Introduction
An important goal in the cognitive neuropsychology of aging is to better understand how normal age-related declines differ from those associated with Alzheimer's disease (AD). AD-related neuropathological changes occur years before apparent clinical symptoms (Bennett et al., Reference Bennett, Schneider, Arvanitakis, Kelly, Aggarwal, Shah and Wilson2006; Morris et al., Reference Morris, Storandt, McKeel, Rubin, Price, Grant and Berg1996; Price & Morris, Reference Price and Morris1999). Recent research has sought to identify pre-clinical markers which distinguish healthy older adults (OA) from those with prodromal AD. Although episodic memory impairments are considered the hallmark of AD and discriminate healthy OA from those with mild dementia of the Alzheimer's type (DAT: Albert, Moss, Blacker, Tanzi, & McArdle, Reference Albert, Moss, Blacker, Tanzi and McArdle2007; Albert, Moss, Tanzi, & Jones, Reference Albert, Moss, Tanzi and Jones2001; Storandt & Hill, Reference Storandt and Hill1989; Storandt, Grant, Miller, & Morris, Reference Storandt, Grant, Miller and Morris2006), declines in attention have recently been shown to also be powerful markers (e.g., Hutchison, Balota, & Duchek, Reference Hutchison, Balota and Duchek2010).
Declines in Attention With Aging and DAT
Declines in attentional control are found in healthy OA and individuals with early stage DAT (see Balota & Faust, Reference Balota and Faust2001, and Perry & Hodges, Reference Perry and Hodges1999, for relevant reviews); selective and divided attention are especially sensitive to early stage DAT-related breakdowns (see Faust & Balota, Reference Faust and Balota2007, for a review). Namely, DAT individuals have trouble maintaining a representation of task demands and inhibiting the intrusion of inappropriate information; this may contribute to episodic memory impairments through continued activation of irrelevant information that disrupts encoding and retrieval processes (Balota et al., Reference Balota, Cortese, Duchek, Adams, Roediger and Yerys1999; Castel, Balota, & McCabe, Reference Castel, Balota and McCabe2009; Craik & Lockhart, Reference Craik and Lockhart1972; Jacoby, Reference Jacoby1999). Vigilance is resistant to normal age-related declines (Tucker, Stern, Basner, & Rakitin, Reference Tucker, Stern, Basner and Rakitin2011) but may show DAT-related deficits under increased task difficulty (Berardi, Parasuraman, & Haxby, Reference Berardi, Parasuraman and Haxby2005).
Attention and Timing
Attention is critical in temporal perception and performance, and so investigations of simple timing tasks may afford a useful paradigm for examining age and AD related changes. One popular information-processing conception of timing posits a pacemaker-accumulator device with an attention-mediated gate (Gibbon, Church, & Meck, Reference Gibbon, Church and Meck1984; Lejeune, Reference Lejeune1998; Rousseau, Picard, & Pitre, Reference Rousseau, Picard and Pitre1984; Zakay & Block, Reference Zakay and Block1997). The degree of attention to time influences the opening and closing of the gate and, subsequently, the number of pulses that pass to the accumulator and represent the target interval. The impact of attentional manipulations on perceived interval length depends on whether attention is disrupted during encoding or response (Brown, Reference Brown1997; Fortin, Rousseau, Bourque, & Kirouac, Reference Fortin, Rousseau, Bourque and Kirouac1993; Macar, Grondin, & Casini, Reference Macar, Grondin and Casini1994; Zakay, Reference Zakay1998). Although the impact on timing variability is less clear (Brown, Reference Brown1997; Perbal, Droit-Volet, Isingrini, & Pouthas, Reference Perbal, Droit-Volet, Isingrini and Pouthas2002), Rakitin (Reference Rakitin2005) examined this relationship using choice time production with 3- and 5-s target durations. Under conditions requiring greater attentional control (stimulus-response incompatibility), the coefficient of variation (COV) increased, especially for the shortest interval. Presumably, non-scalar variability linked to opening and closing of the attentional gate contributed more to the shorter interval's total variability.
