40 Initial Data Preparation and Study Variables.............................................................41 Ceiling and Floor Considerations........................................................................41 Standardization of Scores from the Daily Assessment Battery...........................43 Intraindividual Variability Indices......................................................................44 4 RESULTS......................................................................................................................46 Overview.....................................................................................................................46 Preliminary Analyses..................................................................................................46 Neuropsychological Intake Assessment Data: Participant Neurocognitive Status and Attrition Analysis......................................................................................47 Daily Cognitive Assessment Battery Data..........................................................48 Quality control check: Laboratory to home administration.........................48 Group differences in mean performance over all occasions........................50 Effect of differing delay times for AVLT Delayed Recall..........................54 Distracting environmental variables: Data reduction...................................55 Intraindividual Variability in Memory and Other Cognitive Domains......................56 Aim One and Aim Two.......................................................................................56 Aim One and Aim Two: Review of Analyses.....................................................57 Intraindividual variability differences across groups, based on cognitive status.......................................................................................................57 Data check: Reliability of intrai ndividual variability estimate.....................59 Predicting cognitive status with intraindividual v ariability ..........................61 Intraindividual Variability Over Time: U nderstanding Variability and Performance Relationships..........................................................................................................64 Aim Three.............................................................................................................64 Aim Three: Review of Analyses..........................................................................65 Intraindividual variability over time (occasion), across cognitive status.....65 Relationship between intraindividual va riability and level of performance70 Sources of Intraindividual Variability........................................................................83 vi

PAGE 7

5 DISCUSSION................................................................................................................92 Review of Study Findings..........................................................................................94 Aim One and Aim Two.......................................................................................94 Aim Three..........................................................................................................100 Intraindividual Variability Interrelationships Across Domains........................106 Study Limitations......................................................................................................108 Future Directions......................................................................................................112 Conclusion................................................................................................................115 APPENDIX SAMPLE DAILY WORKBOOK....................................................................................116 REFERENCES................................................................................................................142 BIOGRAPHICAL SKETCH...........................................................................................150 vii

PAGE 8

LIST OF TABLES Table page 3-1. Mean (SD) or N (%) of demographic data................................................................24 3-2. Measures for Neuropsychological Intake Assessment..............................................28 3-3. Measures for Daily Cognitive Assessment (DCAB).................................................31 3-4. Mean performance on the alternative forms of each cognitive task..........................34 3-5. Intercorrelations between mean scores on alternative versions of AVLT List 1......35 3-6. Intercorrelations between mean scores on alternative versions of AVLT Total Score.........................................................................................................................35 3-7. Intercorrelations between mean scores on alternative versions of Backward Digit Span..........................................................................................................................36 3-8. Intercorrelations between mean scores on alternative versions of Symbol Digit......36 3-9. Correlations of demographics and mean IRI scores..................................................45 4-1. Mean performance on neuropsychological measures, by cognitive status................49 4-2. Mean performance on neuropsychological measures, by attrition status..................49 4-3. Comparison of supervised session with first at-home daily session.........................51 4-4. Means (Standard Deviations) for all measures by cognitive status...........................53 4-5. Mean AVLT delay time.............................................................................................54 4-6. Factor loading for distracting environment variables................................................55 4-7. Mean (Standard Deviation) Intraindividual Residual Indices (IRIs).........................59 4-8. Covariation among Intraindividual Variability Indice s over blocks..........................61 4-9. Canonical loadings and classification statistics for discriminant function models...63 4-10. Mean Intraindividual Residual Indices (IRIs) by block and by cognitive status......66 viii

PAGE 9

4-11. Repeated Measures ANOVA: Intraindividual variability predicted by block score and cognitive status..................................................................................................67 4-12. Correlations of mean level of performance and IRI.................................................71 4-13. Regression coefficients for predicting IRI with mean level performance and cognitive status group...............................................................................................72 4-14. Time effects on mean performance..........................................................................76 4-15. Between-person correlations between intraindividual variability and linear slope of performance gains separately for each variable.......................................................81 4-16. Regression coefficients for predicting IRI with linear slope gain in performance and cognitive status group...............................................................................................82 4-17. Intercorrelations of Individual Residual Indices......................................................86 4-18. Coupled variabilities.................................................................................................90 ix

PAGE 10

LIST OF FIGURES Figure page 1-1. Intraindividual variability: Fluctuations in performance.........................................8 1-2. Variability across learning phases.............................................................................16 3-1. Design of the current study........................................................................................37 4-1. Intraindividual variability by blocks..........................................................................69 4-2. Mean performance and intraindividual variability....................................................73 4-3. Growth curves by cognitive status.............................................................................77 4-4. Linear slope and intraindividual variability...............................................................82 x

PAGE 11

Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy UNDERSTANDING THE TRANSITION FROM NORMAL COGNITIVE AGING TO MILD COGNITIVE IMPAIRMENT: COMPARING THE INTRAINDIVIDUAL VARIABILITY IN COGNITIVE FUNCTION By Karin J. M. McCoy December 2004 Chair: Michael Marsiske Cochair: Russell M. Bauer Major Department: Clinical and Health Psychology Intraindividual variability describes fluctuation or transient change in performance, and can be measured by repeated assessment of an ability or trait over a short period of time. Intraindividual variability in biological systems has been demonstrated to indicate systemic compromise (e.g., loss of homeostatic regulation). Consequently, one theory investigators have begun to research is whether intraindividual variability or fluctuation in cognitive performance may be an indicator of cognitive decline. Additionally, a second theory suggests that fluctuations in cognitive performance may be greater during periods of learning acquisition, with a corresponding reduction in variability following the acquisition phase. This study investigated intraindividual variability of cognition on measures of attention, processing speed, working memory, and episodic memory in older adults (over 65 years of age) with and without mild cognitive impairment (MCI), by xi

PAGE 12

assessing performance in these domains daily for 31 days. MCI may be a transitional stage between normal aging and dementia. In particular, the term amnestic MCI describes individuals with focal cognitive impairment in the memory domain; such impairment might foretell future Alzheimers dementia. For this study, MCI is defined as list memory performance 1.5 standard deviations below age-appropriate norms, supplemented by subjective memory complaints and informant report of memory problems, as well as intact cognition in non-memory domains. Results revealed that older adults with amnestic MCI demonstrated a pattern of intraindividual variability and performance level that was consistent with patterns seen in previous studies of cognitively intact or demented individuals. Individuals with MCI demonstrated similar rates of practice-related gain over occasions as did the cognitively intact individuals. However, intraindividual variability was related inversely to performance, for most measures. MCI status was not consistently related to intraindividual variability across the cognitive battery studied. Interrelationships of performance gain slopes with degree of fluctuation did not provide clear evidence for either of the two theories regarding the role of intraindividual variability in practice-related gain or neurocognitive vulnerability. These findings provided evidence that the two types of intraindividual variability described by the current theories in the literature may co-occur in the MCI population. xii

PAGE 13

CHAPTER 1 INTRODUCTION Intraindividual variability (IIV) describes fluctuation or transient change in performance, and can be measured by repeated assessment over a short period of time. A growing literature suggests that inconsistency of cognitive performance, or higher amounts of intraindividual variability, may be particularly meaningful as an indicator of cognitive vulnerability in older adults (Hultsch et al., 2000, Murtha et al., 2002, Walker et al., 2000). This is an intuitively appealing argument, drawing from conventional notions of homeostasis and the importance for self-regulated organisms to show a relatively steady state in most system functions. Interestingly, however, relatively little research has tried to extend the intraindividual variability concept to one of the most cognitively vulnerable populations: cognitively impaired older adults. As the following discussion will demonstrate, this study was designed to investigate the effects of cognitive impairment on intraindividual consistency of performance on cognitive measures. This introduction and review of the literature chapter is organized as follows. First, the distinction between normal cognitive aging and mild cognitive impairment will be considered, focusing primarily on changes in the level of performance on cognitive measures. Second, a review of the emerging body of literature on intraindividual variability, or inconsistency of performance, and aging will follow. Specifically, evidence that increasing cognitive vulnerability is associated with cognitive variability will be investigated. Third, the review will conclude with a 1

PAGE 14

2 discussion of unresolved issues in the current research literature. It is these unresolved issues that motivated the current investigation. The current study, designed to investigate intraindividual variability in cognitive performance in older adults with amnestic mild cognitive impairment (Petersen et al., 2001), was premised on six major findings. First, normal cognitive aging and cognitive impairment are distinguishable by level of performance on cognitive measures. Episodic memory and verbal learning deficits are a prominent source of difference between amnestic mild cognitive impairment (MCI) and unimpaired elders (Bozoki et al., 2001; Chen et al., 2000; Morris et al., 2001; Petersen et al., 2001), although the magnitude and breadth of cognitive differences varies with the number of domains impaired in MCI (Bozoki et al.; Petersen et al., 1999). Second, there is growing evidence in non-impaired elders that intraindividual variability of performance may be an important individual differences characteristic. Persons at lower levels of functioning, and who are older, often show more inconsistency on a variety of cognitive measures (e.g., Anstey, 1999; Bleibeg et al., 1997; Fozard et al., 1994; Rabbitt, Osman & Moore, 2001; Salthouse, 1993; Stuss et al., 1994). Third, in recent studies, persons with more advanced cognitive impairment (dementia) evinced even higher levels of intraindividual variability than either non-demented or physically impaired elders (Hultsch et al., 2000; Murtha et al., 2002). Fourth, in MCI, level of performance may be relatively unaffected in some areas, such as attention and working memory (Mesulam, 2000; Petersen et al., 2001). An unanswered question, however, is whether impairment-related increases in intraindividual variability in functions like attention or working memory might be detectable even before substantial performance level impairments are seen. Fifth, in non-impaired elders,

PAGE 15

3 variability appears to change, qualitatively, over the course of many retest occasions. For some tasks, in which there is learning and performance improvement, individuals show initial performance increments, followed by a stable asymptote. Intraindividual variability before and after reaching asymptote are relatively unrelated to one another; as cognitive responding becomes more strategic and automatic over retest trials, there is a reduction in intraindividual variability (Allaire & Marsiske, 2004). Sixth, the relative absence of strategic learning is a hallmark characteristic of dementia (Bckman et al., 2000; Mesulam, 2000), and is often found in early dementia, pre-dementia, or MCI (Chen et al., 2000; Morris et al., 2001: Petersen et al., 1999). This finding has sparked much debate regarding the definition of MCI as an early stage of dementia, and in fact, experts have opined that most pathological conditions of [amnestic] MCI are likely to be early-stage Alzheimers disease (Petersen et al., 2001, p. 1989). Thus, a question is whether individuals with MCI might show less reduction in intraindividual variability over trials, as well as more stable patterns of individual differences in variability over trials. Cognitive Aging Normal Cognitive Aging Typically, aging is associated with complaints of cognitive impairment, particularly laments of mild to moderate deficits in memory performance compared to functioning at an earlier age. Jorm and colleagues (1994) reported that, in their sample of community dwelling adults aged 70 or older, 62% believed that their memory was worse than earlier in life. Although perceived changes in memory performance are not reliable indicators of objective alterations in cognitive status (Jorm et al., 1997; Turvey et al., 2000), formal neurocognitive assessment in clinical and research settings provides objective evidence that cognitive function is impacted by aging (for a review see

PAGE 16

4 Woodruff-Pak, 1997). Although methodological issues, such as differing findings between cross-sectional and longitudinal research (e.g., Seattle Longitudinal study; Schaie and Willis, 1991; Schaie, 1995), have sparked comment and controversy among investigators, a number of findings appear to be replicable. In particular, consistent evidence of reductions in processing speed (Craik & Salthouse, 2000), and episodic memory performance (new learning and recall; Peterson et al., 1992; Wahlin et al., 1995; for reviews see Anderson & Craik, 2000, and Balota, Dolan & Duchek, 2000) has resulted in general agreement that these domains are particularly susceptible to the effects of normal aging. Development of the diagnostic category age associated memory impairment (AAMI; Crook et al., 1986) reflected initial attempts to describe and delineate the changes in memory function thought to be common in healthy older adults. Specifically, the criteria for AAMI included observable memory test performance one standard deviation below young adult levels of performance. Additional research, critical commentary, and further refinements in the understanding of cognitive functioning of older adults resulted in the presentation of multiple additional diagnostic categories (e.g., age-consistent memory impairment, late-life forgetfulness, and age-related memory decline; see Larrabee, 1996, for a thorough review). Most recent developments in the delineation of age-related effects on memory function include an awareness and focus on individual differences in normal aging, and an investigation of increased interindividual differences in performance and trajectory of change (see Mesulam, 2000). Of primary interest to investigators of the developmental trajectory of cognitive aging is that normal aging and non-normal cognitive aging (e.g., dementia) are generally distinguishable by

PAGE 17

5 level of performance on cognitive measures, even in those domains affected by normal aging (Lishman, 1988; Mesulam, 2000; Petersen et al., 1992). Mild Cognitive Impairment Mild cognitive impairment (MCI; Flicker, Harris, & Reisberg, 1991; Petersen et al., 1999) is the term used to describe non-demented older adults with a slight impairment in cognitive functioning, typically in the memory domain (Celsis, 2000; Petersen et al., 2001). MCI differs from normal age-related changes of cognitive functioning, since the defining criteria require impaired cognitive function when compared to ageand education-matched peers. Petersen and colleagues (2001) indicate that MCI commonly refers to impairment in memory functioning, and that the term would be more accurately labeled amnestic MCI. MCI, specifically amnestic MCI, is considered by many to capture the boundary or transitional condition between normal aging and mild Alzheimers disease (Morris et al., 2001; Peterson et al., 2001). Individuals with amnestic MCI typically progress to Alzheimers Disease at a rate of 10 15% per year (Petersen et al., 1999; Tierney et al., 1996). Morris and colleagues (2001) reported a five year conversion (MCI to Alzheimers disease, based on Clinical Dementia Rating score) of 40 60%, with a nine-year conversion rate in excess of 90%. As investigators have increasingly focused on characterizing reductions in cognitive performance that might capture a pre-clinical phase of dementia, particularly Alzheimers disease (Celsis, 2000), demonstrable and reliable differences in cognitive performance between older adults with normal, intact cognitive functioning, and those with impaired cognition are recognized to be of significant value for clinicians and researchers. Investigations into the predictors of amnestic MCI may provide means for early detection, intervention, and eventually treatment for Alzheimers disease (Knopman

PAGE 18

6 et al., 2001; Petersen et al., 2001). Typically, as with investigations into normal cognitive aging, differences along the normal aging MCI Alzheimers disease continuum are made on the basis of level of performance on neurocognitive measures. However, level of performance comparisons on a single occasion of measurement on objective cognitive measures are not the sole means to a successful and accurate diagnosis of cognitive impairment. In fact, from both a developmental perspective and a neuropsychological approach, repeated assessment (e.g., once every six months) to document change within an individual over time provides a more sensitive indicator of impairment in cognitive functioning especially at early or premorbid stages of a dementing illness (Daly et al., 2000; Elias et al., 2000; Petersen et al., 2001). Substantial research to characterize both normal and impaired cognitive aging follows a basic longitudinal methodology in order to accurately group participants. For example, Schaie and Willis (1991) have observed that the results of a cross-sectional comparison (the level of functioning approach) and 20-year longitudinal follow-up of adults of varied ages supported different conclusions. Several differences between the performance of the young and elderly participants in the cross-sectional arm of the study were a reflection of cohort differences rather than change over time: results of the cross-sectional analyses would have led to incorrect conclusions regarding the nature of cognitive functioning for a number of participants. Although the, albeit often unstated, research goal in the area of cognitive impairment in older adults is to determine a means to accurately diagnose impairment based on a one-time assessment, this approach has been less than completely successful for the diagnosis of many types of cognitive dysfunction in the older years (Knopman et

PAGE 19

7 al., 2001). Delineation of the trajectory of change of functioning in various cognitive domains is crucial for accurate diagnosis and prognosis. Typically, this assessment of change occurs over months or years. Intraindividual variability methodology may provide a means to determine the trajectory of change and/or diagnostically useful patterns in cognitive performance in a much shorter time frame, possibly even prior to observable changes in overall level of performance. Intraindividual Variability Definition of Intraindividual Variability The measurement of intraindividual variability (or fluctuation) in cognitive performance over a shorter period of time may provide useful information in the diagnosis or identification of cognitive dysfunction. Long-term changes in cognitive functioning within an individual, as well as increasing differences in level of cognitive performance between persons may be based on permanent alterations in intraindividual variability (Li & Lindenberger, 1999; Siegler, 1994). Intraindividual variability (IIV), defined as fluctuation or change in performance over a short period of time, provides a measure of an individuals hum around their mean, or "steady-state" level (Nesselroade, 1991). The intensive, repeated measurements necessary to assess intraindividual variability provide a more accurate representation of an individuals true mean performance, as well as a finer grained resolution of fluctuations in performance. Figure 1-1 depicts the hum or variability around the mean performance (indicated by the regression line) for two individuals (Allaire, 2001). Variability in performance from one session to the next has sometimes erroneously been considered to be measurement error, yet, intraindividual variability can be reliably measured, and is of sufficient magnitude to be important. Measurement error and

PAGE 20

8 practice effects can be distinguished from lawful but transient changes in performance from session to session (Hultsch et al., 2000; Li et al., 2001). When assessing cross-occasion intraindividual variability in story recall via weekly assessments for two years, Hertzog et al. (1992) found considerable intraindividual variability in performance across occasions; more than 20% of this variability was reliable variance not associated with practice, alternative forms, or other systematic changes over time. Rabbitt et al. (2001) found systematic changes between sessions independent of circadian variability or practice. 0510152025303515913172125293337414549535761656973778185899397101105109113117Participant 21 IRI = 1.54Participant 39 IRI = 1.99Best Fitting Regression Line Best Fitting Regression Line Figure 1-1. Intraindividual variability: Fluctuations in performance. From Allaire, 2001. Variability in Biological Systems Not only is intraindividual variability measurable, but research findings also suggest that extreme intraindividual variability within a system is an indicator of systemic

PAGE 21

9 dysfunction. Behavioral research into the intraindividual variability of psychologically relevant constructs draws on the extensive research in physiology, biology, medicine, and allied disciplines, where investigators consistently have found that increased variability within a system is associated with increased age and poorer overall functioning (Britton, 1997; Fluckiger et al., 1999; Guimares & Isaacs, 1980; Hausdorf et al., 1997; Higgins, 1997; Pagani, 1999). Intraindividual variability within a fairly narrow range (e.g., homeostatic regulation) is one defining characteristic of biological systems. At the behavioral level, short-term, narrow-ranged intraindividual variability can be considered indicative of adaptive ongoing processes in response to an ever-changing environment (Nesselroade et al., 1996). In contrast, variability characterized by extreme or erratic fluctuations may reflect a breakdown in the homeostatic regulatory system; this type of dramatic variability may be found in the biological or psychological arena. Evidence of such lability in the intraindividual variability of cognitive performance in older adults with mild cognitive impairment would be consistent with the findings in the medical literature regarding the correlation between increased variability and poorer systemic functioning. Variability in Psychological Domains In the psychological literature, studies examining variability in adults have focused on describing the extent to which particular psychological constructs, such as affect and mood, self-esteem, and personality are marked by variability. Similarly, explorations of intraindividual variability in the older adult population have described the correlation of fluctuations in the domains of affect and mood, self-efficacy, and world views and religious beliefs. Kernis et al.s (1993) investigation of variability in self-esteem in adults over a four day period illustrates the unique contribution of intraindividual

PAGE 22

10 variability, as they found that day-to-day deviations in self-esteem correlated with fluctuations of perceived competence and social acceptability independent of mean level. Eizenman et al.s (1997) examination of intraindividual variability in perceived control, locus of control, and perceived competence in a sample of older adults over the course of seven months provides further demonstration of the distinction between variability and mean performance. While the mean of the two aspects of control was not a significant predictor of mortality, the two variability scores for locus of control and perceived control added significantly to the prediction of five year mortality. The predictive utility of variability in this study is consistent with findings from other studies of self-efficacy and self-esteem reporting that variability is a better predictor of certain outcomes than mean level (Butler et al., 1994; Kernis et al., 1993). Variability in Cognitive Functioning of Older Adults Recently, researchers have begun to examine the extent to which cognitive functioning in healthy older adults is characterized by intraindividual variability. Overall, results suggest a significant relationship between increased intraindividual variability and age. For instance, Fozard, Vercruyssen, Reynolds & Hancock (1994) reported that, over an 8-year longitudinal study in which participants aged 17 to 96 years were repeatedly assessed, intraindividual variability in reaction time increased with increasing age. Results reported by Shammi, Bosman, and Stuss (1998) indicated that within-person variability was greater on two psychomotor tasks (i.e., finger tapping and choice reaction time), and a time-estimation tasks for an older sample of participants when compared with younger adults. More recently, MacDonald, Hultsch, & Dixon (2003) reported greater inconsistency of performance on reaction times tasks in older

PAGE 23

11 adults, even after controlling for differences in response speed. Baseline differences in variability were associated with variability 6 years later. Studies focusing solely on intraindividual variability in cognitive performance of older adults (i.e., without younger aged comparison groups) have also provided evidence establishing a link between age and cognitive intraindividual variability. Salthouse (1993) reported reaction times in four independent samples of older adults; results from this study indicated that, across the four samples, increased age was associated with increased intraindividual variability of reaction time over 90 consecutive trials. Similarly, Ansteys (1999) examination of age differences in older adults performance on measures of simple and complex reaction time presented in both the visual and auditory modalities provides evidence that intraindividual variability in performance on some reaction time measures (complex auditory task and simple and complex visual tasks) is positively related to age. Extrapolating from findings that variability in performance appears to increase with chronological age, researchers have speculated that increased intraindividual variability of a behavior or within a measured construct might be an indicator of a more general underlying self-regulatory breakdown in performance in that domain. Extending this notion to cognitive functioning, some investigators theorize that increased intraindividual variability in cognitive functioning is indicative of compromised neurobiological functioning and an increase in neural noise (Bruhn & Parsons, 1977; Hendrickson, 1982; Li and Lindenberger, 1999). If this assumption is correct, then individuals who, on average, perform more poorly on tests of cognitive functioning may exhibit more intraindividual variability, implying a negative relationship between intraindividual

PAGE 24

12 variability and mean performance. Consistent with this hypothesis, Rabbitt, Osman & Moores (2001) examination of covariation between week-to-week intraindividual variability in reaction time over a 36-week period and performance on a test of reasoning ability in older adults indicated that the magnitude of intraindividual variability was greater for the individuals who performed poorly on the battery of intelligence tests. Li and her colleagues (2001) also found evidence of a relationship between cognitive intraindividual variability and mean performance, although their results were somewhat more mixed. Specifically, these authors found that the association between intraindividual variability and mean performance was in the expected negative direction for a memory task, with higher levels of intraindividual variability associated with poorer overall performance. However, a significant correlation between intraindividual variability and mean performance was not found for working memory. Furthermore, a positive correlation between variability and mean performance was found for a spatial memory task, indicating that higher overall spatial memory performance was related to greater intraindividual variability. These results, which contrast with other findings, suggest that intraindividual variability in performance in some cognitive domains may not be associated with vulnerability or impairment in overall level of performance. In fact, recent work by Allaire and Marsiske (2004), consistent with the research presented by Li and colleagues, suggests that variability in performance may also be influenced by strategic learning, such that increased variability during learning and memory may reflect intact, adaptive processes. As an alternative to the hypothesis of a negative relationship between intraindividual variability and mean performance, significant fluctuations or lability in

PAGE 25

13 intraindividual variability may be indicative of compromised neurocognitive functioning independent of overall performance. Past research, particularly in the domains of self-esteem and perceived control (Eizenman et al., 1997; Kernis et al., 1993), which demonstrated that intraindividual variability uniquely predicted outcome independent of mean level, provides evidence that the assessment of intraindividual variability may be valuable in the absence of differential mean performance. Similar findings in research on cognitive functioning in older adults may provide a unique tool to early diagnosis of potential future declines in cognition, as seen in Alzheimers disease, since early or pre-clinical phases of the disorder are characterized by a lack of reduction in the observed level of performance of functioning in a number of domains which will later become impaired. Performance in these domains may be characterized by increased variability prior to the disorder-related reduction in overall level of performance. Variability and Cognitively Impaired Populations To date, there is a paucity of published studies of intraindividual variability with older patient populations, and none reported which examines variability in cognitive functioning in individuals with mild cognitive impairment. One of the few studies of neurologically impaired participants is Shifren, Hooker, Wood, & Nesselroades (1997) examination of variability of mood in older adults diagnosed with Parkinsons disease, via participant completion of a mood questionnaire daily for 70 days. Using dynamic factor analysis, the authors examined dimensionality of mood within individuals over time as well as the extent to which mood was related to itself over time. The number of dimensions (factors) needed to explain the variation in the day-to-day variability in mood varied between different subsets of participants. Similarly, the relationship between

PAGE 26

14 mood on subsequent days was different within individuals, in that mood on one day influenced subsequent daily mood ratings differently for participants. This study highlights intraindividual variability over time of a particular construct typically thought to be highly stable and demonstrates the transactional nature of psychological processes. Additionally, this study demonstrates the utility and feasibility of utilizing an intraindividual variability approach with neurocognitively compromised patient populations. Hultsch et al. (2000) examined intraindividual variability in performance on a reaction time task, as well as in two measures of memory functioning (i.e., word and story recognition) in healthy adults, cognitively intact adults with arthritis, and adults diagnosed with mild dementia. Collapsing across the three groups, the authors found that intraindividual variability on the cognitive tasks was positively associated with the mean latency scores and negatively related to accuracy scores within each measure. That is, more intraindividual variability was associated with slower average latency scores and poorer average accuracy scores. In addition to this correlational evidence, Hultsch and colleagues also found that the participants in the mildly demented group, presumably the group with poorest overall cognitive functioning, exhibited a significantly greater level of intraindividual variability on all three cognitive tasks than the participants in the healthy and arthritic groups. In their comparison of cognitively intact older adults with cognitively impaired elders (patients with Alzheimers disease and patients with frontal lobe dementia), Murtha and colleagues (2002) demonstrated that intraindividual variability of performance on three distinct reaction time measures distinguished the healthy older

PAGE 27

15 adults and the two groups of patients. Walker et al. (2000) utilized attentional measures of intraindividual variability in cognitive performance to distinguish healthy, cognitively intact older adults from patients with Alzheimers disease and patients with Lewy-Body dementia. Variability in cognitive performance on measures of vigilance and reaction time confirmed the clinical assessment of variability and differentiated the groups by diagnostic category. Similar results have been found in non-aging studies, where the cognitive functioning of individuals who had experienced a traumatic brain injury was characterized by higher levels of intraindividual variability (Bleiberg et al.,1997; Hetherington et al., 1996; Stuss et al., 1994). Although, to date, there have been no published reports of intraindividual variability assessment of individuals with mild cognitive impairment, variability assessment may be useful approach for this population. Individuals with MCI suffer from cognitive impairment, suggesting that investigation of intraindividual variability in this group may provide further insights into the transition from normal aging to cognitive impairment. Variability and Learning Over Trials Recent work by Allaire (2001; Allaire & Marsiske, 2004), suggests that temporal patterns in intraindividual variability of cognitively intact older adults may reflect learning over trials. Findings that variability increased prior to the attainment of maximum performance suggest that increased intraindividual variability during the learning phase might reflect acquisition of strategies (a finding consistent with Siegler's (1994) work on strategy learning in childhood), while subsequent reduction in intraindividual variability reflects consistent implementation of the strategy. This is summarized in Figure 1-2. Trial-and-error learning is likely to be adaptive, especially in

PAGE 28

16 conjunction with increased overall level of performance. Excessive variation around an asymptotic, or non-increasing, level of functioning is more likely to reflect inconsistency and vulnerability, since the observed variability is not correlated with increased strategy acquisition or learning. Older adults with memory impairment (e.g., Alzheimers disease) are distinguishable by poor recall on episodic memory tasks, suggesting that learning (e.g., strategy acquisition) is not taking place (Bckman et al., 2000; Chen et al, 2000; Mesulam, 2000; Morris et al., 2001; Petersen et al., 1999). Thus, it is likely that patients Figure 1-2. Variability across learning phases. From Allaire, 2001 Variability is maladaptive and reflects performance inconsistency Consequently, negative associations between level and variability (lower functioning individuals = more variability) Will tend to be observed during periods of stable or asymptotic functioning More likely to be seen for very easy or very hard tasks on which participants are not getting better at over time Variability is adaptive and reflects strategy use Consequently, positive associations between mean performance and variability (better functioning individuals = less variability Will tend to be observed during periods of linear growth in performance More likely to be seen for hard-to-learn and complex tasks; tasks on which learning is possible but mastery has not been achieved Strategy Acquisition Phase Strategy Implementation Phase O ccasions of Measurement Performance

PAGE 29

17 with dementia may demonstrate a unique pattern of intraindividual variability in cognitive performance over time, typified by no noticeable reduction in performance fluctuations. This pattern in performance would likely differ from that expected in cognitively intact older adults, since individuals capable of learning new material woulddemonstrate variability in performance during the strategy acquisition stage, followed by a redu In con e. likely an ers cline even in the absence of changes in overall level of ction in variability during the strategy implementation phase (once task-specific strategies are acquired). Individuals with MCI, especially older adults with amnestic MCI, who are likely atincrease risk of converting to Alzheimers disease, may demonstrate a pattern of intraindividual variability similar to that expected by patients with Alzheimers disease. junction with an expected lack of, or significant reduction in, new learning, theseindividuals may demonstrate a lack of reduction in intraindividual variability over timIn summary, intraindividual variability, or fluctuations in performance around amean, is measurable and provides an unique predictor of outcome in biological and psychological research. Past research has demonstrated that extreme intraindividual variability is an indicator of systemic compromise in biological systems, and is indicator of compromised function in cognition, as patients with neurological disorddemonstrate increased intraindividual variability in cognitive performance. In cognitively intact older adults, intraindividual variability in cognitive performance declines over time, corresponding with practice and strategy acquisition. Increased intraindividual variability is likely in mild cognitive impairment, and may provide a unique predictor of future de

PAGE 30

18 performance, espece with repeated y ormance, there have he likelihood that increased intraindividual variability in an intacta has been no investigation of the pattern of intraindividual variability in the cognitive performance of cognitively impaired individuals as they are exposed to a task repeatedly. ially if observed variability does not decline over tim exposure to the same tasks. Unresolved Issues Motivating the Current Study Variability in Cognitive Impairment Past research into cognitive aging has provided significant evidence that normal cognitive aging, mild cognitive impairment, and progressive dementing disorders are distinguishable by level of performance in specific domains of cognition (e.g., memory).An unanswered question is the degree to which differences in intraindividual variabilitin performance in the same cognitive domains might likewise distinguish these groups. Although increasingly the evidence suggests that variability in performance indicates vulnerability in a system, and may foretell future impairments in perf been no investigations of t cognitive domain might predict future decline in that domain. Variability in Learning Previous research suggests that intraindividual variability in cognitive performance may be an indicator of strategy acquisition during repeated exposures to novel task. A pattern of increased variability during a learning phase, followed by a reduction in variability in performance when overall level of performance reaches an asymptote, has been observed in cognitively intact older adults. To date, there

PAGE 31

CHAPTER 2 STATEMENT OF THE PROBLEM Although investigators have continued to focus on identifying the correlates and predictors of current and future impairment in level of performance in cognitive domains of function, only recently has any attention been paid to the possible contribution of intraindividual variability in performance to the diagnosis or prediction of cognitive decline. The primary concern of this research is to examine the relationship between cognitive impairment and intraindividual variability in cognition in older adults. The Study Aims draw upon what is known about changes in level of performance on neurocognitive measures in older individuals with mild cognitive impairment, and upon a small number of studies that have examined intraindividual variability in cognitive functioning in unimpaired and cognitively impaired elders (e.g., Hultsch et al., 2000; Murtha et al., 2002; Walker et al., 2000). Additionally, the aims reflect recent work which indicates that intraindividual variability during learning may represent strategy acquisition, while variability without concurrent learning (either because individuals have plateaued or because they cannot learn) may reflect compromised function (Allaire & Marsiske, 2004; Li et al., 2001). Finally, the study aims reflect recent innovations in the literature hinting at the potentially differential contribution of varied temporal resolutions of intraindividual variability measurement (Strauss et al., 2002; Walker, 2000). The study investigated whether (a) there exist differences in intraindividual variability between older adults with memory impairment and those whom are 19

PAGE 32

20 cognitively intact, and whether these differences are observable both in memory and in other domains of cognitive functioning, (b) such group differences in variability might be a more sensitive distinguishing characteristic of cognitive impairment than level of performance, especially if non-memory domains revealed heightened intraindividual variability at a time when level of performance had not declined substantially in these domains, and (c) learning-related (e.g., practice effects or level of performance) changes in variability are more characteristic of non-impaired than impaired elders. Thus, the study has three major aims in two general areas. Intraindividual Variability in Memory and Other Cognitive Domains Aim One To investigate whether greater intraindividual variability in memory functioning is seen in older adults with amnestic mild cognitive impairment, compared to cognitively intact elders. Hypothesis One If greater intraindividual variability is an indicator of neurological compromise, then older adults with amnestic mild cognitive impairment (MCI; defined as memory performance 1.5 SD or more below ageand education-appropriate norms at baseline assessment; Petersen et al., 2001) should exhibit more intraindividual variability in memory performance over 31 occasions of measurement when compared to older adults with no memory impairment at baseline. Aim Two To investigate whether cognitive tasks measuring neurocognitive domains that typically show relatively less impairment in level of performance in mild cognitive impairment (e.g., attention, processing speed) nonetheless reflect impairment-related

PAGE 33

21 increases in intraindividual variability in older adults with mild cognitive impairment relative to unimpaired elders. Hypothesis Two If mild cognitive impairment, as indexed by memory dysfunction, reflects the onset of a more general cognitive decline process (i.e., 12 15% of persons with MCI convert to dementia annually; Petersen et al., 2001), and if individual variability is an indicator of cognitive compromise related to cognitive decline, then older adults with memory impairment will also exhibit greater intraindividual variability in other cognitive domains assessed over 31 occasions even before showing mean level impairment. Intraindividual Variability in Learning Aim Three To investigate whether, in conjunction with expected differences in overall learning across trials within session and across sessions, memory-impaired and cognitively intact individuals show different patterns in intraindividual variability on memory tasks over time. Extant research (Allaire & Marsiske, 2002; Li et al., 2001) suggests that, on tasks in which elders show retest-related learning, variability has several properties: (a) it tends to be positively associated with level of performance, (b) it is uncorrelated with the variability observed after individuals reach asymptotic performance, and (c) it is reduced over trials. Hypothesis Three Since mild cognitive impairment (when defined by memory performance) is characterized by increased difficulty of new learning (Bozoki, et al., 2001; Chen et al., 2000; Morris et al., 2001), it is expected that, relative to non-impaired elders, persons with memory impairment will show relatively stable intraindividual variability patterns

PAGE 34

22 throughout the period of study, reflecting little to no strategy acquisition. In contrast, unimpaired elders will be more likely to show decrements in the magnitude of learning-related variability. That is, the performance of the cognitively intact older adults will show a positive relationship between variability and improvements in performance with decrements in the magnitude of learning-related variability over the 31 occasions. As initial performance improves, initial variability will decrease until remaining variability is no longer associated with performance at asymptotic performance.

PAGE 35

CHAPTER 3 METHODS Study Design The current short-term longitudinal study used a mixed betweenand within-subjects design to investigate intraindividual variability on tasks of attention, working memory, and immediate and delayed episodic memory across two groups (older adults who meet criteria for mild cognitive impairment [MCI] vs. cognitively healthy). Performance on these tasks was assessed daily for thirty-one days (an initial assessment in the laboratory followed by thirty daily assessment sessions at-home). This design is an extension of the multivariate, replicated, single-subject, repeated measures design (MRSMR; Jones & Nesselroade, 1990; Nesselroade & Ford, 1985), which traditionally calls for the repeated (e.g., daily, weekly, monthly) multivariate assessment of an individual participant over a finite period of time, replicated over multiple individuals. The design has been extended in the current study by the placement of individuals into one of two groups based on cognitive status prior to the initiation of the repeated measures study. Participants Sixty-eight community-dwelling volunteers aged 65 years and older (age range 65 87 years) participated in the study. Fifteen participants were subsequently classified as having amnestic mild cognitive impairment (MCI) and 53 were classified as cognitively intact (Non-MCI). Details of the consensus procedure by which these classifications were achieved are provided below. Table 3-1 displays the demographic information for 23

PAGE 36

24 the full sample and each group. The only significant difference between the groups on demographic variables was in the sex distribution (p = .033), with a higher proportion of males in the MCI group. The MCI group demonstrated equivalent predicted IQ (as calculated on the North American Adult Reading Test) and a significantly lower mean score on both the Telephone Interview of Cognitive Status (TICS) and the Mini-Mental Status Exam (MMSE). This is consistent with the criteria for MCI group, as outlined below. Table 3-1. Mean (SD) or N (%) of demographic data. Total Sample MCI Non-MCI p value (N= 68) (n = 15) (n = 53) Age 78.01 (5.75) 76.60 (7.76) 74.57 (5.05) .229 Education 16.12 (2.66) 16.33 (2.90) 16.06 (2.62) .725 Sex .033 Males 29 (42.60) 10 (66.70) 19 (35.80) Females 39 (57.40) 5 (33.30) 34 (64.20) Race .717 Caucasian 63 (92.60) 14 (93.30) 49 (92.50) Other 5 (7.4) 1 (6.70) 4 (7.50) TICS 36.10 (3.72) 31.67 (3.64) 37.38 (2.61) .000 MMSE 28.73 (1.24) 27.55 (1.75) 29.05 (0.84) .018 Predicted IQ 113.64 (7.32) 110.69 (5.92) 114.47 (7.51) .077 Note. TICS = Telephone Interview of Cognitive Status, MMSE = Mini-Mental Status Exam, Predicted IQ from NAART (North American Adult Reading Test). Participant Recruitment Participants were recruited from the community, with a main focus on attaining a sample of community dwelling older adults with and without mild cognitive impairment. Primary recruitment for cognitively healthy older adults was through community events, community organizations, and an article with advertising in a senior-oriented magazine (the Senior Times). The second strategy, which served as recruitment for a number of mildly cognitively impaired participants, was to identify individuals who expressed

PAGE 37

25 interest in participating in research via established participant pools maintained by faculty and students associated with the Institute on Aging. Recruitment of cognitively impaired individuals stressed subjective memory complaints, confirmed objectively by the neuropsychological assessment for qualification into the study. As outlined in detail below, participants in the mild cognitive impairment group were required to meet criteria for amnestic MCI (Petersen et al., 2001). Participants were recruited individually and in dyads. Each target participant was asked to identify a research partner (e.g., spouse, in-home caregiver, neighbor, or child) to assist with administration and monitoring of the 30 at-home cognitive assessments. Target participants and research partners were required to consent to participate. Individuals unable to identify a willing partner were matched with other participants in similar circumstances. Participants recruited in dyads (e.g., spouses) were allowed to enroll in the study as both participants and research partners, as long as both met all inclusion and exclusion criteria. Materials were coded so as to provide different cognitive tests to each member of a dyad each day. Inclusion/Exclusion Characteristics All participants All participants were over 65 years of age. Each was screened via telephone in order to assure they met the following criteria by self-report: (i) no severe dementing illness, (ii) no history of closed head injury with loss of consciousness, (iii) no other neurological or major medical illnesses, (iv) no self-reported severe uncorrected vision or hearing impairments, (v) no psychiatric disturbance sufficient to warrant inpatient psychiatric treatment, (vi) no extensive drug or alcohol abuse, and (vii) self and research partner willingness to participate in repeated cognitive evaluations. These criteria,

PAGE 38

26 common in neuropsychological studies of older adults, were utilized to rule out alternative potential etiologies for cognitive impairment and restrict focus to MCI. Additionally, during this initial screening, Telephone Interview of Cognitive Status (TICS; Brandt, Spencer, & Folstein, 1988) score below 30 points (the published cut score for dementia), resulted in exclusion of two individuals. An exception was made for three cases where point deductions were restricted to an item assessing delay verbal recall ability. All participants who met these initial criteria were assessed in person, with the Neuropsychological Intake Assessment measures (see below), and were required to (a) attain a Mini-Mental Status (Folstein et al., 1975) score above 23, (b) perform within one standard deviation of ageand education-adjusted norms on neurocognitive measures in non-memory domains, and (c) demonstrate no evidence of impairment on Activities of Daily Living (ADL), as measured by proxy interview with the Blessed Dementia Scale (Blessed et al., 1968; following criteria for Mild Cognitive Impairment, Petersen et al., 2001). Participants in the cognitively intact group were also required to (d) exhibit intact (within one standard deviation of ageand education-adjusted norms) performance on memory measures (i.e., no impairment in the memory domain). MCI participants In addition to the above criteria (a, b, c), participants in the MCI group were required to demonstrate impaired memory function, measured by a score 1.5 standard deviations below ageand education-adjusted norms on a list learning memory task (Hopkins Verbal Learning Test Revised; HVLT-R; Brandt & Benedict, 2001). The specific scores on the HVTL-R used were Total Recall, Delayed Recall, and Percent Retention. Additionally, participants in this group were required to receive a Clinical

PAGE 39

27 Dementia Rating (CDR; Hughes et al., 1982; Morris, 1993) score below 1.0 (i.e., 0.0 or 0.5). This requirement insured that observed cognitive impairment was mild. Higher scores on the CDR indicate more substantial memory impairment and/or more substantial impact on daily functioning. These basic requirements were confirmed and considered when group assignments were made as part of the Consensus Conference, described below. Measures Participants were exposed to three study phases, an initial telephone screening, a neuropsychological intake assessment, and the daily assessment. While the measures are explained by phase, the phases of study are described in detail in the Procedure section of this chapter. Phase 1: Telephone Screening Initial screening of participants took place using the brief Telephone Interview for Cognitive Status (TIC; Brandt, Spencer, & Folstein, 1988). Using a cut score of 30 points to distinguish demented individuals from cognitively intact individuals, the TICS has a sensitivity of 94% and a specificity of 100% (Brandt, Spencer & Folstein). Thus, the TICS allowed for initial exclusion of demented individuals from the study. Additionally, since individuals with MCI were more challenging to recruit and identify than cognitively intact older adults, performance on the recall memory item of the TICS was utilized to assist in selecting individuals with a greater likelihood of meeting criteria for amnestic mild cognitive impairment. The telephone screening, including exclusion criteria, TICS interview, and description of the study took approximately 30 minutes.

