[1] 13w Cognitively-Impaired-Not-Demented Status Moderates the Time-Varying Association between Finger Tapping Inconsistency and Executive Performance

[1] 161w Objective: Response time inconsistency (RTI) in cognitive performance predicts deleterious health outcomes in late-life; however, RTI estimates are often confounded by additional influences (e.g., individual differences in learning). Finger tapping is a basic sensorimotor measure largely independent of higher-order cognition that may circumvent such confounds of RTI estimates. We examined the within-person coupling of finger-tapping mean and RTI on working memory, and the moderation of these associations by cognitive status. Method: A total of 262 older adults were recruited and classified as controls, cognitively-impaired-not-demented (CIND) unstable or CIND stable. Participants completed finger-tapping and working-memory tasks during multiple weekly assessments, repeated annually for 4 years. Results: Within-person coupling estimates from multilevel models indicated that on occasions when RTI was greater, working-memory response latency was slower for the CIND-stable, but not for the CIND-unstable or control individuals. Conclusions: The finger-tapping task shows potential for minimizing confounds on RTI estimates, and for yielding RTI estimates sensitive to central nervous system function and cognitive status.

[1] 214w The study of response time inconsistency (RTI) in cognitive performance has grown substantially in the past decade, particularly as a predictor of age-related normative and pathological outcomes. RTI reflects short-term, within-person variability in performance largely ignored in service of averaging across all trials to compute estimates of central tendency. Within a number of cognitive tasks (e.g., choice reaction time), increased RTI has shown linkages to age (Vasquez, Binns & Anderson, 2016), impaired cognitive status (Murtha, Cismaru, Waechter & Chertkow, 2002), cognitive function (Grand, Stawski & MacDonald, 2016), brain injury (Stuss, Murphy, Binns & Alexander, 2003) and to more rapid cognitive decline (Bielak, Hultsch, Strauss, MacDonald & Hunter, 2010), independent of average performance. Cognitive aging researchers argue that RTI in cognitive behavioral performance may stem from age-related degradation of the central nervous system (CNS; Anstey et al., 2007;MacDonald, Karlsson, Rieckmann, Nyberg & Bäckman, 2012). Increased RTI may result from a host of functional (MacDonald, Nyberg & Bäckman, 2006), neuromodulatory (e.g., dopamine binding potential, MacDonald, Karlsson, Rieckmann, Nyberg & Bäckman, 2012), and structural changes including white-matter hyperintensities (Walhovd & Fjell, 2007), particularly in the frontal lobes (Bunce, Anstey, Christensen, Dear, Wen & Sachdev, 2007). As such, behavioral RTI may serve as a useful proxy of CNS integrity that is not reflected in estimates of central tendency.

[2] 214w Response time inconsistency for certain tasks, particularly those measuring higher-order constructs (e.g., executive functions), may be subject to a number of additional sources of variance (e.g., learning, practice), and may therefore be confounded as a proxy of CNS integrity. The removal of confounding sources of variance from RTI metrics is a pressing issue in Neuropsychology and Gerontology alike. Failure to do so may result in spurious associations, as increasing variability is positively associated with age due to average slowing or systematic differences in learning (Hultsch, Strauss, Hunter & MacDonald, 2008). Methodological solutions to circumvent confounds when computing variability estimates include statistically controlling for mean group differences (e.g., coefficient of variation [CV]) or statistically partialling confounding influences (e.g., practice-related decreases in RT) and their higher-order interactions prior to computing an index of variability (e.g., intraindividual standard deviation [ISD]). An alternative approach is to use a measure that minimizes higher-order cognitive demands, yet remains sensitive to CNS function. One such measure is the finger-tapping task, which is regularly employed in studies of cognitive aging (e.g., Bielak et al., 2010, Sullivan et al., 2001) and benefits from ease of administration in clinical contexts. Compared with higher-order cognitive tasks, basic sensorimotor tasks minimize the influence of cognitive (e.g., attention) and task (e.g., learning) demands when deriving RTI estimates.