Most studies manipulating attention and timing have involved durations beyond the upper range (3–5 s) of the “psychological present” (see Pöppel, Reference Pöppel2004, for a review) where people have difficulty perceiving two stimuli as part of a unified event, causing timing to rely more on executive control. However, engagement of executive functions may also occur at shorter durations. For example, Michon (Reference Michon1985) argued that 500 ms delineates automatic versus cognitively mediated temporal processes. Likewise, there is differential engagement of brain networks during timing of durations shorter (sensorimotor regions) versus longer (right prefrontal and parietal regions) than 1 s (see Koch, Oliveri, & Caltagirone, Reference Koch, Oliveri and Caltagirone2009, and Lewis & Miall, Reference Lewis and Miall2006, for relevant reviews). The latter brain areas are associated with attention and executive functions (Lewis & Miall, Reference Lewis and Miall2003). In the context of a continuous tapping paradigm, like that used in the current study, slower tapping rates and longer tapping bouts lead to drift, with breaks in patterns of drift occurring at approximately 1000 and 1300 ms (Collier & Ogden, Reference Collier and Ogden2004; Madison, Reference Madison2001). Increased drift is linked to increased response dispersion, and breaks implicate possible shifts in processes used for timing.
Timing, Aging, and AD
A meta-analysis (Block, Zakay, & Hancock, Reference Block, Zakay and Hancock1998) of the aging and time literature found more variable estimates and shorter productions with age when demands on controlled attention were greater at encoding than at test. Of the few studies that have explored the impact of DAT on timing, most have used temporal discrimination with supra-second durations. Although the results are somewhat mixed, DAT appears linked to more variable estimates (Nichelli, Vernneri, Molinari, Tavani, & Grafman, Reference Nichelli, Vernneri, Molinari, Tavani and Grafman1993; Papagno, Allegra, & Cardaci, Reference Papagno, Allegra and Cardaci2004; Rueda & Schmitter-Edgecombe, Reference Rueda and Schmitter-Edgecombe2009).
Simple continuation tapping involves the continuation of a tapping pulse to an external signal after that signal has been removed. This task allows partitioning of timing variability into a component attributable to a putative internal clock and one associated with peripheral motor processes (Wing & Kristofferson, Reference Wing and Kristofferson1973). This paradigm has revealed declines in special populations, such as individuals with prodromal Huntington's disease, using sub-second intervals (Rowe et al., Reference Rowe, Paulsen, Langbehn, Duff, Beglinger, Wang and Moser2010). Similarly, differences were found between non-demented Parkinson's patients on medication and normal, healthy controls on durations of 300 and 600 ms (Harrington, Haaland, & Hermanowicz (Reference Harrington, Haaland and Hermanowicz1998). Although Duchek, Balota, and Ferraro (Reference Duchek, Balota and Ferraro1994) did not replicate this finding with Parkinson's patients using a 550-ms interval, they did find increased clock variability in mild DAT. Neither very mild DAT nor normal aging led to deficiencies in clock or motor variability. Others, however, have found a positive relationship between clock variability and age in this task (Woodruff-Pak & Jaeger, Reference Woodruff-Pak and Jaeger1998).
Stronger age- and DAT-related differences may emerge in this paradigm at slower tapping rates. When tapping at rates ranging from 150 to 1709 ms, OA speed up at the longest interval (McAuley, Jones, Holub, Johnston, & Miller, Reference McAuley, Jones, Holub, Johnston and Miller2006). Likewise, Krampe, Doumas, Lavrysen, and Rapp (Reference Krampe, Doumas, Lavrysen and Rapp2010) found that during tapping at fast (550 ms) and slow (2100 ms) rates, dual tasking led both OA and YA to speed up their tapping at the slow rate, while OA also sped up during the fast rate. Dual tasking also led to increased variability for both groups; importantly, this was magnified at the slow rate for OA. Thus, even with simple repetitive tapping, attention plays a critical role.