PAGE 40

28 Phase 2: Neuropsychological Intake Assessment Following informed consent procedures approved by the University of Florida Institutional Review Board, participants were assessed individually, in a standardized manner, on a selection of commonly used neuropsychological measures, in order to determine appropriate assignment to the amnestic MCI or unimpaired groups. Measures are listed in Table 3-2 (in a thematic grouping, not test order). This assessment took approximately two hours for cognitively intact individuals, and approximately three hours for individuals later deemed to meet criteria for MCI. Table 3-2. Measures for Neuropsychological Intake Assessment. Cognitive Domain Measure Source Mini Mental Status Examination (MMSE) Folstein, Folstein, & McHugh, 1975 Overall Cognitive Functioning North American Adult Reading Test (NAART) Blair & Spreen, 1989 Attention & Working Memory Trail Making Test A & B Reitan, 1992 Hopkins Verbal Learning Test Revised (HVLT-R) Brandt & Benedict, 2001 Rivermead Behavioral Memory Test (prose memory subtest) Wilson, Cockburn & Baddeley, 1985 Memory Brief Visuospatial Memory Test Revised (BVMT-R) Benedict, 1997 Boston Naming Test Second Edition (BNT) Kaplan, Goodglass & Weintraub, 2001 Language Controlled Oral Word Association Benton & Hamsher, 1989 Visuospatial Rey-Osterrieth Complex Figure Copy Rey, 1941 Mood Geriatric Depression Scale (GDS) Yesavage, 1983 Center for Epidemiological Studies Depression Scale (CES-D) Radloff, 1977 Blessed Dementia Scale Blessed, Tomlinson, & Roth, 1968 Functional Assessment Clinical Dementia Rating Scale (CDR) Hughes, Berg, & Danzinger, 1982; Morris, 1993

PAGE 41

29 Rationale for inclusion of measures in Phase 2: Neuropsychological intake assessment. Measures were selected in order to ensure that participants fit the inclusion/exclusion criteria for amnestic mild cognitive impairment, as outlined by Petersen and colleagues (2001). These criteria require adequate assessment of memory functioning, in order to establish that performance on a standardized assessment of new learning and memory falls 1.5 standard deviations below expected for age and education. Three measures of memory were chosen to assess verbal and non-verbal memory. The HVLT-R immediate and delayed memory scores had the greatest weight in determining memory impairment. Additionally, criteria for amnestic MCI require intact functioning in non-memory domains of cognition, thus all participants underwent assessment of attention, working memory, language, and visuospatial skills in order to ensure unimpaired status in these areas. Following standard clinical neuropsychological practice, the measures chosen for this battery allow broad and diverse assessment of the relevant domains within a reasonable time period. Additionally, these measures have been previously utilized in the determination of amnestic MCI (Bozoki, et al., 2001; Petersen, et al., 2001), thus providing appropriate method for sample comparison and generalization purposes. Two individuals did not demonstrate intact cognitive functioning in the non-memory domains during the Neuropsychological Intake Assessment. After consultation with a clinical neuropsychologist, these individuals were provided with referrals for further clinical assessment of cognitive functioning and dropped from the remainder of the study. Both indicated appreciation for the referrals.

PAGE 42

30 As noted above, under inclusion criteria, in order for participants to be considered unimpaired, performance during the Neuropsychological Intake Assessment on all cognitive measures in memory and non-memory domains was required to be within one standard deviation of age and education adjusted norms. Two mood measures were included to assess any contribution depression may have on memory and attentional performance, and rule out depression as a cause for any observed impairments. The Center for Epidemiological Studies Depression Scale (CES-D) was selected in order to screen for clinical depression, while the Geriatric Depression Scale was selected in order to screen for depressive symptoms unique to older adults. Notably, telephone screening questions about psychiatric history adequately screened out individuals diagnosed with clinical depression. None of the individuals in the MCI group scored above the mild depression cut-off score of 16 on the CES-D. In addition to the neuropsychological assessment of the participant, the participants research partner was asked to participate in a semi-structured interview of cognitive symptoms, daily functioning, and activities of daily living. This interview took approximately 15 minutes to complete after informed consent procedures. Phase 3: Daily Cognitive Assessment Battery (DCAB) This battery took participants 10 20 minutes to complete each day and consisted of measures outlined in Table 3-3. Modifications to standard test administration were necessary in order to facilitate in-home assessment. The participants research partner was trained to assist in the administration of this brief battery. Specifically, the research partner was informed as to how to assist with setting the timer as well as directed as to the administration of the Digit Span test and recording of responses. Additional

PAGE 43

31 description of research partner orientation is outlined below. The Appendix provides a copy of a sample Daily Cognitive Assessment Workbook. Table 3-3. Measures for Daily Cognitive Assessment (DCAB). Domain of Function Measure Source Attention & Working Memory Digit Span (Forward and Backward) subtest of WMS-III Wechsler, 1997 Attention & Processing Speed Symbol Digit Substitution and Number Copy (both timed) Smith, 1982 Immediate & Delayed Recall Memory & Learning List Learning (Rey Auditory Verbal Learning Test modified for visual presentation; three learning trials plus delayed recall trial) Rey, 1964 Sleep Sleep Diary (Sleep Time & Efficiency) Lichstein, 1999 Mood Positive and Negative Affect Scale Watson, Clark & Tellegen, 1988 Internal and External Environment Environmental Distractions Questions Stress Ladder Allaire, 2001 Rationale for inclusion of measures. The domains measured and specific tasks within each domain were selected to provide appropriate assessment for the aims of the study. While level of performance on attentional and working memory measures is often similar across cognitively intact older adults and those diagnosed with mild cognitive impairment, it is not clear that intraindividual variability is similarly comparable. Measures in these domains were chosen to investigate the susceptibility of attention and working memory to changes in intraindividual variability. Deficits in new learning and explicit memory are indicative of an amnestic mild cognitive impairment and may be a precursor to a neurocognitive illness such as Alzheimers disease (Morris et al., 2001; Petersen et al., 2001), thus measurement of this domain provided an evaluation of the correlation and inter-relatedness of level of performance and intraindividual variability.

PAGE 44

32 Negative affect has been shown to negatively impact performance on attentional and memory measures, thus, assessment of daily positive and negative affect was included to control for the influence of fluctuations in mood on cognitive performance. Increased age is associated with poor sleep and increased complaints about sleep quality (Bliwise et al., 1992; Ganguli, Reynolds, & Gilby, 1996). Sleep complaints and insomnia have not been consistently shown to be associated with cognitive performance, although both are highly correlated with psychological factors (McCrae, et al., in press). Sleep time and sleep efficiency (sleep time/time in bed) were assessed via Sleep Diaries (Lichstein et al., 1999) for each nights sleep during the Daily Cognitive Assessment protocol in order to investigate the impact of sleep patterns on cognitive functioning. Questions regarding the internal and external test-taking environment allowed for measurement of discomfort (pain, stress, tiredness) as well as distractions (interruptions, noise level, other people around). The specific questions were based on those used in a previous daily battery study (Allaire, 2001). Alternative forms for DCAB. Sixteen comparable alternative forms of the cognitive measures were used, fifteen for the Daily Cognitive Assessment Battery, and one for the initial orientation session. Thirteen alternative AVLT lists have been published (Schmidt, 1996), and the three additional lists required for the current study were designed with words with similar frequency and imageability. The alternative forms of Digit Span and Symbol Digit were developed using simple random substitutions of existing stimuli. This carefully matched, algorithm based process was utilized to ensure comparability across forms. The fifteen alternative forms devised for the at-home study were given twice to each participant, in random order, counterbalanced across

PAGE 45

33 participants in four pre-specified random patterns, in order to randomize any effects of learning or order across participants. Parallelism of alternative forms. In order to determine whether the 16 alternative versions of each cognitive test were parallel, the means for all sixteen versions, averaged over all occasions and participants, were calculated for each cognitive test. Next, a repeated measures analysis of variance (RMANOVA) with one within-subjects factor, alternate form (i.e., workbooks 0 15) was conducted for each of the cognitive tests to determine if mean performance varied as a function of alternate form. Results from this analysis indicated that mean performance, averaged across the 31 occasions of measurement significantly varied for the alternative forms of the AVLT List 1, F(15, 51) = 7.05, p = .000, AVLT Total Score F(15, 51) = 8.56, p = .000, Backward Digit Span, F(15, 48) = 7.59, p = .000, and Symbol Digit, F (15, 51) = 7.64, p = .000. Thus, for all the measures the alternative forms demonstrated significant differences across forms. Post-hoc contrasts revealed idiosyncratic differences between workbook pairs, with different patterns by measure. Since the specific workbook contrasts were not theoretically driven, and since the order of workbooks was randomized across subjects, specific statistical comparisons between workbooks are not shown. Rather, the main point of these ANOVAS is that the workbooks are not completely equivalent, and thus may be a source of occasion-to-occasion variability that is common across participants. However, as can be seen in Table 3-4, the means and standard deviations of the parallel versions of each measure were highly comparable. In addition, an examination of the correlational relationship among the different versions (Tables 3-5 through 3-8) shows

PAGE 46

34 that participants mean performance on the 16 alternative versions was strongly interrelated for all cognitive measures (all correlations were significant, p = .000). Table 3-4. Mean performance on the alternative forms of each cognitive task. Form AVLT List 1 Mean (SD) AVLT Total Recall Mean (SD) Backward Digit Span Mean (SD) Symbol Digit Mean (SD) 0 7.87 (2.92) 30.16 (8.40) 7.31 (2.26) 35.29 (12.11) 1 8.80 (3.06) 32.38 (8.50) 9.33 (1.99) 41.25 (9.50) 2 9.32 (2.82) 33.71 (7.87) 9.05 (2.11) 40.17 (10.92) 3 8.92 (2.47) 32.66 (7.50) 9.19 (1.87) 41.67 (9.52) 4 9.23 (2.90) 33.58 (8.21) 9.35 (1.69) 43.08 (10.68) 5 9.77 (2.73) 35.08 (7.49) 9.43 (1.88) 41.80 (9.15) 6 8.96 (2.90) 32.74 (7.91) 8.88 (1.92) 42.15 (12.17) 7 10.02 (2.96) 35.66 (7.49) 9.20 (1.79) 43.63 (9.13) 8 9.22 (2.83) 33.36 (7.62) 9.15 (2.02) 42.18 (9.47) 9 8.92 (2.61) 32.85 (7.88) 9.09 (2.01) 42.05 (9.32) 10 9.75 (2.75) 35.30 (6.90) 9.05 (2.07) 41.27 (9.82) 11 9.52 (2.79) 34.64 (7.45) 9.40 (1.76) 40.56 (8.75) 12 9.97 (2.89) 35.07 (8.00) 9.51 (1.68) 41.60 (9.31) 13 9.78 (2.69) 34.65 (7.07) 9.21 (1.84) 45.52 (11.25) 14 9.65 (2.85) 35.11 (7.25) 9.73 (1.75) 42.34 (9.85) 15 9.76 (2.71) 34.92 (7.73) 9.54 (1.92) 42.49 (10.23)

PAGE 47

35 Table 3-5. Intercorrelations between mean scores on alternative versions of AVLT List 1. Wkbk 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 1 .580 1 2 .581 .755 1 3 .558 .688 .661 1 4 .551 .741 .759 .827 1 5 .568 .705 .769 .644 .768 1 6 .594 .741 .811 .749 .783 .770 1 7 .565 .627 .769 .622 .766 .776 .817 1 8 .555 .611 .691 .649 .698 .706 .789 .739 1 9 .616 .750 .753 .665 .777 .741 .767 .746 .682 1 10 .588 .741 .744 .709 .804 .744 .751 .729 .768 .742 1 11 .395 .678 .615 .621 .753 .759 .762 .738 .687 .703 .666 1 12 .598 .618 .692 .685 .764 .759 .838 .795 .748 .730 .750 .763 1 13 .596 .721 .767 .756 .821 .794 .843 .772 .740 .770 .763 .769 .853 1 14 .622 .702 .768 .747 .772 .817 .791 .794 .765 .780 .730 .746 .756 .842 1 15 .689 .701 .749 .718 .757 .761 .833 .841 .736 .720 .740 .728 .839 .802 .813 Note. All correlations were significant, p = .000 Table 3-6. Intercorrelations between mean scores on alternative versions of AVLT Total Score. Wkbk 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 1 .808 1 2 .823 .901 1 3 .748 .850 .861 1 4 .804 .892 .895 .904 1 5 .745 .857 .896 .836 .912 1 6 .809 .887 .903 .860 .917 .889 1 7 .748 .826 .884 .828 .891 .859 .894 1 8 .803 .860 .857 .828 .903 .854 .905 .807 1 9 .801 .880 .881 .880 .903 .856 .894 .853 .843 1 10 .776 .857 .848 .835 .914 .867 .886 .858 .875 .848 1 11 .665 .838 .822 .833 .883 .851 .872 .838 .817 .857 .790 1 12 .748 .830 .859 .851 .903 .843 .911 .878 .836 .876 .877 .846 1 13 .768 .867 .889 .871 .916 .880 .905 .863 .852 .887 .868 .845 .884 1 14 .773 .891 .902 .883 .920 .887 .898 .901 .880 .901 .862 .853 .880 .918 1 15 .797 .893 .904 .876 .919 .875 .932 .920 .867 .888 .870 .874 .921 .912 .942 Note. All correlations were significant, p = .000

PAGE 48

36 Table 3-7. Intercorrelations between mean scores on alternative versions of Backward Digit Span. Wkbk 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 1 .541 1 2 .619 .698 1 3 .656 .735 .719 1 4 .553 .700 .699 .797 1 5 .554 .654 .730 .726 .647 1 6 .600 .714 .722 .764 .771 .709 1 7 .696 .796 .805 .840 .793 .782 .791 1 8 .507 .752 .700 .744 .697 .678 .747 .746 1 9 .713 .766 .763 .814 .789 .701 .738 .754 .768 1 10 .701 .791 .705 .852 .766 .684 .765 .816 .713 .826 1 11 .561 .735 .682 .720 .661 .699 .714 .738 .699 .684 .718 1 12 .522 .626 .673 .697 .664 .691 .614 .721 .662 .653 .685 .748 1 13 .623 .724 .706 .789 .763 .751 .784 .790 .742 .753 .771 .759 .750 1 14 .590 .708 .696 .812 .804 .731 .753 .795 .722 .781 .798 .722 .738 .778 1 15 .596 .762 .727 .769 .793 .751 .761 .831 .715 .760 .803 .743 .676 .762 .746 Note. All correlations were significant, p = .000 Table 3-8. Intercorrelations between mean scores on alternative versions of Symbol Digit. Wkbk 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 1 .498 1 2 .521 .907 1 3 .506 .821 .825 1 4 .520 .865 .857 .816 1 5 .459 .783 .788 .880 .812 1 6 .398 .727 .707 .663 .726 .696 1 7 .547 .805 .851 .83 .84 .86 .73 1 8 .571 .780 .815 .829 .790 .847 .685 .866 1 9 .460 .774 .831 .817 .821 .775 .706 .809 .781 1 10 .500 .805 .798 .810 .755 .802 .704 .857 .823 .817 1 11 .508 .761 .810 .868 .751 .875 .641 .881 .798 .757 .822 1 12 .485 .773 .807 .843 .787 .877 .700 .904 .825 .809 .846 .880 1 13 .495 .664 .676 .649 .614 .659 .590 .715 .730 .684 .640 .668 .686 1 14 .474 .886 .854 .855 .885 .808 .724 .856 .801 .810 .797 .838 .835 .697 1 15 .535 .747 .791 .836 .748 .839 .693 .867 .858 .800 .884 .823 .865 .677 .786 Note. All correlations were significant, p = .000

PAGE 49

37 Procedure Overview of Study Phases As can be seen in the flow chart, participant assessment occurred in three phases (Figure 3-1). In Phase 1 and 2, Screening and Group Assignment, potential participants were screened and then assessed for amnestic MCI. This screening occurred as a two-part process, an initial telephone screening and an in-person, individually administered neuropsychological assessment, conducted in order to provide a formal, standardized measure of MCI categorization. The two-part screening process provided an opportunity to assess cognitive status initially via telephone, in order to screen for inappropriate potential participants prior to the time-intensive neuropsychological assessment. Following the neuropsychological assessment, group assignment was confirmed via the consensus conference procedure described below. All cognitive assessment took place in the laboratory following consent procedures as approved by the University of Florida Institutional Review Board. Phase 1 and Phase 2 Telephone Screening and Group Assignment Recruitment & Telephone Screening No MCI MCI Neuropsychological Intake Assessment Phase 3 Orientation Training and Daily Protocol Research Partner or Caregiver Training & Supervised Daily Cognitive A sse ss m e n t No MCI MCI 30 Occasion Daily Cognitive Assessment Protocol Figure 3-1. Design of the current study.

PAGE 50

38 In Phase 3, the Daily Cognitive Assessment, participants and their research partners were asked to attend an orientation session when they were provided with timers and thirty assessment booklets containing the Daily Cognitive Assessment Battery (DCAB; described above), and trained in their use. These sessions were generally held one week after the Neuropsychological Intake Assessment. In a few cases, the neurocognitive assessment and participant orientation sessions were held on the same day. This orientation session took place in the laboratory, or at home (n = 3), after both the participant and research partner were informed of the requirements of the daily protocol and provided informed consent as approved by the University of Florida Institutional Review Board. For the participants, this informed consent step was analogous to the informed consent obtained prior to the Neuropsychological Intake Assessment. This second tier of informed consent procedures was required in order to adequately inform the research partner. Training included explanation of directions and observation of administration, in order to ensure adequate standardization across participantresearch partner dyads. Thus, as part of the training, each research dyad was observed completing a sample Daily Cognitive Assessment Workbook. Research partners were asked to participate in the administration of the Digit Span task by reading the digits aloud and recording the exact response. They were instructed to give all Digit Span items (up to and including a span of seven digits) regardless of performance. Research partners were instructed to read the numbers in the Digit Span sequences at a rate of one number per second. This was modeled for the research partner and participant. Research partners were corrected regarding their rate and pacing during the laboratory setting. Dyads were instructed that one booklet should be completed each day,

PAGE 51

39 at the same time of day, beginning the day after the orientation session. Additional guidelines were that the workbooks were to be completed without assistance, and were to be returned at the end of each week. Since workbooks were numbered and dated, participants were instructed that, should they miss a day, they were to skip that workbook and resume with the one that corresponded to the current date. In addition to providing an opportunity to orient and teach the research participant and research partner about the Daily Cognitive Assessment Workbook, the laboratory session also provided a supervised standard for comparison with the unsupervised home-based Workbooks completed for the Daily Cognitive Assessment protocol (occasions 1 30). Rationale for the research partner administration protocol Past research (Allaire, 2001) indicates that cognitively intact older adults can successfully self-administer a brief cognitive battery. However, an initial concern was that participants with mild cognitive impairment might be unable to consistently do so. Caregiver/partner assistance was thought to be a useful means to ensure protocol standardization as well as to provide a build-in reminder to accomplish the task for participants with memory impairment. For the healthy elders, spousal administration was thought to ensure protocol fidelity and consistency in the social-psychological conditions of test administration. Anecdotally, it appeared from observation of the dyad interactions during the initial laboratory orientation session, that in at least two occasions, the research partner of a cognitive intact participant may have been suffering from mild cognitive impairment. Due to the lack of formalized assessment for the research partner, this cannot be objectively demonstrated.

PAGE 52

40 Notably, when an MCI participant and his or her cognitively intact spouse were both enrolled as participants in the study it was often clear that there was a difference in performance on the Daily Cognitive Assessment Battery tasks between the two. For all dual participant dyads, both participants were directed not to compare performance and not to view the study as a competition. They were instructed to complete their Daily Assessment Workbooks in different locations and at different times of day if they felt any concern about comparing performance. Compliance monitoring On a weekly basis, the workbooks were returned to the researchers in pre-paid, pre-addressed envelopes. Each was reviewed to monitor compliance and completeness of administration. Any weekly sets of workbooks with incomplete or incorrectly completed tasks occurring on more than one occasion resulted in reminder phone calls. Queries regarding administration were minor (e.g., on Digit Span, Do I write down exactly what my partner says?) and were answered by telephone. Non-compliance or incomplete administration resulting in incompletion of the first block (10 occasions) resulted in discontinuation. All participants received reminder phone calls during the first and second weeks, which served as an opportunity to answer questions or address concerns regarding the assessments. Group Membership Assignment: Consensus Conference As indicate above in the discussion of inclusion and exclusion criteria, group placement was confirmed by means of a consensus conference. Participants were classified as probably MCI or unimpaired (Non-MCI) by consensus. Each case was presented to the consensus members (which consisted of a senior researcher in cognitive aging, a clinical neuropsychologist, and three graduate level clinical neuropsychology

PAGE 53

41 students, one at the doctoral level and two at the masters level). For each case, the members of the conference were presented with masked neuropsychological data from the Neuropsychological Intake Assessment, described above, as well as Clinical Dementia rating Scale (CDR) score and a composite score representing subjective complaints about memory. These variables were chosen in order to conform to the Petersen criteria for amnestic MCI (Petersen et al., 2001). If all panel members agreed on group placement, that group assignment was made. If at least one member disagreed, a discussion was held regarding the criteria, and majority vote determined group assignment. Initial Data Preparation and Study Variables Ceiling and Floor Considerations Prior to presenting the analyses and results for the specific aims, issues related to range restriction due to performance at floor (lowest score possible) or ceiling (highest score possible) are described. This is necessary since reduction in range of scores, by repeated performance at ceiling or floor levels, would limit the potential for variability in performance. For cognitive tasks, performance at ceiling levels artificially restricts learning over future trials. If variability is an indicator of learning, the forced absence of learning would, in turn, reduce the relationship between variability and performance gains for those individuals with the best performance. Similarly, for individuals with the worst performance, inappropriately difficult tasks, with performance consistently at floor, would mask any evidence that variability is an indicator of cognitive compromise. In order to eliminate problems due to excessive ceiling or floor scores within an individual participants performance the following analyses were completed. Initially, an investigation of frequencies of floor values for the cognitive variables revealed that the

PAGE 54

42 incidence of floor and near-floor values was < 1% of all occasions for all variables. Floor, or near-floor, values were defined as the lowest possible score plus one item for all variables except AVLT Total Score, for which the sum of the floors for the three lists (e.g., score of six) was considered the floor. Additionally, no individual participant performed at or near floor more than twice over the thirty-one occasions. Thus, no further assessment of floor values was conducted. Similar investigation of frequencies of ceiling values for the cognitive variables revealed that restricted ranges concerns were relevant for the AVLT Total Score, AVLT List 1 Score, AVLT Delay Recall Score, and Digit Span Backward Score. Ceiling values were considered to be the highest score attainable on the measure (AVLT Total Score ceiling = 45; AVLT List 1 Score ceiling = 15; AVLT Delay Recall Score ceiling = 15; Digit Span Backward Score ceiling = 12.) AVLT Total Score, AVLT List 1 Score, and Digit Span Backward Score reached ceiling levels on 3.5% (74/2108), 4.5% (94/2108), and 10% (216/2108) of occasions, respectively. A small number of individual participants (n = 5, n = 6, and n = 5, respectively) accounted for the majority of these ceiling performance values, as the participant reached ceiling performance and remained at that level of performance for the remainder of their sessions. All occasions after reaching ceiling performance were removed from analyses for these participants. Since ceiling performance was reached and maintained near or at the beginning of the third block of occasions (occasions 21 30) these participants do not contribute to Block 3 analyses for AVLT Total Score, AVLT List 1 Score, or Digit Span Backward Score. Notably, these changes had

PAGE 55

43 negligible affect on the mean scores or IRI calculations across all occasions and over the three blocks. Thirty-four percent (711/2108) of occasions were at ceiling performance for AVLT Delay Recall. Thirty-two participants reached ceiling for this variable during the first block (occasions 1 10) of sessions and were removed from the AVLT Delay Recall analyses. Two participants dropped from the AVLT Delay Recall mean score and IRI score calculations were in the MCI group, while the remainder were in the Non-MCI group. Notably, the loss of 30 participants from the cognitive intact group had a significant effect on the overall IRI calculations for that group (e.g., original Mean IRI for Non-MCI = 3.07 (SD = 1.46), Ceiling Corrected Mean IRI for Non-MCI = 4.07 (SD = 0.92)). This is noteworthy since the IRI for the MCI group did not change (Mean IRI = 3.72, SD = 0.72). There was however, no change to the significance of the differences, as the IRI score across the two groups remained non significantly different. Regarding other cognitive variables, Symbol Digit Score was at ceiling for <0.1% of occasions. Notably, AVLT Percent Retained calculations resulted in a restricted range, although values > 100% retention were allowed. No modifications were made to this variable, since the potential range restrictions were inherent to the calculations. Standardization of Scores from the Daily Assessment Battery Measures for the Daily Assessment Battery were modified or newly created for this study, and, as such have no published norms available. However, standardization of scores is useful to facilitate comparisons on the same scale. Therefore, we standardized the scores from the Daily Assessment Battery to a T-score metric, using the mean and standard deviations from this sample. Thus, prior to the calculation of intraindividual reliability indices for the cognitive and non-cognitive measures in the Daily Assessment

PAGE 56

44 Battery, all measures were standardized to T-score metric, with mean = 50, standard deviation = 10. For all daily cognitive measures (memory and non-memory), as well as daily mood measures, standardization was based on the orientation and training in-laboratory session across all participants (including both MCI and Non-MCI participants). Thus, the mean and standard deviation at the first, monitored, session were set to 50 and 10, respectively, for all participants, and were allowed to vary at subsequent occasions within the standardized metric of that first session. The additional daily non-cognitive measures (sleep and distracting environmental variables) were standardized to the first at-home occasion. Intraindividual Variability Indices Calculation of an intraindividual variability index was required in order to perform the relevant statistical analyses for the hypotheses connected with each aim. Results from Allaires (2001) work indicates that the intraindividual standard deviation index (ISD; Hultsch, 2001) is not effective if the construct studied involves growth or learning. An alternative, the intraindividual residual index (IRI), represents the average amount of variability in an individuals performance around a best fitting regression line, and can be used to describe variability when growth or learning is present or absent (Allaire, 2001). The IRI was calculated by first estimating a regression line (incorporating linear and quadratic time trends to reflect growth for each participant) for each participants performance over the 31 occasions for each measure, and obtaining, for each subject, a residual between their actual data point and their estimated value from the regression line. These residuals were squared, summed, and then divided by the number of occasions, to obtain a mean squared residual across all 31 occasions. The square root of this term (i.e., root mean square residual) was calculated in the last step, to create an

PAGE 57

45 intraindividual residual index (IRI). One advantage of this index is that, by taking the square root, it is expressed in the same metric/units as the test/measure that it represents. This procedure was followed to calculate the intraindividual residual index for all relevant measure in the Daily Cognitive Assessment Battery (e.g., AVLT List 1 Score, AVLT Total Score, AVLT Delay Recall Score, AVLT Percent Retained, Backward Digit Span, Symbol Digit Score, PANAS Positive Affect, PANAS Negative Affect, Distracting Environment Factor 1: Discomfort, Distracting Environment Factor 2: Distractions, Total Sleep Time, and Sleep Efficiency). Table 3-9 shows the correlations of the mean IRIs for the cognitive variables with the main demographics of age, education, and sex. Intraindividual variability on AVLT Percent Retained and Backward Digit Span was positively correlated with age, meaning that for these variables, older individuals were more variable in performance from day-to-day. Greater intraindividual variability on AVLT Percent Retained was also significantly correlated with male gender. Table 3-9. Correlations of demographics and mean IRI scores. Age Education Sex AVLT Total Score IRI .229 -.047 -.172 AVLT List 1 IRI -.170 .091 .216 AVLT Delay Recall IRI .128 .136 -.248 AVLT Percent Retained IRI .472** -.098 -.396** Backward Digit Span IRI .368** .079 -.194 Symbol Digit Score IRI .077 -.053 .123

PAGE 58

CHAPTER 4 RESULTS Overview This chapter focuses on the three major aims. First is a descriptive examination of intraindividual variability in memory and non-memory domains, and its relationship to Mild Cognitive Impairment (MCI) status. Secondly, there is a consideration of whether, as some previous research has suggested, intraindividual variability is related to practice-related gain (i.e., that those who experience more gain on a task show more variability during the period of time in which they are improving). Since individuals with MCI are presumed to have less ability to profit from practice, an embedded question is whether this will produce diagnosis-related differences in intraindividual variability over time. Finally, the chapter concludes with an examination of potential sources of intraindividual variability in participants, with a particular focus on coupled daily intraindividual variability. Preliminary Analyses Before the intraindividual variability analyses for the specific aims of the study, a number of preliminary analyses and psychometric checks were conducted. For those data arising from the Neuropsychological Intake Assessment, MCI and Non-MCI participants were compared on the initial neuropsychological measures. These measures are not utilized for the study aims, but were integral to the group assignments and provide a description of the neurocognitive status of the participants upon entry into the study. A small group of individuals who participated in the Neuropsychological Intake Assessment 46

PAGE 59

47 did not continue to the Daily Cognitive Assessment Battery, so a brief attrition analysis is presented. With regards to preliminary analyses for the Daily Cognitive Battery, four specific analyses are presented. First, as mentioned in the Methods chapter, performance during the laboratory session was compared to performance during the initial at-home session, in order to provide a basic quality control check. Secondly, mean performance over 31 occasions on the measures in the Daily Cognitive Assessment Battery were compared for the two groups based on cognitive status in order to confirm expected performance level differences. Thirdly, the AVLT delay time (in minutes) between the third presentation of the word list and the delayed recall of the list was investigated in order to determine if variable delay times were related to performance. Finally, data reduction in the form of factor analysis was conducted for the environmental distraction and discomfort questions. Neuropsychological Intake Assessment Data: Participant Neurocognitive Status and Attrition Analysis. After telephone screening with the TICS, 84 participants met criteria for entrance into the study and were assessed with the full Neuropsychological Intake Assessment. Of these participants, eight declined to continue to participate following the Neuropsychological Assessment, leaving 76 participants who began the Daily Cognitive Assessment Battery. Eight of these participants withdrew prior to completion of the first block of daily at-home assessments (i.e., prior to completion of 10 days of assessments), thus, these participants were discontinued. Of the remaining 68 participants, all except three completed the full 30 days with fewer than three missing occasions. The three who did not fully complete the at-home assessments completed Block 1 (n=3) and Block 2 (n

PAGE 60

48 = 2) and were retained for analyses involving these blocks. Thus, for most analyses study sample n = 68, but for analyses involving only Block 3 the sample n = 65. Table 4-1 shows the mean performance on neuropsychological measures for the participants who completed the study, by cognitive status. As expected, the participants in the MCI group performed significantly worse than the cognitively intact individuals on measures of immediate and delayed memory function. Unexpectedly, the MCI group was also significantly worse on the Boston Naming Test. Table 4-2 shows the mean scores for a selection of measures from the neuropsychological assessment battery for the participants who dropped out before completing at least 10 daily sessions compared with those participants who completed the daily at-home assessments. Variances were not equivalent for the COWA score and the MMSE score, resulting in reduced degrees of freedom for these t-tests. Overall, the individuals who dropped out of the study after the neuropsychological assessment, or during the first few occasions of daily assessment, demonstrated significantly worse immediate and delayed memory that those individuals who remained in the study. The participants who dropped out of the study performed poorer on cognitive tests overall and endorsed more symptoms of depression. Daily Cognitive Assessment Battery Data Quality control check: Laboratory to home administration Table 4-3 shows a comparison of performance during the laboratory session with performance during the first session at home. As can be seen from the t-test and significance values in Table 4-3, the only significant differences between the introductory laboratory session and the first daily at-home session are noted in Backward Digit Span

PAGE 61

49 Table 4-1. Mean performance on neuropsychological measures, by cognitive status. Measure MCI Participants n = 15 M (SD) Non-MCI Participants n = 53 M (SD) Significance Test TICS Score 31.67 (3.64) 37.38 (2.61) t(65) = 6.809; .000 MMSE 27.55 (1.75) 29.05 (0.84) t(11.2) = 2.762; .018 Predicted IQ 110.69 (5.92) 114.47 (7.51) t(66) = 1.795; .077 BVMT Total Score 13.93 (5.60) 22.89 (6.09) t(66) = 5.113; .000 BVMT Delayed Recall (T Score) 38.60 (9.11) 55.32 (8.63) t(66) = 6.546; .000 HVLT Total T Score 40.27 (7.66) 55.79 (7.18) t(66) = 7.288; .000 HVLT Delayed Recall 3.93 (2.25) 9.51 (1.82) t(66) = 9.955; .000 COWA Total 11.33 (2.13) 12.08 (2.19) t(66) = 1.165; .248 BNT Scale score 11.00 (2.17) 13.09 (2.92) t(66) = 2.580; .012 Trails A 10.20 (2.15) 10.96 (2.92) t(66) = 0.941; .350 Trails B 10.93 (1.91) 11.06 (2.96) t(66) = 0.153; .879 CES-D 7.33 (4.88) 5.06 (4.78) t(66) = -1.621; .110 Note. TICS = Telephone Interview of Cognitive Status, MMSE = Mini-Mental Status Examination, BVMT = Brief Visuospatial Memory Test, HVLT = Hopkins Verbal Learning Test, COWA = Controlled Oral Word Association, BNT = Boston Naming Test, CES-D = Center for Epidemiological Studies Depression Scale. Table 4-2. Mean performance on neuropsychological measures, by attrition status. Measure Dropouts N= 16 M (SD) Daily Participants N = 68 M (SD) Significance Test TICS Score 33.00(4.65) 36.10 (3.72) t(81) = -2.854; .005 MMSE 27.18 (2.40) 28.73 (1.24) t(11.2) = -2.082; .061 Predicted IQ 112.55 (6.77) 113.64 (7.32) t(82) = -.542; .590 BVMT Total Score 16.81 (8.41) 20.91 (7.02) t(82) = -2.022; .046 BVMT Delayed Recall (T Score) 48.13 (12.56) 51.63 (11.13) t(82) = -1.107; .272 HVLT Total T Score 44.56 (13.34) 52.37 (9.71) t(82) = -2.683; .009 HVLT Delayed Recall 5.81 (4.02) 8.28 (3.01) t(82) = -2.761; .007 COWA Total 11.06 (3.30) 11.91 (2.18) t(18) = -.981; .339 BNT Scale score 11.13 (3.32) 12.63 (2.89) t(82) = -1.824; .072 Trails A 11.00 (2.94) 10.79 (2.77) t(82) = .265; .792 Trails B 9.06 (2.02) 11.03 (2.75) t(81) = -2.692; .009 CES-D 9.19 (6.18) 5.56 (4.86) t(82) = 2.547; .013 Note. TICS = Telephone Interview of Cognitive Status, MMSE = Mini-Mental Status Examination, BVMT = Brief Visuospatial Memory Test, HVLT = Hopkins Verbal Learning Test, COWA = Controlled Oral Word Association, BNT = Boston Naming Test, CES-D = Center for Epidemiological Studies Depression Scale.

PAGE 62

50 and Distracting Environmental Variables Factor 2: Distractions. These differences are likely due to systematic observed differences between these occasions. During the laboratory session, it was noted that the directions for Backward Digit Span were the most often repeated by the project testers. Administration of this task during the laboratory practice session was often interrupted and resumed in order to clarify directions. Likely, familiarity with the task, better understanding of the directions, and fewer interruptions during administration resulted in improved performance during the first at-home administration. Regarding significant differences noted on the second distracting environmental factor, this factor consists of questions about whether there are distractions while the tasks are being completed, including whether or not there are others around, noise, and interruptions (see below for factor loadings). Notably, during the laboratory session participants responded positively to these questions due to the presence of the examiner/instructor. The first at-home session likely involved fewer distractions as a result. Performance on all other measures was equivalent across the two sessions. Group differences in mean performance over all occasions Table 4-4 shows the mean levels of performance across all 31 occasions for the memory and non-memory cognitive variables for the two groups based on cognitive status. Mean values for the non-cognitive potential covariates are also depicted. For all variables raw and standardized scores are presented. As indicated in the Methods, scores were standardized to T score metric (mean = 50, SD = 10) in order that performance comparisons across variables could be made using the same scale. Results of the independent sample t tests for mean performance across groups appear in the same table.

PAGE 63

51 Table 4-3. Comparison of supervised session with first at-home daily session. Measure Laboratory Session Mean (SD) First Daily Session Mean (SD) Significance Test t(df); p AVLT List 1 7.87 (2.92) 7.71 (2.73) t(134) = .334; .739 AVLT Total Score 30.16 (8.40) 30.43 (8.01) t(134) = -.188; .851 AVLT Delay Recall 10.78 (3.76) 10.46 (4.12) t(133) = .467; .641 AVLT Percent Retained 89.28 (19.33) 82.31 (26.07) t(122) = 1.763; .080 Symbol Digit 35.29 (12.11) 36.53 (9.11) t(132) = -.666; .506 Backward Digit Span 7.31 (2.26) 8.17 (2.26) t(132) = -2.194; .030 PANAS Positive 2.43 (0.58) 2.39 (0.69) t(131) = .354; .724 PANAS Negative 0.48 (0.42) 0.39 (0.44) t(131) = 1.098; .274 Factor 1: Discomfort 47.85 (6.74) 49.82 (7.81) t(134) = -1.574; .118 Factor 2: Distractions 62.61 (9.29) 50.08 (6.97) t(124) = 8.895; .000 Total Sleep Time (min) 449.57 (82.17) 444.45 (97.11) t(131) = .328; .743 Sleep Efficiency 89.60 (10.62) 88.33 (14.55) t(131) = .576; .566 Note Assumptions for Levenes test of equality of variance were not met for AVLT Percent Retained and Distracting Environmental Variables Factor 2: Distractions. AVLT = Rey Auditory Verbal Learning Test; PANAS = Positive and Negative Affect Scale. Note that the t test results are equivalent for mean comparisons of raw scores or standardized scores, thus only one appears in the table for each variable. Multiple individual mean comparisons were performed, since different patterns were expected for the different measures (e.g., the groups were expected to differ on mean performance on memory measures but not on other, non-memory, cognitive measures, or on non-cognitive measures). Bonferroni correction for multiple comparisons was not employed, due to concerns about power. Presentation of the actual significance value in Table 4-4 allows for observation of the likelihood of significance had Bonferroni correction been applied. As expected, the two groups differed significantly on all memory variables, with the Mild Cognitive Impairment (MCI) group performing worse on AVLT Total Score

PAGE 64

52 (t(16.6) = 4.068; p = .001), AVLT List 1 (t(66) = 3.942; p = .000), AVLT Delay Recall (t(15.32) = 3.744; p = .002), and AVLT Percent Retained (t(14.7) = 3.334; p = .005). It is important to note that group assignments were not made on the basis of these AVLT scores, but on the basis of independently measured HVLT-R scores assessed during the Neuropsychological Assessment prior to the Daily Cognitive Battery phase. On the non-memory cognitive variables, the groups differed only on Digit Span Backward Score (t(17.5) = 2.169; p = .044), with the MCI group again performing worse overall. There were no significant differences between the groups on the non-cognitive measures. Assumptions for Levenes test of homogeneity of variances were not met for AVLT Total Score, AVLT Delay Recall, AVLT Percent Retained, Digit Span Backward Score, and PANAS Negative Affect comparisons, resulting in adjusted degrees of freedom for these mean comparisons. Even with Bonferroni adjustment, all the memory differences between groups remained significant; in contrast, the Digit Span Backward was no longer significantly different. Thus, the results generally confirm the categorization by consensus conference, and indicate that the two groups differ significantly in level of performance, specifically on memory measures. The distribution for all cognitive variable means approximated normal distributions, with the exception of the AVLT Percent Retained, which had a kurtosis estimate > |2|. As indicated in the Methods section, a small number of participants reached the ceiling level of performance for four of the variab l es (AVLT Total Score, AVLT List 1, AVLT Delay Recall, and Backward Digit Span). Table 4-4 clearly labels these variables as ceiling corrected, however, for all subsequent tabular presentations, these variables are no longer identified as ceiling corrected, although the corrected values were used for all analyses.

PAGE 65

53 Table 4-4. Means (Standard Deviations) for all measures by cognitive status. Measure MCI n=15 Non-MCI n=53 Significance Test Mean (SD) Mean (SD) t(df); p Memory Measures AVLT Total Score Ceiling Corrected Raw 26.39 (8.58) 35.80 (4.87) Standard 46.45 (9.40) 56.76 (5.34) t(16.6) = 4.068; .001 AVLT List One Ceiling Corrected Raw 7.51 (2.37) 9.76 (1.82) Standard 49.26 (7.83) 56.69 (6.02) t(66) = 3.942; .000 AVLT Delay Recall Ceiling Corrected Raw 6.90 (3.80) 11.10 (1.86) Standard 41.32 (9.15) 51.45 (4.48) t(15.32) = 3.744; .002 AVLT Percent Retained Raw 71.62(25.69) 94.01 (7.59) Standard 43.13 (11.56) 53.20 (3.41) t(14.7) = 3.334; .005 Non-Memory Cognitive Measures Digit Span Backward Score Ceiling Corrected Raw 8.28 (1.90) 9.41 (1.24) Standard 54.59 (8.40) 59.57 (5.48) t(17.5) = 2.169; .044 Symbol Digit Score Raw 40.48 (10.01) 42.21 (7.17) Standard 55.05 (8.03) 56.44 (5.75) t(18.3) = 0.626; .539 Non-Cognitive Measures PANAS: Positive Affect Raw 2.26 (0.61) 2.27 (0.68) Standard 46.44 (10.73) 46.68 (12.00) t(66) = 0.072; .943 PANAS: Negative Affect Raw 0.60 (0.58) 0.33 (0.31) Standard 51.59 (12.50) 45.83 (6.74) t(16.4) = -1.715; .105 Total Sleep Time Minutes Raw 440.02 (35.86) 441.98 (64.13) Standard 49.09 (3.61) 49.28 (6.45) t(66) = 0.113; .910 Sleep Efficiency Raw 87.88 (6.02) 88.63 (7.41) Standard 49.58 (4.21) 50.10 (5.17) t(66) = 0.356; .723 Factor 1: Discomfort 51.07 (5.96) 49.31 (5.83) t(66) = -1.028; .308 Factor 2: Distractions 48.33 (3.46) 50.44 (4.97) t(66) = 1.534; .130 Note: AVLT = Rey Auditory Verbal Learning Task, PANAS = Positive and Negative Affect Scale.