[3] 201w The present study examines whether finger tapping, as a basic measure of sensorimotor function, can yield RTI estimates that reflect a reliable, but potentially less-confounded indicator of cognitive change and status. Here, we evaluated withinperson coupling to determine whether variation in finger-tapping mean (FTM) and RTI was associated with the corresponding variation in working-memory performance, as an index of executive function. This analytic approach is particularly novel as it facilitates the examination of "within-person" associations over time, removing confounds due to between-subject differences (e.g., age, cognitive status), and thus informs the potential functional underpinnings of RTI. We then examined whether these coupling patterns systematically varied as a function of cognitive status. Hypotheses included the expectation of a within-person coupling association between greater finger-tapping inconsistency (FTI) and slower working memory, that cognitive status would moderate this association, and that the most impaired group of individuals (stable cognitively-impairednot-demented (CIND) classification across two assessment occasions) would should the most pronounced association. In keeping with assertions that RTI is a sensitive indicator of CNS integrity (e.g., Bielak et al., 2010;De Frias, Dixon, Fisher & Camicioli, 2007), we anticipated that these effects would be more prominent for inconsistency in finger tapping, relative to mean finger tapping.

[1] 65w Our initial research objective was to assess whether the finger-tapping and working-memory measures exhibited a significant change over time or variance in change. Employing linear mixed models (see Supplementary Material online), we examined the fixed and random effects associated with change in speeded performance over weeks within years for (a) mean finger tapping and (b) mean four-choice RT one-back (BRT). Table 1 summarizes the results.

[2] 126w Mean finger tapping. There was no significant effect of change over time in mean finger-tapping latency across weeks (Week = -0.30, p = .309) or years (Year = -0.05, p = .935), indicating that across participants in the study, the average amount of change was not significant. These non-significant change estimates suggest that there was no long-term mean slowing in finger-tapping speed, and that participants did not benefit from repeated exposure to the task (i.e., there were no practice effects) across weeks or years. Notably, however, the Level-3 random effects (see Table 1) for the weekly (χ 2 = 418.16, p < .001) and yearly slopes (χ 2 = 780.42, p < .001) were significant, indicating significant between-person variance in rates of week-to-week and year-to-year change.

[3] 97w Mean BRT. Significant change in mean BRT latency was observed across both weeks (Week = -93.97, p < .001) and years (Year = -99.30, p < .001), indicating that on average, participants responded more quickly across weeks as well as across years within the study. This observed improvement is consistent with practice effects reported for other intensive measurement-burst design studies. The between-person random effects for the weekly (χ 2 = 436.30, p < .001) and yearly slopes (χ 2 = 805.20, p < .001) emerged as significant, indicating between-person heterogeneity in rates of change in workingmemory performance.

[4] 111w Having explored patterns of change and heterogeneity in rates of change for the finger-tapping and working-memory measures, we proceeded to the primary research focus; examining whether variation in FTM or inconsistency covaried significantly with BRT latency within persons over time, and whether these coupling associations were further moderated by CIND status (control, CIND-UN, CIND-S). A description of the ISD computation for indexing inconsistency can be found in Supplementary Material online. For each coupling model, we evaluated tests of simple slopes assessing whether coupling associations for a given cognitive status subgroup significantly differed from 0, and whether these within-person associations significantly differed between CIND groups (see Supplementary Material online for full description).

[5] 106w Central tendency in tapping speed. First, we specifically tested whether the within-person association of FTM on variation in BRT latency was significant. A significant simple effect for FTM-BRT coupling was observed for the control group (β = 0.73, p < .01), indicating that on occasions when tapping speed was slower by 1 ms relative to a person's own mean, BRT latency slowed by 0.73 ms. However, no significant simple effects for the FTM-BRT coupling association were observed for either the CIND-UN or CIND-S groups. Further, the coupling associations did not significantly differ (all ps > 0.05) between the three CIND groups (see Table 2, group comparisons).

[6] 174w Inconsistency in tapping speed. Next, we assessed whether variation in FTI exhibited a significant within-person association with BRT latency (see Table 2). In contrast to mean tapping speed, no significant simple slope of FTI-BRT coupling was observed for the control participants (β = 1.31, p = .53). However, a significant simple effect of FTI-BRT latency coupling was observed for the most cognitively-impaired (CIND-S) group (β = 7.96, p < .05), indicating that on occasions when an individual exhibited a 1 unit increase (~0.1 SD units) in finger-tapping variability relative to her/his own mean, BRT latency slowed by 7.96 ms. In addition, significant differences in FTI-BRT coupling were observed between the reference control group and the most cognitively-impaired CIND-S subgroup (β = -6.64, p < .05, one-tailed), as well as between the CIND-UN and CIND-S subgroups (β = -10.43, p < .05). These group comparisons indicate that, on occasions when CIND-S participants were more variable in their finger-tapping latencies, they performed more slowly on the BRT task relative to both the control and CIND-UN subgroups.