The current study expands upon work by Duchek and colleagues (1994) to examine whether normal aging and very mild DAT lead to breakdowns on a simple synchronization-continuation tapping task (see Figure 1) when slower tap rates, intended to tax attention, are included. Its apparent simplicity in terms of performance demands and instructions make it well-suited for individuals with cognitive declines. We use intervals (500, 1000, and 1500 ms) well within the “psychological present” to ensure each inter-tap interval (ITI) is perceived as a unified event. Each tap rate condition lasts approximately 3 min. This is longer than the typical bout required for this procedure and should engage attentional mechanisms to maintain an accurate representation of the rate over time. While this differs from traditional attention/vigilance tasks, participants must monitor their internal environment to detect a generated expectancy indicating when to execute a tap.
Schematic of the synchronization-continuation timing task. Participants produced 12 intervals in sync with an auditory pacer, then the tones disappeared and tapping continued while participants tried to maintain the target tapping rate for the rest of the trial. Time on task was roughly equated for the three tapping rate conditions.
Slower tapping rates should be more likely to engage attentional mechanisms. To the degree that dual-task interference serves as a valid model of aging and DAT-related decline, we expect these groups to show increased variability and speed of tapping, especially at the slowest rates, compared to YA (Krampe et al., Reference Krampe, Doumas, Lavrysen and Rapp2010; McAuley et al., Reference McAuley, Jones, Holub, Johnston and Miller2006). These changes may be magnified even in the very earliest stages of the disease, i.e., levels of DAT that are comparable to the cognitive decline seen in MCI (see Storandt et al., Reference Storandt, Grant, Miller and Morris2006). Given recent evidence that variability measures are sensitive to early stage AD (e.g., Duchek, Balota, Tse, Holtzman, & Goate, Reference Duchek, Balota, Tse, Holtzman and Goate2009; Hultsch, MacDonald, Hunter, Levy-Bencheton, & Strauss, 2000; Tse, Balota, Yap, Duchek, & McCabe, Reference Tse, Balota, Yap, Duchek and McCabe2010), we are especially interested in how intra-individual variability in timing differs for the healthy older and very mild DAT groups. Although we report results from a small sample of individuals with mild DAT (Clinical Dementia Rating, CDR, of 1), we are most interested in individuals with very mild DAT (CDR of .5), because they are at the earliest detectable transition from healthy aging to early stage AD (see Storandt et al., Reference Storandt, Grant, Miller and Morris2006). The mild DAT group was examined to ensure that increases in dementia severity do not produce unexpected patterns of results.
Methods
Participants
A total of 282 individuals completed this study, approved by the Institutional Review Board at Washington University; participants gave informed consent before participating. Sixty-six college-aged YA (mean age = 20.27, SD = 1.56) recruited from the Washington University Psychology department undergraduate pool participated for course credit. The remaining 216 older adults were recruited from the Washington University Alzheimer's Disease Research Center (ADRC) and were screened for depression, untreated hypertension, reversible dementias, and other disorders associated with cognitive impairment. Presence and severity of dementia was assessed using the Washington University Clinical Dementia Rating (CDR) scale (Morris, Reference Morris1993; Morris & Fulling, Reference Morris and Fulling1988). CDR scale values and their matching dementia status are: 0 = no dementia, .5 = very mild dementia, 1 = mild dementia, 2 = moderate dementia, and 3 = severe dementia.Footnote 1 Note that individuals classified as CDR .5 in our study scored, on average, 27 on the Mini-Mental State Examination, suggesting they are at the very earliest detectable stages of dementia. Of the OA, 133 were classified as healthy CDR 0, 58 as very mild dementia, or CDR .5, and 25 as mild dementia, or CDR 1 (see Table 1).