PAGE 66

54 Effect of differing delay times for AVLT Delayed Recall The Daily Cognitive Assessment Battery provided a specific structure and order for participants to follow in completing the cognitive tasks each day. The list learning task (modified AVLT) was the first task each day. Upon completion of the third list, participants were directed to record the time. After completing the intervening tasks (Number Copy, Symbol Digit, Digit Span, Sleep Diary, Mood report, and environmental distractions questions), participants were prompted to record the time prior to attempting the delay recall trial. Recorded delay times (from completion of List 3 to initiation of Delay List) ranged from 0 minutes to 365 minutes. Mean, median, and mode delay time for the sample and by cognitive status are outlined in Table 4-5. The difference between the mean delay times for the two groups was non-significant, as the independent samples t test result was: t(472) = -1.503; p = .134. Levenes test for equality of variances was significant, resulting in a reduction in the degrees of freedom for the mean comparison test. Table 4-5. Mean AVLT delay time. Group AVLT Delay time (in minutes) Mean ( SD ) Median Mode Entire Sample (n = 68) 16.45 (15.73) 15.00 10.00 MCI (n = 15) 17.82 (21.44) 15.00 10.00 Non-MCI (n = 53) 16.10 (13.88) 14.00 10.00 A linear mixed models analysis to determine if delay time is related to performance revealed that the fixed effect of AVLT Delay Time was unrelated to the dependent variable of AVLT Delay Recall, F(1, 63) = 3.54; p = .064; parameter estimate = -.086). Similarly, linear mixed models analysis to determine if cognitive status interacted with

PAGE 67

55 delay time to influence performance was non-significant for the interaction of AVLT Delay Time and Cognitive Status, F(1, 60) = 0.057; p = .812, parameter estimate = -.026, indicating that the two groups did not differ in the relationship between length of delay and performance on the delay recall trial. Distracting environmental variables: Data reduction Participants were asked to use a Likert scale on a daily basis to report pain level, stress level, tiredness level, noise level, and light level. Additionally, they were asked to indicate the number of interruptions while completing the workbook and to note whether or not other people were around while they were working. A principal axis factor analysis with promax rotation revealed three distinct factors. Factor loadings can be seen in Table 4-6. Pain level, stress level, and tiredness level loaded on the first factor (explaining 27.77% of the variance, which was labeled Discomfort. Noise level, number of interruptions, and positive response to people being around loaded on the second factor (explaining 20.40% of the variance), labeled Distractions. Light level loaded on a third factor (explaining 14.50% of variance). Since there was little to no variation in response to light level (all responses were excellent or very good), and since this indicator did not relate strongly to other measures, this item-specific factor was not used in further analyses. Table 4-6. Factor loading for distracting environment variables. Discomfort Distractions Light Level Stress Level .997 Tiredness Level .538 Pain Level .340 Noise Level .568 People Nearby .429 Number Interruptions .364 Light Level .319

PAGE 68

56 Intraindividual Variability in Memory and Other Cognitive Domains The initial analyses of mean performance over all daily occasions, above, confirmed that the two groups differed significantly on level of performance on measures of memory function. One focus of the current study was the degree to which the two groups might differ on intraindividual variability in memory and non-memory (e.g., working memory, attention, and processing speed) domains. The inquiry is captured in the first two aims. Aim One and Aim Two The first specific aim of the study was to investigate whether greater intraindividual variability in memory functioning is seen in older adults with amnestic mild cognitive impairment, compared to cognitively intact elders. The second specific aim, conceptually tied closely with the first aim, was to investigate whether cognitive tasks measuring neurocognitive domains that typically show relatively less impairment in level of performance in mild cognitive impairment (e.g., attention, processing speed) nonetheless reflect increases in intraindividual variability in older adults with mild cognitive impairment relative to unimpaired elders. Aim two also considered the extent to which impaired and non-impaired groups could be differentiated on the basis of their within-person variability in non-memory measures. The underlying question was whether, even before the emergence of more general group differences in level of performance, group differences in performance variability might serve as an early warning indicator of impending general cognitive compromise. As part of this aim, we sought to investigate whether intraindividual variability might independently contribute to the prediction of cognitive status group membership.

PAGE 69

57 Aim One and Aim Two: Review of Analyses Intraindividual variability differences across groups, based on cognitive status Aim one: Memory intraindividual variability. The variables of interest included the Intraindividual Residual Indices (IRI) f or initial verbal encoding (AVLT List 1), total recall (AVLT Total Score = AVLT List 1 + List 2 + List 3), long delay recall (AVLT Delay Recall), and percent retained over delay (AVLT Percent Retained = AVLT Delay Recall/ Higher of AVLT List 2 or List 3). Recall from the description of the IRI calculation in the Methods chapter that the Intraindividual Residual Index, or IRI, is the average amount of variability in an individual participants performance around a best fitting regression line over the days of measurement. In order to investigate differences in intraindividual variability on memory performance between the two groups over the thirty-one occasions of measurement, multiple independent samples t-tests were utilized to compare the mean IRI values across groups based on cognitive status. Mean and standard deviations for these IRI values by cognitive status are shown in Table 4-7. Results of the independent sample t tests for the IRIs of the memory measures appear in the same table. For AVLT List 1 (t(66) = 2.124, p = .037) the Non-MCI group demonstrated greater intraindividual variability, while for AVLT Percent Retained, the MCI group demonstrated greater intraindividual variability in performance over occasions (t(66) = -3.277, p =.002). There were no significant differences in intraindividual variability between groups on the other memory variables. Aim two: Non-memory cognitive intraindividual variability. For the investigation of intraindividual variability across the two groups on non-memory cognitive measures, the variables of interest included the intraindividual variability estimates for the non-memory cognitive measures assessed during the Daily Cognitive

PAGE 70

58 Assessment Battery. These are: Digit Span Backward Score and Symbol Digit Score. As with aim one, in order to investigate differences in intraindividual variability on non-memory cognitive performance between the two groups over the thirty-one occasions of measurement, multiple independent samples t-tests were utilized to compare the mean IRI values across groups based on cognitive status. Mean and standard deviations for the IRI of the non-memory cognitive variables by cognitive status are shown in Table 4-7. Results of the independent s a mple t-tests appear in the same table. There were no significant differences between groups in variability on the non-memory cognitive variables. Additionally, for sake of completeness, the IRIs for the non-cognitive measures appear in Table 4-7. There were no significant differences between the two groups on the non-cognitive variables assessing positive and negative affect, sleep time and efficiency, or those assessing environmental distraction and discomfort. For both sets of analyses, assessing the intraindividual variability differences in the memory variables (focus of aim one) as well as the non-memory cognitive variables (aim two) across cognitive status, multiple mean comparisons were made without use of Bonferroni correction for family-wise error. Exact significance values appear in Table 4-7 in order to assist with determining whether these IRI differences would continue to be highly significant had Bonferroni corrections been used. Were Bonferroni corrections to have been done, only the AVLT Percent Retained group difference would have been judged to be significantly different from zero. The distributions for all cognitive variable IRIs approximated normal distributions, with the exception of Symbol Digit IRI, which has skewness and kurtosis estimates > |2|.

PAGE 71

59 Table 4-7. Mean (Standard Deviation) Intraindividual Residual Indices (IRIs). Measure Total Group MCI Non-MCI Significance Test n=15 n=53 M(SD) M(SD) M(SD) t(df); p Memory Measures AVLT Total Score 3.21 (0.76) 3.22 (0.69) 3.20 (0.79) t(66) = -.076; .940 AVLT List 1 4.85 (1.08) 4.34 (1.11) 5.00 (1.04) t(66) = 2.124; .037 AVLT Delay Recall a 3.94(0.86) 3.72 (0.72) 4.07 (0.92) t(34) = 1.161; .254 AVLT Per. Retained 4.61 (2.44) 6.31 (2.77) 4.13(2.12) t(66) = -3.277;.002 Non-Memory Cognitive Measures Digit Span Backward 4.40 (0.92) 4.50 (0.79) 4.37 (0.95) t(66) = -.502; .617 Symbol Digit Score 3.44 (2.08) 2.98 (0.76) 3.57 (2.31) t(66) = .960; .341 Non-Cognitive Measures PANAS: Positive 5.56 (2.61) 5.22 (2.69) 5.65 (2.61) t(66) = .555; .581 PANAS: Negative 5.31 (2.98) 6.05 (3.47) 5.10 (2.84) t(63)= -1.059; .294 Total Sleep Time 5.19 (2.17) 4.60 (2.21) 5.35 (2.15) t(66) = 1.194; .237 Sleep Efficiency 4.42 (2.90) 3.81 (2.59) 4.60 (2.98) t(66) = .925; .358 Factor 1: Discomfort 3.74 (1.66) 3.69 (1.82) 3.76 (1.63) t(66) = .148; .883 Factor 2: Distractions 5.59 (3.08) 4.61 (1.92) 5.86 (3.30) t(66) = 1.396; .167 Note. AVLT = Rey Auditory Verbal Learning Task, PANAS = Positive and Negative Affect Scale. a Due to ceiling corrections, AVLT Delay Recall n = 36, 13, and 23 for the three samples. Data check: Reliability of intraindividual variability estimate A question that emerged, given the fairly minimal group differences in cognitive intraindividual variability, was whether unreliability of the IRI index might be responsible for the absence of group differences, since day-to-day fluctuations in performance on an instrument could be a reflection of random noise or measurement

PAGE 72

60 imprecision. However, previously, Allaire (2001) reported that measures of test-retest stability of the IRI were uniformly high in his sample. The variability for a particular measure in one block of occasions was generally positively and significantly related to variability in adjacent blocks of occasions. We employed the same approach in the present study. In order to confirm that the cognitive intraindividual variability estimates, calculated via the Intraindividual Residual Index (IRI), represented consistent, trait-like properties of the participants, a correlational analysis was performed over the three time-ordered blocks (i.e., Occasions 0-10, or Block 1; Occasions 11-20, or Block 2; and Occasions 21-30, or Block 3). If intraindividual variability is trait-like, and not just noise, the IRIs would be strongly and positively interrelated across blocks. The by-block relationships among the IRIs for each of the cognitive measures are shown in Table 4-8. For the memory variables, AVLT Total Score, AVLT List One, and AVLT Percent Retained show significant and positively correlated IRIs across the blocks, although the correlations are generally fairly modest (less than 0.40). Higher correlations (greater than 0.70) between blocks were observed for the AVLT Percent Retained measure. AVLT Delay Recall IRI demonstrated a different pattern, with few significant correlations. Among the non-memory cognitive variables, the Symbol Digit Score IRI correlations were positive and significant across all three blocks (again of relatively modest magnitude). Backward Digit Span IRI correlations across blocks were smaller, although still positive. As will be considered in the Discussion section, a question is whether range restriction in some measures constrained reliability.

PAGE 73

61 The findings also replicate those reported by Allaire (2001) in another way, in that there is a quasi-simplex pattern to the obtained test-retest stabilities of the IRI measure. That is, adjacent blocks (e.g., Blocks 1 and 2, Blocks 2 and 3) tend to be more highly correlated than distal blocks (i.e., Blocks 1 and 3). As is considered in the Discussion, this may be an index of qualitative transformations in the meaning of intraindividual variability over the course of practice. Table 4-8. Covariation among Intraindivi dual Variability Indices over blocks. Variable Block Block 1 (IRI) Block 2 (IRI) Block 3 (IRI) AVLT Total Score Block 1 (IRI) 1.00 --Block 2 (IRI) .343** 1.00 -Block 3 (IRI) .216 .384** 1.00 AVLT List One Block 1 (IRI) 1.00 --Block 2 (IRI) .218 1.00 -Block 3 (IRI) .287* .249* 1.00 AVLT Delay Recall Block 1 (IRI) 1.00 --Block 2 (IRI) .366* 1.00 -Block 3 (IRI) -.020 .214 1.00 AVLT Percent Retained Block 1 (IRI) 1.00 --Block 2 (IRI) .736** 1.00 -Block 3 (IRI) .636** .514** 1.00 Backward Digit Span Block 1 (IRI) 1.00 --Block 2 (IRI) .278* 1.00 -Block 3 (IRI) .227 .178 1.00 Symbol Digit Score Block 1 (IRI) 1.00 --Block 2 (IRI) .410** 1.00 -Block 3 (IRI) .303* .468** 1.00 = p < .05; ** = p < .01 Note: AVLT = Rey Auditory Verbal Learning Task; IRI = Intraindividual Residual Index; Block 1 = occasions 0-10; Block 2 = occasions 11-20; Block 3 = occasions 21-30. Predicting cognitive status with intraindividual variability As indicated, a follow-up question, based on an expect e d finding of differences in intraindividual variability across groups on the non-memory measures, had been initially

PAGE 74

62 posed. This question asked whether there would be a unique contribution of the non-memory cognitive variable IRI differences in predicting group membership. Notably, the IRI differences on the non-memory cognitive scores were not significant, so group membership was not predicted with solely the non-cognitive variables. Rather, in order to investigate whether memory and non-memory cognitive IRIs added to the predictive ability of the mean performance on memory and non-memory measures, discriminant function analysis was utilized. Table 4-9 shows the models used in the analyses and the canonical loadings and classification rates for each of the models. A four step approach was used. First, only the expected group distinction variables (i.e., memory level, AVLT Total Score and AVLT Percent Retained) were used to predict group (MCI, Non-MCI) membership. It is important to note that these AVLT scores represent the average AVLT performance from the daily assessments, and are independent of the memory measures administered at pretest (and used as part of the consensus conference). Eighty-five percent (58 out of 68) of the participants were correctly classified. The canonical loadings (Model One) showed that the two variables were equally important in making assignments (AVLT Total Score loading = .49; AVLT Percent Retained loading = .58). In step two, the model examined whether intraindividual variability might add additional group distinction information beyond that of level of performance in verbal memory. Thus, the next analysis examined the added benefit of the IRIs for the two memory variables. Results (Table 4-9, Model 2) showed that this model correctly classified one fewer participant (84%), as a Non-MCI participant was classified in the

PAGE 75

63 MCI group; in other words, memory variability was not helpful in further distinguishing the groups. In step three, we examined whether mean level differences in other non-memory domains would aid in distinguishing between the MCI and non-MCI groups. Model 3 (Table 4-9) shows that the added benefit of adding the mean level of performance for Backward Digit Span and Symbol Digit Score still does not exceed the classification attained by the memory performance variable alone. Table 4-9. Canonical loadings and classification statistics for discriminant function models. Model 1 Model 2 Model 3 Model 4 AVLT Total Score .49 .86 .88 .94 AVLT Percent Retained .58 .62 .58 .48 AVLT Total Score IRI .08 .13 .21 AVLT Percent Retained IRI .55 -.21 -.21 Digit Span Backward .09 .11 Symbol Digit Score .53 .53 Digit Span Backward IRI .14 Symbol Digit Score IRI .17 % Correctly Classified 85.3 83.8 85.3 83.8 % Sensitivity 60.0 60.0 60.0 53.3 % Specificity 92.5 90.6 92.5 92.5 In step four, we examined whether variability differences in these other non-memory domains might aid in distinguishing between the MCI and Non-MCI groups. Similar to the addition of the memory IRI values, as shown in Model 4 (Table 4-9) the addition of the IRIs for Backward Digit Span and Symbol Digit Score reduce the correct classification by one participant (in this case, one MCI participant was classified Non

PAGE 76

64 MCI). In summary, cognitive status was best predicted by performance on memory measures alone. Intraindividual Variability Over Time: Understanding Variability and Performance Relationships The results conveyed above demonstrated that over all thirty-one occasions of measurement the two groups differed in overall performance in memory. With regards to intraindividual variability, the two groups differed significantly only on percent of information retained after a short delay (AVLT Percent Retained), with the MCI group showing greater intraindividual variability overall. Group assignment was best predicted by performance on memory measures, with day-to-day variability in performance adding no additional predictive ability. The next main focus of the current study was to determine the relationship between performance and intraindividual variability, the details of which are outlined below. Aim Three The third specific aim of the study proposed to investigate the relationships between cognitive status, performance over 31 days, and intraindividual variability in daily performance. Specifically, the stated aim was to investigate whether, in conjunction with expected differences in overall practice-related improvement across sessions, memory-impaired and cognitively intact individuals show different patterns in intraindividual variability on tasks over time. Extant research (Allaire & Marsiske, 2002; Li et al., 2001) suggests that, on tasks in which cognitively intact elders show retest-related learning, variability has several properties; (a) it tends to be positively associated with level of performance, (b) it is uncorrelated with the variability observed after individuals reach asymptotic performance, and (c) it is reduced over occasions.

PAGE 77

65 Aim Three: Review of Analyses Intraindividual variability over time (occasion), across cognitive status Are there changes in intraindividual variability over time? The previously reported results demonstrated that the two groups based on cognitive status did not differ in intraindividual variability when averaged across the thirty-one occasions of measurement, except on AVLT Percent Retained. However, those findings did not address whether day-to-day variability changed over time, and if so, whether these changes might differ by group. In other words, whether intraindividual variability during earlier sessions was greater or less than intraindividual variability during later sessions, and whether the groups differ in the pattern of variability across early or late sessions. In order to demonstrate differences between the two g roups with respect to intraindividual variability changes over time, the thirty-one sessions were divided into three equivalent blocks of sessions, labeled Block 1 (occasions 0 10), Block 2 (occasions 11 20), and Block 3 (occasions 21 30). Table 4-10 depicts the mean IRIs for the cognitive variables for each of these three time-ordered blocks. A visual scan of this table reveals that for most variables there appears to be a trend, over the blocks, towards a reduction in intraindividual variability for each group. However, the comparisons of interest for the means in Table 4-10 are not only whether within each group there are trends across the blocks, but also whether the groups differ within each block. Any potential interaction of group and block was also of interest. Thus, in order to assess all of these comparisons, repeated measures analyses examining the presence of cognitive status Group and Block effects on variability were conducted, as described below and presented in Table 4-11.

PAGE 78

66 Table 4-10. Mean Intraindividual Residual Indices (IRIs) by block and by cognitive status. Measure MCI Non-MCI M(SD) M(SD) Memory Measures AVLT Total Score Block 1 3.28 (0.90) 3.38 (0.95) Block 2 3.33 (0.80) 3.14 (1.14) Block 3 3.11 (0.87) 2.71 (1.16) AVLT List 1 Block 1 4.33 (1.20) 5.06 (1.51) Block 2 4.45 (1.56) 4.80 (1.37) Block 3 4.22 (1.21) 4.63 (1.60) AVLT Delay Recall Block 1 4.25 (1.17) 3.96 (0.97) Block 2 3.25 (1.09) 3.94 (1.21) Block 3 3.44 (0.90) 4.03 (1.62) AVLT Percent Retained Block 1 7.00 (3.13) 4.46 (2.08) Block 2 5.28 (3.14) 3.99 (2.68) Block 3 5.43 (2.86) 3.24 (2.69) Non-Memory Cognitive Measures Digit Span Backward Score Block 1 4.50 (0.91) 4.34 (1.25) Block 2 4.46 (1.34) 4.57 (1.37) Block 3 4.43 (1.48) 3.90 (1.29) Symbol Digit Score Block 1 2.99 (0.77) 4.11 (4.10) Block 2 3.14 (1.24) 3.10(1.14) Block 3 2.65 (1.18) 2.75 (0.86) The blocks were utilized in multiple repeated measures ANOVAs, with Block IRI (Block 1: sessions 0 10; Block 2: sessions 11 20; Block 3: sessions 21 30) as the within subjects variable, and Group (Cognitive status) as the between subjects variable, in order to evaluate whether intraindividual variability was reduced over time in the

PAGE 79

67 cognitively intact but not MCI participants. Results indicated that main effects of Block were significant for AVLT Total Score, AVLT Delayed Recall, AVLT Percent Retained, and Symbol Digit Score, indicating a reduction in intraindividual variability over occasions. The linear trend of the Block main effect was significant for only AVLT Total Score [F(1,63) = 7.090, p = .010] and AVLT Percent Retained [F(1, 64) = 14.007, p = .001]. No quadratic effects of Block were significant. Main effects for Group (Cognitive Status) were significant for AVLT Percent Retained. As with the t tests above, participants with MCI showed substantially more intraindividual variability on this measure. Notably, the Block and Group interaction was not significant, except for AVLT Delayed Recall, suggesting that the MCI and Non-MCI groups did not differ in the slope of variability reduction over blocks, except for this one measure. For the AVLT Delay Recall score, the Non-MCI group showed little reduction in variability over Blocks, while the MCI group showed substantial variability reduction between Blocks 1 and 2. Table 4-11 shows the F-statistics, significance values, observed power, and effect sizes for these repeated measures ANOVAs. As power is low for most main and interaction effects, Figure 4-1 depicts the observed patterns of intraindividual variability across blocks for the two groups on all variables. Note that the y-axes for the graphs are all on the same scale (e.g., three units), even though they represent different ranges of units. Table 4-11. Repeated Measures ANOVA: Intraindividual variability predicted by block score and cognitive status. AVLT Total Score Source Df F p power 2 Between subjects Group 1 .76 .388 .137 .012 Within subjects Block 2 4.10 .019 .718 .061 Group x Block 2 .58 .560 .145 .009

PAGE 80

68 Table 4-11. continued AVLT List 1 Source df F p power 2 Between subjects Group 1 2.53 .117 .347 .039 Within subjects Block 2 .71 .493 .168 .011 Group x Block 2 .18 .834 .078 .003 AVLT Delay Recall Source df F p power 2 Between subjects Group 1 .88 .355 .149 .026 Within subjects Block 2 4.28 .023 .705 .211 Group x Block 2 4.15 .025 .690 .206 AVLT Percent Retained Source df F p power 2 Between subjects Group 1 8.08 .006 .800 .112 Within subjects Block 2 10.89 .000 .988 .257 Group x Block 2 1.73 .185 .351 .052 Backward Digit Span Source df F p power 2 Between subjects Group 1 .88 .352 .152 .014 Within subjects Block 2 .44 .645 .120 .007 Group x Block 2 1.51 .225 .317 .024 Symbol Digit Score Source df F p power 2 Between subjects Group 1 .62 .433 .122 .010 Within subjects Block 2 3.67 .031 .656 .104 Group x Block 2 .67 .514 .158 .021 Note Power computed using alpha = .05.

PAGE 81

69 123block 2.02.53.03.54.04.55.0Intraindividual Residual Index cog stat:Non-MCIMCI AVLT Total Score IRI by Cognitive Status 123block 3.54.04.55.05.56.06.5Intraindividual Residual Index cog stat:Non-MCIMCI AVLT List 1 IRI by Cognitive Status 123block 2.02.53.03.54.04.55.0Intraindividual Residual Index cog stat:Non-MCIMCI AVLT Delay Recall IRI by Cognitive Status 123block 456Intraindividual Residual Index cog stat:Non-MCIMCI AVLT Percent Retained IRI by Cognitive Status Figure 4-1. Intraindividual variability by blocks. Plots for (a) AVLT Total Score IRI, (b) AVLT List 1 IRI, (c) AVLT Delay Recall IRI, (d) AVLT Percent Retained IRI, (e) Backward Digit Span IRI, and (f) Symbol Digit Score IRI by cognitive status.

PAGE 82

70 123block 2.53.03.54.04.55.05.5Intraindividual Residual Index cog stat:Non-MCIMCI Backwards Digit Span IRI by Cognitive Status 123block 2.02.53.03.54.04.55.0Intraindividual Residual Index cog stat:Non-MCIMCi Symbol Digit IRI by Cognitive Status Figure 4-1. continued. Relationship between intraindividual variability and level of performance Subsequent to the above consideration of the patterns in intraindividual variability over time and across groups, the next relationship of interest is that of level of performance and day-to-day variability in performance. Several analyses were undertaken in order to investigate the relationship between level of performance and intraindividual variability for the participants as a whole as well as separately based on cognitive status. Previous research, conducted primarily with reaction time measures, suggests that poorer performers show more variability (i.e., the variability = vulnerability perspective). Allaire and Marsiskes results (e.g., 2004), however, have suggested that, in accuracy data, better performers are actually more variable, particularly in the early rapid acquisition phases of a practice curve. Is intraindividual variability positively or negatively associated with level of performance? As outlined in Table 4-12, Pearson product-moment correlations between the intraindividual residual index (calculated over all 31 occasions) and mean level of

PAGE 83

71 performance over all 31 occasions for the cognitive variables revealed significant correlations for AVLT Total, AVLT List 1, AVLT Percent Retained, and Backward Digit Span Score. With the exception of AVLT List 1, these significant correlations were negative, indicating that greater intraindividual variability was associated with lower levels of performance on the AVLT Total Score, AVLT Percent Retained, and Backward Digit Span. For AVLT List 1, the relationship was reversed, such that greater intraindividual variability was associated with higher performance on the first trial of the list learning task, a pattern identical to that reported by Allaire (2001) with the identical measure. Table 4-12. Correlations of mean level of performance and IRI. Variable Correlation Mean over 31 Occasions with IRI p AVLT Total -.253 .037 AVLT List 1 .424 .000 AVLT Delay .004 .984 AVLT Per. Ret. -.727 .000 Back Digit Span Score -.432 .000 Symbol Digit Score -.185 .131 Does the relationship between intraindividual variability and mean level of performance differ by cognitive status? Following the initial correlations of overall means scores and IRIs for the cognitive variables, the next set of analyses examined whether the mean-variability relationship was different for MCI and Non-MCI groups. Thus, the relationship between variability, performance, and cognitive status was investigated for the same variables via multiple univariate regressions taking the following form: IRI = Mean Performance (over 31 occasions) + Cognitive Status Group (1 = MCI, 0 = Non-MCI) + Mean x Group Interaction + error. All regression equations were significant, except for Symbol Digit Score. Results in Table 4-13 show that both

PAGE 84

72 main effects (Mean Score and Cognitive Status Group) as well as the interaction are significant for AVLT Total Score, AVLT Delay Recall, AVLT Percent Retained and Backward Digit Span, suggesting that level of performance was differentially related to variability by cognitive status. For AVLT List 1 and Symbol Digit Score only the Mean Score main effect was significant. The significant, and negative, main effects for Mean level of performance again indicated that for AVLT Total Score, AVLT Delay Recall, AVLT Percent Retained, Backward Digit Span, and Symbol Digit Score that the higher initial performance was, the lower the intraindividual variability, and are (as expected) identical to the results obtained in the correlational analysis above. The negative main effects for Group indicate, unexpectedly given our hypotheses, that when level of performance is controlled, the MCI group (Group = 1) shows less intraindividual variability than the Non-MCI group (Group = 0). Table 4-13. Regression coefficients for predicting IRI with mean level performance and cognitive status group. Variable (Standardized Beta Weights) F R 2 Mean Group Mean x Group AVLT Total Score 6.62** .237 -.48 ** -.43 ** .40 ** AVLT List 1 5.16** .195 .25 -.13 .10 AVLT Delay Recall 6.65** .384 -.58 ** -.62 ** .59 ** AVLT Percent Retained 28.88** .575 -.60 ** -.32 ** .31 ** Backward Digit Span 7.18** .252 -.49 ** -.28 .27 Symbol Digit Score 1.77 .077 -.25 -.18 .16 p < .05; ** p < .01 Notably, the interaction is significant for all variables where both main effects (Mean, Group) were significant. This complicates the interpretation somewhat, since with a significant interaction, AVLT Total Score, AVLT Delay Recall, AVLT Percent Retained, and Backward Digit Span evince a pattern suggesting that individuals in the Non-MCI group who perform at the lower overall levels of performance have higher

PAGE 85

73 individual variability indices (consistent with theories on strategy acquisition and practice-related intraindividual variability). In contrast, participants in the Non-MCI group who are performing at the higher end of the range, have lower variability. However, in the MCI group, higher performers have higher intraindividual variability, while lower performers have le ss variability. These patterns are clearer in Figure 4-2, which shows the scatterplots of the relations hip between IRI and mean score for the two cognitive status groups, with average best-fit ting lines superimposed, for all the memory and non-memory cognitive variables. 10.0020.0030.0040.00AVLT Total Score MEAN 2.004.006.008.0010.0012.00AVLT Total Score IRI cog stat:Non-MCIMCI 2.004.006.008.0010.0012.0014.00AVLT List 1 MEAN 2.004.006.008.0010.0012.00AVLT List 1 IRI cog stat:Non-MCIMCI Figure 4-2. Mean performance and intraindividual variability. Plots for (a) AVLT Total Score Mean and IRI, (b) AVLT List 1 Mean and IRI, (c) AVLT Delay Recall Mean and IRI, (d) AVLT Percent Retained Mean and IRI, (e) Backward Digit Span Mean and IRI, and (f) Symbol Digit Score Mean and IRI by cognitive status.

PAGE 86

74 0.002.004.006.008.0010.0012.0014.00AVLT Delay Recall MEAN 0.002.004.006.008.0010.0012.00AVLT Delay Recall IRI cog stat:Non-MCIMCI 0.200.400.600.801.00AVLT Percent Retained MEAN 2.004.006.008.0010.0012.00AVLT Percent Retained IRI cog stat:Non-MCIMCI 5.006.007.008.009.0010.0011.0012.00Backward Digit Span MEAN 2.004.006.008.0010.0012.00Backwards Digit Span IRI cog stat:Non-MCIMCI 10.0020.0030.0040.0050.0060.00Symbol Digit Score MEAN 2.004.006.008.0010.0012.00Symbol Digit Score IRI cog stat:Non-MCIMCI Figure 4-2. continued. Are there practice related gains in level of performance evidence across occasions? Do these gains differ by group? Do improvements reach asymptotic levels of performance? Does this differ by groups? Previously presented results have shown intraindividual variability changes over time, via patterns present over the three time-ordered blocks of occasions. Additionally, results describing the relationship between intraindividual variability and performance have been presented. However, in

PAGE 87

75 order to more clearly understand the interactions across the two groups between intraindividual variability and performance, it is crucial, at this point, to investigate whether there are practice-related gains in performance across the occasions of measurement. If so, the changes in intraindividual variability might be more clearly understood in the context of strategic acquisition and practice-related learning. Additionally, since previous work has theorized that intraindividual variability is positively associated with strategy acquisition through out performance gains, with a loss of relationship when asymptotic performance is reached, it is important to determine if performance over time reaches asymptote, and if so, if the groups differ in whether performance gains reach a plateau. The most effective method to answer these questions in one approach was via latent growth curves, estimated via simplified mixed model analysis. Mixed effects modeling allows for a determination of whether variables are related to each other in a fixed way for all participants (fixed effects), or in differing ways for different participants (random effects). Thus, we can determine if Time (linear and quadratic effects of occasion) influences performance, and if so, whether the influences differ based on Group (cognitive status). With regard to the question of performance gains, the effect of interest is whether, on average, a linear effect of Time was detected. With regard to the question of whether, on average, individuals leveled off, or reached asymptotes, the critical effect of interest is whether, on average, a quadratic effect of Time was detectable. In the models that follow, only the fixed (average) effects of Time are considered. The results are analogous to those of a traditional repeated measures analysis, but they have the advantage of using all available data, and not just data from those participants with no missing data at all occasions (i.e., listwise deletion).

PAGE 88

76 As seen in Table 4-14, results from the mixed effects analyses fell into the same pattern for AVLT Total Score, AVLT List 1, AVLT Delay Recall, AVLT Percent Retained and Backward Digit Span. The growth curves for all the cognitive variables can be seen in Figure 4-3. For the noted variables, the linear and quadratic effects of Time and the main effect of Group were significant. The interactions of Group with linear and quadratic effects of time were not significant. This means that for these variables, while the two groups differed on overall level of performance, the growth curves followed a similar trajectory. This trajectory included both linear and quadratic growth, indicating that, on average, performance gains were apparent, and performance asymptotes were reached (the quadratic main effect). An asymptote reflects a leveling off of performance gains, the presence of which allows for further analyses of the relationship of intraindividual variability and slope of performance gains, below. The remaining cognitive variable, Symbol Digit Score demonstrated a different pattern of effects following the mixed model analyses. Symbol Digit Score reflects linear and quadratic effects of time, but no main effect of Group, indicating that the growth curves, while containing a performance asymptote, are nearly identical for the two groups. Table 4-14. Time effects on mean performance. AVLT Total Score AVLT List 1 Fixed Effects Estimate F p Estimate F P Intercept 33.14 1654.16 .000 8.55 940.81 .000 Workbook a 9.82 .000 a 5.06 .000 Time .20 277.96 .000 .09 219.29 .000 Time 2 -.01 11.47 .001 .00 11.22 .001 Group -8.94 27.15 .000 -1.96 11.18 .001 Time x Group -.03 1.32 .251 -.02 2.33 .128 Time 2 x Group .00 .01 .943 .00 .02 .877

PAGE 89

77 Table 4-14. continued AVLT Delay Recall AVLT Percent Retained Fixed Effects Estimate F p Estimate F P Intercept 12.17 1026.57 .000 .922 2316.07 .000 Workbook a 8.26 .000 a 4.29 .000 Time .06 121.99 .000 .00 16.95 .000 Time 2 .00 16.20 .000 .00 8.32 .004 Group -5.00 39.91 .000 -.24 34.16 .000 Time x Group .00 .09 .764 .00 1.55 .213 Time 2 x Group .00 .86 .353 .00 .28 .594 Backward Digit Span Symbol Digit Score Fixed Effects Estimate F p Estimate F P Intercept 8.85 1877.86 .000 37.13 1140.69 .000 Workbook a 4.00 .000 a 6.27 .000 Time .05 144.11 .000 .31 202.08 .000 Time 2 .00 10.09 .002 -.01 -5.71 .017 Group -1.16 7.39 .008 -2.20 .87 .355 Time x Group .00 .32 .572 .04 .72 .396 Time 2 x Group .00 .33 .560 .00 .28 .600 a. There are 15 separate workbook estimates for each cognitive variable. 024681012141618202224262830Occasion 25.0030.0035.0040.00Mean Predicted Values cog stat:Non-MCIMCI Growth Curves for AVLT Total Recall 024681012141618202224262830Occasion 6.007.008.009.0010.0011.00Mean Predicted Values cog stat:Non-MCIMCI Growth Curves for AVLT List One Figure 4-3. Growth curves by cognitive status. Plots for (a) AVLT Total Score, (b) AVLT List 1, (c) AVLT Delay Recall, (d) AVLT Percent Retained (e) Backward Digit Span, and (f) Symbol Digit Score by cognitive status.

PAGE 90

78 024681012141618202224262830Occasion 6.008.0010.0012.0014.00Mean Predicted Values cog stat:Non-MCIMCI Growth Curves for AVLT Delay Recall 024681012141618202224262830Occasion 0.700.800.901.00Mean Predicted Values cog stat:Non-MCIMCI Growth Curves for AVLT Percent Retained 024681012141618202224262830Occasion 7.007.508.008.509.009.5010.00Mean Predicted Values cog stat:Non-MCIMCI Growth Curves for Backward Digit Span 024681012141618202224262830Occasion 34.0036.0038.0040.0042.0044.0046.00Mean Predicted Values cog stat:Non-MCIMCI Growth Curves for Symbol Digit Score Figure 4-3. continued Is intraindividual variability related to slope of practice-related improvement in performance? Is the amount of gain (linear slope) related to intraindividual variability? Does this relationship differ by groups? From the previous analysis, it is clear that there are distinctive trends in the slope of performance over time. The next question to be answered is whether intraindividual variability is related to performance

PAGE 91

79 gains. Table 4-15 illustrates the correlations between linear slope of performance gain (i.e., the slope coefficient obtained when performance for each participant is regressed on occasion) and intraindividual variability for all 31 occasions and separately for the three time-ordered blocks. The linear slopes for the individual blocks were obtained by conducting individual regression equations, separately for each participant, and separately for each block. We were then able to save the linear slope estimates for each participant, and correlate those linear slopes with the IRIs calculated for each participant in the same block. Thus, for each participant, on each cognitive variable, we obtained three IRI scores (Block 1, 2 and 3) and three linear slope estimates (Block 1, 2 and 3). First, we examined the relationship between overall performance gains (i.e., linear slope over all 31 occasions) and overall variability (i.e., IRI calculated over 31 occasions). Across all 31 sessions, performance gains (linear slope) and intraindividual variability were significantly and positive correlated for AVLT Percent Retained. Thus, greater gains were associated with greater intraindividual variability throughout the 31 occasions for this variable. For Symbol Digit Score the overall relationship between performance gains and intraindividual variability was reversed, such that fewer gains were associated with greater intraindividual variability. For none of the other cognitive variables assessed (AVLT Total, AVLT List 1, AVLT Delay Recall or Backward Digit Span) was the overall (31-occasion) relationship between variability and slope of improvement significantly different from zero. The analyses were then conducted separately for each of the three blocks of occasions, to answer the question of whether slope and variability might be differentially related at different points of the 31-occasion practice curve. The rationale for this

PAGE 92

80 analysis emerged from the quadratic growth curves presented above. Since it appeared that most of the gain occurred in the earlier sessions, with a leveling off of performance gains during later occasions, if gain and variability are positively related, we might have expected this relationship to be particularly strong during early Blocks. For Symbol Digit Number Correct there was indeed a stronger relations h ip between linear slope and variability in Block 1. However, as with the overall (31-occasion) relationship reported above, this relationship was negative, such that more variability was associated with smaller gain slopes. There was no relationship between gain and variability for Blocks 2 or 3. For most of the other measures, an inconsistent and scattershot pattern of correlations was detectable. For the AVLT Total Score, there was a positive relationship between gain (linear slope) and variability, but for Block 2 only. For AVLT List One, a negative relationship was found between gain and variability, but only for Block 1; a similar negative relationship was found for Backward Digit Span, but in Block 3 only. In order to determine if the relationships between performance gain and intraindividual variability remained constant for both groups, this was investigated via multiple univariate regressions taking the form: IRI = Linear Slope (over 31 occasions) + Cognitive Status Group (Non-MCI = 0, MCI = 1) + Slope x Group Interaction + error. Given the relatively haphazard pattern of results by block in the preceding analyses, this analysis examined only the overall (31 occasions) relationship between variability (IRI) and gain (linear improvement slope). Results in Table 4-16 show that Slope main effect was significant and negative for AVLT List 1, indicating that for this variable, the higher the performance gain, the lower the intraindividual variability. This is consistent

PAGE 93

81 Table 4-15. Between-person correlations between intraindividual variability and linear slope of performance gains separately for each variable. Variable Overall IRI Block 1 IRI Block 2 IRI Block 3 IRI AVLT Total Score Overall Slope .073 ---Block 1 Slope --.086 --Block 2 Slope -.192 .322** -Block 3 Slope -.022 -.157 -.227 AVLT List One Overall Slope -.053 ---Block 1 Slope --.264* --Block 2 Slope -.047 .187 -Block 3 Slope -.052 .079 .094 AVLT Delay Recall Overall Slope .117 ---Block 1 Slope -.131 --Block 2 Slope -.134 .152 -Block 3 Slope --.135 .252 .080 AVLT Percent Retained Overall Slope .281* ---Block 1 Slope --.139 --Block 2 Slope -.176 .222 -Block 3 Slope -.092 .018 -.129 Backward Digit Span Overall Slope -.073 ---Block 1 Slope --.013 --Block 2 Slope --.169 -.065 -Block 3 Slope --.205 .024 -.248* Symbol Digit Score Overall Slope -.319** ---Block 1 Slope --.265* --Block 2 Slope --.046 -.087 -Block 3 Slope -.001 -.162 -.014 p< .05; ** p< .01 with the correlational pattern seen above in Block 1 AVLT Percent Retained demonstrated a significantly positive main effect of Slope, which implies that the reverse pattern held for Percent Retained, i.e., that greater performance gains were associated with greater intraindividual variability. Again, this is consistent with the correlations discussed above. Notably, AVLT Percent Retained evinced a significant interaction, suggesting that the pattern is not the same in the two groups (i.e., that the positive relationship between gain and variability is really only true for persons without MCI).

PAGE 94

82 The only variable to demonstrate a significant main effect for Group was Symbol Digit Score, such that persons with MCI demonstrated less intraindividual variability. However, the Non-MCI group may have artificial raised intraindividual variability due to three outliers (see graph, Figure 4-4). The patterns of interaction are clearer in Figure 4-4, which shows the scatterplots of the relationship between IRI and linear slope for the two cognitive status groups, with average best-fitting lines superimposed, for all the memory and non-memory variables. Table 4-16. Regression Coefficients for predicting IRI with linear slope gain in performance and cognitive status group. Variable (Standardized Beta Weights) F R 2 Slope Group Slope x Group AVLT Total Score .49 .023 -.07 .00 .13 AVLT List 1 1.62 .027 -.26* -.08 .03 AVLT Delay Recall .71 .063 -.18 -.09 .13 AVLT Percent Retained 6.45** .232 .30* .03 -.26* Backward Digit Span .24 .011 .03 -.08 .03 Symbol Digit Score 2.81 .116 -.13 -.32** .06 p < .05; ** p < .01 -4.00-3.00-2.00-1.000.001.002.003.00AVLT Total Score Linear Slope 2.004.006.008.0010.0012.00AVLT Total Score IRI cog stat:Non-MCIMCI -4.00-3.00-2.00-1.000.001.002.003.00AVLT List 1 Linear Slope 2.004.006.008.0010.0012.00AVLT List 1 IRI cog stat:Non-MCIMCI Figure 4-4. Linear slope and intraindividual variability. Plots for (a) AVLT Total Score Slope and IRI, (b) AVLT List 1 Slope and IRI, (c) AVLT Delay Recall Slope and IRI, (d) AVLT Percent Retained Slope and IRI, (e) Backward Digit Span Slope and IRI, and (f) Symbol Digit Score Slope and IRI by cognitive status.

PAGE 95

83 -4.00-2.000.002.004.00AVLT Delay Recall Linear Slope 2.004.006.008.0010.0012.00AVLT Delay Recall IRI cog stat:Non-MCIMCI -7.50-5.00-2.500.002.505.00AVLT Percent Retained Linear Slope 2.004.006.008.0010.0012.00AVLT Percent Retained IRI cog stat:Non-MCIMCI -2.000.002.004.006.008.00Backward Digit Span Linear Slope 2.004.006.008.0010.0012.00Backwards Digit Span IRI cog stat:Non-MCIMCI -4.00-2.000.002.004.00Symbol Digit Score Linear Slope 2.004.006.008.0010.0012.00Symbol Digit Score IRI cog stat:Non-MCIMCI Figure 4-4. continued Sources of Intraindividual Variability To this point we have described intraindividual variability alone and have described patterns of relationships between intraindividual variability in performance,

PAGE 96

84 overall performance, and performance gains (e.g., linear trends of slope). These considerations have reflected relationships within each variable. The section that follows considers the relationships of intraindividual variability across the variables, over the 31 occasions of measurement. First, correlations between the IRIs are presented in order to answer questions regarding the nature and degree of relationships across the Intraindividual Residual Indices. Specifically, the correlations provide an assessment of how over all participants, and over all occasions, the tendency to be variable, or to fluctuate, on one variable is related to the tendency to fluctuate on another. The correlations are followed by presentation of ana l yses utilizing the mixed effects modeling approach. This approach answers questions regarding the nature and degree of relationships between intraindividual variability across variables within each day. In other words, the mixed model analysis provides information as to whether up days on one variable are related to up days on another. The approach has been described as an analysis of coupling of variables on a day-to-day basis. How is intraindividual variability on one measure related to intraindividual variability on another measure? How are variabilities coupled? The correlational relationship between the intraindividual variability coefficients (IRI) for sub-indices from our three cognitive tasks (AVLT List Memory, Backward Digit Span, and Symbol Digit) and the three non-cognitive measures (PANAS, Sleep indicators, and Environmental Distractors) are provided in Table 4-17. The magnitude and direction of the correlations between the IRIs for the three cognitive measures were not uniform. In general, IRIs for indices calculated within the same measure (cognitive or non-cognitive) were positive.

PAGE 97

85 An exception was between AVLT List 1 and AVLT Percent Retained. This IRI correlation was significant and negative, indicating that participants with higher intraindividual variability on one measure also had lower intraindividual variability on the other measure. Between cognitive tests, there were few significant associations in intraindividual variability, except for Backward Digit Span, which was negatively correlated with AVLT List 1. That is, individuals who evinced more variability on Backward Digit Span evinced less variability on AVLT List 1. Between the non-cognitive measures, there were several significantly positive correlations, indicating that greater intraindividual variability on one was associated with greater intraindividual variability on the other. Such correlations were found between PANAS Positive Affect and Environmental Discomfort, PANAS Positive Affect and Environmental Distractions, PANAS Negative Affect and Sleep Efficiency, and PANAS Negative Affect and Environmental Discomfort. Finally, the correlations across the cognitive and non-cognitive measures were positive, when significant, meaning that individuals who showed greater intraindividual variability on the cognitive measure also showed greater intraindividual variability on the non-cognitive measure. These significant correlations were between AVLT List 1 and Sleep Time, AVLT List 1 and Environmental Distractors, AVLT Learning and Sleep Time, and Symbol Digit Score and Environmental Distractors.