[1] 218w Overall, our findings support claims that (a) inconsistency derived from sensorimotor tasks with minimal cognitive demands may help circumvent known confounds in the computation of variability estimates and (b) FTI is more sensitive than mean performance for predicting cognitive outcomes, particularly for more impaired subgroups. With regard to the first claim, across 4 years and up to 17 assessments, no systematic change was observed for mean finger-tapping speed-for either the full sample or within cognitive status subgroups. In contrast, mean latency for the BRT task decreased (i.e., showed improvement) over time (weeks and years), suggesting that practice effects may be more apparent in high-order cognitive tasks relative to basic sensorimotor tasks. The observed stability in mean tapping speed across RT trials for even the shortest retest interval (weeks) of the measurement-burst design supports our hypothesis that a sensorimotor task with minimal cognitive demands would exhibit modest (if any) practice effects. As a consequence, systematic trends in mean tapping speed would not subsequently confound computations of performance inconsistency. The lack of long-term change for the tapping speed measure suggests that careful RT task selection represents an important consideration for circumventing known confounds that influence RTI estimates obtained from more complex cognitive processing tasks. Tapping not only has these advantages, but is also easy to administer in a clinical context.

[2] 228w Pertaining to the latter claim, our findings also suggest that inconsistency across finger-tapping trials is more sensitive than mean finger-tapping performance for predicting cognitive performance and status. On occasions when RTI in finger tapping was higher, BRT performance was slower, but only for the most at-risk, stable CIND group. In contrast, the mean tapping-BRT coupling association did not significantly differ between CIND subgroups. As hypothesized, a dose-response effect was observed for the coupling of tapping variability and working-memory performance, whereby the most pronounced withinperson association was observed for the most cognitively-impaired individuals. Notably, our classification of CIND subgroups differentiated consistency of impairment across a 2-year time period, yielding a relatively robust classification of cognitive status. These findings are consistent with previous research demonstrating that inconsistency in performance is a stable, endogenous characteristic, associated with the aging process, and predictive of cognitive, behavioral, and neurological functioning (Bielak et al., 2010;Burton, Strauss, Hultsch, Moll & Hunter, 2006;Grand et al., 2016;Hultsch, MacDonald, & Dixon, 2002;Hultsch, Macdonald, Hunter, Levy-Bencheton & Strauss, 2000;MacDonald, Hultsch & Dixon, 2003, 2008;MacDonald, Nyberg & Bäckman, 2006;Vasquez et al., 2016). Furthermore, our demonstration of significant within-person coupling of FTI and working memory underscores the importance of examining short-term variation and covariation in indicators of CNS integrity to examine the utility, function, and impact of such dynamic indicators for understanding cognitive and neuropsychological processes (Stawski, Sliwinski & Hofer, 2013).

[3] 110w Participants were not followed past the reported 4-year time frame and as such, conversion to dementia as well as differential diagnoses (e.g., Alzheimer's dementia, vascular dementia) cannot be confirmed. In classifying cognitive status, stable impairment for the CIND-S classification was based on below-average performance on any of the cognitive benchmark tasks and not necessarily on the same task at Years 1 and 2. Hence, drawing etiology-specific claims with regard to CIND subtypes (e.g., amnestic vs. non-amnestic) is beyond the scope of the reported results. Given the information available, we employed a distributional approach to classification (i.e., CIND), rather than formal clinical diagnosis, which may limit the generalizability of our findings.

[4] 109w Investigators seeking to employ RTI metrics to elucidate CNS integrity in clinical populations may consider employing the finger-tapping task, as impaired performance appears to map onto early changes in CNS function (e.g., speed of processing). The task instructions are relatively simple and the task itself is relatively intuitive. Further, the administration time to collect sufficient trials for computing RTI is minimal. Examining the time-varying covariation between finger-tapping RTI and other facets of EF (e.g., inhibition) may further elucidate the sensitivity and specificity of sensorimotor RTI as a proxy for CNS integrity, and may help inform models of how lower level processes (e.g., sensorimotor processing) differentially subserve facets of EF.

[1] 120w Overall, our results demonstrate the potential utility of the finger-tapping task in yielding a reliable estimate of RTI that is sensitive to CNS function and in particular, cognitive status. In contrast to higher-order cognitive tasks commonly employed in research on variability, the finger-tapping task exhibited no practice effects and is thus less confounded for deriving RTI estimates. The task is also easily administered due to its overall simplicity, rendering it of additional clinical utility (e.g., tracking early changes in CNS-related pathology for treatment planning) across a range of populations. Overall, the task is suitable, both empirically and pragmatically, to detect relatively subtle changes in underlying CNS integrity and shows promise in facilitating early identification of cognitive decline in older adults.