Psychometric means (SD) as a function of group
Note. * p < .05 indicates a significant difference when compared against healthy OA. +p < .05 indicates a significant difference when compared against very mildly demented (CDR .5) individuals. Tests of memory included the forward and backward digit span, the logical and associate memory components from the Wechsler Memory Scale (WMS; Wechsler & Stone, Reference Wechsler and Stone1973), as well as free recall from the Selective Reminding Test (SRT Free; Grober, Buschke, Crystal, Bang, & Dresner, Reference Grober, Buschke, Crystal, Bang and Dresner1988). Psychomotor processing speed was assessed with the Digit Symbol subtest of the Wechsler Adult Intelligences Scale (WAIS) (Wechsler, Reference Wechsler1955). Visual perceptual-motor performance was examined with Parts A and B of the Trail Making test (Armitage, Reference Armitage1945). The Word Fluency Test S-P (Thurstone & Thurstone, Reference Thurstone and Thurstone1949), the Animal Fluency Test (Goodglass & Kaplan, Reference Goodglass and Kaplan1983b) and The Boston Naming Test (Goodglass & Kaplan, Reference Goodglass and Kaplan1983a) evaluated semantic/lexical retrieval. Statistical analysis of differences in education and psychometric performance controlled for age. The number of male participants and the number of right-hand dominant individuals are reported for gender and handedness. MMSE = Mini-Mental State Examination.
Participants completed simple synchronization-continuation tapping. Some were excluded from further analysis due to (1) difficulty understanding instructions; (2) finger pain, wrist pain or numbness; or (3) changing the response finger during tapping. This eliminated 2 YA, 14 healthy OA, 12 CDR .5, and 3 CDR 1 individuals, with the primary causes due to pain or motor issues; difficulty with instructions eliminated only 1 YA, 3 OA, 2 CDR .5, and 1 CDR 1 adult(s). Of the remaining participants, the OA and CDR .5 groups did not differ in age, p = .380. However, age differences emerged between the CDR .5 and CDR 1 groups, t(65) = −2.14, p = .036, and between the OA and CDR 1 groups, t(138) = −3.00, p = .003; hence, age was controlled in the analyses comparing these groups.
Apparatus
An IBM-compatible computer running E-prime software (Schneider, Eschman, & Zuccolotto, Reference Schneider, Eschman and Zuccolotto2002) was used to control stimulus presentation and collect data. Stimuli were displayed on a 15-inch monitor.
Psychometric Testing
All OA were administered a separate 2-hr standard neuropsychological battery with an experimenter blind to their CDR score. The psychometric tests included in this battery are reported in Table 1.
Continuous Tapping Task
Participants repetitively tapped at three rates (500, 1000, and 1500 ms). At the start of a trial they heard a 50-ms, 1000 Hz repeating tone and were asked to synchronize their taps with these tones until they disappeared (after 12 ITIs). Participants continued tapping at the same rate until STOP appeared on the computer screen. Taps were made using the index finger of the dominant hand on the space bar of the keyboard.Footnote 2 Participants completed three short practice trials at a rate of 1250 ms. The total number of unpaced ITIs differed for each target rate (500 ms: 296 ITIs, 1000 ms: 142 ITIs, 1500 ms: 90 ITIs) to roughly equate time on task (∼3 min). No explicit performance feedback was given.
An exit questionnaire assessed whether participants had problems completing the task. If they did, as noted earlier, they were eliminated from further analyses. Slightly less than half of all the participants (128 individuals) received the same presentation order of tapping rates (1000 ms followed by 500 ms then 1500 ms) in an effort to examine individual differences. However, we also report data from a counterbalanced set of participants. Thus, in all analyses, we control for tap rate order.Footnote 3
Data Analysis
Psychometrics
Table 1 provides information about the gender, dominant handedness, and psychometric performance of our OA groups. We performed univariate comparisons, controlling for age, to evaluate group differences, which are reported in the table.