PAGE 98

Table 4-17. Intercorrelations of Individual Residual Indices. Measure 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 1. AVLT Total Score .18 .52** .38** -.06 -.01 -.02 -.08 .06 .02 .02 -.01 2. AVLT List 1 1 1 .23 1 -.18 -.36** -.25* -.16 .14 -.05 .24* .15 .09 .28** 3. AVLT Delay Recall 1 .76** .08 .04 .01 -.14 .01 .03 -.10 .12 4. AVLT Percent Retained 1 .05 .14 -.09 .02 -.09 .01 -.03 .10 5. Backward Digit Span 1 .13 -.08 .09 -.12 .00 -.10 -.04 6. Symbol Digit Score .44** .01 .00 .13 .19 .18 .39** 7. PANAS: Positive 1 .37** .18 .22 .62** .30* 8. PANAS: Negative 1 .24 .27* .41** .24 9. Total Sleep 1 .78** .18 .09 10. Sleep Efficiency 1 .32** .00 11. F1: Discomfort 12. F2: Distractions

PAGE 99

87 While the correlations provide overall relationships between the IRIs across variables, as noted above, mixed effects models allow for the determination of how two or more variables, assessed repeatedly, might move together on a day-by-day basis. This allows for an evaluation of the degree to which higher or lower days on one measure might predict higher or lower days on another. In the mixed effects models used for these analyses, the variables of interest were centered, that is, for each participant, each daily score less that individuals mean score was used. Thus, individual day-to-day variability remains, but level of performance differences are controlled. As a result, only fixed effects (those that depict how the variables are related for all participants) are of interest in these models. Random effects (which involve individual differences in performance levels) were eliminated from the models to aid in interpretation. The model for each variable of interest was run in a step-wise fashion, initially including all possible coupled variables (e.g., other cognitive measures as well as non-cognitive measures). Variables from the same measure were not used, since, for example, AVLT Total Score can be almost perfectly predicted from AVLT List 1, due to the linear calculations that relate these variables. In each subsequent model for each variable, non-significant effects were removed until the best fitting model, with only the significantly coupled variables remained. Table 4-18 shows the results for the final models for each variable of interest. Although there were no significant Group effects, the parameter estimate, F statistic, and significance value for the Group Fixed effects appear in each table to illustrate this. All analyses were conducted allowing for the linear and quadratic effects of time as well as any potential effects of different versions (e.g., workbook version). These were universally significant, but are not shown, to simply the presentation.

PAGE 100

88 With regards to the memory variable, day-to-day variability in positive affect, as measured on the PANAS Positive Affect scale, along with fluctuation s in environmental distractors (Factor 2) and discomfort (Factor 1), are reliably related to the intraindividual variability observed on the AVLT Total Score and AVLT Delay Recall (Results in Table 414). Similarly, positive affect (PANAS Positive Affect) and environmental discomfort (Factor 1) were coupled with AVLT List 1, while environmental distractions and discomfort (Factors 1 and 2) were coupled with AVLT Percent Retained. The similar patterns of relationships on the memory variables reflects the inter-relationships of the variables on the memory measure. With regards to the non-memory cognitive measures, Symb o l Digit Score was not significantly coupled with any other variables, but day-to-day fluctuations in Backward Digit Span moved with fluctuations in environmental distractors (Factor 1), likely due to the attentional needs of this working memory task. In turn, the non-cognitive variables revealed coupled variabilities. Intraindividual variability in AVLT Total score, environmental discomfort (Factor 2) and negative affect (PANAS Negative Affect) moved in concert with daily fluctuations in positive affect on the PANAS. In contrast, day-to-day variability of negative affect on the PANAS was coupled with environmental discomfort (Factor 1), and total sleep time. Total time asleep reflected the same pattern, as negative affect and environmental discomfort were significantly coupled. Sleep efficiency (calculated as amount of time in asleep out of time in bed) moved in concert with a number of other variables including Symbol Digit, PANAS positive and negative affect, and environmental discomfort and distractions. Finally, the two factors related to the testing environment (both internal and

PAGE 101

89 external) revealed their own coupling patterns. Factor 1, which related to discomfort including pain and stress level, moved, on a day-to-day basis, with AVLT Total score, Backward Digit Span, PANAS Positive and Negative Affect, Factor 2 (environmental distractions) and Total time asleep. The second factor, related to distractions such as noise, interruptions, and others in the testing room, was coupled with Symbol Digit, PANAS Negative Affect, and Factor 1 (environmental discomfort).

PAGE 102

90 Table 4-18. Coupled variabilities. AVLT Total Score AVLT List 1 Fixed Effects Estimate F p Fixed Effects Estimate F p Group -.10 .22 .639 Group -.02 .058 .810 PANAS Pos .39 4.54 .033 PANAS Pos .21 5.29 .022 F1 Discomfort -.12 53.04 .000 F1 Discomfort -.03 14.13 .000 F2 Distract -.03 5.35 .021 AVLT Delay Recall AVLT Percent Retained Fixed Effects Estimate F p Fixed Effects Estimate F p Group -.05 .11 .743 Group .00 .00 .974 PANAS Pos .31 5.10 .024 F1 Discomfort .00 11.58 .001 F1 Discomfort -.07 29.97 .000 F2 Distract .00 17.39 .000 F2 Distract -.03 13.82 .000 Backward Digit Span Symbol Digit Score Fixed Effects Estimate F p Fixed Effects Estimate F p Group .00 .01 .916 Group -.04 .01 .915 F1 Distract -.01 6.46 .011 -PANAS Positive Affect PANAS Negative Affect Fixed Effects Estimate F p Fixed Effects Estimate F p Group .00 .00 .973 Group .01 .29 .592 AVLT Total .01 4.65 .031 F1 Discomfort .01 87.15 .000 PANAS Neg -.08 7.79 .005 Total Sleep .00 17.08 .000 F1 Discomfort -.02 75.92 .000 Total Sleep Time Sleep Efficiency Fixed Effects Estimate F p Fixed Effects Estimate F p Group -.14 .00 .968 Group .26 .29 .593 PANAS Neg -18.68 16.93 .000 Symbol Digit -.09 6.32 .012 F1 Discomfort -1.59 24.69 .000 PANAS Pos 1.96 19.29 .000 PANAS Neg -3.94 42.70 .000 F1 Discomfort -.17 15.40 .000 F2 Distract -.08 8.24 .004

PAGE 103

91 Table 4-18. continued Factor 1: Discomfort Factor 2: Distractions Fixed Effects Estimate F p Fixed Effects Estimate F p Group -.15 .26 .609 Group .00 .00 .999 AVLT Total -.20 47.00 .000 Symbol Digit -.06 4.80 .029 Back Digit Sp -.20 5.36 .021 PANAS Neg 1.01 4.27 .039 PANAS Pos -2.08 73.24 .000 F1 Discomfort .11 10.77 .001 PANAS Neg 2.70 70.33 .000 F2 Distract .04 6.77 .009 Total Sleep -.01 16.83 .000 Note: PANAS = Positive and Negative Affect Scale.

PAGE 104

CHAPTER 5 DISCUSSION The prevailing question in this study was whether intraindividual variability in cognition within a sample of older adults serves as a meaningful indicator of underlying cognitive status and learning potential. This chapter will review the major study findings, attempting to highlight study contributions to the growing literature on cognitive variability in cognitively intact and impaired older adults. After considering some of the underlying theoretical and methodological issues, a brief review of the study limitations accompanies the discussion of findings. The chapter concludes with suggestions for future directions of study as well as further analyses of the available data. This study attempted to integrate recent findings in the emerging field of intraindividual variability research. As noted previously, explorations of intraindividual variability in cognitive functions in older adults have, to date, focused on two distinct areas. One, through investigation of intraindividual variability in cognitively intact community-dwelling older adults, provides evidence and theory that intraindividual variability in cognitive functioning is directly related to strategy acquisition and performance improvement (Allaire, 2001; Allaire & Marsiske, 2004). Similar to findings in work with childrens developmental changes (Siegler, 1994), this arm of research suggests that intraindividual variability is heightened when individuals are exploring cognitive strategies for new tasks. As useful and successful strategies are identified, these are chosen in a consistent manner to attack the task. Thus, performance becomes 92

PAGE 105

93 not only more successful, but also more consistent, resulting in concurrent declines in intraindividual variability. The second distinct area of intraindividual variability research has focused on the patterns of intraindividual variability observable in older adults with declining cognitive function, specifically those diagnosed with dementia (Hultsch et al., 2000, Murtha et al., 2002, Walker et al., 2000). To date, such explorations have been generally restricted to assessments of intraindividual variability in latency of response on processing speed and recognition memory tasks, although work by Li and colleagues (2001) included an investigation of recall memory in addition to latency and vigilance tasks. This branch of research has provided evidence that older adults with dementia perform worse and respond with greater intraindividual variability in performance than older adults without cognitive impairment. Additionally, findings suggest that this observed increase in individualized performance fluctuations appears to be unique to cognitive impairment, as older adults with solely physical limitations did not present with greater intraindividual variability (Hultsch et al.). Such findings have expanded to incorporate earlier speculations that intraindividual variability is a unique indicator of neurologic compromise (Li & Lindenberger, 1999) and may foretell impending decrements in performance on cognitive tasks. In this vein, Murtha and colleagues explored intraindividual variability as an indictor of frontal lobe dementia, while Walker et al. reported that variability in performance on attention and vigilance tasks discriminated between participants with intact cognition and those with Lewy-Body dementia and Alzheimers disease.

PAGE 106

94 This study attempted to integrate these disparate ideas on the meaning and causality of intraindividual variability in cognitive functioning. As a group, older adults with mild cognitive impairment of the amnestic type are, by definition, generally cognitively intact with the exception of focal impairment evidenced by performance on immediate and delayed memory tasks. This group has been thought to contain a substantial subset of individuals who are in transition from normal function to a degenerative cognitive decline (Celsis, 2000; Petersen, et al., 1999; Tierney et al., 1996). Thus, individuals with mild cognitive impairment represent a group for whom the two hypotheses about intraindividual variability might apply. In fact, the specific question for this group would be: does their intraindividual variability in performance reflect strategy acquisition, with reductions in variability as performance improves, or does their intraindividual variability reflect cognitive compromise in memory systems? If intraindividual variability is reflective of cognitive compromise, is it focal (i.e., observed only in the domain memory which has experienced decline), or more pervasive, such that increases in intraindividual variability in non-memory cognitive domains would occur prior to observable decrements in performance? One would expect such non-memory effects under the assumption that persons with MCI are on a pathway to more global cognitive impairment, and heightened inconsistency in non-memory tasks serve as an early warning indicator of impending cognitive decline beyond the existing memory dysfunction. Review of Study Findings Aim One and Aim Two Above, in the introductory literature review, six major findings from extant research were presented to form the basis for the current study aims and hypotheses. The

PAGE 107

95 study findings are summarized with regard to the three aims and these six major notions, four of which relate to aims one and two. 1. Normal cognitive aging and cognitive impairment are distinguishable by level of performance on cognitive measures. In the current study, participants were assessed on a brief neuropsychological battery prior to group designation. Scores were utilized during the consensus conference to determine group assignment. Group comparisons on these measures confirmed the amnestic nature of the mild cognitive impairment in the MCI sample, as participants adjudged to belong to the MCI group had significantly lower scores on submeasures of the HVLT-R and BVMT-R. Interestingly, one non-memory measure (BNT) also showed significant differences between groups. Notably, the MCI group, while performing slightly worse on the BNT, performed in the average range on this measure. This group difference might be due to overly verbal, highly educated individuals in the intact group (although in this study, MCI and non-MCI participants did not differ in education), or emergent non-memory reductions in cognitive function in the amnestic MCI group. All other cognitive, non-memory measures (Trails A, Trails B, COWA, NAART, REY-O) did not differ significantly between groups. Of course, these comparisons may not be true tests of MCI/Non-MCI differences, since these tests were also examined in the consensus conferences in which participant group assignment was determined. Thus, these group differences in level of cognitive performance were independently confirmed by analysis of participants overall mean performance, over 31 days, from the daily variability phase. The distributions for the mean level of performance over all occasions fit a normal distribution, with the exception of AVLT

PAGE 108

96 Percent Retained, which was skewed. All submeasures from the daily AVLT showed significantly lower mean level of performance in participants with MCI. Prior to Bonferroni correction, Backward Digit Span also showed a significant difference, but this difference was no longer significant after correction, again confirming the specifically amnestic nature of the impairment in the MCI group. Thus, in this sample, normal cognitive aging and mild cognitive impairment could be differentiated based level of performance on cognitive tasks in the memory domain. 2. Persons at lower levels of functioning, and who are older, often show more inconsistency on a variety of cognitive measures. 3. In addition, persons with cognitive impairment (i.e., dementia) evince even higher levels of intraindividual variability than non-demented elders. These two notions from the existing literature are related and tie directly with aim one. Since previous research relied on demented elders, the pivotal question of aim one was whether individuals with mild cognitive impairment demonstrated greater intraindividual variability in cognitive functioning on tasks assessing memory performance. Like Li and colleagues (2001), who found different relationships for different measures in their older adult sample, a different MCI-related pattern was found for these measures. In fact, individuals in the mild cognitive impairment group demonstrated greater intraindividual variability only on the percent of material retained over a delay (AVLT Percent Retained). While this variable is highly related to the determination of impaired status (e.g., impaired performance on delayed recall on list learning task at baseline), other measures of memory functioning were expected to show the same pattern.

PAGE 109

97 In fact, on the first trial of the list learning task (AVLT List 1) in the Daily Cognitive Assessment Battery, individuals without cognitive impairment performed with greater intraindividual variability. This finding suggests that individuals with intact cognition were more variable in attention and initial encoding (processes the first trial is most presumed to capture). These dissociated results may actually provide a bridge between the two hypotheses regarding intraindividual variability in older adults. While this is admittedly highly speculative, perhaps, for the individuals with intact cognition, intraindividual variability is associated with strategy development over trials, which primarily comes into play during the initial trial learning, while for individuals with cognitive impairment, such strategic gain is less likely. In such a case, variability is more of a negative, reflecting inconsistency, rather than an aspect of performance gain. Clearly, the relationship between cognitive functioning and intraindividual variability on memory tasks is complex. Thus, further exploration of the relationship of intraindividual variability, level of performance, and performance gain over occasions for the two groups based on cognitive status was conducted in order to clarify the findings and will be reviewed below (aim three). Although the two groups did not differ in age, it is of interest to note that in the overall sample greater age was more associated with intraindividual variability on AVLT Percent Retained and Backward Digit Span, consistent with earlier findings that increased age is correlated with increased variability. 4. In MCI, level of performance may be relatively unaffected in some areas, such as attention, working memory, and processing speed. This finding from the literature prompted the question for study aim two: will impairment-related increases in

PAGE 110

98 intraindividual variability in unimpaired domains be detectable even before substantial performance level impairments are seen? Aim two focused on the differences in intraindividual variability in cognitive function between the two groups on the non-memory cognitive measures in the daily battery. Initial hypotheses were that intraindividual variability in cognitively impaired older adults would be a harbinger of future performance losses. Thus, it was expected that intraindividual variability differences would clearly distinguish the two groups, even in the absence of differences in level of performance. Again, findings did not support hypotheses. In fact, there were no significant differences between the two groups on overall intraindividual variability in performance on the non-memory measures. Additionally, intraindividual variability was not a useful predictor of group membership in a series of hierarchical discriminant function analyses. Likely, with groups known to be similar in overall performance in these non-memory cognitive domains, differences in intraindividual variability may be more subtle, possibly reflecting patterns in the relationship between level of performance and intraindividual variability over time. This was investigated as part of aim three (below). In summary, aims one and two investigated the relationship of cognitive status and overall intraindividual variability on memory and non-memory domains of cognitive functioning, with the expectation that individuals with MCI would show increased variability. Only one variable clearly followed this pattern, AVLT Percent Retained, or percent of information recall after a delay. It is noteworthy that both the mean performance level and the degree of intraindividual variability of this variable capture differences between the groups. The mean distribution of AVLT Percent Retained was

PAGE 111

99 noted to be skewed, however the distribution of the IRI generally followed a normal distribution, as did those of all IRIs, with the exception of Symbol Digit. Due to the fairly consistent lack of group differences in intraindividual variability, a check on the reliability of the intraindividual residual indices (IRI) was done by examining intercorrelations between IRI measured in three successive blocks of occasions (Occasions 0-10, 11-20, 21-30). In general the variables showed modest but positive relationships across the blocks, with the exception of AVLT Delay Recall. The modest degree of intercorrelations suggested that while the fluctuations in performance are related, systematic differences across blocks might be a factor in reducing the reliability of the IRIs over time. First, the low correlations across blocks might reflect unreliability. It is important to remember that intraindividual variability includes not only reliable inconsistency in participants, but also true unsystematic noise or measurement error. Allaire and Marsiske (2004), for example, had 40 occasions in each of their three blocks; here, we have only 10 or 11. Thus, the number of occasions may have constrained reliability of the IRI as well. Second, however, the low correlation across blocks might reflect the fact that if the meaning of intraindividual variation changed for one or both groups this would impact the overall relationships between the IRIs and the groups (i.e., that subjects had moved from a preto post-asymptotic phase, and that variability had shifted from a learning and strategy acquisition manifestation to an inconsistency manifestation). Thus, more complex relationships between cognitive status, intraindividual variability, level of performance, and performance gains over time were explored in aim three.

PAGE 112

100 Aim Three The final two of the six research findings outlined in the literature review resulted in the study questions driving aim three. 5. In non-impaired elders variability appears to change, qualitatively, over the course of many retest occasions. For tasks in which there is learning and performance improvement, variability is reduced once responding reaches a stable performance asymptote, leading to the notion that intraindividual variability is due to strategy acquisition during performance gains. 6. The relative absence of strategic learning is a hallmark characteristic of dementia, and is often found in early dementia, pre-dementia, or MCI. In answering the questions behind aim three, we obtained further insight into any differences between the two groups in level and degree of intraindividual variability, as these questions focused on the relationship of level of performance, improvement in performance over time, and patterns of intraindividual variability over time. This aim was addressed in several segments, asking whether (a) variability changed over occasions, (b) there were observable performance gains over the occasions, (c) performance gains reached an asymptotic level of performance, (d) variability was correlated with performance gains (slope) before asymptote and not after asymptote was reached (on average, for the groups), and whether (e) these patterns were observable only in the group of cognitively intact individuals. Since a key aspect to the diagnostic label of mild cognitive impairment is reduced performance on tasks of learning and memory, and intraindividual variability was proposed to be positively related to strategic learning changes (i.e., subjects would be more variable during the period of maximal gain), it was hypothesized that a presumed lack of performance gains in persons with MCI might result in a lack of concurrent

PAGE 113

101 change in intraindividual variability. A further implication is that persons with MCI would have flatter gain slopes than persons without MCI. Additionally, following the work of Hultsch and others, it was hypothesized that persons with MCI might show more inconsistency during the asymptotic part of their practice-related curves, reflecting a greater difficulty in maintaining stable performance. (A.) Intraindividual variability change over occasions: The repeated measures ANOVA of the individual residual indices (IRI) over the three time-ordered blocks of occasions provided strong evidence that intraindividual variability changed over time. Specifically, memory (AVLT Total Recall, AVLT Delayed Recall, AVLT Percent Retained) and attention and processing speed (Symbol Digit Score) intraindividual variability indices changed over time for both groups. All changes were significant reductions in intraindividual variability over time, and approximately equivalent for both groups, with the exception of AVLT Delay Recall. The pattern for the intraindividual variability for this variable was that the two groups were at approximately the same degree of intraindividual variability initially, but only the MCI group showed significant reduction in variability over time. In summary, both groups displayed evidence of IRI reduction over time. This may reflect strategic performance gains that occurred in both groups. (B.) Performance gains over occasions: A key question beyond the issue of whether IRI changed over time was whether level of performance changed over time for the two groups. Significant linear fixed effects of time in the simplified mixed models analysis provided clear evidence that practice-related performance gains were made over time for both groups. In fact, the growth curves demonstrated that the performance gains

PAGE 114

102 were almost identical for the two groups, but at persistently different levels of performance (i.e., Non-MCI participants outperforming those with MCI at all points), for all variables except Symbol Digit Score. This suggests that the group distinction is clear with regards to performance, but that MCI status in this study was unrelated to the potential for improvement, as reflected in the identical practice effects over time. (C.) Asymptote of performance gains: In order to determine if intraindividual variability was reduced as performance gains diminished, quadratic trends were also examined in the growth models, with strong quadratic effects taken as evidence that participants, on average, reached an asymptote in performance level. The mixed model analyses of growth also revealed quadratic trends to the growth curves for both groups over all variables. The growth curves again revealed the parallel nature of both the performance gains and the leveling off of gains at the asymptote for both groups. (D.) Correlation of performance gains and intraindividual variability: An crucial question for this aim, and one that is closely tied with previous work, is that of whether there was a positive or negative association between level of performance and intraindividual variability. When significant, level of performance and intraindividual variability in cognition were negatively associated for all measures except the first trial of the list learning task (AVLT List 1), a pattern that is consistent with the majority of previous studies (Hultsch et al., 2000; Li et al., 2001; Rabbitt, Osman & Moore, 2001). In fact, this replicates a fairly common finding in the intraindividual variability literature to date. This negative association indicates that higher level of performance is associated with lower intraindividual variability and vice versa. Across the entire sample, poorer performers demonstrated greater intraindividual variability. This is consistent with the

PAGE 115

103 hypothesis that greater intraindividual variability is due to, and reflective of, cognitive compromise. Notably, as we saw above, over the occasions, intraindividual variability decreased over time while performance improved over time. This relationship, which is consistent with idea of variability being associated with strategy acquisition, is also consistent with the finding of a negative correlation between variability and performance if the better performers have no need for strategy acquisition. Note an important implication: better performers will have no need for strategy acquisition when the tests used are very easy for them, either because they are already at ceiling, or because the demands of the task are so clear that they quickly educe the requirements of the task and experience rapid performance gain. A reverse relationship was found for AVLT List 1. Notably, AVLT List 1 was also the variable for which the Non-MCI group demonstrated greater variability. These findings are consistent, since the Non-MCI group performed better, on average. Again, we can speculate that, for those who showed performance gains, greater intraindividual variability on this measure of initial encoding and attention provided the necessary strategy acquisition to improve over time. This argument is less satisfactory when we reconsider that both groups demonstrated similar performance gain curves. It must be noted that, specific to this AVLT List 1, Allaire and Marsiske (2004) found the identical pattern of positive associations between level and variability (they also used the exact same lists as were used in this study). Thus, there seems to be something consistently different about the AVLT List 1 task, that differs from our remaining tasks. Notably, the regression analyses investigating the variability-performance relationship by group revealed significant interaction effects for several variables. Thus,

PAGE 116

104 for AVLT Total Score, AVLT Delay Recall, AVLT Percent Retained, and Backward Digit Span poorer performers in the Non-MCI group had greater variability and better performers had lower intraindividual variability in this group. In the MCI group, for most variables, intraindividual variability was stable across all levels of performance. Although the Symbol Digit score interaction was not significant, the plot shows a similar pattern with the exception of three outliers. This pattern implies that for the MCI participants, performance was non-contingent on intraindividual variability. This finding provokes speculation as to the degree to which heterogeneity in the MCI group impacted the ability to distinguish individuals who might be variable due to cognitive compromise vs. those whose fluctuations in performance might be reflective of strategy acquisition. The linear slope by variability correlations across blocks demonstrated how performance gains (slope) were related to intraindividual variability over time. Few block intercorrelations were significant. Those that were significant and near-significant indicated that for some participants, improvements in performance during the first block of sessions was associated with lower levels of intraindividual variability, while for some participants, performance gains in later trials were positively associated with intraindividual variability. It is difficult to speculate about the origins of the mixed correlational findings obtained. One speculation is that the average linear gain slopes in the study might have been driven by different groups of subjects at different points in the study. Perhaps the positive slopes in Block 1 might be attributable to early bloomers, those who performed well initially, had no need of new strategies, and therefore demonstrated little variability in their performance. Perhaps the positive slopes in Block 2, in contrast,

PAGE 117

105 might be attributable to late bloomers or those who needed some sessions to learn the task, did engage in strategy acquisition as they developed skills on the tasks. Their trial and error learning resulted in day-to-day fluctuations in performance. The final block of sessions may represent the, on average, post-asymptotic phase of performance for most participants, resulting in the expected near zero correlation between performance and variability. Here, intraindividual variability may be due to factors that are less cognitively related, as the known tasks might require fewer cognitive demands in general. An alternative method of differentiating groups, such as using the idea that participants might be better described by late versus early bloomers is also suggested in the finding that the two cognitive status groups did not differ on strategy use on the AVLT List 1 at the first and last occasion. (Domenech et al., 2004). In other words, the MCI grouping does not appear to define a qualitatively different group. Overall, the findings suggest that although the two groups are clearly distinguished by level of performance on memory measures, the degree and nature of performance gains and intraindividual variability over the thirty-one occasions is not different across groups. Notably, both groups demonstrated similar performance gains and similar decrements in intraindividual variability over time. One finding that clearly differed between the groups was that the Non-MCI participants demonstrated a stronger relationship between level of performance and degree of variability. That is, for the Non-MCI group, low performance was related to greater variability and vice versa, while for the MCI group, for most variables, intraindividual variability was similar for most levels of performance. This is not just a statistical power issue, but is clearly reflected in the

PAGE 118

106 differential slopes of the scatter plots for the two groups. This finding suggests that the MCI group, as a whole, may be less predictable. Nonetheless, the overall similarities in the performance gains and intraindividual variability indices across the groups clearly indicate that the individuals adjudged as having MCI benefited equally from repeated exposures to the tasks. This provides additional evidence that while individuals with MCI demonstrate overall lower level of performance than their age and educationally matched cognitively intact peers, they are capable of improvement and, as a group, do not qualify for a dementia diagnosis. It is likely that the individuals in the MCI group are heterogeneous, as this maybe somewhat inherent in the Petersen definition (Petersen, 2001). Groups defined primarily by impaired memory performance probably include a mix of true pre-clinical dementia participants and low normal elders, a fact affirmed by high rates of conversion from MCI-to-normal in some studies. While this problem plagues the entire MCI literature, it may be particularly acute in a study of subtle differences between groups, as in variability. Intraindividual Variability Interrelationships Across Domains A final consideration in the review of findings is that of the interrelationships of the intraindividual residual indices. Correlations between the IRIs across cognitive and non-cognitive domains provide evidence as to how the variables might be similar or different with regards to inconsistency. Expressed differently, a strong positive correlation in the IRI of two variables would suggest that persons tendency to be variable on one measure is strongly related with their tendency to be variable on a second measure.

PAGE 119

107 Intraindividual residual indices within each measure were positively correlated (e.g., most of the subindices of from the AVLT were positively correlated), with the exception of AVLT List 1 and AVLT Percent Retained. The significant negative correlation between these two variables suggests that the greater fluctuations in the number of items recalled on List 1, the lower the variability in the percent of material recalled after the delay; this continues the pattern of findings that suggests that AVLT List 1 is different in some way from all of the other variables. The cross-domain intercorrelations of IRI indices (cognitive variables with non-cognitive variables) were positive when significant, suggesting that individuals may have a somewhat trait-like tendency towards greater variability across multiple tasks. This is more supportive of intraindividual variability being due to a common neurologic or biologic etiology. A rather different analysis strategy was employed to look at the issue of coupled variability. Unlike correlated IRIs, which simply address the question of whether the tendency to be inconsistent on one variable is associated with inconsistency in another, coupled variability analyses examineon a day by day basiswhether the tendency to have up days in one measure (i.e., days above or below ones personal mean) are associated with the tendency to have up or down days in another measure. Thus, the correlated IRI analysis addresses whether the size of the variability band around ones personal learning curve shows stable interindividual ranks across multiple measures; the coupling analysis asks whether (within subjects) two variables move together or have shared trajectories over time.

PAGE 120

108 To investigate coupling of variables trajectories, mixed effects modeling was again utilized. The models identified sets of variables that seemed to be coupled from day-to-day. For example, AVLT Total Score was found to move with PANAS Positive Affect, Environmental Discomfort and Environmental Distractors. This meant that on a specific day when an individual participant reported improved positive affect, reduced discomfort and pain, and reduced distractors, then performance on AVLT Total Score was greater. An interesting overall finding in these analyses of coupled variabilities is the degree to which affect and environmental conditions are coupled with cognitive performance. These analyses do not provide information as to causality, so we can speculate as to whether affect impacts performance on cognitive measures or whether performance impacts reported affect (i.e., reduced positive affect was a reaction to the veridical perception of poor AVLT performance). Likely environmental distractors impact cognitive performance, yet, knowledge of poor performance could result in over reporting of distractors (i.e., shifting the blame for perceived poor performance to environmental noise). Regardless, the findings suggest that intraindividual variability in cognition may be thought of as state-like, reflecting lawful, predictable performance variations that capture personally above and below average days. Some studies have begun to examine this coupled variability notion, but to date studies have examined couplings within the cognitive system (e.g., Sliwinski, Hofer, & Hall, 2003). The focus on non-cognitive predictors, and prediction in persons with and without MCI, is a unique contribution of this study. Study Limitations Studies with multivariate, replicated, single-subject, repeated measures designs are rare, as are studies investigating intraindividual variability in cognition on recall

PAGE 121

109 memory tasks using daily variability as the temporal resolution. To date there have been few, if any, studies investigating cognitive intraindividual variability in individuals with mild cognitive impairment. The present study utilized a unique self-administered paradigm to allow for a number of assessment occasions over time, while not burdening cognitively impaired individuals unduly. However, several limitations of the study, outlined below, are directly related to these design strengths. One key limitation to this study is the sample. Specifically, the individuals who participated in this study were healthy, highly educated elders, many of whom expressed heightened concern over their memory function. The education level of this sample mirrors that of many recently published investigations of MCI. Challenges recruiting individuals with mild cognitive impairment, due to the difficulty identifying them prior to the neuropsychological intake assessment and interview, resulted in a decreased sample size for this group as well as a lack of ethnically diverse participants. The sample size may have resulted in reduced power, since, in some analyses trends were observed that did not reach significance. Additionally, the MCI group differed from the Non-MCI group in the proportion of males. A increased number of males in the MCI group might be a result of the study requirement for a research partner. Anecdotally, it appeared that a number of women concerned about their spouses cognition jointly volunteered themselves and their spouse for the study. In addition, demographically, a higher proportion of male older adults are partnered, therefore the recruitment requirement of an available research partner resulted in more males in both groups than might be typically found in a study with older adult participants.

PAGE 122

110 A greater proportion of those who attrited from the study fell in the MCI group. The differences between the study sample and the drop-out group revealed that those who discontinued the study were performing at a slightly lower level than the overall study sample, and more similarly to those in the final MCI group. This has the effect of reducing the generalizability of the study, since the sample is somewhat self-selected towards the less impaired individuals. Another limitation of the study is inherent to the protocol. The assessments were daily, self-administered workbooks. Although participants completed the cognitive assessments with a supervising research partner, it is still impossible to determine whether they adhered to time limits and no peeking rules. It is unknown whether they received coaching from an outside source, and whether they completed one workbook a day. However, evidence comparing the supervised laboratory session with the first at-home session suggests that at least initially, participants were consistent in their execution of the daily tasks. Anecdotally, commentary from participants after the conclusion of the daily assessments suggests that they were serious about completing the workbooks properly and expected that compliance would result in improved performance over time. Notably, an unexpected result of the self-administered workbooks was that the AVLT Delay Time (between last study and delay recall) varied more than expected between participants. During pre-testing of the workbooks, the delay time was found to be approximately 15 minutes, however, participants reported delay times of as short as five minutes and as long as 6 hours. Mixed effects modeling revealed that overall, there were no significant effects of delay time on AVLT Delay Recall performance. However, the range of delay times likely impact the number of ceiling responses, which in turn

PAGE 123

111 likely restricted the range of variability. This was more evident in the Non-MCI group, as a greater proportion of these participants reached ceiling on AVLT Delayed Recall. The workbooks themselves, although of necessity provided to participants in alternative forms, may have contributed variance to the day-to-day performance due to lack of completely parallel forms. While intercorrelations were high, and means across workbooks for each measure were similar, the RMANOVA results indicated that the forms were not parallel. Notably, this was due, for the most part, to the workbook used during the laboratory session. Likely the circumstances of that session contributed to the significantly difference performance as compared to that of the other workbooks (although, overall the laboratory session performance was not significantly different from the first at-home session). This workbook effect contributes to variability, however, it does so across all participants so it does not explain individual differences in the occasion-to-occasion variability, which was the focus of this study. Thus, covaring on workbook effects would not alter the individual differences analyses inherent to this study, with the exception that the intercorrelations between the IRIs might be increased. Inclusion of the workbook effects confounds method and trait variance, so a benefit of calculating the IRIs covaried for workbook effects would be the strengthening of the intercorrelations (by removing workbook-related fluctuations from the IRI calculation, and thereby disattenuating for measurement error), however, performing these analyses would hamper the comparability of this studys results with those in the existing literature. To date, intraindividual variability is often calculated as the within subjects standard deviation, without covaring for growth over time or for alternative forms.

PAGE 124

112 The effects of restricted range on the variables were addressed via analysis and removal, when necessary, of the floor and ceiling scores. However, the number of values near-ceiling, especially on the immediate and delayed memory tasks, particularly for the cognitively intact group indicated that a restriction of range likely impact the potential for intraindividual variability in performance over occasions. While attempting to provide tasks that were neither too difficult for the impaired participants nor too easy for the cognitively intact individuals, apparently the latter condition was not met. Future work with computer adapted tasks might aid in providing adaptive tasks for each ability level. Future Directions The findings from the current study have provided additional insight into the nature of cognitive intraindividual variability and the relationship of intraindividual variability and level of performance in cognitive intact and cognitively impaired individuals. They raise a number of new questions and ideas for future exploration. One direction for future work includes new or follow-up analyses for the current data. Investigating time lags in the coupled intraindividual variability analyses would allow comparisons of the impact of affect or discomfort on the following days cognitive performance, or vice versa. Further investigation of the relationship of reported affect and cognitive performance might provide insights regarding subtle effects of fluctuations in mood. Individuals with known mood disorders were excluded from the study, due to previous findings that depression and anxiety have significant negative impacts on cognitive functioning. The current finding of coupled daily fluctuations in mood and cognition might represent a more subtle aspect of this strong finding in the literature. As indicated earlier, recomputing the IRIs with workbook version covaried might be a useful follow-up analysis in order to confirm that IRI intercorrelations are strengthen

PAGE 125

113 once the method and trait variance are separated. Similarly, time of day effects and presence of spouse in the study might be similarly co-varied in order to select only trait variance. Finally, a subset of these participants also volunteered for a concurrent study of explicit memory function and correlates of hippocampal volumes (Leritz, 2004). Collaboration on recruitment between the investigators has provided a potential future opportunity for correlation of hippocampal volumes and intraindividual variability indices. Such an analysis might clarify the relationships between performance and variability in the heterogeneous MCI group, as the hippocampal volumes provide a key biological marker for memory decline. An alternative method to investigating the impact on a different MCI group definition would be to use the neuropsychological measures in a strict actuarial (quantitative, norm-based) manner, rather than the commonly accepted method of consensus conference determination. This alternative method might provide a differential, easily replicable, and quantifiable group distinction based on normative scores, which might clarify the findings regarding intraindividual variability and performance. The most informative expansion of the current work would be for a long-term follow-up. Since mild cognitive impairment is a fairly new designation, and current usage allows it to be applied to individuals who may convert to dementia as well as to individuals for whom the future holds no degenerative decline in cognitive functioning, it would be beneficial to follow participants to see if they decline. Outcome information would assist in clarifying the group differences and allow for more subtle understanding of the nuances of the intraindividual variability and performance interactions. Additionally, given that both groups experienced similar gains in level of performance

PAGE 126

114 over the course of the study, it would be valuable to see if this intensive learning experience has an impact on future impairment or functioning. For this, a control group would be required. Another useful follow-up to the current study would be to examine intraindividual variability in cognitive functioning in a more heterogeneous, representative sample. For example, examining cognitive intraindividual variability in a sample that included individuals with mild cognitive impairment from a lower educational background, with fewer cognitive reserves, might provide further evidence of the role of intraindividual variability in neurologic compromise. Notably, individuals with impairment beyond that of the mild cognitive impairment categorization would likely be unable to complete this study due to the ongoing cognitive demands and lack of rewarding feedback. An additional direction for future investigations would be to explore the impact of temporal resolution on intraindividual variability. Cognitive variability studies, such as this one, which focus on day-to-day variability, might be assessing a different aspect of brain and cognitive function than those which assess moment-to-moment variability via reaction time trials. Comparisons of the two would be a useful addition to the current field. An expansion to weekly or monthly temporal resolutions might also be of interest, however, if these studies include individuals with neurocognitive impairment, such resolutions would be confounded with degenerative decline in function over time. As mentioned briefly above, task difficulty had a significant impact on range of scores, as a number of participants reached ceiling performance on the immediate and delayed memory tasks. Computerized testing, either in the laboratory, via handheld computers, or on-line utilizing the internet, would allow for adaptive testing so that

PAGE 127

115 participants could continually improve without reaching ceiling. Additionally, for impaired individuals, adaptive testing allows discontinuation at a reasonable point without added frustration for the participant. Conclusion In summary, in this study older adults with amnestic MCI demonstrated a pattern of intraindividual variability and performance level that was consistent with patterns seen in previous studies of cognitively intact or demented individuals. Individuals with MCI demonstrated similar rates of practice-related gain over occasions as did the cognitively intact individuals. However, intraindividual variability usually was related inversely to performance. MCI status was not consistently related to intraindividual variability across the cognitive battery studied. Interrelationships of performance gain slopes with reductions of degree of fluctuation did not provide clear evidence for either of the two theories regarding the role of intraindividual variability in practice-related gain or neurocognitive vulnerability. These findings provide evidence that both theories of intraindividual variability might co-occur in the MCI population. Further investigations to clarify these relationships are warranted.

PAGE 128

APPENDIX SAMPLE DAILY WORKBOOK

PAGE 129

117 Word List 1 Set your timer for :30 minutes. Once you have set the timer for :30 minutes, start the timer and turn the page to study the list of words.

PAGE 130

118 Study the words below until you hear the timer alarm. baby ocean palace lip bar dress steam coin rock army building friend storm village cell When 1 1/2 minutes is over, and the timer sounds, please STOP STUDYING and turn to the next page.

PAGE 131

119 Now, set your timer for :30 minute. Once the timer is set, start the timer and go to the next page and write down as many words as you can remember in the spaces provided. DO NOT LOOK BACK AT THE LIST OF WORDS! DO NOT L OOK BACK AT THE LIST OF WORDS!

PAGE 132

120 __________________________ __________________________ __________________________ __________________________ __________________________ __________________________ __________________________ __________________________ __________________________ __________________________ __________________________ __________________________ __________________________ __________________________ __________________________ When you hear the timer alarm, do not write any more words. Turn to the next page.

PAGE 133

121 Word List 2 Set your timer for :30 minutes. Once you have set the timer for :30 minutes, start the timer and turn the page to study the list of words.

PAGE 134

122 Study the words below until you hear the timer alarm. baby ocean palace lip bar dress steam coin rock army building friend storm village cell When 1 1/2 minutes is over, and the timer sounds, please STOP STUDYING and turn to the next page.

PAGE 135

123 Now, set your timer for :30 minute. Once the timer is set, start the timer and go to the next page and write down as many words as you can remember in the spaces provided. DO NOT LOOK BACK AT THE LIST OF WORDS!

PAGE 136

124 DO NOT L OOK BACK AT THE LIST OF WORDS! __________________________ __________________________ __________________________ __________________________ __________________________ __________________________ __________________________ __________________________ __________________________ __________________________ __________________________ __________________________ __________________________ __________________________ __________________________ When you hear the timer alarm, do not write any more words. Turn to the next page.

PAGE 137

125 Word List 3 Set your timer for :30 minutes. Once you have set the timer for :30 minutes, start the timer and turn the page to study the list of words.

PAGE 138

126 Study the words below until you hear the timer alarm. baby ocean palace lip bar dress steam coin rock army building friend storm village cell When 1 1/2 minutes is over, and the timer sounds, please STOP STUDYING and turn to the next page.

PAGE 139

127 Now, set your timer for :30 minute. Once the timer is set, start the timer and go to the next page and write down as many words as you can remember in the spaces provided. DO NOT LOOK BACK AT THE LIST OF WORDS!

PAGE 140

128 DO NOT L OOK BACK AT THE LIST OF WORDS! __________________________ __________________________ __________________________ __________________________ __________________________ __________________________ __________________________ __________________________ __________________________ __________________________ __________________________ __________________________ __________________________ __________________________ __________________________ When you hear the timer alarm, do not write any more words. Turn to the next page.

PAGE 141

129 Before you continue, please record the time: Time: ____ ____ : ____ ____ circle: am pm

PAGE 142

130 Directions: Please ask your research partner to read the DIGIT SPAN Forward and DIGIT SPAN Backward Test to you now. Once you have completed the DIGIT SPAN Tests, please return to this booklet and turn the page.

PAGE 143

131 Symbol Digit On the next page are boxes to be filled in with the matching digit, like you see in the key on the top of the page. Set the timer for :30 minutes. Keep working until you hear the alarm. Go in order, without skipping any boxes. Start the timer when you are ready to begin.

PAGE 144

132 KEY 1 2 3 4 5 6 7 8 9 4 1 3 9 2 5 9 6 3 1

PAGE 145

133 Number Copy On the next page are boxes to be filled in with the same digit you see in the top of the box. Set the timer for :30 minutes. Keep working until you hear the alarm. Go in order, without skipping any boxes. Start the timer when you are ready to begin. If you finish before the timer sounds the alarm, please stop the timer and record the time left, just as it appears on the timer.