[1] 149w This study was approved by the University of Victoria Human Research Ethics Board and was conducted in accordance with institutional guidelines. Community-dwelling individuals were recruited from Victoria, Canada to Project Mental Inconsistency in Normals and Dementia (MIND) through media solicitations for volunteers who might be experiencing cognitive complaints, but who were not diagnosed by their physician with a neurological condition. Participants were excluded if they had (a) a major medical illness (e.g., dementia), (b) severe sensory impairment (e.g., difficulty writing or pressing buttons), (c) current drug or alcohol abuse, (d) current psychiatric diagnosis, (e) significant cognitive impairment (i.e., Mini-Mental Status Examination score below 24), or (f) English as a second language. Data from 262 participants (180 women, 82 men) aged 64-92 years (m = 73.84, SD = 5.88) were included for present analyses. Participants' level of education ranged from 7 to 24 years (m = 15.18, SD = 3.07).

[2] 107w Participants were enrolled after satisfying inclusion criteria. Initial Year 1 group testing (neuropsychological assessment) was held at the university, with subsequent individual sessions (reaction time tasks) conducted in participants' homes. For Year 1 of the study, participants completed five biweekly home sessions, which varied across days of the week and times of the day. Testing was repeated annually for another 4 years, with participants completing four (rather than five) biweekly testing sessions for Years 2 through 4. The attrition rates were 11.0%, 3.5%, and 4.5% of the sample between Years 1-2, 2-3, and 3-4, respectively. See Bielak and colleagues (2010) for a full description of testing procedures.

[3] 83w Cognitively-impaired-not-demented status was classified based on participants' performance across five cognitive benchmark tasks (perceptual speed [Digit Symbol], inductive reasoning [Letter Series], verbal fluency [Similarities], crystallized ability [Vocabulary], and episodic memory [Word Recall]). Normative data were available from an independent study of adults 65-94 years of age recruited from the same demographic, as part of a separate, independent study. For a complete description of the normative sample and benchmark tasks, see Vandermorris, Hultsch, Hunter, MacDonald and Strauss (2011) and Dixon and De Frias (2004).

[4] 184w Based on performance across the cognitive benchmark tasks at Years 1 and 2, participants were classified as healthy control (n = 162, m age = 73.79, SD age = 5.77), unstable CIND (CIND-UN; n = 55, m age = 73.80, SD age = 5.89) or stable CIND (CIND-S; n = 45, m age = 74.04, SD age = 6.35). The CIND status was determined when a participant scored >1.5 SD below his or her age-and education-matched peers across one or more of the cognitive benchmark tasks (Vandermorris et al., 2011;Winblad, Palmer, Kivipelto, Jelic, Fratiglioni, Wahlund et al., 2004). CIND-UN participants were those who demonstrated cognitive impairment at either Year 1 or Year 2 (but not both), whereas CIND-S participants were those who demonstrated cognitive impairment at both Year 1 and Year 2. Healthy control participants did not demonstrate cognitive impairment at Year 1 or Year 2 (Dixon & De Frias, 2004). Groups did not differ in terms of chronological age, sex distribution, depressive affect or number of chronic illnesses. Controls had more years of education than CIND-UN and CIND-S individuals, who did not differ.

[5] 83w Finger tapping. Participants were instructed to tap a response key as quickly as possible using first their right hand, then their left, for a total of 47 taps per hand. Each finger tap led to a letter (L or R) appearing on the screen. Participants completed a practice run of 24 taps followed by a test run of 47 taps per hand, prior to completing other experimental tasks. Latencies were recorded as the time elapsed between consecutive finger taps (Bielak et al., 2010).

[6] 95w Four-choice RT one-back. Participants were presented with four plus signs displayed in a horizontal row along with a response input device containing four spatially mapped keys. On each trial, following a 1,000-ms delay, a box replaced one of the plus signs. For each trial, participants were asked to respond to the location of the box on the "previous" trial. Instructions emphasized speed, however participants were instructed to remain as accurate as possible. A total of 10 practice trials and 61 test trials were administered, with latency computed as the outcome measure (Bielak et al., 2010).

[7] 72w Response time inconsistency. As an index of RTI, residualized ISD estimates were computed in order to partial systematic within-(i.e., trial) and between-subject (i.e., age group) sources of variance in mean RT. Having removed these systematic confounds, the resulting RTI estimates were not conflated with mean age differences in response speed or individual differences in developmental change or learning to learn. See Hultsch and colleagues (2008) for a full description of this procedure.