Continuous tapping
Several performance measures of unpaced tapping were assessed across the complete trial (time on task was equated) and across the first 90 ITIs for each tap rate (hence, equating the number of tap events). The former set of data may be more sensitive to breakdowns in vigilance linked to time on task. Accuracy measures included (1) the accuracy index (AI, Baudouin, Vanneste, Pouthas, & Isingrini, Reference Baudouin, Vanneste, Pouthas and Isingrini2006), a relative timing measure calculated by dividing the mean tapping rate by the target rate (values less than 1 indicate faster tapping than the target rate and values greater than 1 indicate slower tapping), and (2) absolute error (AE), calculated by averaging the absolute difference between each produced and target ITI. The AI is useful for comparing accuracy across rates, and has been proposed to index the integrity of the decision rule used to compare different temporal representations (Gallistel & Gibbon, Reference Gallistel and Gibbon2000; Gibbon & Fairhurst, Reference Gibbon and Fairhurst1994; Malapani & Fairhurst, Reference Malapani and Fairhurst2002). We also included a measure of drift, or people's ability to maintain the representation of the target interval over time; larger negative slopes indicate a faster loss of the representation. This performance change across the tapping series was calculated by estimating the slope of a linear regression across each individual's trimmed tapping intervals for a particular rate and dividing it by the target rate. Resultant values were multiplied by 100 for ease of interpretation. Variability measures included (1) the standard deviation (SD) and (2) coefficient of variation (COV), which involves dividing the SD by mean tapping rate. The COV is often considered a measure of timing sensitivity and allows for comparisons of variability to be made across tap rates.Footnote 4
For all dependent measures, we conducted mixed-factor ANOVAs, controlling for tap rate order, with group as the between-participants and tap rate as the within-participants factor. A set of orthogonal planned contrasts evaluating expected group differences and interactions, using one-tailed tests were performed. These explored simple effects comparing pairs of groups. The first (AGE) contrast (1, −1, 0, 0) compared the YA and healthy OA groups, with the expectation that YA would show better timing performance. The second (DAT) contrast (0 1 −1 0) was a focused analyses of the OA and very mild CDR .5 groups to determine if very mild DAT led to reliably worse performance. When significant effects involving this second contrast emerged we conducted Bonferroni-corrected post hoc univariate tests (critical p = .017) comparing these two groups at each tap rate to explore the specific conditions impacted by the earliest detectable stages of Alzheimer's disease. The final (SEVERITY) contrast (0 0 1 −1) compared the very mild and mild DAT groups to determine how disease severity impacts performance. Of critical interest were the first two contrasts for revealing normal age-related and DAT-specific declines in simple repetitive tapping.
Where sphericity was violated in omnibus tests we applied the Huynh-Feldt correction. Effect sizes are reported as partial eta-squared.Footnote 5 Outliers were identified within each duration on each timing measure by standardizing the values within each group and searching for scores beyond ±3 standard deviations from the mean. Analyses of each performance measure excluded the relevant outliers.Footnote 6
Results
Timing Measures: Omnibus Comparisons
Results from the omnibus analyses comparing all participant groups across the entire tapping trial are reported in Table 2. Since the analyses across the first 90 ITIs showed the same pattern of significance they are not included in Table 2. Likewise, for all analyses below, results across the first 90 ITIs are only reported when they differ from those across the full trial. For all timing measures, we found significant main effects of tap rate and group (all ps < .01). All interactions were also significant (all ps < .01), except for the standard deviation measure (p = .303). Figures 2–6 show the timing performance means for each group across the first 90 ITIs (Panel A) and the entire tapping trial (Panel B). Significant differences between the healthy OA and CDR .5 groups are indicated by asterisks.
Continuous Tapping—ANOVA table of omnibus results for each dependent measure (columns) across the full trial
Note. MS is the mean squared for the indicated effect or error. Degrees of freedom (df) are reported using the Huynh-Feldt correction for violations of sphericity and rounded to the nearest whole number. *** p < .001, ** p < .01, * p < .05, one-tailed.
AI = accuracy index; AE = absolute error; ANOVA = analysis of variance; BS = between subjects; COV = coefficient of variation; G = Group; SD = standard deviation; TR = tap rate; WS = within subjects.
Mean tapping accuracy index for all four participant groups across the unpaced intervals (A) for the first 90 ITIs and (B) for the full trial for each tapping rate. Error bars are ±1 standard error. The asterisk (*) indicates that in the comparison between just the healthy older adult group and the CDR .5 group, the groups differed at the indicated tap rate. YA = younger adults; OA = healthy older adults; CDR = Clinical Dementia Rating.