PAGE 146

134 4 1 3 9 2 5 9 6 3 1 4 7 5 2 8 4 1 3 9 2 5 9 6 3 1 5 7 1 5 4 6 7 3 5 9 1 2 8 2 3 8 9 4 7 8 1 3 4 6 9 2 3 5 7 1 5 8 1 2 9 5 4 7 1 8 7 2 1 5 4 4 3 7 8 4 1 7 2 9 7 6 3 1 8 6 6 7 9 3 1 4 9 2 5 4 7 3 2 5 7 3 5 4 6 9 4 5 8 3 6 5 1 9 2 1 If you finish all the boxes before the time ends, write the remaining time here: __ __:__ __

PAGE 147

135 SLEEP DIARY Please answer the following questions for YESTERDAY Provide the information to describe your sleep yesterday and last night. Yesterdays day: __________ _____________ Yesterdays date: __________ _____________ 1. If you napped yesterday, how long did you nap, in minutes? 2. What time did you enter bed for the purpose of sleeping last night? 3. Counting from the time you wished to fall asleep, how many minutes did it take you to fall asleep? 4. How many times did you awaken during the night? 5. What is the total number of minutes you were awake during the middle of the night once you fell asleep? This does not include the time it took to fall asleep at the beginning of the night, or the time you spent awake in bed before getting out of bed in the morning. 6. What time did you wake up for the last time this morning? 7. What time did you actually get out of bed this morning? 8. Pick ONE number to indicate your overall QUALITY RATING or satisfaction with your sleep. [1 = very poor; 2 = poor; 3 = fair; 4 = good; 5 = excellent] 9. List any sleep medication or alcohol taken at or near bedtime, and give the amount and time taken. 1. NAP (mins): ____________ 2. BEDTIME: ____________ 3. TIME TO FALL ASLEEP (mins): ____________ 4. AWAKENINGS: ____________ 5. WAKE TIME (middle of night): ____________ 6. FINAL WAKE-UP: ____________ 7. OUT OF BED: ____________ 8. QUALITY RATING: _____________ 9. BEDTIME MEDICATION (amount and time: ________________

PAGE 148

136 (1) Number of interrupt ions while completing this workbook: ___________ (2) Rate how tired you feel right now: 1 2 3 4 5 6 7 Not Tired Extremely Tired (3) On a scale of zero to ten, where zero means no pain and ten equals the worst possible pain, what is your current pain level ? |----------|----------|-----------|----------|----------|----------| ----------|----------|---------|---------| 0 1 2 3 4 5 6 7 8 9 10 No pain Mild Moderate Severe Worst Possible Pain (3a) If you are experiencing pain, please circle the l ocation of your pain by circling all the bodily locations that apply: Neck Elbows Hips Feet/Toes Back Wrists Knees Shoulders Ankles Head Hand/Fingers Other: ___________ (4) Were other people (other than your research partner) around while you were doing the workbook? Yes No (Circle One) If you circled YES please answer questions 4a and 4b. (4a) Can you see them? Ye s No (Circle One) (4b) Can you hear them? Ye s No (Circle One) (5) In general how noisy was it while you did the workbook? 1 2 3 4 5 6 7 Completely Quiet Extremely Noisy (6) Please rate how good the lighting wa s while you were do ing the workbook? 1 2 3 4 5 6 7 Excellent Extremely Poor

PAGE 149

137 PANAS Please indicate the extent to which you have experienced the following emotions in the last 24 hours by checking the appropriate box. interested [ ] Not at all [ ] A Little [ ] Moderately [ ] Quite a bit [ ] Very much distressed [ ] Not at all [ ] A L ittle [ ] Moderate ly [ ] Quite a b it [ ] Very much excited [ ] Not at all [ ] A L ittle [ ] Moderate ly [ ] Quite a b it [ ] Very much upset [ ] Not at all [ ] A L ittle [ ] Moderate ly [ ] Quite a b it [ ] Very much strong [ ] Not at all [ ] A L ittle [ ] Moderate ly [ ] Quite a b it [ ] Very much guilty [ ] Not at all [ ] A L ittle [ ] Moderate ly [ ] Quite a b it [ ] Very much scared [ ] Not at all [ ] A L ittle [ ] Moderate ly [ ] Quite a b it [ ] Very much hostile [ ] Not at all [ ] A L ittle [ ] Moderate ly [ ] Quite a b it [ ] Very much enthusiastic [ ] Not at all [ ] A L ittle [ ] Moderate ly [ ] Quite a b it [ ] Very much When you have completed this page, please proceed to the next page.

PAGE 150

138 proud [ ] Not at all [ ] A Little [ ] Moderate ly [ ] Quite a b it [ ] Very much irritable [ ] Not at all [ ] A L ittle [ ] Moderate ly [ ] Quite a b it [ ] Very much alert [ ] Not at all [ ] A L ittle [ ] Moderate ly [ ] Quite a b it [ ] Very much ashamed [ ] Not at all [ ] A L ittle [ ] Moderate ly [ ] Quite a b it [ ] Very much inspired [ ] Not at all [ ] A L ittle [ ] Moderate ly [ ] Quite a b it [ ] Very much nervous [ ] Not at all [ ] A L ittle [ ] Moderate ly [ ] Quite a b it [ ] Very much determined [ ] Not at all [ ] A L ittle [ ] Moderate ly [ ] Quite a b it [ ] Very much attentive [ ] Not at all [ ] A L ittle [ ] Moderate ly [ ] Quite a b it [ ] Very much jittery [ ] Not at all [ ] A L ittle [ ] Moderate ly [ ] Quite a b it [ ] Very much active [ ] Not at all [ ] A L ittle [ ] Moderate ly [ ] Quite a b it [ ] Very much afraid [ ] Not at all [ ] A L ittle [ ] Moderate ly [ ] Quite a b it [ ] Very much When you have completed this page, please proceed to the next page.

PAGE 151

139 Set your timer for :00 minutes. Once the timer is set, start the timer and go to the next page and write down as many words as you can remember from the list you studied earlier. PLEASE DO NOT LOOK BACK AT THE LIST. Before you continue, please record the time: Time: ____ ____ : _____ ____ circle: am pm

PAGE 152

140 DO NOT LOOK BACK AT THE LIST OF WORDS! _______________________ _______________________ _______________________ _______________________ _______________________ _______________________ _______________________ _______________________ _______________________ _______________________ _______________________ _______________________ _______________________ _______________________ _______________________ When you hear the timer alarm, do not write any more words. Please turn over to the last page.

PAGE 153

141 How stressed are you feeling? Please circle one number below to describe how you feel now. (1 = not at all stressed, or feeling "very well" 5 = feeling very stressed or panicked, or "very bad") Thank you! You are done for today!

PAGE 154

REFERENCES Allaire, J. C. (2001). Performance gains and fluctuations: The effects of cognitive practice and the rate of intraindividual cognitive variability in older adults. Doctoral Dissertation, Wayne State University. Allaire, J. C., & Marsiske, M. (2004). Intraindividual Variability may not always indicate Vulnerability in Elders' Cognitive Performance. Manuscript under review (Psychology and Aging) Anderson, N. D., & Craik, F. I. M. (2000). Memory in the aging brain. In E. Tulving & F. I. M. Craik (Eds.) The Oxford Handbook of Memory (p. 411-425.) New York: Oxford University Press. Anstey, K. J. (1999) Sensory motor variables and forced expiratory volume as correlates of speed, accuracy, and variability in reaction time performance in late adulthood. Aging, Neuropsychology, and Cognition, 6, 84-95. Bckman, L., Small, B. J., Wahlin, ., & Larsson, M. (2000). Cognitive functioning in very old age. In F. I. M. Craik & T. A. Salthouse, (Eds.) The Handbook of Aging and Cognition (2nd Ed.). (p. 499-558.) Mahwah, NJ: Lawrence Erlbaum. Balota, D. A., Dolan, P. O., & Duchek, J. M. (2000). Memory changes in healthy older adults. In E. Tulving & F. I. M. Craik (Eds.) The Oxford Handbook of Memory (p. 395 409.) New York: Oxford University Press. Benedict, R. H. B. (1997). Brief Visuospatial Memory Test Revised Odessa FL: Psychological Assessment Resources, Inc. Benton, A., & Hamsher, K. (1989). Multilingual Aphasia Examination Iowa City: AJA Associates. Bezeau, S., & Graves, R. (2001). Statistical power and effect sizes of clinical neuropsychological research. Journal of Clinical and Experimental Neuropsychology, 23 399-406. Blair, J. R., & Spreen, O. (1989). Predicting pre-morbid IQ: A revision of the National Adult Reading Test. The Clinical Neuropsychologist, 3 129-136. Bleibeg, J., Garmoe, W. S., Halpern, E. L., Reeves, D. L., & Nadler, J. D. (1997). Consistency of within-day and across-day performance after mild brain injury. Neuropsychiatry, Neuropsychology, and Behavioral Neurology, 10 247-253. 142

PAGE 155

143 Blessed, G., Tomlinson, B. E., & Roth, M. (1968). Quantitative measures of dementia and senile change. British Journal of Psychiatry, 114 797-811. Bliwise, D. L., King, A. C., Harris, R. B., & Haskell, W. L. (1992). Prevalence of self-reported poor sleep in a healthy population aged 50-65. Social Science and Medicine, 34 49-55. Bozoki, A., Giordani, B., Heidebrink, J. L., Berent, S., & Foster, N. L. (2001). Mild Cognitive Impairments predict dementia in nondemented elderly patients with memory loss. Archives of Neurology, 58 411-416. Brandt, J., Spencer, M., & Folstein, M. (1988). The Telephone Interview for Cognitive Status. Neuropsychiatry, Neuropsychology, and Behavioral Neurology, 2 111117. Brandt, J., & Benedict, R. H. B. (2001). Hopkins Verbal Learning Test Revised: Professional Manual Odessa, FL: Psychological Assessment Resources. Britton, J. (1997). Measurement of peak flow variability in community populations: Methodology. European Journal of Respiration, 10 42-44. Bruhn, P., & Parsons, O. A. (1977). Reaction time variability in epileptic and brain damaged patients. Cortex, 13 373-384. Butler, A. C., Hokanson, J. E., & Flynn, H. A. (1994). A comparison of self-esteem lability and low trait self-esteem as vulnerability factors for depression. Journal of Personality and Social Psychology, 66 166-177. Carstensen, L.L., Pasupathi, M., Mayr, U., & Nesselroade, J. R. (2000). Emotional experience in everyday life across the adult life span. Journal of Personality and Social Psychology, 79 644-655. Celsis, P. (2000). Age-related cognitive decline, mild cognitive impairment or preclinical Alzheimers disease? Annals of Medicine, 32 6-14. Chen, P., Ratcliff, G., Belle, S. H., Cauley, J. A., DeKosky, S. T., & Ganguli, M. (2000). Cognitive tests that best discriminate between presymptomatic AD and those who remain nondemented. Neurology, 55 1847-1853. Cohen, J. (1988). Statistical Power Analysis for the Behavioral Sciences (2nd Ed.). Hillsdale, NJ: Lawrence Erlbaum. Craik, F. I. M., & Salthouse, T. A. (2000). The Handbook of Aging and Cognition (2nd Ed.). Mahwah: Lawrence Erlbaum. Crook, T. H., Feher, E. P., & Larrabee, G. J. (1992). Assessment of memory complaints in age-associated memory impairment: The MAC-Q. International Psychogeriatrics, 4 165-176.

PAGE 156

144 Crook, T. H., Bartus, R. T., Ferris, S. H., Whitehouse, P., Cohen., G. D., & Gershon, S. (1986). Age-Associated Memory Impairment: Proposed diagnostic criteria and measures of clinical change. Report of a National Institute of Mental Health work group. Developmental Neuropsychology, 2 261-276. Daly, E., Zaitchik, D., Copeland, M., Schmahmann, J., Gunther, J., & Albert, M. (2000). Predicting conversion to Alzheimers disease using standardized clinical information. Archives of Neurology, 57 675-680. Domenech, A.L., Marsiske, M., McCoy, K.J.M., Aiken, A.T., & Cook, S.E. (2004). Primacy and recency effects in memory after extensive practice. Poster presentation at the Cognitive Aging Conference, Atlanta, GA. Eizenman, D. R., Nessleroade, J. R., Featherman, D. L., & Rowe, J. W. (1997). Intraindividual variability in perceived control in an older sample: The MacArthur Successful Aging Studies. Psychology and Aging, 12 489-502. Elias, M. F., Beiser, A., Wolf, P. A., Au, R., White, R. F., & DAgostino, R. B. (2000). The preclinical phase of Alzheimer Disease: A 22-year prospective study of the Framingham cohort. Archives of Neurology, 57 808-813. Flicker, C., Ferris, S., & Reisberg, B. (1991). Mild cognitive impairment in the elderly: Predictors of dementia. Neurology, 41 1006-1009. Fluckiger, L., Boivin, J. M., Quillot, D., Jeandel, C., & Zannad, F. (1999). Differential effects of aging on heart rate variability and blood pressure variability. Journal of Gerontology: Biological Sciences, 54A B219-B224. Folstein, M. F., Folstein, S. E. & McHugh, P. R. (1975). Mini Mental State: A practical method for grading the cognitive states of patients for the clinician. Journal of Psychiatric Research, 12 189-198. Fozard, J. L., Vercruyssen, M., Reynolds, S. L., & Hancock, P. A. (1994). Age differences and changes in reaction time: The Baltimore Longitudinal Study of Aging. Journal of Gerontology, 49 179-189. Fuentes, K., Hunter, M. A., Strauss, E., & Hultsch, D. F. (2001). Intraindividual variability in cognitive performance in persons with Chronic Fatigue Syndrome. The Clinical Neuropsychologist, 15 210-227. Ganguli, M., Reynolds, C. F., & Gilby, J. E. (1996) Prevalence and persistence of sleep complaints in a rural older community sample: the MoVIES project. Journal of the American Geriatrics Society, 44 778-84. Guimares, R. M., & Isaacs, B. (1980). Characteristics of the gait in older people who fall. International Rehabilitation Medicine, 2 177-280.

PAGE 157

145 Hausdorff, J. M., Edelberg, H. K., Mitchell, S. L., Goldberger, A. L., & Wei, J. Y. (1997). Increased gait unsteadiness in community-dwelling elderly. Archives of Physical Medicine Rehabilitation, 78 278-283. Hendrickson, A. E. (1982). The biological basis of intelligence. Part I: Theory. In H. J. Eysenck (Ed.), A Model for Intelligence. (pp. 151-196). Berlin: Springer-Verlag. Hertzog, C., Dixon, R. A., & Hultsch, D. F. (1992). Intraindividual change in text recall of the elderly. Brain and Language, 42 248-269. Hetherington, C. R., Stuss, D.T., & Finlayson, M. A. J. (1996). Reaction time and variability 5 and 10 years after traumatic brain injury. Brain Injury, 10 473-486. Higgins, B. F. (1979). Peak expiratory flow variability in the general population. European Journal of Respiration, 10 45-48. Hughes, C. P., Berg., L., & Danzinger, W. L. (1982). A new scale for staging of dementia. British Journal of Psychiatry, 140 526-535. Hultsch, D. F., MacDonald, S. W. S., Hunter, M. A., Levy-Bencheton, J., & Strauss, E. (2000). Intraindividual variability in cognitive performance in older adults: Comparison of adults with mild dementia, adults with arthritis, and healthy adults. Neuropsychology, 14 588-598. Hultsch, D. F., MacDonald, S. W. S., & Dixon, R. A. (2002). Variability in reaction time performance of younger and older adults. Journal of Gerontology: Psychological Sciences, 57B P101-P115. Ivnik, R. J., Malec, J. F., & Smith, G. E. (1996). Neuropsychological test norms above age 55 COWAT, MAE Token, WRAT-R Reading, AMNART, Stroop, TMT, and JLO. The Clinical Neuropsychologist, 10 262-378. Jones, C.J., & Nesselroade, R.J. (1990). Multivariate, replicated, single-subject designs and P-technique factor analysis: A selective review of intraindividual change studies. Experimental Aging Research, 16 171-183. Jorm. A. F., Christensen, H., Henderson, A. S., Korten, A. E., Mackinnon, A. J., & Scott, R. (1994). Complaints of cognitive decline in the elderly: A comparison of reports by subjects and informants in a community survey. Psychological Medicine, 34 365-374. Jorm, A. F., Christensen, H., Korten, A. E., Henderson, A. S., Jacomb, P. A., & Mackinnon, A. (1997). Do cognitive complaints either predict future cognitive decline or reflect past cognitive decline? A longitudinal study of an elderly community sample. Psychological Medicine, 27 91-98. Kaplan, E.F., Goodglass, H., & Weintraub, S. (2001). The Boston Naming Test (2nd Ed.) Baltimore: Lippincott.

PAGE 158

146 Kaplan, H. I. & Sadock, B. J. (1998). Synopsis of Psychiatry (8th Ed.). Baltimore: Williams & Williams. Kernis, M. H., Cornell, D. P., Sun, C. R., Berry, A., & Harlow, T. (1993). Theres more to self-esteem than whether it is high or low: The importance of stability of self-esteem. Journal of Personality and Social Psychology, 65 1190-1204. Knopman, D. S., DeKosky, S. T., Cummings, J. L., Chui, H., Corey-Bloom, J., Relkin, N., Small, G. W., Miller, B., & Stevens, J. C. (2001) Practice parameter: Diagnosis of dementia. Neurology, 56 1143-1153. Lang, F. R., Featherman, D. L., & Nesselroade, J. R. (1997). Social self-efficacy and short-term variability in social relationships: The MacArthur Successful Aging Studies. Psychology and Aging, 12 657-666. Larrabee, G. J. (1996). Age-Associated Memory Impairment: Definition and psychometric characteristics. Aging, Neuropsychology, and Cognition, 3, 118-131. Li, S. C., Aggen, S. H., Nesselroade, J. R., & Baltes, P. B. (2001). Short-term fluctuations in elderly peoples sensorimotor functioning predict text and spatial memory performance. Gerontology, 47 100-116. Li, S. C., Lindenberger, U. (1999). Cross-level unification: A computational exploration of the link between deterioration of neurotransmitter systems and dedifferentiation of cognitive abilities in old age. In L.-G. Nilsson (Ed). Cognitive Neuroscience of Memory (pp. 103-146). Seattle: Hogrefe & Huber. Lichstein, K. L., Riedel, B. W., & Means, M. K. (1999) Psychological treatment of late-life insomnia. In: Schulz, R., Maddox, G., Lawton, M. P. (Eds). Annual Review of Gerontology and Geriatrics: Vol. 18. Focus on Interventions Research with Older Adults (pp. 74-110). New York: Springer. Lishman W. A. (1988). Organic Psychiatry: The Psychological Consequences of Cerebral Disorder, 3rd Ed. Oxford: Blackwell Science. MacDonald, S.W.S., Hultsch, D.F., & Dixon, R.A. (2003). Performance variability is related to change in cognition: Evidence from the Victoria Longitudinal Study. Psychology and Aging, 18 510-523. May, C. P., Hasher, L., & Stoltzfus, E. R. (1993). Optimal time of day and the magnitude of age differences in memory. Psychological Science 4, 326-330. McCrae, C. S., Wilson, N. M., Lichstein, K. L., Durrence, H. H., Taylor, D. J., Bush, A. J., & Riedel, B. W. (in press). Young old and Old old poor sleepers with and without insomnia complaints. Journal of Psychosomatic Research. Mesulam, M-M. (2000). Principles of Behavioral and Cognitive Neurology2 nd Ed New York: Oxford University Press.

PAGE 159

147 Morris, J. C. (1993). The Clinical Dementia Rating (CDR): Current version and scoring rules. Neurology, 43 2412-2414. Morris, J. C., Storandt., M., Miller, J.P., McKeel., D. W., Price, J. L., Rubin., E. H., & Berg, L. (2001). Mild Cognitive Impairment represents early-stage Alzheimers Disease. Archives of Neurology, 58 397-405. Murtha, S., Cismaru, R., Waechter, R., & Chertkow, H. (2002). Increased variability accompanied frontal lob damage in dementia. Journal of the International Neuropsychological Society, 8 360-372. Nesselroade, J. R. (1991). Interindividual differences in intraindividual change. In J. L. Horn & L. M. Collins (Eds.) Best Methods for the Analysis of Change: Recent Advances, Unanswered Questions, Future Directions (pp. 92-107). Washington DC: American Psychological Association. Nesselroade, J. R., Featherman, D. L., Aggen, S. H., & Rowe, J. W. (1996). Short-term variability in physical performance and physiological attributes of older adults. Unpublished manuscript: University of Virginia. Nesselroade, J.R., & Ford, D.H. (1985). P-technique comes of age: Multivariate, replicated, single subject designs for research on older adults. Research on Aging, 7 46-80. Pagani, M. (1999). Cardiovascular oscillations: From a physiological curiosity to a probe of aging and health? Journal of Gerontology: Biological Sciences, 54A B225-B226. Petersen, R. C., Doody, R., Kurz, A., Mohs, R. C., Morris, J. C., Rabins, P.V., Ritchie, K., Rossor, M., Thal, L., & Wingblad, B. (2001). Current concepts in Mild Cognitive Impairment. Archives of Neurology, 58 1985-1992. Peterson, R. C., Smith, G., Kokmen, E., Ivnik, R. J., & Tangalos, E. G. (1992). Memory function in normal aging. Neurology, 42 396-401. Petersen, R. C., Smith, G. E., Waring, S. C., Ivnik, R. J., Tangalos, E. G., & Kokmen, E. (1999). Mild Cognitive Impairment: Clinical characterization and outcome. Archives of Neurology, 56 303-308. Rabbitt, P., Osman, P., & Moore, B. (2001). There are stable individual differences in performance variability, both from moment to moment and day to day. Quarterly Journal of Experimental Psychology 54A 981-1003. Radloff, L.S. (1977). The CES-D Scale: A self report depression scale for research in the general population. Applied Psychological Measurement, 1, 385-401. Reitan, R. M. (1992). Trail Making Test: Manual for Administration and Scoring Tucson: Reitan Neuropsychology Lab.

PAGE 160

148 Rey, A. (1941). Lexamen psychologique dans les cas dencephalopathie tramatique. Archives de Psychologie, 28, 215285. Rey, A. (1964). Lexamen clinique en psychologie Paris: Presses universitaires de France. Salthouse, T. A. (1993). Attentional blocks are not responsible for age-related slowing. Journal of Gerontology: Psychological Sciences, 48 263-270. Schaie, K. W. & Willis, S. L. (1991). Adult personality and psychomotor performance: Cross-sectional and longitudinal analyses. Journals of Gerontology, 46, P275-P284. Schmidt, M. (1996). Rey Auditory Verbal Learning Test Los Angeles: Western Psychological Services. Shammi, P., Bosman, E., & Stuss, D. T. (1998) Aging and variability in performance. Aging, Neuropsychology, and Cognition, 5 1-13. Shifren, K., Hooker. K, Wood, P., & Nesselroade, J. (1997). Structure and variation of mood in individuals with Parkinsons Disease: A dynamic factor analysis. Psychology and Aging, 12, 328-339. Siegler, R. S. (1994). Cognitive variability: A key to understanding cognitive development. Current Directions in Psychological Science, 3 1-5. Sliwinski, M. J., Hofer, S. M., & Hall, C.(2003). Correlated and coupled cognitive change in older adults with and without preclinical dementia. Psychology and Aging, 18, 672-683. Smith, A. (1982). Symbol Digit Modalities Test Revised: Manual Los Angeles: Western Psychological Services. Strauss, E., Slick, D. J., Levy-Bencheton, J., Hunter, M., MacDonald, S.W.S., & Hultsch, D. F. (2002). Intraindividual variability as an indicator of malingering in head injury. Archives of Clinical Neuropsychology, 17 423-444. Stuss, D. T., Pogue, J., Buckle, L, & Bondar, J. (1994). Characterization of stability of performance in patients with traumatic brain injury: Variability and consistency on reaction time tests. Neuropsychology, 8 316-324. Tierney, M. C., Szalai, J. P., Snow, W. C., Fisher, R. H., Nores, A., Nadon, G., Dunn, E., & St. George Hyslop, P. H. (1996). Prediction of probable Alzheimers disease in memory-impaired patients: A prospective longitudinal study. Neurology, 46 661-665.

PAGE 161

149 Turvey, C. L., Schultz, S., Arndt, S., Wallace, R. B., & Herzog, R. (2000). Memory complaint in a community sample aged 70 and older. Journal of the American Geriatrics Society, 48 1435-1441. Wahlin, ., Bckman, L., & Winblad, B. (1995). Free recall and recognition of slowly and rapidly presented words in very old age: A community-based study. Experimental Aging Research, 21 251-271. Walker, M.P., Ayre, G. A., Cummings, J. L., Wesnes, K., McKeith, I.G., OBrien, J.T., & Ballard, C.G. (2000). Quantifying fluctuation in dementia with Lewy bodies, Alzheimers disease, and vascular dementia. Neurology, 54, 1616-1624. Watson, D., Clark, L.A., & Tellegen, A. (1988). Development and validation of brief measures of positive and negative affect: The PANAS scales. Journal of Personality and Social Psychology, 54 1063-1070. Wechsler, D. (1997). Wechsler Memory Scale Third Edition: Manual San Antonio: The Psychological Corporation. Wilkinson, R. T., & Allison, S. (1989). Age and simple reaction time: Decade differences for 5,325 subjects. Journal of Gerontology, 44 29-35. Wilson, B., Cockburn, J., & Baddeley, A. (1985). The Rivermead Behavioural Memory Test Bury St. Edmunds, England: Thames Valley Test Co. Woodruff-Pak, D. S. (1997). The Neuropsychology of Aging Malden, MA: Blackwell. Yesavage, J.A. (1983). Development and validation of a Geriatric Depression Screening Scale: A preliminary report. Journal of Psychiatric Research, 17 37-49.

PAGE 162

BIOGRAPHICAL SKETCH Karin J. M. McCoy was born in Albany, New York. She earned her Bachelor of Arts degree with a concentration in psychology from Cornell University in 1989. A Master of Science degree in clinical psychology was awarded to Ms. McCoy from the University of Florida in 1999. Her research interests include cognitive functioning in older adults and cognitive correlates of medical and forensic decisional capacity. 150


Permanent Link: http://ufdc.ufl.edu/UFE0008421/00001

Material Information

Title: Understanding the Transition from Normal Cognitive Aging to Mild Cognitive Impairment: Comparing the Intraindividual Variability in Cognitive Function
Physical Description: xii, 150 p.
Language: English
Creator: Mccoy, Karin Johanna M. ( Dissertant )
Michael Marsiskel. ( Transcriber )
Russell M. Bauer. ( Thesis advisor )
Abrams, Lise ( Reviewer )
Bowers, Dawn ( Reviewer )
Dede, Duane ( Reviewer )
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2004
Copyright Date: 2004

Subjects

Subjects / Keywords: Department of Clinical and Health Psychology thesis, Ph.D   ( local )
Aging   ( lcsh )
Research   ( lcsh )
Dissertations, Academic -- UF -- College of Public Health and Health Professions -- Department of Clinical and Health Psychology   ( local )
Cognition -- Physiology   ( lcsh )
Dementia -- Etiology   ( lcsh )
Dementia -- Pathophysiology   ( lcsh )

Notes

Abstract: Understanding the Transition from Normal Cognitive Aging to Mild Cognitive Impairment: Comparing the Intraindividual Variability in Cognitive Function. Intraindividual variability describes fluctuation or transient change in performance, and can be measured by repeated assessment of an ability or trait over a short period of time. Intraindividual variability in biological systems has been demonstrated to indicate systemic compromise (e.g., loss of homeostatic regulation). Consequently, one theory investigators have begun to research is whether intraindividual variability or fluctuation in cognitive performance may be an indicator of cognitive decline. Additionally, a second theory suggests that fluctuations in cognitive performance may be greater during periods of learning acquisition, with a corresponding reduction in variability following the acquisition phase. This study investigated intraindividual variability of cognition on measures of attention, processing speed, working memory, and episodic memory in older adults (over 65 years of age) with and without mild cognitive impairment (MCI), by assessing performance in these domains daily for 31 days. MCI may be a transitional stage between normal aging and dementia. In particular, the term amnestic MCI describes individuals with focal cognitive impairment in the memory domain; such impairment might foretell future Alzheimer?s dementia. For this study, MCI is defined as list memory performance 1.5 standard deviations below age-appropriate norms, supplemented by subjective memory complaints and informant report of memory problems, as well as intact cognition in non-memory domains. Results revealed that older adults with amnestic MCI demonstrated a pattern of intraindividual variability and performance level that was consistent with patterns seen in previous studies of cognitively intact or demented individuals. Individuals with MCI demonstrated similar rates of practice-related gain over occasions as did the cognitively intact individuals. However, intraindividual variability was related inversely to performance, for most measures. MCI status was not consistently related to intraindividual variability across the cognitive battery studied. Interrelationships of performance gain slopes with degree of fluctuation did not provide clear evidence for either of the two theories regarding the role of intraindividual variability in practice-related gain or neurocognitive vulnerability. These findings provided evidence that the two types of intraindividual variability described by the current theories in the literature may co-occur in the MCI population.
Abstract: Cognition, intraindividual, mci, variability
General Note: Title from title page of source document.
General Note: Document formatted into pages; contains 162 pages.
General Note: Includes vita.
Thesis: Thesis (Ph.D.)--University of Florida, 2004.
Bibliography: Includes bibliographical references.
General Note: Text (Electronic thesis) in PDF format.

Record Information

Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
System ID: UFE0008421:00001

Permanent Link: http://ufdc.ufl.edu/UFE0008421/00001

Material Information

Title: Understanding the Transition from Normal Cognitive Aging to Mild Cognitive Impairment: Comparing the Intraindividual Variability in Cognitive Function
Physical Description: xii, 150 p.
Language: English
Creator: Mccoy, Karin Johanna M. ( Dissertant )
Michael Marsiskel. ( Transcriber )
Russell M. Bauer. ( Thesis advisor )
Abrams, Lise ( Reviewer )
Bowers, Dawn ( Reviewer )
Dede, Duane ( Reviewer )
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2004
Copyright Date: 2004

Subjects

Subjects / Keywords: Department of Clinical and Health Psychology thesis, Ph.D   ( local )
Aging   ( lcsh )
Research   ( lcsh )
Dissertations, Academic -- UF -- College of Public Health and Health Professions -- Department of Clinical and Health Psychology   ( local )
Cognition -- Physiology   ( lcsh )
Dementia -- Etiology   ( lcsh )
Dementia -- Pathophysiology   ( lcsh )

Notes

Abstract: Understanding the Transition from Normal Cognitive Aging to Mild Cognitive Impairment: Comparing the Intraindividual Variability in Cognitive Function. Intraindividual variability describes fluctuation or transient change in performance, and can be measured by repeated assessment of an ability or trait over a short period of time. Intraindividual variability in biological systems has been demonstrated to indicate systemic compromise (e.g., loss of homeostatic regulation). Consequently, one theory investigators have begun to research is whether intraindividual variability or fluctuation in cognitive performance may be an indicator of cognitive decline. Additionally, a second theory suggests that fluctuations in cognitive performance may be greater during periods of learning acquisition, with a corresponding reduction in variability following the acquisition phase. This study investigated intraindividual variability of cognition on measures of attention, processing speed, working memory, and episodic memory in older adults (over 65 years of age) with and without mild cognitive impairment (MCI), by assessing performance in these domains daily for 31 days. MCI may be a transitional stage between normal aging and dementia. In particular, the term amnestic MCI describes individuals with focal cognitive impairment in the memory domain; such impairment might foretell future Alzheimer?s dementia. For this study, MCI is defined as list memory performance 1.5 standard deviations below age-appropriate norms, supplemented by subjective memory complaints and informant report of memory problems, as well as intact cognition in non-memory domains. Results revealed that older adults with amnestic MCI demonstrated a pattern of intraindividual variability and performance level that was consistent with patterns seen in previous studies of cognitively intact or demented individuals. Individuals with MCI demonstrated similar rates of practice-related gain over occasions as did the cognitively intact individuals. However, intraindividual variability was related inversely to performance, for most measures. MCI status was not consistently related to intraindividual variability across the cognitive battery studied. Interrelationships of performance gain slopes with degree of fluctuation did not provide clear evidence for either of the two theories regarding the role of intraindividual variability in practice-related gain or neurocognitive vulnerability. These findings provided evidence that the two types of intraindividual variability described by the current theories in the literature may co-occur in the MCI population.
Abstract: Cognition, intraindividual, mci, variability
General Note: Title from title page of source document.
General Note: Document formatted into pages; contains 162 pages.
General Note: Includes vita.
Thesis: Thesis (Ph.D.)--University of Florida, 2004.
Bibliography: Includes bibliographical references.
General Note: Text (Electronic thesis) in PDF format.

Record Information

Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
System ID: UFE0008421:00001


This item has the following downloads:


Full Text












UNDERSTANDING THE TRANSITION FROM
NORMAL COGNITIVE AGING TO MILD COGNITIVE IMPAIRMENT:
COMPARING THE INTRAINDIVIDUAL VARIABILITY
IN COGNITIVE FUNCTION















By

KARIN J. M. MCCOY


A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA


2004

































Copyright 2004

by

Karin J. M. McCoy

































This dissertation is dedicated to my fiance, Joseph Barker. His love, generosity, and
goodness of spirit supported me through all the hurdles of graduate school.















ACKNOWLEDGMENTS

I would like to express my deepest appreciation to my dissertation chair, Michael

Marsiske. He has patiently and thoughtfully provided prodding and direction throughout

the inception, production, and presentation of this work, as well as giving graciously and

generously of his time and intellect in his role as a mentor. Special recognition goes to

Russell Bauer, since, in addition to his role as my cochair, he has been my mentor in

neuropsychology throughout my graduate career. In conjunction with my chair and

cochair, my committee members, Lise Abrams, Dawn Bowers, and Duane Dede, have

provided me with guidance, support, and friendship. Each has contributed greatly to my

development as a scholar. I would like to thank the FITMIND team, Sarah Cook,

Adrienne Aiken, Amber Domenech, and Jaclyn Pittman, for their tireless assistance with

data collection and entry. The FITMIND participants deserve special thanks for their

commitment to this project. After all the phone calls and hours of assessment, most

returned to our laboratory to provide feedback on their experiences. My work, and the

FITMIND project itself, was supported by the University of Florida Provost Fellowship

in Aging Research and the America Psychological Association Division 20/ Retirement

Research Foundation Research Proposal Award. Finally, my parents, Guy and Maja

McCoy, deserve all my gratitude, love, and appreciation for their unwavering support on

this journey.
















TABLE OF CONTENTS
Page

A C K N O W L E D G M E N T S ......... .................................................................................... iv

L IS T O F T A B L E S ........ .. ................................................................. .... .. .. v iii

LIST OF FIGURES ............................... ... ...... ... ................. .x

ABSTRACT .............. .......................................... xi

CHAPTER

1 IN TR OD U CTION .................................... ................ .... ..... .. ...... ........ ..

C ognitive A going .................................................. 3
N orm al Cognitive A going ......................................................... .............. 3
M ild Cognitive Im pairm ent........................................................ ............... 5
Intraindividual V ariability ................................................. ............................... 7
Definition of Intraindividual Variability .................... ......... ..... .......... 7
Variability in Biological Systems............. ... ............ ........................ ............... 8
Variability in Psychological Domains.......................... ...... ...................9
Variability in Cognitive Functioning of Older Adults .....................................10
Variability and Cognitively Impaired Populations .....................................13
Variability and Learning Over Trials ...................................... ............... 15
Unresolved Issues M otivating the Current Study...................................................18
Variability in Cognitive Impairment .............. ............................................. 18
Variability in Learning ....................... ..... ...... .. ............... .18

2 STATEM ENT OF THE PROBLEM .................................... .......................... ......... 19

Intraindividual Variability in Memory and Other Cognitive Domains ....................20
A im O ne ................................................................... 20
H y p o th e sis O n e ............................................................................................. 2 0
A im T w o .........................................................2 0
Hypothesis Two ................... ... ......... ......... ........21
Intraindividual V ariability in Learning............................................... 21
Aim Three............................................ 21
H hypothesis Three ........................... .......... ....................... .. ............ 21





v









3 M E T H O D S ....................................................... 23

Stu dy D design .....................................................2 3
P participants ........................................23
Participant Recruitment ................... ................. ...... ............... 24
Inclusion/Exclusion Characteristics ........................................ .....................25
A ll p a rticip an ts ....................................................................................... 2 5
M C I participants........ ................................................ .......... ............. 26
Measures .......... ........ ......................... ................ 27
Phase 1: Telephone Screening.... ...................... .. ..................27
Phase 2: Neuropsychological Intake Assessment..................... .....................28
Phase 3: Daily Cognitive Assessment Battery (DCAB)..................................30
P ro c e d u re .............................................................................................................. 3 7
O verview of Study Phases..................... ........... .. .. ............... ... 37
Rationale for the research partner administration protocol..........................39
Compliance monitoring ................. ........ .................................. 40
Group Membership Assignment: Consensus Conference................................40
Initial Data Preparation and Study Variables .................................. ............... 41
Ceiling and Floor Considerations ..................... ...... ......... .................. 41
Standardization of Scores from the Daily Assessment Battery.........................43
Intraindividual V ariability Indices ........................................... .....................44

4 R E SU L T S ..............................................................................46

O v e rv ie w ................................................... ........................................................... 4 6
P relim inary A naly ses ....................................................................... ............. ..... 46
Neuropsychological Intake Assessment Data: Participant Neurocognitive Status
and Attrition Analysis. ............. ....................................... 47
Daily Cognitive Assessment Battery Data .................................. ....................48
Quality control check: Laboratory to home administration .......................48
Group differences in mean performance over all occasions ......................50
Effect of differing delay times for AVLT Delayed Recall..............................54
Distracting environmental variables: Data reduction................................55
Intraindividual Variability in Memory and Other Cognitive Domains ....................56
A im O ne and A im Tw o .............................................. ............................. 56
Aim One and Aim Two: Review of Analyses................. ......... ............. 57
Intraindividual variability differences across groups, based on cognitive
status s ........................................................................ .5 7
Data check: Reliability of intraindividual variability estimate.....................59
Predicting cognitive status with intraindividual variability ........................61
Intraindividual Variability Over Time: Understanding Variability and Performance
Relationships .................... ........................ 64
A im Three............................................. 64
Aim Three: Review of Analyses .................... ........ ........................65
Intraindividual variability over time (occasion), across cognitive status.....65
Relationship between intraindividual variability and level of performance 70
Sources of Intraindividual V ariability ............................ ............ ... .................83











5 D ISC U S SIO N ...............................................................92

R eview of Study Findings ................................................. ............................... 94
A im O ne and A im T w o ........................................ ..........................................94
Aim Three......................................................100
Intraindividual Variability Interrelationships Across Domains. .....................106
Stu dy L im station s........ ..................................................................... ......... .... .. 108
Future D directions .................................................................. ........................ 112
Conclusion ..................... ...... .................... 115

APPENDIX

SAM PLE DAILY W ORKBOOK .......................................... ............... ............... 116

REFEREN CES ........................................ .. .... ........ .............. .. 142

BIOGRAPHICAL SKETCH ............................................... ............... 150
















LIST OF TABLES


Table page

3-1. M ean (SD) or N (%) of demographic data. .................................... .................24

3-2. Measures for Neuropsychological Intake Assessment...................... ...............28

3-3. Measures for Daily Cognitive Assessment (DCAB) .............. .....................31

3-4. Mean performance on the alternative forms of each cognitive task..........................34

3-5. Intercorrelations between mean scores on alternative versions of AVLT List 1. ....35

3-6. Intercorrelations between mean scores on alternative versions of AVLT Total
S core. ......... ......... ......... .................................... ........................... 3 5

3-7. Intercorrelations between mean scores on alternative versions of Backward Digit
S p a n ............................................................................. 3 6

3-8. Intercorrelations between mean scores on alternative versions of Symbol Digit...... 36

3-9. Correlations of demographics and mean IRI scores...............................................45

4-1. Mean performance on neuropsychological measures, by cognitive status ...............49

4-2. Mean performance on neuropsychological measures, by attrition status ................49

4-3. Comparison of supervised session with first at-home daily session. ........................51

4-4. Means (Standard Deviations) for all measures by cognitive status...........................53

4-5. M ean A VLT delay tim e ............................ ................ .................... ............... 54

4-6. Factor loading for distracting environment variables ........................ ............55

4-7. Mean (Standard Deviation) Intraindividual Residual Indices (IRIs).......................59

4-8. Covariation among Intraindividual Variability Indices over blocks..........................61

4-9. Canonical loadings and classification statistics for discriminant function models. ..63

4-10. Mean Intraindividual Residual Indices (IRIs) by block and by cognitive status......66









4-11. Repeated Measures ANOVA: Intraindividual variability predicted by block score
and cognitive status .......................................... ............... .... ....... 67

4-12. Correlations of mean level of performance and IRI...........................................71

4-13. Regression coefficients for predicting IRI with mean level performance and
cognitive status group.............. .. ....................... .........72

4-14. Time effects on mean performance. .............................................. ............... 76

4-15. Between-person correlations between intraindividual variability and linear slope of
performance gains separately for each variable. ................... ................... .......... 81

4-16. Regression coefficients for predicting IRI with linear slope gain in performance and
cognitive status s group ............ ... ......................................................... .... .. ... .. 82

4-17. Intercorrelations of Individual Residual Indices. .............................................. 86

4-18. Coupled variabilities ........................................... ........................ ............... 90
















LIST OF FIGURES

Figure p

1-1. Intraindividual variability: "Fluctuations" in performance. ........................................8

1-2. Variability across learning phases. ..........................................................................16

3-1. D design of the current study.............................................. .............................. 37

4-1. Intraindividual variability by blocks...................................... ........................ 69

4-2. Mean performance and intraindividual variability. ................................................73

4-3. Grow th curves by cognitive status....................................... .......................... 77

4-4. Linear slope and intraindividual variability....................... .................... 82














Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy

UNDERSTANDING THE TRANSITION FROM
NORMAL COGNITIVE AGING TO MILD COGNITIVE IMPAIRMENT:
COMPARING THE INTRAINDIVIDUAL VARIABILITY
IN COGNITIVE FUNCTION


By

Karin J. M. McCoy

December 2004

Chair: Michael Marsiske
Cochair: Russell M. Bauer
Major Department: Clinical and Health Psychology

Intraindividual variability describes fluctuation or transient change in performance,

and can be measured by repeated assessment of an ability or trait over a short period of

time. Intraindividual variability in biological systems has been demonstrated to indicate

systemic compromise (e.g., loss of homeostatic regulation). Consequently, one theory

investigators have begun to research is whether intraindividual variability or fluctuation

in cognitive performance may be an indicator of cognitive decline. Additionally, a

second theory suggests that fluctuations in cognitive performance may be greater during

periods of learning acquisition, with a corresponding reduction in variability following

the acquisition phase. This study investigated intraindividual variability of cognition on

measures of attention, processing speed, working memory, and episodic memory in older

adults (over 65 years of age) with and without mild cognitive impairment (MCI), by









assessing performance in these domains daily for 31 days. MCI may be a transitional

stage between normal aging and dementia. In particular, the term amnestic MCI

describes individuals with focal cognitive impairment in the memory domain; such

impairment might foretell future Alzheimer's dementia. For this study, MCI is defined as

list memory performance 1.5 standard deviations below age-appropriate norms,

supplemented by subjective memory complaints and informant report of memory

problems, as well as intact cognition in non-memory domains. Results revealed that

older adults with amnestic MCI demonstrated a pattern of intraindividual variability and

performance level that was consistent with patterns seen in previous studies of

cognitively intact or demented individuals. Individuals with MCI demonstrated similar

rates of practice-related gain over occasions as did the cognitively intact individuals.

However, intraindividual variability was related inversely to performance, for most

measures. MCI status was not consistently related to intraindividual variability across the

cognitive battery studied. Interrelationships of performance gain slopes with degree of

fluctuation did not provide clear evidence for either of the two theories regarding the role

of intraindividual variability in practice-related gain or neurocognitive vulnerability.

These findings provided evidence that the two types of intraindividual variability

described by the current theories in the literature may co-occur in the MCI population.














CHAPTER 1
INTRODUCTION

Intraindividual variability (IIV) describes fluctuation or transient change in

performance, and can be measured by repeated assessment over a short period of time. A

growing literature suggests that inconsistency of cognitive performance, or higher

amounts of intraindividual variability, may be particularly meaningful as an indicator of

cognitive vulnerability in older adults (Hultsch et al., 2000, Murtha et al., 2002, Walker et

al., 2000). This is an intuitively appealing argument, drawing from conventional notions

of homeostasis and the importance for self-regulated organisms to show a relatively

"steady state" in most system functions. Interestingly, however, relatively little research

has tried to extend the intraindividual variability concept to one of the most cognitively

vulnerable populations: cognitively impaired older adults.

As the following discussion will demonstrate, this study was designed to

investigate the effects of cognitive impairment on intraindividual consistency of

performance on cognitive measures. This introduction and review of the literature

chapter is organized as follows. First, the distinction between normal cognitive aging and

mild cognitive impairment will be considered, focusing primarily on changes in the level

of performance on cognitive measures. Second, a review of the emerging body of

literature on intraindividual variability, or inconsistency of performance, and aging will

follow. Specifically, evidence that increasing cognitive vulnerability is associated with

cognitive variability will be investigated. Third, the review will conclude with a









discussion of unresolved issues in the current research literature. It is these unresolved

issues that motivated the current investigation.

The current study, designed to investigate intraindividual variability in cognitive

performance in older adults with amnesticc" mild cognitive impairment (Petersen et al.,

2001), was premised on six major findings. First, normal cognitive aging and cognitive

impairment are distinguishable by level of performance on cognitive measures. Episodic

memory and verbal learning deficits are a prominent source of difference between

amnestic mild cognitive impairment (MCI) and unimpaired elders (Bozoki et al., 2001;

Chen et al., 2000; Morris et al., 2001; Petersen et al., 2001), although the magnitude and

breadth of cognitive differences varies with the number of domains impaired in MCI

(Bozoki et al.; Petersen et al., 1999). Second, there is growing evidence in non-impaired

elders that intraindividual variability of performance may be an important individual

differences characteristic. Persons at lower levels of functioning, and who are older,

often show more inconsistency on a variety of cognitive measures (e.g., Anstey, 1999;

Bleibeg et al., 1997; Fozard et al., 1994; Rabbitt, Osman & Moore, 2001; Salthouse,

1993; Stuss et al., 1994). Third, in recent studies, persons with more advanced cognitive

impairment (dementia) evinced even higher levels of intraindividual variability than

either non-demented or physically impaired elders (Hultsch et al., 2000; Murtha et al.,

2002). Fourth, in MCI, level of performance may be relatively unaffected in some areas,

such as attention and working memory (Mesulam, 2000; Petersen et al., 2001). An

unanswered question, however, is whether impairment-related increases in intraindividual

variability in functions like attention or working memory might be detectable even before

substantial performance level impairments are seen. Fifth, in non-impaired elders,









variability appears to change, qualitatively, over the course of many retest occasions. For

some tasks, in which there is learning and performance improvement, individuals show

initial performance increments, followed by a stable asymptote. Intraindividual

variability before and after reaching asymptote are relatively unrelated to one another; as

cognitive responding becomes more strategic and automatic over retest trials, there is a

reduction in intraindividual variability (Allaire & Marsiske, 2004). Sixth, the relative

absence of strategic learning is a hallmark characteristic of dementia (Backman et al.,

2000; Mesulam, 2000), and is often found in early dementia, pre-dementia, or MCI (Chen

et al., 2000; Morris et al., 2001: Petersen et al., 1999). This finding has sparked much

debate regarding the definition of MCI as an early stage of dementia, and in fact, experts

have opined that "most pathological conditions of amnesticc] MCI are likely to be early-

stage Alzheimer's disease" (Petersen et al., 2001, p. 1989). Thus, a question is whether

individuals with MCI might show less reduction in intraindividual variability over trials,

as well as more stable patterns of individual differences in variability over trials.

Cognitive Aging

Normal Cognitive Aging

Typically, aging is associated with complaints of cognitive impairment,

particularly laments of mild to moderate deficits in memory performance compared to

functioning at an earlier age. Jorm and colleagues (1994) reported that, in their sample of

community dwelling adults aged 70 or older, 62% believed that their memory was worse

than earlier in life. Although perceived changes in memory performance are not reliable

indicators of objective alterations in cognitive status (Jorm et al., 1997; Turvey et al.,

2000), formal neurocognitive assessment in clinical and research settings provides

objective evidence that cognitive function is impacted by aging (for a review see









Woodruff-Pak, 1997). Although methodological issues, such as differing findings

between cross-sectional and longitudinal research (e.g., Seattle Longitudinal study;

Schaie and Willis, 1991; Schaie, 1995), have sparked comment and controversy among

investigators, a number of findings appear to be replicable. In particular, consistent

evidence of reductions in processing speed (Craik & Salthouse, 2000), and episodic

memory performance (new learning and recall; Peterson et al., 1992; Wahlin et al., 1995;

for reviews see Anderson & Craik, 2000, and Balota, Dolan & Duchek, 2000) has

resulted in general agreement that these domains are particularly susceptible to the effects

of normal aging.

Development of the diagnostic category 'age associated memory impairment'

(AAMI; Crook et al., 1986) reflected initial attempts to describe and delineate the

changes in memory function thought to be common in healthy older adults. Specifically,

the criteria for AAMI included observable memory test performance one standard

deviation below young adult levels of performance. Additional research, critical

commentary, and further refinements in the understanding of cognitive functioning of

older adults resulted in the presentation of multiple additional diagnostic categories (e.g.,

age-consistent memory impairment, late-life forgetfulness, and age-related memory

decline; see Larrabee, 1996, for a thorough review). Most recent developments in the

delineation of age-related effects on memory function include an awareness and focus on

individual differences in normal aging, and an investigation of increased interindividual

differences in performance and trajectory of change (see Mesulam, 2000). Of primary

interest to investigators of the developmental trajectory of cognitive aging is that normal

aging and "non-normal" cognitive aging (e.g., dementia) are generally distinguishable by









level of performance on cognitive measures, even in those domains affected by normal

aging (Lishman, 1988; Mesulam, 2000; Petersen et al., 1992).

Mild Cognitive Impairment

Mild cognitive impairment (MCI; Flicker, Harris, & Reisberg, 1991; Petersen et al.,

1999) is the term used to describe non-demented older adults with a slight impairment in

cognitive functioning, typically in the memory domain (Celsis, 2000; Petersen et al.,

2001). MCI differs from normal age-related changes of cognitive functioning, since the

defining criteria require impaired cognitive function when compared to age- and

education-matched peers. Petersen and colleagues (2001) indicate that MCI commonly

refers to impairment in memory functioning, and that the term would be more accurately

labeled amnesticc MCI". MCI, specifically amnestic MCI, is considered by many to

capture the boundary or transitional condition between normal aging and mild

Alzheimer's disease (Morris et al., 2001; Peterson et al., 2001). Individuals with

amnestic MCI typically progress to Alzheimer's Disease at a rate of 10 15% per year

(Petersen et al., 1999; Tiemey et al., 1996). Morris and colleagues (2001) reported a five

year conversion (MCI to Alzheimer's disease, based on Clinical Dementia Rating score)

of 40 60%, with a nine-year conversion rate in excess of 90%.

As investigators have increasingly focused on characterizing reductions in

cognitive performance that might capture a pre-clinical phase of dementia, particularly

Alzheimer's disease (Celsis, 2000), demonstrable and reliable differences in cognitive

performance between older adults with normal, intact cognitive functioning, and those

with impaired cognition are recognized to be of significant value for clinicians and

researchers. Investigations into the predictors of amnestic MCI may provide means for

early detection, intervention, and eventually treatment for Alzheimer's disease (Knopman









et al., 2001; Petersen et al., 2001). Typically, as with investigations into normal cognitive

aging, differences along the normal aging MCI Alzheimer's disease continuum are

made on the basis of level of performance on neurocognitive measures.

However, level of performance comparisons on a single occasion of measurement

on objective cognitive measures are not the sole means to a successful and accurate

diagnosis of cognitive impairment. In fact, from both a developmental perspective and a

neuropsychological approach, repeated assessment (e.g., once every six months) to

document change within an individual over time provides a more sensitive indicator of

impairment in cognitive functioning especially at early or premorbid stages of a

dementing illness (Daly et al., 2000; Elias et al., 2000; Petersen et al., 2001). Substantial

research to characterize both normal and impaired cognitive aging follows a basic

longitudinal methodology in order to accurately group participants. For example, Schaie

and Willis (1991) have observed that the results of a cross-sectional comparison (the

level of functioning approach) and 20-year longitudinal follow-up of adults of varied ages

supported different conclusions. Several differences between the performance of the

young and elderly participants in the cross-sectional arm of the study were a reflection of

cohort differences rather than change over time: results of the cross-sectional analyses

would have led to incorrect conclusions regarding the nature of cognitive functioning for

a number of participants.

Although the, albeit often unstated, research goal in the area of cognitive

impairment in older adults is to determine a means to accurately diagnose impairment

based on a one-time assessment, this approach has been less than completely successful

for the diagnosis of many types of cognitive dysfunction in the older years (Knopman et









al., 2001). Delineation of the trajectory of change of functioning in various cognitive

domains is crucial for accurate diagnosis and prognosis. Typically, this assessment of

change occurs over months or years. Intraindividual variability methodology may

provide a means to determine the trajectory of change and/or diagnostically useful

patterns in cognitive performance in a much shorter time frame, possibly even prior to

observable changes in overall level of performance.

Intraindividual Variability

Definition of Intraindividual Variability

The measurement of intraindividual variability (or fluctuation) in cognitive

performance over a shorter period of time may provide useful information in the

diagnosis or identification of cognitive dysfunction. Long-term changes in cognitive

functioning within an individual, as well as increasing differences in level of cognitive

performance between persons may be based on permanent alterations in intraindividual

variability (Li & Lindenberger, 1999; Siegler, 1994). Intraindividual variability (IIV),

defined as fluctuation or change in performance over a short period of time, provides a

measure of an individual's "hum" around their mean, or "steady-state" level

(Nesselroade, 1991). The intensive, repeated measurements necessary to assess

intraindividual variability provide a more accurate representation of an individual's true

mean performance, as well as a "finer grained" resolution of fluctuations in performance.

Figure 1-1 depicts the "hum" or variability around the mean performance (indicated by

the regression line) for two individuals (Allaire, 2001).

Variability in performance from one session to the next has sometimes erroneously

been considered to be measurement error, yet, intraindividual variability can be reliably

measured, and is of sufficient magnitude to be important. Measurement error and










practice effects can be distinguished from lawful but transient changes in performance

from session to session (Hultsch et al., 2000; Li et al., 2001). When assessing cross-

occasion intraindividual variability in story recall via weekly assessments for two years,

Hertzog et al. (1992) found considerable intraindividual variability in performance across

occasions; more than 20% of this variability was reliable variance not associated with

practice, alternative forms, or other systematic changes over time. Rabbitt et al. (2001)

found systematic changes between sessions independent of circadian variability or

practice.



35


Participant 21 IRI = 1 54


iL i) (N (N (N c o) cR ) r r CO CO CO r;: rQ) OT co 0c ) 0) m O m D 2

Figure 1-1. Intraindividual variability: "Fluctuations" in performance. From Allaire,
2001.


Variability in Biological Systems

Not only is intraindividual variability measurable, but research findings also

suggest that extreme intraindividual variability within a system is an indicator of systemic









dysfunction. Behavioral research into the intraindividual variability of psychologically

relevant constructs draws on the extensive research in physiology, biology, medicine, and

allied disciplines, where investigators consistently have found that increased variability

within a system is associated with increased age and poorer overall functioning (Britton,

1997; Fluckiger et al., 1999; Guimares & Isaacs, 1980; Hausdorf et al., 1997; Higgins,

1997; Pagani, 1999). Intraindividual variability within a fairly narrow range (e.g.,

homeostatic regulation) is one defining characteristic of biological systems. At the

behavioral level, short-term, narrow-ranged intraindividual variability can be considered

indicative of adaptive ongoing processes in response to an ever-changing environment

(Nesselroade et al., 1996). In contrast, variability characterized by extreme or erratic

fluctuations may reflect a breakdown in the homeostatic regulatory system; this type of

dramatic variability may be found in the biological or psychological arena. Evidence of

such liability in the intraindividual variability of cognitive performance in older adults

with mild cognitive impairment would be consistent with the findings in the medical

literature regarding the correlation between increased variability and poorer systemic

functioning.

Variability in Psychological Domains

In the psychological literature, studies examining variability in adults have focused

on describing the extent to which particular psychological constructs, such as affect and

mood, self-esteem, and personality are marked by variability. Similarly, explorations of

intraindividual variability in the older adult population have described the correlation of

fluctuations in the domains of affect and mood, self-efficacy, and world views and

religious beliefs. Kemis et al.'s (1993) investigation of variability in self-esteem in

adults over a four day period illustrates the unique contribution of intraindividual









variability, as they found that day-to-day deviations in self-esteem correlated with

fluctuations of perceived competence and social acceptability independent of mean level.

Eizenman et al.'s (1997) examination of intraindividual variability in perceived control,

locus of control, and perceived competence in a sample of older adults over the course of

seven months provides further demonstration of the distinction between variability and

mean performance. While the mean of the two aspects of control was not a significant

predictor of mortality, the two variability scores for locus of control and perceived

control added significantly to the prediction of five year mortality. The predictive utility

of variability in this study is consistent with findings from other studies of self-efficacy

and self-esteem reporting that variability is a better predictor of certain outcomes than

mean level (Butler et al., 1994; Kernis et al., 1993).

Variability in Cognitive Functioning of Older Adults

Recently, researchers have begun to examine the extent to which cognitive

functioning in healthy older adults is characterized by intraindividual variability.

Overall, results suggest a significant relationship between increased intraindividual

variability and age. For instance, Fozard, Vercruyssen, Reynolds & Hancock (1994)

reported that, over an 8-year longitudinal study in which participants aged 17 to 96 years

were repeatedly assessed, intraindividual variability in reaction time increased with

increasing age. Results reported by Shammi, Bosman, and Stuss (1998) indicated that

within-person variability was greater on two psychomotor tasks (i.e., finger tapping and

choice reaction time), and a time-estimation tasks for an older sample of participants

when compared with younger adults. More recently, MacDonald, Hultsch, & Dixon

(2003) reported greater inconsistency of performance on reaction times tasks in older









adults, even after controlling for differences in response speed. Baseline differences in

variability were associated with variability 6 years later.

Studies focusing solely on intraindividual variability in cognitive performance of

older adults (i.e., without younger aged comparison groups) have also provided evidence

establishing a link between age and cognitive intraindividual variability. Salthouse

(1993) reported reaction times in four independent samples of older adults; results from

this study indicated that, across the four samples, increased age was associated with

increased intraindividual variability of reaction time over 90 consecutive trials.

Similarly, Anstey's (1999) examination of age differences in older adults' performance

on measures of simple and complex reaction time presented in both the visual and

auditory modalities provides evidence that intraindividual variability in performance on

some reaction time measures (complex auditory task and simple and complex visual

tasks) is positively related to age.

Extrapolating from findings that variability in performance appears to increase with

chronological age, researchers have speculated that increased intraindividual variability

of a behavior or within a measured construct might be an indicator of a more general

underlying self-regulatory breakdown in performance in that domain. Extending this

notion to cognitive functioning, some investigators theorize that increased intraindividual

variability in cognitive functioning is indicative of compromised neurobiological

functioning and an increase in "neural noise" (Bruhn & Parsons, 1977; Hendrickson,

1982; Li and Lindenberger, 1999). If this assumption is correct, then individuals who, on

average, perform more poorly on tests of cognitive functioning may exhibit more

intraindividual variability, implying a negative relationship between intraindividual









variability and mean performance. Consistent with this hypothesis, Rabbitt, Osman &

Moore's (2001) examination of covariation between week-to-week intraindividual

variability in reaction time over a 36-week period and performance on a test of reasoning

ability in older adults indicated that the magnitude of intraindividual variability was

greater for the individuals who performed poorly on the battery of intelligence tests. Li

and her colleagues (2001) also found evidence of a relationship between cognitive

intraindividual variability and mean performance, although their results were somewhat

more mixed. Specifically, these authors found that the association between

intraindividual variability and mean performance was in the expected negative direction

for a memory task, with higher levels of intraindividual variability associated with poorer

overall performance. However, a significant correlation between intraindividual

variability and mean performance was not found for working memory. Furthermore, a

positive correlation between variability and mean performance was found for a spatial

memory task, indicating that higher overall spatial memory performance was related to

greater intraindividual variability. These results, which contrast with other findings,

suggest that intraindividual variability in performance in some cognitive domains may

not be associated with vulnerability or impairment in overall level of performance. In

fact, recent work by Allaire and Marsiske (2004), consistent with the research presented

by Li and colleagues, suggests that variability in performance may also be influenced by

strategic learning, such that increased variability during learning and memory may reflect

intact, adaptive processes.

As an alternative to the hypothesis of a negative relationship between

intraindividual variability and mean performance, significant fluctuations or liability in









intraindividual variability may be indicative of compromised neurocognitive functioning

independent of overall performance. Past research, particularly in the domains of self-

esteem and perceived control (Eizenman et al., 1997; Kernis et al., 1993), which

demonstrated that intraindividual variability uniquely predicted outcome independent of

mean level, provides evidence that the assessment of intraindividual variability may be

valuable in the absence of differential mean performance.

Similar findings in research on cognitive functioning in older adults may provide a

unique tool to early diagnosis of potential future declines in cognition, as seen in

Alzheimer's disease, since early or pre-clinical phases of the disorder are characterized

by a lack of reduction in the observed level of performance of functioning in a number of

domains which will later become impaired. Performance in these domains may be

characterized by increased variability prior to the disorder-related reduction in overall

level of performance.

Variability and Cognitively Impaired Populations

To date, there is a paucity of published studies of intraindividual variability with

older patient populations, and none reported which examines variability in cognitive

functioning in individuals with mild cognitive impairment. One of the few studies of

neurologically impaired participants is Shifren, Hooker, Wood, & Nesselroade's (1997)

examination of variability of mood in older adults diagnosed with Parkinson's disease,

via participant completion of a mood questionnaire daily for 70 days. Using dynamic

factor analysis, the authors examined dimensionality of mood within individuals over

time as well as the extent to which mood was related to itself over time. The number of

dimensions (factors) needed to explain the variation in the day-to-day variability in mood

varied between different subsets of participants. Similarly, the relationship between









mood on subsequent days was different within individuals, in that mood on one day

influenced subsequent daily mood ratings differently for participants. This study

highlights intraindividual variability over time of a particular construct typically thought

to be highly stable and demonstrates the transactional nature of psychological processes.

Additionally, this study demonstrates the utility and feasibility of utilizing an

intraindividual variability approach with neurocognitively compromised patient

populations.

Hultsch et al. (2000) examined intraindividual variability in performance on a

reaction time task, as well as in two measures of memory functioning (i.e., word and

story recognition) in healthy adults, cognitively intact adults with arthritis, and adults

diagnosed with mild dementia. Collapsing across the three groups, the authors found that

intraindividual variability on the cognitive tasks was positively associated with the mean

latency scores and negatively related to accuracy scores within each measure. That is,

more intraindividual variability was associated with slower average latency scores and

poorer average accuracy scores. In addition to this correlational evidence, Hultsch and

colleagues also found that the participants in the mildly demented group, presumably the

group with poorest overall cognitive functioning, exhibited a significantly greater level of

intraindividual variability on all three cognitive tasks than the participants in the healthy

and arthritic groups.

In their comparison of cognitively intact older adults with cognitively impaired

elders (patients with Alzheimer's disease and patients with frontal lobe dementia),

Murtha and colleagues (2002) demonstrated that intraindividual variability of

performance on three distinct reaction time measures distinguished the healthy older









adults and the two groups of patients. Walker et al. (2000) utilized attentional measures

of intraindividual variability in cognitive performance to distinguish healthy, cognitively

intact older adults from patients with Alzheimer's disease and patients with Lewy-Body

dementia. Variability in cognitive performance on measures of vigilance and reaction

time confirmed the clinical assessment of variability and differentiated the groups by

diagnostic category. Similar results have been found in non-aging studies, where the

cognitive functioning of individuals who had experienced a traumatic brain injury was

characterized by higher levels of intraindividual variability (Bleiberg et al.,1997;

Hetherington et al., 1996; Stuss et al., 1994).

Although, to date, there have been no published reports of intraindividual

variability assessment of individuals with mild cognitive impairment, variability

assessment may be useful approach for this population. Individuals with MCI suffer

from cognitive impairment, suggesting that investigation of intraindividual variability in

this group may provide further insights into the transition from normal aging to cognitive

impairment.

Variability and Learning Over Trials

Recent work by Allaire (2001; Allaire & Marsiske, 2004), suggests that temporal

patterns in intraindividual variability of cognitively intact older adults may reflect

learning over trials. Findings that variability increased prior to the attainment of

maximum performance suggest that increased intraindividual variability during the

learning phase might reflect acquisition of strategies (a finding consistent with Siegler's

(1994) work on strategy learning in childhood), while subsequent reduction in

intraindividual variability reflects consistent implementation of the strategy. This is

summarized in Figure 1-2. Trial-and-error learning is likely to be adaptive, especially in








16



conjunction with increased overall level of performance. Excessive variation around an


asymptotic, or non-increasing, level of functioning is more likely to reflect inconsistency


and vulnerability, since the observed variability is not correlated with increased strategy


acquisition or learning.


Older adults with memory impairment (e.g., Alzheimer's disease) are


distinguishable by poor recall on episodic memory tasks, suggesting that learning (e.g.,


strategy acquisition) is not taking place (Backman et al., 2000; Chen et al, 2000;


Mesulam, 2000; Morris et al., 2001; Petersen et al., 1999). Thus, it is likely that patients


Strategy
Acquisition Phase

* Variability is adaptive
and reflects strategy use
* Consequently, positive
associations between
mean performance and
variability (better
functioning individu
less variability
* Will tend to b served
during perigs of linear
growth i performance
* More lely to be seen for
ha -to-learn and
mplex tasks; tasks on
which learning is possible
but "mastery" has not
been achieved


Strategy
Implementation
Phase
* Variability is maladaptive
and reflects performance
inconsistency
* Consequently, negative
associations between
level and variability
(lower functioning
individuals= more
variability)
* Will tend to be observed
during periods of stable
or asymptotic functioning
* More likely to be seen for
very easy or very hard
tasks on which
participants are not
getting better at over time


Occasions of Measurement


Figure 1-2. Variability across learning phases. From Allaire, 2001









with dementia may demonstrate a unique pattern of intraindividual variability in

cognitive performance over time, typified by no noticeable reduction in performance

fluctuations. This pattern in performance would likely differ from that expected in

cognitively intact older adults, since individuals capable of learning new material would

demonstrate variability in performance during the strategy acquisition stage, followed by

a reduction in variability during the strategy implementation phase (once task-specific

strategies are acquired).

Individuals with MCI, especially older adults with amnestic MCI, who are likely at

increase risk of converting to Alzheimer's disease, may demonstrate a pattern of

intraindividual variability similar to that expected by patients with Alzheimer's disease.

In conjunction with an expected lack of, or significant reduction in, new learning, these

individuals may demonstrate a lack of reduction in intraindividual variability over time.

In summary, intraindividual variability, or fluctuations in performance around a

mean, is measurable and provides an unique predictor of outcome in biological and

psychological research. Past research has demonstrated that extreme intraindividual

variability is an indicator of systemic compromise in biological systems, and is likely an

indicator of compromised function in cognition, as patients with neurological disorders

demonstrate increased intraindividual variability in cognitive performance. In

cognitively intact older adults, intraindividual variability in cognitive performance

declines over time, corresponding with practice and strategy acquisition. Increased

intraindividual variability is likely in mild cognitive impairment, and may provide a

unique predictor of future decline even in the absence of changes in overall level of









performance, especially if observed variability does not decline over time with repeated

exposure to the same tasks.

Unresolved Issues Motivating the Current Study

Variability in Cognitive Impairment

Past research into cognitive aging has provided significant evidence that normal

cognitive aging, mild cognitive impairment, and progressive dementing disorders are

distinguishable by level of performance in specific domains of cognition (e.g., memory).

An unanswered question is the degree to which differences in intraindividual variability

in performance in the same cognitive domains might likewise distinguish these groups.

Although increasingly the evidence suggests that variability in performance indicates

vulnerability in a system, and may foretell future impairments in performance, there have

been no investigations of the likelihood that increased intraindividual variability in an

intact cognitive domain might predict future decline in that domain.

Variability in Learning

Previous research suggests that intraindividual variability in cognitive

performance may be an indicator of strategy acquisition during repeated exposures to a

novel task. A pattern of increased variability during a learning phase, followed by a

reduction in variability in performance when overall level of performance reaches an

asymptote, has been observed in cognitively intact older adults. To date, there has been

no investigation of the pattern of intraindividual variability in the cognitive performance

of cognitively impaired individuals as they are exposed to a task repeatedly.














CHAPTER 2
STATEMENT OF THE PROBLEM

Although investigators have continued to focus on identifying the correlates and

predictors of current and future impairment in level of performance in cognitive domains

of function, only recently has any attention been paid to the possible contribution of

intraindividual variability in performance to the diagnosis or prediction of cognitive

decline. The primary concern of this research is to examine the relationship between

cognitive impairment and intraindividual variability in cognition in older adults.

The Study Aims draw upon what is known about changes in level of performance

on neurocognitive measures in older individuals with mild cognitive impairment, and

upon a small number of studies that have examined intraindividual variability in

cognitive functioning in unimpaired and cognitively impaired elders (e.g., Hultsch et al.,

2000; Murtha et al., 2002; Walker et al., 2000). Additionally, the aims reflect recent

work which indicates that intraindividual variability during learning may represent

strategy acquisition, while variability without concurrent learning (either because

individuals have plateaued or because they cannot learn) may reflect compromised

function (Allaire & Marsiske, 2004; Li et al., 2001). Finally, the study aims reflect recent

innovations in the literature hinting at the potentially differential contribution of varied

temporal resolutions of intraindividual variability measurement (Strauss et al., 2002;

Walker, 2000).

The study investigated whether (a) there exist differences in intraindividual

variability between older adults with memory impairment and those whom are









cognitively intact, and whether these differences are observable both in memory and in

other domains of cognitive functioning, (b) such group differences in variability might be

a more sensitive distinguishing characteristic of cognitive impairment than level of

performance, especially if non-memory domains revealed heightened intraindividual

variability at a time when level of performance had not declined substantially in these

domains, and (c) learning-related (e.g., practice effects or level of performance) changes

in variability are more characteristic of non-impaired than impaired elders. Thus, the

study has three major aims in two general areas.

Intraindividual Variability in Memory and Other Cognitive Domains

Aim One

To investigate whether greater intraindividual variability in memory functioning is

seen in older adults with amnestic mild cognitive impairment, compared to cognitively

intact elders.

Hypothesis One

If greater intraindividual variability is an indicator of neurological compromise,

then older adults with amnestic mild cognitive impairment (MCI; defined as memory

performance 1.5 SD or more below age- and education-appropriate norms at baseline

assessment; Petersen et al., 2001) should exhibit more intraindividual variability in

memory performance over 31 occasions of measurement when compared to older adults

with no memory impairment at baseline.

Aim Two

To investigate whether cognitive tasks measuring neurocognitive domains that

typically show relatively less impairment in level of performance in mild cognitive

impairment (e.g., attention, processing speed) nonetheless reflect impairment-related









increases in intraindividual variability in older adults with mild cognitive impairment

relative to unimpaired elders.

Hypothesis Two

If mild cognitive impairment, as indexed by memory dysfunction, reflects the onset

of a more general cognitive decline process (i.e., 12 15% of persons with MCI convert

to dementia annually; Petersen et al., 2001), and if individual variability is an indicator of

cognitive compromise related to cognitive decline, then older adults with memory

impairment will also exhibit greater intraindividual variability in other cognitive domains

assessed over 31 occasions even before showing mean level impairment.

Intraindividual Variability in Learning

Aim Three

To investigate whether, in conjunction with expected differences in overall learning

across trials within session and across sessions, memory-impaired and cognitively intact

individuals show different patterns in intraindividual variability on memory tasks over

time. Extant research (Allaire & Marsiske, 2002; Li et al., 2001) suggests that, on tasks in

which elders show retest-related learning, variability has several properties: (a) it tends to

be positively associated with level of performance, (b) it is uncorrelated with the

variability observed after individuals reach asymptotic performance, and (c) it is reduced

over trials.

Hypothesis Three

Since mild cognitive impairment (when defined by memory performance) is

characterized by increased difficulty of new learning (Bozoki, et al., 2001; Chen et al.,

2000; Morris et al., 2001), it is expected that, relative to non-impaired elders, persons

with memory impairment will show relatively stable intraindividual variability patterns






22


throughout the period of study, reflecting little to no strategy acquisition. In contrast,

unimpaired elders will be more likely to show decrements in the magnitude of learning-

related variability. That is, the performance of the cognitively intact older adults will

show a positive relationship between variability and improvements in performance with

decrements in the magnitude of learning-related variability over the 31 occasions. As

initial performance improves, initial variability will decrease until remaining variability is

no longer associated with performance at asymptotic performance.














CHAPTER 3
METHODS

Study Design

The current short-term longitudinal study used a mixed between- and within-

subjects design to investigate intraindividual variability on tasks of attention, working

memory, and immediate and delayed episodic memory across two groups (older adults

who meet criteria for mild cognitive impairment [MCI] vs. cognitively healthy).

Performance on these tasks was assessed daily for thirty-one days (an initial assessment

in the laboratory followed by thirty daily assessment sessions at-home). This design is an

extension of the multivariate, replicated, single-subject, repeated measures design

(MRSMR; Jones & Nesselroade, 1990; Nesselroade & Ford, 1985), which traditionally

calls for the repeated (e.g., daily, weekly, monthly) multivariate assessment of an

individual participant over a finite period of time, replicated over multiple individuals.

The design has been extended in the current study by the placement of individuals into

one of two groups based on cognitive status prior to the initiation of the repeated

measures study.

Participants

Sixty-eight community-dwelling volunteers aged 65 years and older (age range 65

- 87 years) participated in the study. Fifteen participants were subsequently classified as

having amnestic mild cognitive impairment (MCI) and 53 were classified as cognitively

intact (Non-MCI). Details of the consensus procedure by which these classifications

were achieved are provided below. Table 3-1 displays the demographic information for









the full sample and each group. The only significant difference between the groups on

demographic variables was in the sex distribution (p = .033), with a higher proportion of

males in the MCI group. The MCI group demonstrated equivalent predicted IQ (as

calculated on the North American Adult Reading Test) and a significantly lower mean

score on both the Telephone Interview of Cognitive Status (TICS) and the Mini-Mental

Status Exam (MMSE). This is consistent with the criteria for MCI group, as outlined

below.


Table 3-1. Mean (SD) or N (%) of demographic data.
Total Sample MCI Non-MCI p value
(N= 68) (n= 15) (n = 53)
Age 78.01 (5.75) 76.60 (7.76) 74.57 (5.05) .229
Education 16.12 (2.66) 16.33 (2.90) 16.06 (2.62) .725
Sex .033
Males 29 (42.60) 10(66.70) 19(35.80)
Females 39 (57.40) 5 (33.30) 34 (64.20)
Race .717
Caucasian 63 (92.60) 14 (93.30) 49 (92.50)
Other 5 (7.4) 1(6.70) 4 (7.50)
TICS 36.10 (3.72) 31.67 (3.64) 37.38 (2.61) .000
MMSE 28.73 (1.24) 27.55 (1.75) 29.05 (0.84) .018
Predicted IQ 113.64 (7.32) 110.69 (5.92) 114.47 (7.51) .077
Note. TICS = Telephone Interview of Cognitive Status, MMSE = Mini-Mental Status
Exam, Predicted IQ from NAART (North American Adult Reading Test).

Participant Recruitment

Participants were recruited from the community, with a main focus on attaining a

sample of community dwelling older adults with and without mild cognitive impairment.

Primary recruitment for cognitively healthy older adults was through community events,

community organizations, and an article with advertising in a senior-oriented magazine

(the Senior Times). The second strategy, which served as recruitment for a number of

mildly cognitively impaired participants, was to identify individuals who expressed









interest in participating in research via established participant pools maintained by faculty

and students associated with the Institute on Aging. Recruitment of cognitively impaired

individuals stressed subjective memory complaints, confirmed objectively by the

neuropsychological assessment for qualification into the study. As outlined in detail

below, participants in the mild cognitive impairment group were required to meet criteria

for amnesticc MCI" (Petersen et al., 2001).

Participants were recruited individually and in dyads. Each target participant was

asked to identify a "research partner" (e.g., spouse, in-home caregiver, neighbor, or child)

to assist with administration and monitoring of the 30 at-home cognitive assessments.

Target participants and research partners were required to consent to participate.

Individuals unable to identify a willing partner were matched with other participants in

similar circumstances. Participants recruited in dyads (e.g., spouses) were allowed to

enroll in the study as both participants and research partners, as long as both met all

inclusion and exclusion criteria. Materials were coded so as to provide different

cognitive tests to each member of a dyad each day.

Inclusion/Exclusion Characteristics

All participants

All participants were over 65 years of age. Each was screened via telephone in

order to assure they met the following criteria by self-report: (i) no severe dementing

illness, (ii) no history of closed head injury with loss of consciousness, (iii) no other

neurological or major medical illnesses, (iv) no self-reported severe uncorrected vision or

hearing impairments, (v) no psychiatric disturbance sufficient to warrant inpatient

psychiatric treatment, (vi) no extensive drug or alcohol abuse, and (vii) self and research

partner willingness to participate in repeated cognitive evaluations. These criteria,









common in neuropsychological studies of older adults, were utilized to rule out

alternative potential etiologies for cognitive impairment and restrict focus to MCI.

Additionally, during this initial screening, Telephone Interview of Cognitive Status

(TICS; Brandt, Spencer, & Folstein, 1988) score below 30 points (the published cut score

for dementia), resulted in exclusion of two individuals. An exception was made for three

cases where point deductions were restricted to an item assessing delay verbal recall

ability.

All participants who met these initial criteria were assessed in person, with the

Neuropsychological Intake Assessment measures (see below), and were required to (a)

attain a Mini-Mental Status (Folstein et al., 1975) score above 23, (b) perform within one

standard deviation of age- and education-adjusted norms on neurocognitive measures in

non-memory domains, and (c) demonstrate no evidence of impairment on Activities of

Daily Living (ADL), as measured by proxy interview with the Blessed Dementia Scale

(Blessed et al., 1968; following criteria for Mild Cognitive Impairment, Petersen et al.,

2001). Participants in the cognitively intact group were also required to (d) exhibit intact

(within one standard deviation of age- and education-adjusted norms) performance on

memory measures (i.e., no impairment in the memory domain).

MCI participants

In addition to the above criteria (a, b, c), participants in the MCI group were

required to demonstrate impaired memory function, measured by a score 1.5 standard

deviations below age- and education-adjusted norms on a list learning memory task

(Hopkins Verbal Learning Test Revised; HVLT-R; Brandt & Benedict, 2001). The

specific scores on the HVTL-R used were Total Recall, Delayed Recall, and Percent

Retention. Additionally, participants in this group were required to receive a Clinical









Dementia Rating (CDR; Hughes et al., 1982; Morris, 1993) score below 1.0 (i.e., 0.0 or

0.5). This requirement insured that observed cognitive impairment was mild. Higher

scores on the CDR indicate more substantial memory impairment and/or more substantial

impact on daily functioning. These basic requirements were confirmed and considered

when group assignments were made as part of the Consensus Conference, described

below.

Measures

Participants were exposed to three study phases, an initial telephone screening, a

neuropsychological intake assessment, and the daily assessment. While the measures are

explained by phase, the phases of study are described in detail in the Procedure section of

this chapter.

Phase 1: Telephone Screening

Initial screening of participants took place using the brief Telephone Interview for

Cognitive Status (TIC; Brandt, Spencer, & Folstein, 1988). Using a cut score of 30

points to distinguish demented individuals from cognitively intact individuals, the TICS

has a sensitivity of 94% and a specificity of 100% (Brandt, Spencer & Folstein). Thus,

the TICS allowed for initial exclusion of demented individuals from the study.

Additionally, since individuals with MCI were more challenging to recruit and identify

than cognitively intact older adults, performance on the recall memory item of the TICS

was utilized to assist in selecting individuals with a greater likelihood of meeting criteria

for amnestic mild cognitive impairment. The telephone screening, including exclusion

criteria, TICS interview, and description of the study took approximately 30 minutes.









Phase 2: Neuropsychological Intake Assessment

Following informed consent procedures approved by the University of Florida

Institutional Review Board, participants were assessed individually, in a standardized

manner, on a selection of commonly used neuropsychological measures, in order to

determine appropriate assignment to the amnestic MCI or unimpaired groups. Measures

are listed in Table 3-2 (in a thematic grouping, not test order). This assessment took

approximately two hours for cognitively intact individuals, and approximately three

hours for individuals later deemed to meet criteria for MCI.

Table 3-2. Measures for Neuropsychological Intake Assessment.
Cognitive Domain Measure Source
Overall Cognitive Mini Mental Status Examination Folstein, Folstein, &
Functioning (MMSE) McHugh, 1975
North American Adult Reading Test Blair & Spreen, 1989
(NAART)
Attention & Trail Making Test A & B Reitan, 1992
Working Memory
Memory Hopkins Verbal Learning Test Brandt & Benedict, 2001
Revised (HVLT-R)
Rivermead Behavioral Memory Test Wilson, Cockburn &
(prose memory subtest) Baddeley, 1985
Brief Visuospatial Memory Test Benedict, 1997
Revised (BVMT-R)
Language Boston Naming Test Second Kaplan, Goodglass &
Edition (BNT) Weintraub, 2001
Controlled Oral Word Association Benton & Hamsher, 1989
Visuospatial Rey-Osterrieth Complex Figure Copy Rey, 1941
Mood Geriatric Depression Scale (GDS) Yesavage, 1983
Center for Epidemiological Studies Radloff, 1977
Depression Scale (CES-D)
Functional Blessed Dementia Scale Blessed, Tomlinson, &
Assessment Roth, 1968
Clinical Dementia Rating Scale Hughes, Berg, &
(CDR) Danzinger, 1982; Morris,
1993









Rationale for inclusion of measures in Phase 2: Neuropsychological intake

assessment. Measures were selected in order to ensure that participants fit the

inclusion/exclusion criteria for amnestic mild cognitive impairment, as outlined by

Petersen and colleagues (2001). These criteria require adequate assessment of memory

functioning, in order to establish that performance on a standardized assessment of new

learning and memory falls 1.5 standard deviations below expected for age and education.

Three measures of memory were chosen to assess verbal and non-verbal memory. The

HVLT-R immediate and delayed memory scores had the greatest weight in determining

memory impairment.

Additionally, criteria for amnestic MCI require intact functioning in non-memory

domains of cognition, thus all participants underwent assessment of attention, working

memory, language, and visuospatial skills in order to ensure unimpaired status in these

areas. Following standard clinical neuropsychological practice, the measures chosen for

this battery allow broad and diverse assessment of the relevant domains within a

reasonable time period. Additionally, these measures have been previously utilized in the

determination of amnestic MCI (Bozoki, et al., 2001; Petersen, et al., 2001), thus

providing appropriate method for sample comparison and generalization purposes.

Two individuals did not demonstrate intact cognitive functioning in the non-

memory domains during the Neuropsychological Intake Assessment. After consultation

with a clinical neuropsychologist, these individuals were provided with referrals for

further clinical assessment of cognitive functioning and dropped from the remainder of

the study. Both indicated appreciation for the referrals.









As noted above, under inclusion criteria, in order for participants to be considered

unimpaired, performance during the Neuropsychological Intake Assessment on all

cognitive measures in memory and non-memory domains was required to be within one

standard deviation of age and education adjusted norms.

Two mood measures were included to assess any contribution depression may have

on memory and attentional performance, and rule out depression as a cause for any

observed impairments. The Center for Epidemiological Studies Depression Scale (CES-

D) was selected in order to screen for clinical depression, while the Geriatric Depression

Scale was selected in order to screen for depressive symptoms unique to older adults.

Notably, telephone screening questions about psychiatric history adequately screened out

individuals diagnosed with clinical depression. None of the individuals in the MCI group

scored above the mild depression cut-off score of 16 on the CES-D.

In addition to the neuropsychological assessment of the participant, the

participant's research partner was asked to participate in a semi-structured interview of

cognitive symptoms, daily functioning, and activities of daily living. This interview took

approximately 15 minutes to complete after informed consent procedures.

Phase 3: Daily Cognitive Assessment Battery (DCAB)

This battery took participants 10 20 minutes to complete each day and consisted

of measures outlined in Table 3-3. Modifications to standard test administration were

necessary in order to facilitate in-home assessment. The participant's research partner

was trained to assist in the administration of this brief battery. Specifically, the research

partner was informed as to how to assist with setting the timer as well as directed as to

the administration of the Digit Span test and recording of responses. Additional









description of research partner orientation is outlined below. The Appendix provides a

copy of a sample Daily Cognitive Assessment Workbook.


Table 3-3. Measures for Daily Cognitive Assessment (DCAB).
Domain of Function Measure

Attention & Working Digit Span (Forward and Backward)
Memory subtest of WMS-III
Attention & Processing Symbol Digit Substitution and Number
Speed Copy (both timed)
Immediate & Delayed List Learning (Rey Auditory Verbal
Recall Memory & Learning Test modified for visual
Learning presentation; three learning trials plus
delayed recall trial)
Sleep Sleep Diary (Sleep Time & Efficiency)
Mood Positive and Negative Affect Scale

Internal and External Environmental Distractions Questions
Environment Stress Ladder


Source

Wechsler, 1997

Smith, 1982

Rey, 1964



Lichstein, 1999
Watson, Clark &
Tellegen, 1988
Allaire, 2001


Rationale for inclusion of measures. The domains measured and specific tasks

within each domain were selected to provide appropriate assessment for the aims of the

study. While level of performance on attentional and working memory measures is often

similar across cognitively intact older adults and those diagnosed with mild cognitive

impairment, it is not clear that intraindividual variability is similarly comparable.

Measures in these domains were chosen to investigate the susceptibility of attention and

working memory to changes in intraindividual variability. Deficits in new learning and

explicit memory are indicative of an amnestic mild cognitive impairment and may be a

precursor to a neurocognitive illness such as Alzheimer's disease (Morris et al., 2001;

Petersen et al., 2001), thus measurement of this domain provided an evaluation of the

correlation and inter-relatedness of level of performance and intraindividual variability.









Negative affect has been shown to negatively impact performance on attentional

and memory measures, thus, assessment of daily positive and negative affect was

included to control for the influence of fluctuations in mood on cognitive performance.

Increased age is associated with poor sleep and increased complaints about sleep

quality (Bliwise et al., 1992; Ganguli, Reynolds, & Gilby, 1996). Sleep complaints and

insomnia have not been consistently shown to be associated with cognitive performance,

although both are highly correlated with psychological factors (McCrae, et al., in press).

Sleep time and sleep efficiency (sleep time/time in bed) were assessed via Sleep Diaries

(Lichstein et al., 1999) for each night's sleep during the Daily Cognitive Assessment

protocol in order to investigate the impact of sleep patterns on cognitive functioning.

Questions regarding the internal and external test-taking environment allowed for

measurement of discomfort (pain, stress, tiredness) as well as distractions (interruptions,

noise level, other people around). The specific questions were based on those used in a

previous daily battery study (Allaire, 2001).

Alternative forms for DCAB. Sixteen comparable alternative forms of the

cognitive measures were used, fifteen for the Daily Cognitive Assessment Battery, and

one for the initial orientation session. Thirteen alternative AVLT lists have been

published (Schmidt, 1996), and the three additional lists required for the current study

were designed with words with similar frequency and imageability. The alternative

forms of Digit Span and Symbol Digit were developed using simple random substitutions

of existing stimuli. This carefully matched, algorithm based process was utilized to

ensure comparability across forms. The fifteen alternative forms devised for the at-home

study were given twice to each participant, in random order, counterbalanced across









participants in four pre-specified random patterns, in order to randomize any effects of

learning or order across participants.

Parallelism of alternative forms. In order to determine whether the 16 alternative

versions of each cognitive test were parallel, the means for all sixteen versions, averaged

over all occasions and participants, were calculated for each cognitive test. Next, a

repeated measures analysis of variance (RMANOVA) with one within-subjects factor,

alternate form (i.e., workbooks 0 15) was conducted for each of the cognitive tests to

determine if mean performance varied as a function of alternate form. Results from this

analysis indicated that mean performance, averaged across the 31 occasions of

measurement significantly varied for the alternative forms of the AVLT List 1, F(15, 51)

= 7.05, p = .000, AVLT Total Score F(15, 51) = 8.56, p = .000, Backward Digit Span,

F(15, 48) = 7.59, p = .000, and Symbol Digit, F (15, 51) = 7.64, p = .000. Thus, for all

the measures the alternative forms demonstrated significant differences across forms.

Post-hoc contrasts revealed idiosyncratic differences between workbook pairs, with

different patterns by measure. Since the specific workbook contrasts were not

theoretically driven, and since the order of workbooks was randomized across subjects,

specific statistical comparisons between workbooks are not shown. Rather, the main

point of these ANOVAS is that the workbooks are not completely equivalent, and thus

may be a source of occasion-to-occasion variability that is common across participants.

However, as can be seen in Table 3-4, the means and standard deviations of the parallel

versions of each measure were highly comparable. In addition, an examination of the

correlational relationship among the different versions (Tables 3-5 through 3-8) shows









that participants' mean performance on the 16 alternative versions was strongly

interrelated for all cognitive measures (all correlations were significant, p = .000).



Table 3-4. Mean performance on the alternative forms of each cognitive task.


Form AVLT
List 1
Mean (SD)
0 7.87 (2.92)
1 8.80(3.06)
2 9.32 (2.82)
3 8.92 (2.47)
4 9.23 (2.90)
5 9.77 (2.73)
6 8.96 (2.90)
7 10.02 (2.96)
8 9.22 (2.83)
9 8.92(2.61)
10 9.75 (2.75)
11 9.52(2.79)
12 9.97 (2.89)
13 9.78 (2.69)
14 9.65 (2.85)
15 9.76(2.71)


AVLT
Total Recall
Mean (SD)
30.16(8.40)
32.38 (8.50)
33.71 (7.87)
32.66 (7.50)
33.58 (8.21)
35.08 (7.49)
32.74 (7.91)
35.66 (7.49)
33.36 (7.62)
32.85 (7.88)
35.30 (6.90)
34.64 (7.45)
35.07 (8.00)
34.65 (7.07)
35.11 (7.25)
34.92 (7.73)


Backward
Digit Span
Mean (SD)
7.31 (2.26)
9.33 (1.99)
9.05(2.11)
9.19(1.87)
9.35 (1.69)
9.43 (1.88)
8.88 (1.92)
9.20(1.79)
9.15(2.02)
9.09 (2.01)
9.05 (2.07)
9.40(1.76)
9.51 (1.68)
9.21 (1.84)
9.73 (1.75)
9.54 (1.92)


Symbol Digit

Mean (SD)
35.29 (12.11)
41.25 (9.50)
40.17 (10.92)
41.67 (9.52)
43.08 (10.68)
41.80 (9.15)
42.15 (12.17)
43.63 (9.13)
42.18 (9.47)
42.05 (9.32)
41.27 (9.82)
40.56 (8.75)
41.60 (9.31)
45.52 (11.25)
42.34 (9.85)
42.49 (10.23)










Table 3-5. Intercorrelations between mean scores on alternative versions of AVLT List 1.

Wkbk 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
1 .580 1
2 .581 .755 1
3 .558 .688 .661 1
4 .551 .741 .759 .827 1
5 .568 .705 .769 .644 .768 1
6 .594 .741 .811 .749 .783 .770 1
7 .565 .627 .769 .622 .766 .776 .817 1
8 .555 .611 .691 .649 .698 .706 .789 .739 1
9 .616 .750 .753 .665 .777 .741 .767 .746 .682 1
10 .588 .741 .744 .709 .804 .744 .751 .729 .768 .742 1
11 .395 .678 .615 .621 .753 .759 .762 .738 .687 .703 .666 1
12 .598 .618 .692 .685 .764 .759 .838 .795 .748 .730 .750 .763 1
13 .596 .721 .767 .756 .821 .794 .843 .772 .740 .770 .763 .769 .853 1
14 .622 .702 .768 .747 .772 .817 .791 .794 .765 .780 .730 .746 .756 .842 1
15 .689 .701 .749 .718 .757 .761 .833 .841 .736 .720 .740 .728 .839 .802 .813
Note. All correlations were significant,p = .000



Table 3-6. Intercorrelations between mean scores on alternative versions of AVLT Total Score.
Wkbk 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
1 .808 1
2 .823 .901 1
3 .748 .850 .861 1
4 .804 .892 .895 .904 1
5 .745 .857 .896 .836 .912 1
6 .809 .887 .903 .860 .917 .889 1
7 .748 .826 .884 .828 .891 .859 .894 1
8 .803 .860 .857 .828 .903 .854 .905 .807 1
9 .801 .880 .881 .880 .903 .856 .894 .853 .843 1
10 .776 .857 .848 .835 .914 .867 .886 .858 .875 .848 1
11 .665 .838 .822 .833 .883 .851 .872 .838 .817 .857 .790 1
12 .748 .830 .859 .851 .903 .843 .911 .878 .836 .876 .877 .846 1
13 .768 .867 .889 .871 .916 .880 .905 .863 .852 .887 .868 .845 .884 1
14 .773 .891 .902 .883 .920 .887 .898 .901 .880 .901 .862 .853 .880 .918 1
15 .797 .893 .904 .876 .919 .875 .932 .920 .867 .888 .870 .874 .921 .912 .942
Note. All correlations were significant,p = .000











Table 3-7. Intercorrelations between mean scores on alternative versions of Backward Digit
Span.
Wkbk 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
1 .541 1
2 .619 .698 1
3 .656 .735 .719 1
4 .553 .700 .699 .797 1
5 .554 .654 .730 .726 .647 1
6 .600 .714 .722 .764 .771 .709 1
7 .696 .796 .805 .840 .793 .782 .791 1
8 .507 .752 .700 .744 .697 .678 .747 .746 1
9 .713 .766 .763 .814 .789 .701 .738 .754 .768 1
10 .701 .791 .705 .852 .766 .684 .765 .816 .713 .826 1
11 .561 .735 .682 .720 .661 .699 .714 .738 .699 .684 .718 1
12 .522 .626 .673 .697 .664 .691 .614 .721 .662 .653 .685 .748 1
13 .623 .724 .706 .789 .763 .751 .784 .790 .742 .753 .771 .759 .750 1
14 .590 .708 .696 .812 .804 .731 .753 .795 .722 .781 .798 .722 .738 .778 1
15 .596 .762 .727 .769 .793 .751 .761 .831 .715 .760 .803 .743 .676 .762 .746
Note. All correlations were significant,p = .000



Table 3-8. Intercorrelations between mean scores on alternative versions of Symbol Digit.
Wkbk 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
1 .498 1
2 .521 .907 1
3 .506 .821 .825 1
4 .520 .865 .857 .816 1
5 .459 .783 .788 .880 .812 1
6 .398 .727 .707 .663 .726 .696 1
7 .547 .805 .851 .83 .84 .86 .73 1
8 .571 .780 .815 .829 .790 .847 .685 .866 1
9 .460 .774 .831 .817 .821 .775 .706 .809 .781 1
10 .500 .805 .798 .810 .755 .802 .704 .857 .823 .817 1
11 .508 .761 .810 .868 .751 .875 .641 .881 .798 .757 .822 1
12 .485 .773 .807 .843 .787 .877 .700 .904 .825 .809 .846 .880 1
13 .495 .664 .676 .649 .614 .659 .590 .715 .730 .684 .640 .668 .686 1
14 .474 .886 .854 .855 .885 .808 .724 .856 .801 .810 .797 .838 .835 .697 1
15 .535 .747 .791 .836 .748 .839 .693 .867 .858 .800 .884 .823 .865 .677 .786
Note. All correlations were significant,p = .000










Procedure

Overview of Study Phases

As can be seen in the flow chart, participant assessment occurred in three phases

(Figure 3-1). In Phase 1 and 2, "Screening and Group Assignment", potential

participants were screened and then assessed for amnestic MCI. This screening occurred

as a two-part process, an initial telephone screening and an in-person, individually

administered neuropsychological assessment, conducted in order to provide a formal,

standardized measure of MCI categorization. The two-part screening process provided

an opportunity to assess cognitive status initially via telephone, in order to screen for

inappropriate potential participants prior to the time-intensive neuropsychological

assessment. Following the neuropsychological assessment, group assignment was

confirmed via the consensus conference procedure described below. All cognitive

assessment took place in the laboratory following consent procedures as approved by the

University of Florida Institutional Review Board.


Phase 1 and Phase 2
Telephone Screening and Group Assignment
No MCI
Recruitment & Neuropsychological N MC
Telephone Intake Assessment
Screening MCI


Phase 3
Orientation Training and Daily Protocol
Research Partner or 30 Occasion Daily
No MCI Caregiver Training Cognitive
& Supervised Assessment
MCI Daily Cognitive Protocol
Assessment


Figure 3-1. Design of the current study.









In Phase 3, the Daily Cognitive Assessment, participants and their research

partners were asked to attend an orientation session when they were provided with timers

and thirty assessment booklets containing the Daily Cognitive Assessment Battery

(DCAB; described above), and trained in their use. These sessions were generally held

one week after the Neuropsychological Intake Assessment. In a few cases, the

neurocognitive assessment and participant orientation sessions were held on the same

day. This orientation session took place in the laboratory, or at home (n = 3), after both

the participant and research partner were informed of the requirements of the daily

protocol and provided informed consent as approved by the University of Florida

Institutional Review Board. For the participants, this informed consent step was

analogous to the informed consent obtained prior to the Neuropsychological Intake

Assessment. This second tier of informed consent procedures was required in order to

adequately inform the research partner. Training included explanation of directions and

observation of administration, in order to ensure adequate standardization across

participant- research partner dyads. Thus, as part of the training, each research dyad was

observed completing a "sample" Daily Cognitive Assessment Workbook. Research

partners were asked to participate in the administration of the Digit Span task by reading

the digits aloud and recording the exact response. They were instructed to give all Digit

Span items (up to and including a span of seven digits) regardless of performance.

Research partners were instructed to read the numbers in the Digit Span sequences at a

rate of one number per second. This was modeled for the research partner and

participant. Research partners were corrected regarding their rate and pacing during the

laboratory setting. Dyads were instructed that one booklet should be completed each day,









at the same time of day, beginning the day after the orientation session. Additional

guidelines were that the workbooks were to be completed without assistance, and were to

be returned at the end of each week. Since workbooks were numbered and dated,

participants were instructed that, should they miss a day, they were to skip that workbook

and resume with the one that corresponded to the current date.

In addition to providing an opportunity to orient and teach the research participant

and research partner about the Daily Cognitive Assessment Workbook, the laboratory

session also provided a supervised standard for comparison with the "unsupervised"

home-based Workbooks completed for the Daily Cognitive Assessment protocol

(occasions 1 30).

Rationale for the research partner administration protocol

Past research (Allaire, 2001) indicates that cognitively intact older adults can

successfully self-administer a brief cognitive battery. However, an initial concern was

that participants with mild cognitive impairment might be unable to consistently do so.

Caregiver/partner assistance was thought to be a useful means to ensure protocol

standardization as well as to provide a build-in reminder to accomplish the task for

participants with memory impairment. For the healthy elders, spousal administration was

thought to ensure protocol fidelity and consistency in the social-psychological conditions

of test administration. Anecdotally, it appeared from observation of the dyad interactions

during the initial laboratory orientation session, that in at least two occasions, the

research partner of a cognitive intact participant may have been suffering from mild

cognitive impairment. Due to the lack of formalized assessment for the research partner,

this cannot be objectively demonstrated.









Notably, when an MCI participant and his or her cognitively intact spouse were

both enrolled as participants in the study it was often clear that there was a difference in

performance on the Daily Cognitive Assessment Battery tasks between the two. For all

dual participant dyads, both participants were directed not to compare performance and

not to view the study as a competition. They were instructed to complete their Daily

Assessment Workbooks in different locations and at different times of day if they felt any

concern about comparing performance.

Compliance monitoring

On a weekly basis, the workbooks were returned to the researchers in pre-paid, pre-

addressed envelopes. Each was reviewed to monitor compliance and completeness of

administration. Any weekly sets of workbooks with incomplete or incorrectly completed

tasks occurring on more than one occasion resulted in reminder phone calls. Queries

regarding administration were minor (e.g., on Digit Span, "Do I write down exactly what

my partner says?") and were answered by telephone. Non-compliance or incomplete

administration resulting in incompletion of the first block (10 occasions) resulted in

discontinuation. All participants received reminder phone calls during the first and

second weeks, which served as an opportunity to answer questions or address concerns

regarding the assessments.

Group Membership Assignment: Consensus Conference

As indicate above in the discussion of inclusion and exclusion criteria, group

placement was confirmed by means of a consensus conference. Participants were

classified as probably MCI or unimpaired (Non-MCI) by consensus. Each case was

presented to the consensus members (which consisted of a senior researcher in cognitive

aging, a clinical neuropsychologist, and three graduate level clinical neuropsychology









students, one at the doctoral level and two at the master's level). For each case, the

members of the conference were presented with masked neuropsychological data from

the Neuropsychological Intake Assessment, described above, as well as Clinical

Dementia rating Scale (CDR) score and a composite score representing subjective

complaints about memory. These variables were chosen in order to conform to the

Petersen criteria for amnestic MCI (Petersen et al., 2001). If all panel members agreed on

group placement, that group assignment was made. If at least one member disagreed, a

discussion was held regarding the criteria, and majority vote determined group

assignment.

Initial Data Preparation and Study Variables

Ceiling and Floor Considerations

Prior to presenting the analyses and results for the specific aims, issues related to

range restriction due to performance at floor (lowest score possible) or ceiling (highest

score possible) are described. This is necessary since reduction in range of scores, by

repeated performance at ceiling or floor levels, would limit the potential for variability in

performance. For cognitive tasks, performance at ceiling levels artificially restricts

learning over future trials. If variability is an indicator of learning, the forced absence of

learning would, in turn, reduce the relationship between variability and performance

gains for those individuals with the best performance. Similarly, for individuals with the

worst performance, inappropriately difficult tasks, with performance consistently at floor,

would mask any evidence that variability is an indicator of cognitive compromise.

In order to eliminate problems due to excessive ceiling or floor scores within an

individual participant's performance the following analyses were completed. Initially, an

investigation of frequencies of floor values for the cognitive variables revealed that the









incidence of floor and near-floor values was < 1% of all occasions for all variables. Floor,

or near-floor, values were defined as the lowest possible score plus one item for all

variables except AVLT Total Score, for which the sum of the floors for the three lists

(e.g., score of six) was considered the floor. Additionally, no individual participant

performed at or near floor more than twice over the thirty-one occasions. Thus, no

further assessment of floor values was conducted.

Similar investigation of frequencies of ceiling values for the cognitive variables

revealed that restricted ranges concerns were relevant for the AVLT Total Score, AVLT

List 1 Score, AVLT Delay Recall Score, and Digit Span Backward Score. Ceiling values

were considered to be the highest score attainable on the measure (AVLT Total Score

ceiling = 45; AVLT List 1 Score ceiling = 15; AVLT Delay Recall Score ceiling = 15;

Digit Span Backward Score ceiling = 12.)

AVLT Total Score, AVLT List 1 Score, and Digit Span Backward Score reached

ceiling levels on 3.5% (74/2108), 4.5% (94/2108), and 10% (216/2108) of occasions,

respectively. A small number of individual participants (n = 5, n = 6, and n = 5,

respectively) accounted for the majority of these ceiling performance values, as the

participant reached ceiling performance and remained at that level of performance for the

remainder of their sessions. All occasions after reaching ceiling performance were

removed from analyses for these participants. Since ceiling performance was reached

and maintained near or at the beginning of the third block of occasions (occasions 21 -

30) these participants do not contribute to Block 3 analyses for AVLT Total Score,

AVLT List 1 Score, or Digit Span Backward Score. Notably, these changes had









negligible affect on the mean scores or IRI calculations across all occasions and over the

three blocks.

Thirty-four percent (711/2108) of occasions were at ceiling performance for AVLT

Delay Recall. Thirty-two participants reached ceiling for this variable during the first

block (occasions 1 10) of sessions and were removed from the AVLT Delay Recall

analyses. Two participants dropped from the AVLT Delay Recall mean score and IRI

score calculations were in the MCI group, while the remainder were in the Non-MCI

group. Notably, the loss of 30 participants from the cognitive intact group had a

significant effect on the overall IRI calculations for that group (e.g., original Mean IRI

for Non-MCI = 3.07 (SD = 1.46), Ceiling Corrected Mean IRI for Non-MCI = 4.07 (SD

= 0.92)). This is noteworthy since the IRI for the MCI group did not change (Mean IRI =

3.72, SD = 0.72). There was however, no change to the significance of the differences, as

the IRI score across the two groups remained non significantly different.

Regarding other cognitive variables, Symbol Digit Score was at ceiling for <0.1%

of occasions. Notably, AVLT Percent Retained calculations resulted in a restricted range,

although values > 100% retention were allowed. No modifications were made to this

variable, since the potential range restrictions were inherent to the calculations.

Standardization of Scores from the Daily Assessment Battery

Measures for the Daily Assessment Battery were modified or newly created for this

study, and, as such have no published norms available. However, standardization of

scores is useful to facilitate comparisons on the same scale. Therefore, we standardized

the scores from the Daily Assessment Battery to a T-score metric, using the mean and

standard deviations from this sample. Thus, prior to the calculation of intraindividual

reliability indices for the cognitive and non-cognitive measures in the Daily Assessment









Battery, all measures were standardized to T-score metric, with mean = 50, standard

deviation = 10. For all daily cognitive measures (memory and non-memory), as well as

daily mood measures, standardization was based on the orientation and training in-

laboratory session across all participants (including both MCI and Non-MCI

participants). Thus, the mean and standard deviation at the first, monitored, session were

set to 50 and 10, respectively, for all participants, and were allowed to vary at subsequent

occasions within the standardized metric of that first session. The additional daily non-

cognitive measures (sleep and distracting environmental variables) were standardized to

the first at-home occasion.

Intraindividual Variability Indices

Calculation of an intraindividual variability index was required in order to

perform the relevant statistical analyses for the hypotheses connected with each aim.

Results from Allaire's (2001) work indicates that the intraindividual standard deviation

index (ISD; Hultsch, 2001) is not effective if the construct studied involves growth or

learning. An alternative, the intraindividual residual index (IRI), represents the average

amount of variability in an individual's performance around a best fitting regression line,

and can be used to describe variability when growth or learning is present or absent

(Allaire, 2001). The IRI was calculated by first estimating a regression line

(incorporating linear and quadratic time trends to reflect growth for each participant) for

each participant's performance over the 31 occasions for each measure, and obtaining, for

each subject, a residual between their actual data point and their estimated value from the

regression line. These residuals were squared, summed, and then divided by the number

of occasions, to obtain a mean squared residual across all 31 occasions. The square root

of this term (i.e., root mean square residual) was calculated in the last step, to create an









intraindividual residual index (IRI). One advantage of this index is that, by taking the

square root, it is expressed in the same metric/units as the test/measure that it represents.

This procedure was followed to calculate the intraindividual residual index for all

relevant measure in the Daily Cognitive Assessment Battery (e.g., AVLT List 1 Score,

AVLT Total Score, AVLT Delay Recall Score, AVLT Percent Retained, Backward Digit

Span, Symbol Digit Score, PANAS Positive Affect, PANAS Negative Affect, Distracting

Environment Factor 1: Discomfort, Distracting Environment Factor 2: Distractions, Total

Sleep Time, and Sleep Efficiency). Table 3-9 shows the correlations of the mean IRIs for

the cognitive variables with the main demographics of age, education, and sex.

Intraindividual variability on AVLT Percent Retained and Backward Digit Span was

positively correlated with age, meaning that for these variables, older individuals were

more variable in performance from day-to-day. Greater intraindividual variability on

AVLT Percent Retained was also significantly correlated with male gender.

Table 3-9. Correlations of demographics and mean IRI scores.
Age Education Sex
AVLT Total Score IRI .229 -.047 -.172
AVLT List 1 IRI -.170 .091 .216
AVLT Delay Recall IRI .128 .136 -.248
AVLT Percent Retained IRI .472** -.098 -.396**
Backward Digit Span IRI .368** .079 -.194
Symbol Digit Score IRI .077 -.053 .123














CHAPTER 4
RESULTS

Overview

This chapter focuses on the three major aims. First is a descriptive examination of

intraindividual variability in memory and non-memory domains, and its relationship to

Mild Cognitive Impairment (MCI) status. Secondly, there is a consideration of whether,

as some previous research has suggested, intraindividual variability is related to practice-

related gain (i.e., that those who experience more gain on a task show more variability

during the period of time in which they are improving). Since individuals with MCI are

presumed to have less ability to profit from practice, an embedded question is whether

this will produce diagnosis-related differences in intraindividual variability over time.

Finally, the chapter concludes with an examination of potential sources of intraindividual

variability in participants, with a particular focus on "coupled" daily intraindividual

variability.

Preliminary Analyses

Before the intraindividual variability analyses for the specific aims of the study, a

number of preliminary analyses and psychometric checks were conducted. For those data

arising from the Neuropsychological Intake Assessment, MCI and Non-MCI participants

were compared on the initial neuropsychological measures. These measures are not

utilized for the study aims, but were integral to the group assignments and provide a

description of the neurocognitive status of the participants upon entry into the study. A

small group of individuals who participated in the Neuropsychological Intake Assessment









did not continue to the Daily Cognitive Assessment Battery, so a brief attrition analysis is

presented.

With regards to preliminary analyses for the Daily Cognitive Battery, four specific

analyses are presented. First, as mentioned in the Methods chapter, performance during

the laboratory session was compared to performance during the initial at-home session, in

order to provide a basic quality control check. Secondly, mean performance over 31

occasions on the measures in the Daily Cognitive Assessment Battery were compared for

the two groups based on cognitive status in order to confirm expected performance level

differences. Thirdly, the AVLT delay time (in minutes) between the third presentation of

the word list and the delayed recall of the list was investigated in order to determine if

variable delay times were related to performance. Finally, data reduction in the form of

factor analysis was conducted for the environmental distraction and discomfort questions.

Neuropsychological Intake Assessment Data: Participant Neurocognitive Status and
Attrition Analysis.

After telephone screening with the TICS, 84 participants met criteria for entrance

into the study and were assessed with the full Neuropsychological Intake Assessment. Of

these participants, eight declined to continue to participate following the

Neuropsychological Assessment, leaving 76 participants who began the Daily Cognitive

Assessment Battery. Eight of these participants withdrew prior to completion of the first

block of daily at-home assessments (i.e., prior to completion of 10 days of assessments),

thus, these participants were discontinued. Of the remaining 68 participants, all except

three completed the full 30 days with fewer than three missing occasions. The three who

did not fully complete the at-home assessments completed Block 1 (n=3) and Block 2 (n









= 2) and were retained for analyses involving these blocks. Thus, for most analyses study

sample n = 68, but for analyses involving only Block 3 the sample n = 65.

Table 4-1 shows the mean performance on neuropsychological measures for the

participants who completed the study, by cognitive status. As expected, the participants

in the MCI group performed significantly worse than the cognitively intact individuals on

measures of immediate and delayed memory function. Unexpectedly, the MCI group

was also significantly worse on the Boston Naming Test.

Table 4-2 shows the mean scores for a selection of measures from the

neuropsychological assessment battery for the participants who dropped out before

completing at least 10 daily sessions compared with those participants who completed the

daily at-home assessments. Variances were not equivalent for the COWA score and the

MMSE score, resulting in reduced degrees of freedom for these t-tests. Overall, the

individuals who dropped out of the study after the neuropsychological assessment, or

during the first few occasions of daily assessment, demonstrated significantly worse

immediate and delayed memory that those individuals who remained in the study. The

participants who dropped out of the study performed poorer on cognitive tests overall and

endorsed more symptoms of depression.

Daily Cognitive Assessment Battery Data

Quality control check: Laboratory to home administration

Table 4-3 shows a comparison of performance during the laboratory session with

performance during the first session at home. As can be seen from the t-test and

significance values in Table 4-3, the only significant differences between the introductory

laboratory session and the first daily at-home session are noted in Backward Digit Span









Table 4-1. Mean performance on neuropsychological measures,


Measure MCI Non-MCI Significance Test
Participants Participants
n= 15 n= 53
M (SD) M (SD)
TICS Score 31.67 (3.64) 37.38 (2.61) t(65) = 6.809; .000
MMSE 27.55 (1.75) 29.05 (0.84) t(11.2) = 2.762; .018
Predicted IQ 110.69 (5.92) 114.47 (7.51) t(66) = 1.795; .077
BVMT Total Score 13.93 (5.60) 22.89 (6.09) t(66) = 5.113; .000
BVMT Delayed 38.60 (9.11) 55.32 (8.63) t(66) = 6.546; .000
Recall (T Score)
HVLT Total T Score 40.27 (7.66) 55.79 (7.18) t(66) = 7.288; .000
HVLT Delayed Recall 3.93 (2.25) 9.51 (1.82) t(66) = 9.955; .000
COWA Total 11.33 (2.13) 12.08 (2.19) t(66) = 1.165; .248
BNT Scale score 11.00(2.17) 13.09 (2.92) t(66) = 2.580; .012
Trails A 10.20 (2.15) 10.96 (2.92) t(66) = 0.941; .350
Trails B 10.93 (1.91) 11.06 (2.96) t(66) = 0.153; .879
CES-D 7.33 (4.88) 5.06 (4.78) t(66) = -1.621; .110
Note. TICS = Telephone Interview of Cognitive Status, MMSE = Mini-Mental Status
Examination, BVMT = Brief Visuospatial Memory Test, HVLT = Hopkins Verbal
Learning Test, COWA = Controlled Oral Word Association, BNT = Boston Naming
Test, CES-D = Center for Epidemiological Studies Depression Scale.

Table 4-2. Mean performance on neuropsychological measures, by attrition status.
Measure Dropouts Daily Significance Test
N= 16 Participants
N= 68
M (SD) M (SD)
TICS Score 33.00(4.65) 36.10 (3.72) t(81)= -2.854; .005
MMSE 27.18 (2.40) 28.73 (1.24) t(11.2) =-2.082; .061
Predicted IQ 112.55 (6.77) 113.64 (7.32) t(82) = -.542; .590
BVMT Total Score 16.81 (8.41) 20.91 (7.02) t(82) = -2.022; .046
BVMT Delayed 48.13 (12.56) 51.63 (11.13) t(82)= -1.107; .272
Recall (T Score)
HVLT Total T Score 44.56 (13.34) 52.37 (9.71) t(82) = -2.683; .009
HVLT Delayed Recall 5.81 (4.02) 8.28 (3.01) t(82) = -2.761; .007
COWA Total 11.06 (3.30) 11.91 (2.18) t(18) = -.981; .339
BNT Scale score 11.13 (3.32) 12.63 (2.89) t(82) = -1.824; .072
Trails A 11.00 (2.94) 10.79 (2.77) t(82) = .265; .792
Trails B 9.06 (2.02) 11.03 (2.75) t(81) = -2.692; .009
CES-D 9.19 (6.18) 5.56 (4.86) t(82) = 2.547; .013
Note. TICS = Telephone Interview of Cognitive Status, MMSE = Mini-Mental Status
Examination, BVMT = Brief Visuospatial Memory Test, HVLT = Hopkins Verbal
Learning Test, COWA = Controlled Oral Word Association, BNT = Boston Naming
Test, CES-D = Center for Epidemiological Studies Depression Scale.


bv cognitive status.









and Distracting Environmental Variables Factor 2: Distractions. These differences are

likely due to systematic observed differences between these occasions. During the

laboratory session, it was noted that the directions for Backward Digit Span were the

most often repeated by the project testers. Administration of this task during the

laboratory practice session was often interrupted and resumed in order to clarify

directions. Likely, familiarity with the task, better understanding of the directions, and

fewer interruptions during administration resulted in improved performance during the

first at-home administration. Regarding significant differences noted on the second

distracting environmental factor, this factor consists of questions about whether there are

distractions while the tasks are being completed, including whether or not there are others

around, noise, and interruptions (see below for factor loadings). Notably, during the

laboratory session participants responded positively to these questions due to the

presence of the examiner/instructor. The first at-home session likely involved fewer

distractions as a result. Performance on all other measures was equivalent across the two

sessions.

Group differences in mean performance over all occasions

Table 4-4 shows the mean levels of performance across all 31 occasions for the

memory and non-memory cognitive variables for the two groups based on cognitive

status. Mean values for the non-cognitive potential covariates are also depicted. For all

variables raw and standardized scores are presented. As indicated in the Methods, scores

were standardized to T score metric (mean = 50, SD = 10) in order that performance

comparisons across variables could be made using the same scale. Results of the

independent sample t tests for mean performance across groups appear in the same table.









Table 4-3. Comparison of supervised session with first at-home daily session.
Measure Laboratory First Daily Significance Test
Session Session t(df); p
Mean (SD) Mean (SD)

AVLT List 1 7.87 (2.92) 7.71 (2.73) t(134) = .334; .739
AVLT Total Score 30.16 (8.40) 30.43 (8.01) t(134) = -.188; .851
AVLT Delay Recall 10.78 (3.76) 10.46 (4.12) t(133) = .467; .641

AVLT Percent 89.28 (19.33) 82.31 (26.07) t(122) = 1.763; .080
Retained
Symbol Digit 35.29 (12.11) 36.53 (9.11) t(132) = -.666; .506
Backward Digit Span 7.31 (2.26) 8.17 (2.26) t(132) = -2.194; .030
PANAS Positive 2.43 (0.58) 2.39 (0.69) t(131) = .354; .724
PANAS Negative 0.48 (0.42) 0.39 (0.44) t(131) = 1.098; .274
Factor 1: Discomfort 47.85 (6.74) 49.82 (7.81) t(134) = -1.574; .118
Factor 2: Distractions 62.61 (9.29) 50.08 (6.97) t(124) = 8.895; .000
Total Sleep Time (min) 449.57 (82.17) 444.45 (97.11) t(131) = .328; .743
Sleep Efficiency 89.60 (10.62) 88.33 (14.55) t(131) = .576; .566
Note. Assumptions for Levene's test of equality of variance were not met for AVLT
Percent Retained and Distracting Environmental Variables Factor 2: Distractions. AVLT
= Rey Auditory Verbal Learning Test; PANAS = Positive and Negative Affect Scale.


Note that the t test results are equivalent for mean comparisons of raw scores or

standardized scores, thus only one appears in the table for each variable. Multiple

individual mean comparisons were performed, since different patterns were expected for

the different measures (e.g., the groups were expected to differ on mean performance on

memory measures but not on other, non-memory, cognitive measures, or on non-

cognitive measures). Bonferroni correction for multiple comparisons was not employed,

due to concerns about power. Presentation of the actual significance value in Table 4-4

allows for observation of the likelihood of significance had Bonferroni correction been

applied.

As expected, the two groups differed significantly on all memory variables, with

the Mild Cognitive Impairment (MCI) group performing worse on AVLT Total Score









(t(16.6)= 4.068; p = .001), AVLT List 1 (t(66)= 3.942; p= .000), AVLT Delay Recall

(t(15.32) = 3.744; p = .002), and AVLT Percent Retained (t(14.7) = 3.334; p = .005). It is

important to note that group assignments were not made on the basis of these AVLT

scores, but on the basis of independently measured HVLT-R scores assessed during the

Neuropsychological Assessment prior to the Daily Cognitive Battery phase. On the non-

memory cognitive variables, the groups differed only on Digit Span Backward Score

(t(17.5) = 2.169; p = .044), with the MCI group again performing worse overall. There

were no significant differences between the groups on the non-cognitive measures.

Assumptions for Levene's test of homogeneity of variances were not met for AVLT

Total Score, AVLT Delay Recall, AVLT Percent Retained, Digit Span Backward Score,

and PANAS Negative Affect comparisons, resulting in adjusted degrees of freedom for

these mean comparisons. Even with Bonferroni adjustment, all the memory differences

between groups remained significant; in contrast, the Digit Span Backward was no longer

significantly different. Thus, the results generally confirm the categorization by

consensus conference, and indicate that the two groups differ significantly in level of

performance, specifically on memory measures. The distribution for all cognitive

variable means approximated normal distributions, with the exception of the AVLT

Percent Retained, which had a kurtosis estimate > 121.

As indicated in the Methods section, a small number of participants reached the

ceiling level of performance for four of the variables (AVLT Total Score, AVLT List 1,

AVLT Delay Recall, and Backward Digit Span). Table 4-4 clearly labels these variables

as ceiling corrected, however, for all subsequent tabular presentations, these variables are

no longer identified as "ceiling corrected", although the corrected values were used for all

analyses.












Table 4-4. Means (Standard Deviations) for all measures by cognitive status.


Measure MCI n=15 Non-MCI n=53 Significance Test
Mean (SD) Mean (SD) t(df); p
Memory Measures


AVLT Total Score Ceiling Corrected
Raw 26.39 (8.58)
Standard 46.45 (9.40)
AVLT List One Ceiling Corrected
Raw 7.51(2.37)
Standard 49.26 (7.83)
AVLT Delay Recall Ceiling Corrected
Raw 6.90 (3.80)
Standard 41.32 (9.15)
AVLT Percent Retained
Raw 71.62(25.69)
Standard 43.13 (11.56)


35.80 (4.87)
56.76 (5.34)

9.76 (1.82)
56.69 (6.02)

11.10 (1.86)
51.45 (4.48)

94.01 (7.59)
53.20 (3.41)


t(16.6) = 4.068; .001


t(66)= 3.942; .000


t(15.32) = 3.744; .002


t(14.7) = 3.334; .005


Non-Memory Cognitive Measures


Digit Span Backward Score Ceiling Corrected
Raw 8.28 (1.90)
Standard 54.59 (8.40)


Symbol Digit Score
Raw
Standard


40.48 (10.01)
55.05 (8.03)


9.41 (1.24)
59.57 (5.48)

42.21 (7.17)
56.44 (5.75)


t(17.5)= 2.169; .044


t(18.3) = 0.626; .539


PANAS: Positive Affect
Raw
Standard
PANAS: Negative Affect
Raw
Standard
Total Sleep Time Minutes
Raw
Standard
Sleep Efficiency
Raw
Standard
Factor 1: Discomfort
Factor 2: Distractions


Non-Cognitive Measures


2.26 (0.61)
46.44 (10.73)

0.60 (0.58)
51.59 (12.50)

440.02 (35.86)
49.09 (3.61)

87.88 (6.02)
49.58 (4.21)
51.07 (5.96)
48.33 (3.46)


2.27 (0.68)
46.68 (12.00)

0.33 (0.31)
45.83 (6.74)

441.98 (64.13)
49.28 (6.45)

88.63 (7.41)
50.10 (5.17)
49.31 (5.83)
50.44 (4.97)


t(66) = 0.072; .943


t(16.4)= -1.715; .105


t(66) = 0.113; .910


t(66)
t(66)
t(66)


0.356; .723
-1.028; .308
1.534; .130


Positive and Negative


Note: AVLT = Rey Auditory Verbal Learning Task, PANAS
Affect Scale.









Effect of differing delay times for AVLT Delayed Recall

The Daily Cognitive Assessment Battery provided a specific structure and order

for participants to follow in completing the cognitive tasks each day. The list learning

task (modified AVLT) was the first task each day. Upon completion of the third list,

participants were directed to record the time. After completing the intervening tasks

(Number Copy, Symbol Digit, Digit Span, Sleep Diary, Mood report, and environmental

distractions questions), participants were prompted to record the time prior to attempting

the delay recall trial. Recorded delay times (from completion of List 3 to initiation of

Delay List) ranged from 0 minutes to 365 minutes. Mean, median, and mode delay time

for the sample and by cognitive status are outlined in Table 4-5. The difference between

the mean delay times for the two groups was non-significant, as the independent samples

t test result was: t(472) = -1.503;p = .134. Levene's test for equality of variances was

significant, resulting in a reduction in the degrees of freedom for the mean comparison

test.

Table 4-5. Mean AVLT delay time.
AVLT Delay time (in minutes)
Group
Mean ( SD ) Median Mode
Entire Sample (n = 68) 16.45 (15.73) 15.00 10.00

MCI (n = 15) 17.82 (21.44) 15.00 10.00

Non-MCI (n = 53) 16.10(13.88) 14.00 10.00


A linear mixed models analysis to determine if delay time is related to performance

revealed that the fixed effect of AVLT Delay Time was unrelated to the dependent

variable of AVLT Delay Recall, F(1, 63) = 3.54; p = .064; parameter estimate = -.086).

Similarly, linear mixed models analysis to determine if cognitive status interacted with









delay time to influence performance was non-significant for the interaction of AVLT

Delay Time and Cognitive Status, F(1, 60) = 0.057; p = .812, parameter estimate = -.026,

indicating that the two groups did not differ in the relationship between length of delay

and performance on the delay recall trial.

Distracting environmental variables: Data reduction

Participants were asked to use a Likert scale on a daily basis to report pain level,

stress level, tiredness level, noise level, and light level. Additionally, they were asked to

indicate the number of interruptions while completing the workbook and to note whether

or not other people were around while they were working. A principal axis factor

analysis with promax rotation revealed three distinct factors. Factor loadings can be seen

in Table 4-6. Pain level, stress level, and tiredness level loaded on the first factor

(explaining 27.77% of the variance, which was labeled "Discomfort". Noise level,

number of interruptions, and positive response to people being around loaded on the

second factor (explaining 20.40% of the variance), labeled "Distractions". Light level

loaded on a third factor (explaining 14.50% of variance). Since there was little to no

variation in response to light level (all responses were excellent or very good), and since

this indicator did not relate strongly to other measures, this item-specific factor was not

used in further analyses.

Table 4-6. Factor loading for distracting environment variables.
Discomfort Distractions Light Level
Stress Level .997
Tiredness Level .538
Pain Level .340
Noise Level .568
People Nearby .429
Number Interruptions .364
Light Level .319









Intraindividual Variability in Memory and Other Cognitive Domains

The initial analyses of mean performance over all daily occasions, above,

confirmed that the two groups differed significantly on level of performance on measures

of memory function. One focus of the current study was the degree to which the two

groups might differ on intraindividual variability in memory and non-memory (e.g.,

working memory, attention, and processing speed) domains. The inquiry is captured in

the first two aims.

Aim One and Aim Two

The first specific aim of the study was to investigate whether greater intraindividual

variability in memory functioning is seen in older adults with amnestic mild cognitive

impairment, compared to cognitively intact elders. The second specific aim, conceptually

tied closely with the first aim, was to investigate whether cognitive tasks measuring

neurocognitive domains that typically show relatively less impairment in level of

performance in mild cognitive impairment (e.g., attention, processing speed) nonetheless

reflect increases in intraindividual variability in older adults with mild cognitive

impairment relative to unimpaired elders. Aim two also considered the extent to which

impaired and non-impaired groups could be differentiated on the basis of their within-

person variability in non-memory measures. The underlying question was whether, even

before the emergence of more general group differences in level of performance, group

differences in performance variability might serve as an "early warning indicator" of

impending general cognitive compromise. As part of this aim, we sought to investigate

whether intraindividual variability might independently contribute to the prediction of

cognitive status group membership.









Aim One and Aim Two: Review of Analyses

Intraindividual variability differences across groups, based on cognitive status

Aim one: Memory intraindividual variability. The variables of interest

included the Intraindividual Residual Indices (IRI) for initial verbal encoding (AVLT List

1), total recall (AVLT Total Score = AVLT List 1 + List 2 + List 3), long delay recall

(AVLT Delay Recall), and percent retained over delay (AVLT Percent Retained = AVLT

Delay Recall/ Higher of AVLT List 2 or List 3). Recall from the description of the IRI

calculation in the Methods chapter that the Intraindividual Residual Index, or IRI, is the

average amount of variability in an individual participant's performance around a best

fitting regression line over the days of measurement. In order to investigate differences

in intraindividual variability on memory performance between the two groups over the

thirty-one occasions of measurement, multiple independent samples t-tests were utilized

to compare the mean IRI values across groups based on cognitive status.

Mean and standard deviations for these IRI values by cognitive status are shown

in Table 4-7. Results of the independent sample t tests for the IRIs of the memory

measures appear in the same table. For AVLT List 1 (t(66) = 2.124, p = .037) the Non-

MCI group demonstrated greater intraindividual variability, while for AVLT Percent

Retained, the MCI group demonstrated greater intraindividual variability in performance

over occasions (t(66) = -3.277, p =.002). There were no significant differences in

intraindividual variability between groups on the other memory variables.

Aim two: Non-memory cognitive intraindividual variability. For the

investigation of intraindividual variability across the two groups on non-memory

cognitive measures, the variables of interest included the intraindividual variability

estimates for the non-memory cognitive measures assessed during the Daily Cognitive









Assessment Battery. These are: Digit Span Backward Score and Symbol Digit Score.

As with aim one, in order to investigate differences in intraindividual variability on non-

memory cognitive performance between the two groups over the thirty-one occasions of

measurement, multiple independent samples t-tests were utilized to compare the mean

IRI values across groups based on cognitive status.

Mean and standard deviations for the IRI of the non-memory cognitive variables

by cognitive status are shown in Table 4-7. Results of the independent sample t-tests

appear in the same table. There were no significant differences between groups in

variability on the non-memory cognitive variables. Additionally, for sake of

completeness, the IRIs for the non-cognitive measures appear in Table 4-7. There were

no significant differences between the two groups on the non-cognitive variables

assessing positive and negative affect, sleep time and efficiency, or those assessing

environmental distraction and discomfort.

For both sets of analyses, assessing the intraindividual variability differences in

the memory variables (focus of aim one) as well as the non-memory cognitive variables

(aim two) across cognitive status, multiple mean comparisons were made without use of

Bonferroni correction for family-wise error. Exact significance values appear in Table 4-

7 in order to assist with determining whether these IRI differences would continue to be

highly significant had Bonferroni corrections been used. Were Bonferroni corrections to

have been done, only the AVLT Percent Retained group difference would have been

judged to be significantly different from zero. The distributions for all cognitive variable

IRIs approximated normal distributions, with the exception of Symbol Digit IRI, which

has skewness and kurtosis estimates > 121.










Table 4-7. Mean (Standard Deviation) Intraindividual Residual Indices (IRIs).


Measure Total Group MCI Non-MCI Significance Test
n=15 n=53
M(SD) M(SD) M(SD) t(df); p
Memory Measures


AVLT Total Score

AVLT List 1

AVLT Delay Recalla

AVLT Per. Retained



Digit Span Backward

Symbol Digit Score



PANAS: Positive

PANAS: Negative

Total Sleep Time

Sleep Efficiency

Factor 1: Discomfort

Factor 2: Distractions


3.21 (0.76) 3.22 (0.69) 3.20 (0.79)

4.85(1.08) 4.34(1.11) 5.00(1.04)

3.94(0.86) 3.72 (0.72) 4.07 (0.92)

4.61 (2.44) 6.31(2.77) 4.13(2.12)

Non-Memory Cognitive Measures

4.40 (0.92) 4.50 (0.79) 4.37 (0.95)

3.44 (2.08) 2.98 (0.76) 3.57 (2.31)

Non-Cognitive Measures

5.56 (2.61) 5.22 (2.69) 5.65 (2.61)

5.31 (2.98) 6.05 (3.47) 5.10(2.84)

5.19(2.17) 4.60(2.21) 5.35 (2.15)

4.42 (2.90) 3.81 (2.59) 4.60 (2.98)

3.74(1.66) 3.69 (1.82) 3.76(1.63)

5.59 (3.08) 4.61 (1.92) 5.86 (3.30)


Note. AVLT = Rey Auditory Verbal Learning Task, PANAS = Positive and Negative Affect
Scale. a Due to ceiling corrections, AVLT Delay Recall n = 36, 13, and 23 for the three samples.


Data check: Reliability of intraindividual variability estimate

A question that emerged, given the fairly minimal group differences in cognitive

intraindividual variability, was whether unreliability of the IRI index might be

responsible for the absence of group differences, since day-to-day fluctuations in

performance on an instrument could be a reflection of random "noise" or measurement


t(66)

t(66)

t(34)

t(66)


t(66)

t(66)



t(66)

t(63):

t(66)

t(66)

t(66)

t(66)


--.076; .940

S2.124; .037

= 1.161; .254

=-3.277;.002



--.502; .617

S.960; .341



=.555; .581

-1.059; .294

= 1.194; .237

S.925; .358

S.148; .883

= 1.396; .167









imprecision. However, previously, Allaire (2001) reported that measures of test-retest

stability of the IRI were uniformly high in his sample. The variability for a particular

measure in one block of occasions was generally positively and significantly related to

variability in adjacent blocks of occasions. We employed the same approach in the

present study.

In order to confirm that the cognitive intraindividual variability estimates,

calculated via the Intraindividual Residual Index (IRI), represented consistent, trait-like

properties of the participants, a correlational analysis was performed over the three time-

ordered blocks (i.e., Occasions 0-10, or Block 1; Occasions 11-20, or Block 2; and

Occasions 21-30, or Block 3). If intraindividual variability is trait-like, and not just

"noise", the IRIs would be strongly and positively interrelated across blocks. The by-

block relationships among the IRIs for each of the cognitive measures are shown in Table

4-8. For the memory variables, AVLT Total Score, AVLT List One, and AVLT Percent

Retained show significant and positively correlated IRIs across the blocks, although the

correlations are generally fairly modest (less than 0.40). Higher correlations (greater than

0.70) between blocks were observed for the AVLT Percent Retained measure. AVLT

Delay Recall IRI demonstrated a different pattern, with few significant correlations.

Among the non-memory cognitive variables, the Symbol Digit Score IRI correlations

were positive and significant across all three blocks (again of relatively modest

magnitude). Backward Digit Span IRI correlations across blocks were smaller, although

still positive. As will be considered in the Discussion section, a question is whether

range restriction in some measures constrained reliability.









The findings also replicate those reported by Allaire (2001) in another way, in that

there is a "quasi-simplex" pattern to the obtained test-retest stabilities of the IRI measure.

That is, adjacent blocks (e.g., Blocks 1 and 2, Blocks 2 and 3) tend to be more highly

correlated than distal blocks (i.e., Blocks 1 and 3). As is considered in the Discussion,

this may be an index of qualitative transformations in the meaning of intraindividual

variability over the course of practice.

Table 4-8. Covariation among Intraindividual Variability Indices over blocks.

Variable Block Block 1 Block 2 Block 3
(IRI) (IRI) (IRI)

AVLT Total Score Block 1 (IRI) 1.00
Block 2 (IRI) .343** 1.00
Block 3 (IRI) .216 .384** 1.00
AVLT List One Block 1 (IRI) 1.00
Block 2 (IRI) .218 1.00
Block 3 (IRI) .287* .249* 1.00
AVLT Delay Recall Block 1 (IRI) 1.00
Block 2 (IRI) .366* 1.00
Block 3 (IRI) -.020 .214 1.00
AVLT Percent Retained Block 1 (IRI) 1.00
Block 2 (IRI) .736** 1.00
Block 3 (IRI) .636** .514** 1.00
Backward Digit Span Block 1 (IRI) 1.00
Block 2 (IRI) .278* 1.00
Block 3 (IRI) .227 .178 1.00
Symbol Digit Score Block 1 (IRI) 1.00
Block 2 (IRI) .410** 1.00
Block 3 (IRI) .303* .468** 1.00
*=p <.05; ** =p <.01
Note: AVLT = Rey Auditory Verbal Learning Task; IRI = Intraindividual Residual
Index; Block 1 = occasions 0-10; Block 2 = occasions 11-20; Block 3 = occasions 21-30.


Predicting cognitive status with intraindividual variability

As indicated, a follow-up question, based on an expected finding of differences in

intraindividual variability across groups on the non-memory measures, had been initially









posed. This question asked whether there would be a unique contribution of the non-

memory cognitive variable IRI differences in predicting group membership. Notably, the

IRI differences on the non-memory cognitive scores were not significant, so group

membership was not predicted with solely the non-cognitive variables.

Rather, in order to investigate whether memory and non-memory cognitive IRIs

added to the predictive ability of the mean performance on memory and non-memory

measures, discriminant function analysis was utilized. Table 4-9 shows the models used

in the analyses and the canonical loadings and classification rates for each of the models.

A four step approach was used. First, only the expected group distinction variables

(i.e., memory level, AVLT Total Score and AVLT Percent Retained) were used to predict

group (MCI, Non-MCI) membership. It is important to note that these AVLT scores

represent the average AVLT performance from the daily assessments, and are

independent of the memory measures administered at pretest (and used as part of the

consensus conference). Eighty-five percent (58 out of 68) of the participants were

correctly classified. The canonical loadings (Model One) showed that the two variables

were equally important in making assignments (AVLT Total Score loading = .49; AVLT

Percent Retained loading = .58).

In step two, the model examined whether intraindividual variability might add

additional group distinction information beyond that of level of performance in verbal

memory. Thus, the next analysis examined the added benefit of the IRIs for the two

memory variables. Results (Table 4-9, Model 2) showed that this model correctly

classified one fewer participant (84%), as a Non-MCI participant was classified in the









MCI group; in other words, memory variability was not helpful in further distinguishing

the groups.

In step three, we examined whether mean level differences in other non-memory

domains would aid in distinguishing between the MCI and non-MCI groups. Model 3

(Table 4-9) shows that the added benefit of adding the mean level of performance for

Backward Digit Span and Symbol Digit Score still does not exceed the classification

attained by the memory performance variable alone.

Table 4-9. Canonical loadings and classification statistics for discriminant function
models.
Model 1 Model 2 Model 3 Model 4
AVLT Total Score .49 .86 .88 .94
AVLT Percent Retained .58 .62 .58 .48
AVLT Total Score IRI .08 .13 .21
AVLT Percent Retained IRI .55 -.21 -.21
Digit Span Backward .09 .11
Symbol Digit Score .53 .53
Digit Span Backward IRI .14
Symbol Digit Score IRI .17
% Correctly Classified 85.3 83.8 85.3 83.8
% Sensitivity 60.0 60.0 60.0 53.3
% Specificity 92.5 90.6 92.5 92.5


In step four, we examined whether variability differences in these other non-

memory domains might aid in distinguishing between the MCI and Non-MCI groups.

Similar to the addition of the memory IRI values, as shown in Model 4 (Table 4-9) the

addition of the IRIs for Backward Digit Span and Symbol Digit Score reduce the correct

classification by one participant (in this case, one MCI participant was classified Non-









MCI). In summary, cognitive status was best predicted by performance on memory

measures alone.

Intraindividual Variability Over Time:

Understanding Variability and Performance Relationships

The results conveyed above demonstrated that over all thirty-one occasions of

measurement the two groups differed in overall performance in memory. With regards to

intraindividual variability, the two groups differed significantly only on percent of

information retained after a short delay (AVLT Percent Retained), with the MCI group

showing greater intraindividual variability overall. Group assignment was best predicted

by performance on memory measures, with day-to-day variability in performance adding

no additional predictive ability. The next main focus of the current study was to

determine the relationship between performance and intraindividual variability, the

details of which are outlined below.

Aim Three

The third specific aim of the study proposed to investigate the relationships

between cognitive status, performance over 31 days, and intraindividual variability in

daily performance. Specifically, the stated aim was to investigate whether, in

conjunction with expected differences in overall practice-related improvement across

sessions, memory-impaired and cognitively intact individuals show different patterns in

intraindividual variability on tasks over time. Extant research (Allaire & Marsiske, 2002;

Li et al., 2001) suggests that, on tasks in which cognitively intact elders show retest-

related learning, variability has several properties; (a) it tends to be positively associated

with level of performance, (b) it is uncorrelated with the variability observed after

individuals reach asymptotic performance, and (c) it is reduced over occasions.









Aim Three: Review of Analyses

Intraindividual variability over time (occasion), across cognitive status

Are there changes in intraindividual variability over time? The previously

reported results demonstrated that the two groups based on cognitive status did not differ

in intraindividual variability when averaged across the thirty-one occasions of

measurement, except on AVLT Percent Retained. However, those findings did not

address whether day-to-day variability changed over time, and if so, whether these

changes might differ by group. In other words, whether intraindividual variability during

earlier sessions was greater or less than intraindividual variability during later sessions,

and whether the groups differ in the pattern of variability across early or late sessions.

In order to demonstrate differences between the two groups with respect to

intraindividual variability changes over time, the thirty-one sessions were divided into

three equivalent blocks of sessions, labeled Block 1 (occasions 0 10), Block 2

(occasions 11 20), and Block 3 (occasions 21 30). Table 4-10 depicts the mean IRIs

for the cognitive variables for each of these three time-ordered blocks. A visual scan of

this table reveals that for most variables there appears to be a trend, over the blocks,

towards a reduction in intraindividual variability for each group. However, the

comparisons of interest for the means in Table 4-10 are not only whether within each

group there are trends across the blocks, but also whether the groups differ within each

block. Any potential interaction of group and block was also of interest. Thus, in order

to assess all of these comparisons, repeated measures analyses examining the presence of

cognitive status Group and Block effects on variability were conducted, as described

below and presented in Table 4-11.









Table 4-10. Mean Intraindividual Residual Indices (IRIs) by block and by cognitive
status.


Measure MCI Non-MCI
M(SD) M(SD)
Memory Measures

AVLT Total Score
Block 1 3.28 (0.90) 3.38 (0.95)
Block 2 3.33 (0.80) 3.14 (1.14)
Block 3 3.11 (0.87) 2.71 (1.16)

AVLT List 1
Block 1 4.33 (1.20) 5.06 (1.51)
Block 2 4.45 (1.56) 4.80 (1.37)
Block 3 4.22 (1.21) 4.63 (1.60)

AVLT Delay Recall
Block 1 4.25 (1.17) 3.96 (0.97)
Block 2 3.25 (1.09) 3.94 (1.21)
Block 3 3.44 (0.90) 4.03 (1.62)

AVLT Percent Retained
Block 1 7.00 (3.13) 4.46 (2.08)
Block 2 5.28 (3.14) 3.99 (2.68)
Block 3 5.43 (2.86) 3.24 (2.69)

Non-Memory Cognitive Measures

Digit Span Backward Score
Block 1 4.50 (0.91) 4.34 (1.25)
Block 2 4.46 (1.34) 4.57 (1.37)
Block 3 4.43 (1.48) 3.90 (1.29)

Symbol Digit Score
Block 1 2.99 (0.77) 4.11(4.10)
Block 2 3.14(1.24) 3.10(1.14)
Block 3 2.65 (1.18) 2.75 (0.86)



The blocks were utilized in multiple repeated measures ANOVAs, with Block IRI

(Block 1: sessions 0 10; Block 2: sessions 11 20; Block 3: sessions 21 30) as the

within subjects variable, and Group (Cognitive status) as the between subjects variable, in

order to evaluate whether intraindividual variability was reduced over time in the









cognitively intact but not MCI participants. Results indicated that main effects of Block

were significant for AVLT Total Score, AVLT Delayed Recall, AVLT Percent Retained,

and Symbol Digit Score, indicating a reduction in intraindividual variability over

occasions. The linear trend of the Block main effect was significant for only AVLT Total

Score [F(1,63) = 7.090, p = .010] and AVLT Percent Retained [F(1, 64) = 14.007, p =

.001]. No quadratic effects of Block were significant. Main effects for Group (Cognitive

Status) were significant for AVLT Percent Retained. As with the t tests above,

participants with MCI showed substantially more intraindividual variability on this

measure. Notably, the Block and Group interaction was not significant, except for AVLT

Delayed Recall, suggesting that the MCI and Non-MCI groups did not differ in the slope

of variability reduction over blocks, except for this one measure. For the AVLT Delay

Recall score, the Non-MCI group showed little reduction in variability over Blocks, while

the MCI group showed substantial variability reduction between Blocks 1 and 2. Table

4-11 shows the F-statistics, significance values, observed power, and effect sizes for

these repeated measures ANOVAs. As power is low for most main and interaction

effects, Figure 4-1 depicts the observed patterns of intraindividual variability across

blocks for the two groups on all variables. Note that the y-axes for the graphs are all on

the same scale (e.g., three units), even though they represent different ranges of units.

Table 4-11. Repeated Measures ANOVA: Intraindividual variability
predicted by block score and cognitive status.
AVLT Total Score
Source Df F p power r2
Between subjects
Group 1 .76 .388 .137 .012
Within subjects
Block 2 4.10 .019 .718 .061
Group x Block 2 .58 .560 .145 .009









Table 4-11. continued
AVLT List 1
Source df F p power r2
Between subjects
Group 1 2.53 .117 .347 .039
Within subjects
Block 2 .71 .493 .168 .011
Group x Block 2 .18 .834 .078 .003


AVLT Delay Recall
Source df F p power r2
Between subjects
Group 1 .88 .355 .149 .026
Within subjects
Block 2 4.28 .023 .705 .211
Group x Block 2 4.15 .025 .690 .206


AVLT Percent Retained
2
Source df F p power r2
Between subjects
Group 1 8.08 .006 .800 .112
Within subjects
Block 2 10.89 .000 .988 .257
Group x Block 2 1.73 .185 .351 .052


Backward Digit Span
Source df F p power q2
Between subjects
Group 1 .88 .352 .152 .014
Within subjects
Block 2 .44 .645 .120 .007
Group x Block 2 1.51 .225 .317 .024


Symbol Digit Score
Source df F p power q2
Between subjects
Group 1 .62 .433 .122 .010
Within subjects
Block 2 3.67 .031 .656 .104
Group x Block 2 .67 .514 .158 .021
Note. Power computed using alpha = .05.






















AVLT Total Score IRI by Cognitive Status


1 2
block


cog stat
- Non-MCI
- MCI


black
block


AVLT Delay Recall IRI by Cognitive Status


AVLT Percent Retained IRI by Cognitive Status


cog stat
- Non-MCI
- MCI


cog stat
- Non-MCI
- MCI


2 3
block


black
block


Figure 4-1. Intraindividual variability by blocks. Plots for (a) AVLT Total Score IRI, (b)

AVLT List 1 IRI, (c) AVLT Delay Recall IRI, (d) AVLT Percent Retained

IRI, (e) Backward Digit Span IRI, and (f) Symbol Digit Score IRI by

cognitive status.


cog stat
- Non-MCI
- MCI


AVLT List 1 IRI by Cognitive Status








70



Backwards Digit Span IRI by Cognitive Status
Symbol Digit IRI by Cognitive Status

cog stat 50 cog stat
Non-MCI Non-MCI
MCI MC,
X50- 45-

45- 40-

40- 35-

S35- 30-

-30- 25

25- 20-
1 2 3 1 2 3
block block


Figure 4-1. continued.


Relationship between intraindividual variability and level of performance

Subsequent to the above consideration of the patterns in intraindividual variability


over time and across groups, the next relationship of interest is that of level of


performance and day-to-day variability in performance. Several analyses were


undertaken in order to investigate the relationship between level of performance and


intraindividual variability for the participants as a whole as well as separately based on


cognitive status. Previous research, conducted primarily with reaction time measures,


suggests that poorer performers show more variability (i.e., the "variability =


vulnerability" perspective). Allaire and Marsiske's results (e.g., 2004), however, have


suggested that, in accuracy data, better performers are actually more variable, particularly


in the early "rapid acquisition" phases of a practice curve.


Is intraindividual variability positively or negatively associated with level of


performance? As outlined in Table 4-12, Pearson product-moment correlations between


the intraindividual residual index (calculated over all 31 occasions) and mean level of









performance over all 31 occasions for the cognitive variables revealed significant

correlations for AVLT Total, AVLT List 1, AVLT Percent Retained, and Backward Digit

Span Score. With the exception of AVLT List 1, these significant correlations were

negative, indicating that greater intraindividual variability was associated with lower

levels of performance on the AVLT Total Score, AVLT Percent Retained, and Backward

Digit Span. For AVLT List 1, the relationship was reversed, such that greater

intraindividual variability was associated with higher performance on the first trial of the

list learning task, a pattern identical to that reported by Allaire (2001) with the identical

measure.

Table 4-12. Correlations of mean level of performance and IRI.
Correlation
Mean over 31
Variable Occasions with IRI p
AVLT Total -.253 .037
AVLT List 1 .424 .000
AVLT Delay .004 .984
AVLT Per. Ret. -.727 .000
Back Digit Span Score -.432 .000
Symbol Digit Score -.185 .131

Does the relationship between intraindividual variability and mean level of

performance differ by cognitive status? Following the initial correlations of overall

means scores and IRIs for the cognitive variables, the next set of analyses examined

whether the mean-variability relationship was different for MCI and Non-MCI groups.

Thus, the relationship between variability, performance, and cognitive status was

investigated for the same variables via multiple univariate regressions taking the

following form: IRI = Mean Performance (over 31 occasions) + Cognitive Status Group

(1 = MCI, 0 = Non-MCI) + Mean x Group Interaction + error. All regression equations

were significant, except for Symbol Digit Score. Results in Table 4-13 show that both









main effects (Mean Score and Cognitive Status Group) as well as the interaction are

significant for AVLT Total Score, AVLT Delay Recall, AVLT Percent Retained and

Backward Digit Span, suggesting that level of performance was differentially related to

variability by cognitive status. For AVLT List 1 and Symbol Digit Score only the Mean

Score main effect was significant. The significant, and negative, main effects for Mean

level of performance again indicated that for AVLT Total Score, AVLT Delay Recall,

AVLT Percent Retained, Backward Digit Span, and Symbol Digit Score that the higher

initial performance was, the lower the intraindividual variability, and are (as expected)

identical to the results obtained in the correlational analysis above. The negative main

effects for Group indicate, unexpectedly given our hypotheses, that when level of

performance is controlled, the MCI group (Group = 1) shows less intraindividual

variability than the Non-MCI group (Group = 0).

Table 4-13. Regression coefficients for predicting IRI with mean level performance and
cognitive status group.
Variable B (Standardized Beta Weights)
F R2 Mean Group Mean x Group
AVLT Total Score 6.62** .237 -.48 ** -.43 ** .40 **
AVLT List 1 5.16** .195 .25 -.13 .10
AVLT Delay Recall 6.65** .384 -.58** -.62** .59**
AVLT Percent Retained 28.88** .575 -.60** -.32** .31 **
Backward Digit Span 7.18** .252 -.49** -.28* .27*
Symbol Digit Score 1.77 .077 -.25 -.18 .16
*p<.05; **p< .01

Notably, the interaction is significant for all variables where both main effects

(Mean, Group) were significant. This complicates the interpretation somewhat, since

with a significant interaction, AVLT Total Score, AVLT Delay Recall, AVLT Percent

Retained, and Backward Digit Span evince a pattern suggesting that individuals in the

Non-MCI group who perform at the lower overall levels of performance have higher









73




individual variability indices (consistent with theories on strategy acquisition and



practice-related intraindividual variability). In contrast, participants in the Non-MCI



group who are performing at the higher end of the range, have lower variability.



However, in the MCI group, higher performers have higher intraindividual variability,



while lower performers have less variability. These patterns are clearer in Figure 4-2,



which shows the scatterplots of the relationship between IRI and mean score for the two



cognitive status groups, with average best-fitting lines superimposed, for all the memory



and non-memory cognitive variables.


I I I I
1000 2000 3000 4000
AVLT Total Score MEAN


cog stat
0 Non-MCI
0 MCI


1200-


1000-



800-
--

I- 600-



400-



200-


0
8
5OO
0 00 0
6o
90 0

o oC0


oo
0
0




200 400 600 800 1000 1200 1400
AVLT List 1 MEAN


Figure 4-2. Mean performance and intraindividual variability. Plots for (a) AVLT Total

Score Mean and IRI, (b) AVLT List 1 Mean and IRI, (c) AVLT Delay Recall

Mean and IRI, (d) AVLT Percent Retained Mean and IRI, (e) Backward Digit

Span Mean and IRI, and (f) Symbol Digit Score Mean and IRI by cognitive

status.


1200-



1000-



- 800-
0
-


o 600-
I-



400-



200-


cog stat
0 Non-MCI
0 MCI








74



cog stat 12cog stat
O Non-MCI O2 Non-MCI
0 MCI 0 MCI

1000- 1000-
0
0

S800- 0- O O

o o\ \ O
S0 -0

0o

O OO 00- 8 0 o

200- O O
0 00 8 oO0




2000- O
0
CO
000 0


AVLT Delay Recall MEAN AVLT Percent Retained MEAN



cog stat
1200- oN MCI ONon-MCI
OMCI Q

000- 1000
00- 800-





0) 0
S600- O 600-
400 O.O 0 O O 0 O 0
.~t^o o os Q o
i O 0 O OO

l 400- 00 400- Oo O
0 o"o0 O 0 0 0 00 0
So o oo8 .r

o o o 00o o,
200 o 0 o 0



500 600 700 800 900 10 00 11 00 1200 1000 2000 3000 4000 5000 6000
Backward Digit Span MEAN Symbol Digit Score MEAN



Figure 4-2. continued.


Are there practice related gains in level of performance evidence across


occasions? Do these gains differ by group? Do improvements reach asymptotic


levels of performance? Does this differ by groups? Previously presented results


have shown intraindividual variability changes over time, via patterns present over the


three time-ordered blocks of occasions. Additionally, results describing the relationship


between intraindividual variability and performance have been presented. However, in
0 0 0 0 00 00 0 0 0s nleelofpefrmnc eiene crs

ocaios D heegan dferb gopD i00 00t eahasmtoi

leves ofperfrmane? Des tis dfferby goups Preiousy 0 0 'Q reulp

haveshon inraidivdualvarabiity hanes ver ime vi patern prsen08 00th

200 0 200- 0 00 _9ocains ddtonlyrsutsdsciin herlainsi









order to more clearly understand the interactions across the two groups between

intraindividual variability and performance, it is crucial, at this point, to investigate

whether there are practice-related gains in performance across the occasions of

measurement. If so, the changes in intraindividual variability might be more clearly

understood in the context of strategic acquisition and practice-related learning.

Additionally, since previous work has theorized that intraindividual variability is

positively associated with strategy acquisition through out performance gains, with a loss

of relationship when asymptotic performance is reached, it is important to determine if

performance over time reaches asymptote, and if so, if the groups differ in whether

performance gains reach a plateau. The most effective method to answer these questions

in one approach was via latent growth curves, estimated via simplified mixed model

analysis. Mixed effects modeling allows for a determination of whether variables are

related to each other in a fixed way for all participants (fixed effects), or in differing ways

for different participants (random effects). Thus, we can determine if Time (linear and

quadratic effects of occasion) influences performance, and if so, whether the influences

differ based on Group (cognitive status). With regard to the question of performance

gains, the effect of interest is whether, on average, a linear effect of Time was detected.

With regard to the question of whether, on average, individuals "leveled off', or reached

asymptotes, the critical effect of interest is whether, on average, a quadratic effect of

Time was detectable. In the models that follow, only the fixed (average) effects of Time

are considered. The results are analogous to those of a traditional repeated measures

analysis, but they have the advantage of using all available data, and not just data from

those participants with no missing data at all occasions (i.e., listwise deletion).









As seen in Table 4-14, results from the mixed effects analyses fell into the same

pattern for AVLT Total Score, AVLT List 1, AVLT Delay Recall, AVLT Percent

Retained and Backward Digit Span. The growth curves for all the cognitive variables can

be seen in Figure 4-3. For the noted variables, the linear and quadratic effects of Time

and the main effect of Group were significant. The interactions of Group with linear and

quadratic effects of time were not significant. This means that for these variables, while

the two groups differed on overall level of performance, the growth curves followed a

similar trajectory. This trajectory included both linear and quadratic growth, indicating

that, on average, performance gains were apparent, and performance asymptotes were

reached (the quadratic main effect). An asymptote reflects a leveling off of performance

gains, the presence of which allows for further analyses of the relationship of

intraindividual variability and slope of performance gains, below.

The remaining cognitive variable, Symbol Digit Score demonstrated a different

pattern of effects following the mixed model analyses. Symbol Digit Score reflects linear

and quadratic effects of time, but no main effect of Group, indicating that the growth

curves, while containing a performance asymptote, are nearly identical for the two

groups.

Table 4-14. Time effects on mean performance.
AVLT Total Score AVLT List 1
Fixed Effects Estimate F p Estimate F P
Intercept 33.14 1654.16 .000 8.55 940.81 .000
Workbook a 9.82 .000 a 5.06 .000
Time .20 277.96 .000 .09 219.29 .000
Time2 -.01 11.47 .001 .00 11.22 .001
Group -8.94 27.15 .000 -1.96 11.18 .001
Time x Group -.03 1.32 .251 -.02 2.33 .128
Time2 x Group .00 .01 .943 .00 .02 .877










Table 4-14. continued


AVLT Delay Recall
Fixed Effects Estimate F p
Intercept 12.17 1026.57
Workbook a 8.26


Time
Time2
Group
Time x Group
Time2 x Group


.06
.00
-5.00
.00
.00


121.99
16.20
39.91
.09
.86


.000
.000
.000
.000
.000
.764
.353


AVLT Percent Retained
Estimate F P
.922 2316.07
a 4.29
.00 16.95
.00 8.32
-.24 34.16
.00 1.55
.00 .28


Fixed Effects
Intercept
Workbook
Time
Time2
Group
Time x Group
Time2 x Group


Backward Digit Span
Estimate F p
8.85 1877.86
a 4.00
.05 144.11
.00 10.09
-1.16 7.39
.00 .32
.00 .33


Symbol Digit Score


.000
.000
.000
.002
.008
.572
.560


Estimate
37.13
a
.31
-.01
-2.20
.04
.00


1140.69
6.27
202.08
-5.71
.87
.72
.28


a. There are 15 separate workbook estimates for each cognitive variable.


4000-


S3500-
-


S3000-
-.


2500-


Growth Curves for AVLT Total Recall















S a Occasion ,
Occasion


cog stat
- Non-MCI
- MCI


11 00-


1000-
-
s
900-

S800-

700-

600-


Growth Curves for AVLT List One
















Occasion


Figure 4-3. Growth curves by cognitive status. Plots for (a) AVLT Total Score, (b)
AVLT List 1, (c) AVLT Delay Recall, (d) AVLT Percent Retained (e)
Backward Digit Span, and (f) Symbol Digit Score by cognitive status.


.000
.000
.000
.004
.000
.213
.594


.000
.000
.000
.017
.355
.396
.600


cog stat
- Non-MCI
- MCI


















Growth Curves for AVLT Delay Recall








"-I
















a Occasion 'V "c
Occasion


Growth Curves for Backward Digit Span


cog stat
- Non-MCI
- MCI


Growth Curves for AVLT Percent Retained


cog stat
Non-MCI
MCI


S/ / Occasion '
Occasion


Growth Curves for Symbol Digit Score


4600-
cog stat
- Non-MCI
- MCI
4400-


4200-
-


.u 4000-

0.
S3800-


3600-


3400-


Figure 4-3. continued








Is intraindividual variability related to slope of practice-related improvement



in performance? Is the amount of gain (linear slope) related to intraindividual



variability? Does this relationship differ by groups? From the previous analysis, it is



clear that there are distinctive trends in the slope of performance over time. The next



question to be answered is whether intraindividual variability is related to performance


1400-




12 00-
-



, 10 00-




800-




600-


cog stat
- Non-MCI
- MCI


Occasion


Occasion









gains. Table 4-15 illustrates the correlations between linear slope of performance gain

(i.e., the slope coefficient obtained when performance for each participant is regressed on

occasion) and intraindividual variability for all 31 occasions and separately for the three

time-ordered blocks. The linear slopes for the individual blocks were obtained by

conducting individual regression equations, separately for each participant, and

separately for each block. We were then able to save the linear slope estimates for each

participant, and correlate those linear slopes with the IRIs calculated for each participant

in the same block. Thus, for each participant, on each cognitive variable, we obtained

three IRI scores (Block 1, 2 and 3) and three linear slope estimates (Block 1, 2 and 3).

First, we examined the relationship between overall performance gains (i.e., linear

slope over all 31 occasions) and overall variability (i.e., IRI calculated over 31

occasions). Across all 31 sessions, performance gains (linear slope) and intraindividual

variability were significantly and positive correlated for AVLT Percent Retained. Thus,

greater gains were associated with greater intraindividual variability throughout the 31

occasions for this variable. For Symbol Digit Score the overall relationship between

performance gains and intraindividual variability was reversed, such that fewer gains

were associated with greater intraindividual variability. For none of the other cognitive

variables assessed (AVLT Total, AVLT List 1, AVLT Delay Recall or Backward Digit

Span) was the overall (31-occasion) relationship between variability and slope of

improvement significantly different from zero.

The analyses were then conducted separately for each of the three blocks of

occasions, to answer the question of whether slope and variability might be differentially

related at different points of the 31-occasion practice curve. The rationale for this









analysis emerged from the quadratic growth curves presented above. Since it appeared

that most of the gain occurred in the earlier sessions, with a leveling off of performance

gains during later occasions, if gain and variability are positively related, we might have

expected this relationship to be particularly strong during early Blocks.

For Symbol Digit Number Correct there was indeed a stronger relationship between

linear slope and variability in Block 1. However, as with the overall (31-occasion)

relationship reported above, this relationship was negative, such that more variability was

associated with smaller gain slopes. There was no relationship between gain and

variability for Blocks 2 or 3.

For most of the other measures, an inconsistent and "scattershot" pattern of

correlations was detectable. For the AVLT Total Score, there was a positive relationship

between gain (linear slope) and variability, but for Block 2 only. For AVLT List One, a

negative relationship was found between gain and variability, but only for Block 1; a

similar negative relationship was found for Backward Digit Span, but in Block 3 only.

In order to determine if the relationships between performance gain and

intraindividual variability remained constant for both groups, this was investigated via

multiple univariate regressions taking the form: IRI = Linear Slope (over 31 occasions) +

Cognitive Status Group (Non-MCI = 0, MCI = 1) + Slope x Group Interaction + error.

Given the relatively haphazard pattern of results by block in the preceding analyses, this

analysis examined only the overall (31 occasions) relationship between variability (IRI)

and gain (linear improvement slope). Results in Table 4-16 show that Slope main effect

was significant and negative for AVLT List 1, indicating that for this variable, the higher

the performance gain, the lower the intraindividual variability. This is consistent










Table 4-15. Between-person correlations between intraindividual variability and linear
slope of performance gains separately for each variable.

Variable Overall Block 1 Block 2 Block 3
IRI IRI IRI IRI
AVLT Total Score Overall Slope .073
Block 1 Slope -- -.086
Block 2 Slope -- .192 .322** --
Block 3 Slope -- .022 -.157 -.227
AVLT List One Overall Slope -.053 --
Block 1 Slope -- -.264* -
Block 2 Slope -- .047 .187
Block 3 Slope -- .052 .079 .094
AVLT Delay Recall Overall Slope .117
Block 1 Slope -- .131
Block 2 Slope -- .134 .152
Block 3 Slope --- .135 .252 .080
AVLT Percent Overall Slope .281* --
Retained
Block 1 Slope -- -.139
Block 2 Slope -- .176 .222
Block 3 Slope -- .092 .018 -.129
Backward Digit Span Overall Slope -.073
Block 1 Slope -- -.013
Block 2 Slope -- -.169 -.065
Block 3 Slope -- -.205 .024 -.248*
Symbol Digit Score Overall Slope -.319**
Block 1 Slope -- -.265* -
Block 2 Slope -- -.046 -.087 --
Block 3 Slope -- .001 -.162 -.014
*p<.05; ** p< .01

with the correlational pattern seen above in Block 1. AVLT Percent Retained

demonstrated a significantly positive main effect of Slope, which implies that the reverse

pattern held for Percent Retained, i.e., that greater performance gains were associated

with greater intraindividual variability. Again, this is consistent with the correlations

discussed above. Notably, AVLT Percent Retained evinced a significant interaction,

suggesting that the pattern is not the same in the two groups (i.e., that the positive

relationship between gain and variability is really only true for persons without MCI).







82


The only variable to demonstrate a significant main effect for Group was Symbol Digit


Score, such that persons with MCI demonstrated less intraindividual variability.


However, the Non-MCI group may have artificial raised intraindividual variability due to


three outliers (see graph, Figure 4-4). The patterns of interaction are clearer in Figure 4-


4, which shows the scatterplots of the relationship between IRI and linear slope for the


two cognitive status groups, with average best-fitting lines superimposed, for all the


memory and non-memory variables.


Table 4-16. Regression Coefficients for predicting IRI with linear slope gain in
performance and cognitive status group.
Variable B (Standardized Beta Weights)
F R2 Slope Group Slope x Group
AVLT Total Score .49 .023 -.07 .00 .13
AVLT List 1 1.62 .027 -.26* -.08 .03
AVLT Delay Recall .71 .063 -.18 -.09 .13
AVLT Percent Retained 6.45** .232 .30* .03 -.26*
Backward Digit Span .24 .011 .03 -.08 .03
Symbol Digit Score 2.81 .116 -.13 -.32** .06
*p<.05; **p< .01


cog stat cog stat
1200- Non-MCI 1200- NonMCI
0 MCI 0 MCI

1000- 1000-


800- 800-
o 0 0
u 0
00 o
0 000- 600- 0 0o 0 O

oo0 O O
400-O O O 0 000 -O


-400 -300 -200 -100 0 100 20 30 400 -300 -200 00 0 10 20 300



AVLT Total Score Linear Slope AVLT List 1 Linear Slope


Figure 4-4. Linear slope and intraindividual variability. Plots for (a) AVLT Total Score
Slope and IRI, (b) AVLT List 1 Slope and IRI, (c) AVLT Delay Recall Slope
200- 0 6 0 200-
0

-400 -300 -200 -100 000 100 200 300 -400 -300 -200 -100 000 100 200 300
AVLT Total Score Linear Slope AVLT List 1 Linear Slope


Figure 4-4. Linear slope and intraindividual variability. Plots for (a) AVLT Total Score
Slope and IRI, (b) AVLT List 1 Slope and IRI, (c) AVLT Delay Recall Slope
and IRI, (d) AVLT Percent Retained Slope and IRI, (e) Backward Digit Span
Slope and IRI, and (f) Symbol Digit Score Slope and IRI by cognitive status.






















cog stat
O Non-MCI
O MCI


1000-




S800-






I4
600-




< 400-




200-


-400 -200 000 200
AVLT Delay Recall Linear Slope


0
0
00 0 0
O0 0
0 00 000







0
0



-200 000 200 400 600 800
Backward Digit Span Linear Slope


cog stat
O Non-MCI
O MCI


1200-




1000-


a

S800-
0



4 600-




400-


1200-




1000-
-


800-
SO




S600-
a



400-




200-


-400 -200 000 200 400
Symbol Digit Score Linear Slope


Figure 4-4. continued








Sources of Intraindividual Variability



To this point we have described intraindividual variability alone and have




described patterns of relationships between intraindividual variability in performance,


0
00

I I I I I 1
-750 -500 -250 000 250 500
AVLT Percent Retained Linear Slope


0 0 0 0
0

0 00


o o o





0 00
00 O 0

O O


cog stat
O Non-MCI
0 MCI


cog stat
O Non-MCI
O MCI


1200-




1000-




C 800-
0.


S600-




400-




200-


0
0
0
00
0 0
Oo









overall performance, and performance gains (e.g., linear trends of slope). These

considerations have reflected relationships within each variable.

The section that follows considers the relationships of intraindividual variability

across the variables, over the 31 occasions of measurement. First, correlations between

the IRIs are presented in order to answer questions regarding the nature and degree of

relationships across the Intraindividual Residual Indices. Specifically, the correlations

provide an assessment of how over all participants, and over all occasions, the tendency

to be variable, or to fluctuate, on one variable is related to the tendency to fluctuate on

another.

The correlations are followed by presentation of analyses utilizing the mixed

effects modeling approach. This approach answers questions regarding the nature and

degree of relationships between intraindividual variability across variables within each

day. In other words, the mixed model analysis provides information as to whether "up"

days on one variable are related to "up" days on another. The approach has been

described as an analysis of "coupling" of variables on a day-to-day basis.

How is intraindividual variability on one measure related to intraindividual

variability on another measure? How are variabilities "coupled"? The correlational

relationship between the intraindividual variability coefficients (IRI) for sub-indices from

our three cognitive tasks (AVLT List Memory, Backward Digit Span, and Symbol Digit)

and the three non-cognitive measures (PANAS, Sleep indicators, and Environmental

Distractors) are provided in Table 4-17. The magnitude and direction of the correlations

between the IRIs for the three cognitive measures were not uniform. In general, IRIs for

indices calculated within the same measure (cognitive or non-cognitive) were positive.









An exception was between AVLT List 1 and AVLT Percent Retained. This IRI

correlation was significant and negative, indicating that participants with higher

intraindividual variability on one measure also had lower intraindividual variability on

the other measure.

Between cognitive tests, there were few significant associations in intraindividual

variability, except for Backward Digit Span, which was negatively correlated with AVLT

List 1. That is, individuals who evinced more variability on Backward Digit Span

evinced less variability on AVLT List 1. Between the non-cognitive measures, there

were several significantly positive correlations, indicating that greater intraindividual

variability on one was associated with greater intraindividual variability on the other.

Such correlations were found between PANAS Positive Affect and Environmental

Discomfort, PANAS Positive Affect and Environmental Distractions, PANAS Negative

Affect and Sleep Efficiency, and PANAS Negative Affect and Environmental

Discomfort.

Finally, the correlations across the cognitive and non-cognitive measures were

positive, when significant, meaning that individuals who showed greater intraindividual

variability on the cognitive measure also showed greater intraindividual variability on the

non-cognitive measure. These significant correlations were between AVLT List 1 and

Sleep Time, AVLT List 1 and Environmental Distractors, AVLT Learning and Sleep

Time, and Symbol Digit Score and Environmental Distractors.





7.
-.02

.14

.01

-.09

-.08

.01

1


8. 9.
-.08 .06

-.05 .24*

-.14 .01

.02 -.09

.09 -.12

.00 .13

.37** .18

1 .24

1


10.
.02

.15

.03

.01

.00

.19

.22

.27*

.78**

1


Table 4-17. Intercorrelations of Individual Residual Indices.
Measure 2. 3. 4. 5. 6.
1. AVLT Total .18 .52** .38** -.06 -.01
Score
2. AVLT List 1 1 -.18 -.36** -.25* -.16

3. AVLT 1 .76** .08 .04
Delay Recall
4. AVLT Percent 1 .05 .14
Retained
5. Backward 1 .13
Digit Span
6. Symbol Digit .44**
Score
7. PANAS:
Positive
8. PANAS:
Negative
9. Total Sleep

10. Sleep
Efficiency
11. Fl:
Discomfort
12. F2:
Distractions


12.
-.01

.28**

.12

.10

-.04

.39**

.30*

.24

.09

.00

.23

1


11.
.02

.09

-.10

-.03

-.10

.18

.62**

.41**

.18

.32**

1









While the correlations provide overall relationships between the IRIs across

variables, as noted above, mixed effects models allow for the determination of how two

or more variables, assessed repeatedly, might move together on a day-by-day basis. This

allows for an evaluation of the degree to which higher or lower days on one measure

might predict higher or lower days on another. In the mixed effects models used for

these analyses, the variables of interest were "centered", that is, for each participant, each

daily score less that individual's mean score was used. Thus, individual day-to-day

variability remains, but level of performance differences are controlled. As a result, only

fixed effects (those that depict how the variables are related for all participants) are of

interest in these models. Random effects (which involve individual differences in

performance levels) were eliminated from the models to aid in interpretation. The model

for each variable of interest was run in a step-wise fashion, initially including all possible

"coupled" variables (e.g., other cognitive measures as well as non-cognitive measures).

Variables from the same measure were not used, since, for example, AVLT Total Score

can be almost perfectly predicted from AVLT List 1, due to the linear calculations that

relate these variables. In each subsequent model for each variable, non-significant effects

were removed until the best fitting model, with only the significantly coupled variables

remained. Table 4-18 shows the results for the final models for each variable of interest.

Although there were no significant Group effects, the parameter estimate, F statistic, and

significance value for the Group Fixed effects appear in each table to illustrate this. All

analyses were conducted allowing for the linear and quadratic effects of time as well as

any potential effects of different versions (e.g., workbook version). These were

universally significant, but are not shown, to simply the presentation.









With regards to the memory variable, day-to-day variability in positive affect, as

measured on the PANAS Positive Affect scale, along with fluctuations in environmental

distractors (Factor 2) and discomfort (Factor 1), are reliably related to the intraindividual

variability observed on the AVLT Total Score and AVLT Delay Recall (Results in Table

4- 14). Similarly, positive affect (PANAS Positive Affect) and environmental discomfort

(Factor 1) were coupled with AVLT List 1, while environmental distractions and

discomfort (Factors 1 and 2) were coupled with AVLT Percent Retained. The similar

patterns of relationships on the memory variables reflects the inter-relationships of the

variables on the memory measure.

With regards to the non-memory cognitive measures, Symbol Digit Score was not

significantly coupled with any other variables, but day-to-day fluctuations in Backward

Digit Span moved with fluctuations in environmental distractors (Factor 1), likely due to

the attentional needs of this working memory task.

In turn, the non-cognitive variables revealed "coupled" variabilities.

Intraindividual variability in AVLT Total score, environmental discomfort (Factor 2) and

negative affect (PANAS Negative Affect) moved in concert with daily fluctuations in

positive affect on the PANAS. In contrast, day-to-day variability of negative affect on

the PANAS was coupled with environmental discomfort (Factor 1), and total sleep time.

Total time asleep reflected the same pattern, as negative affect and environmental

discomfort were significantly coupled. Sleep efficiency (calculated as amount of time in

asleep out of time in bed) moved in concert with a number of other variables including

Symbol Digit, PANAS positive and negative affect, and environmental discomfort and

distractions. Finally, the two factors related to the testing environment (both internal and