Timing Measures: Planned Contrasts
Accuracy index
Figure 2 displays the AI. AI was fairly stable across all tapping rates for YA, while aging was linked to an increased rate of tapping as target rate decreased. These observations were supported by a significant AGE contrast effect, F(1,235) = 15.35, p < .001, ηp 2 = .061, and an AGE × tap rate interaction, F(2,234) = 7.05, p < .001, ηp 2 = .057. Importantly, the DAT contrast analysis revealed a significant DAT × tap rate interaction, F(2,234) = 3.09, p = .024, ηp 2 = .026, and a DAT contrast effect, F(1,235) = 5.07, p = .013, ηp 2 = .021, indicating that DAT led to faster tapping above and beyond that seen with normal aging. Post hoc tests confirmed that the only reliable group difference occurred at the 1500-ms rate (p = .002). For the SEVERITY contrast, we found a significant contrast × tap rate interaction, F(2,234) = 5.77, p = .002, ηp 2 = .047, likely driven by opposite patterns in tapping performance at 500 ms. However, the overall contrast effect was not significant (p = .236).
Absolute error
As demonstrated in Figure 3, OA typically showed greater AE than YA, especially at slower tapping rates. This was supported by a significant AGE contrast × tap rate interaction, F(2,233) = 12.77, p < .001, ηp 2 = .099, and a significant AGE effect, F(1,234) = 26.85, p < .001, ηp 2 = .103. Importantly, the DAT contrast analysis revealed a greater increase in AE as tap rate slowed for the CDR .5 group compared to healthy OA, supported by a DAT × tap rate interaction, F(2,233) = 4.69, p = .005, ηp 2 = .039. The DAT contrast effect also reached significance, F(1,234) = 7.53, p = .004, ηp 2 = .031, with a reliable effect only at the 1500-ms rate (p = .005). The SEVERITY contrast × tap rate interaction reached significance, F(2,233) = 6.31, p = .001, ηp 2 = .051 as did the SEVERITY contrast, F(1,234) = 4.79, p = .015, ηp 2 = .020, indicating an increase in AE with increased DAT severity.
Mean absolute error for all four participant groups across the unpaced intervals (A) for the first 90 ITIs and (B) for the full trial for each tapping rate. Error bars are ±1 standard error. The asterisk (*) indicates that in the comparison between just the healthy older adult group and the CDR .5 group, the groups differed at the indicated tap rate. YA = younger adults; OA = healthy older adults; CDR = Clinical Dementia Rating.
Slope
Figure 4 illustrates that while YA showed little performance change across trials, OA showed increasingly negative slopes as tap rate slowed, supported by a significant AGE contrast × tap rate interaction across the first 90 ITIs, F(2,231) = 3.08, p = .024, ηp 2 = .026, but not across the full trial (p = .090). The AGE contrast effect did emerge, F(1,232) = 6.42, p = .006, ηp 2 = .027. The DAT contrast was significant, F(1,232) = 3.56, p = .031, ηp 2 = .015, due to the CDR .5 group being more prone to drift than the healthy OA group. The DAT × tap rate interaction reached significance when evaluated across the full trial (with reliance on vigilance at a premium), F(2,231) = 3.07, p = .024, ηp 2 = .026, but not across the first 90 ITIs, (p = .166). There were no reliable group differences at any tap rate. There was no interaction of the SEVERITY contrast × tap rate interaction (p = .388), but the contrast effect did emerge across the full trial F(1,232) = 2.81, p = .048, ηp 2 = .012, but not the first 90 ITIs (p = .153). Namely, mild DAT tended to show greater drift than very mild DAT individuals.
Transformed slope across the unpaced timing intervals (A) for the first 90 ITIs and (B) for the full trial for each tapping rate in all four participant groups. Error bars are ±1 standard error. YA = younger adults; OA = healthy older adults; CDR = Clinical Dementia Rating.
Standard deviation
Figure 5 shows the SD results. We found neither an AGE contrast (p = .425), nor an AGE × tap rate interaction (p = .442); normal aging was not associated with increased SD. There was a DAT contrast × tap rate interaction across the full trials, F(2,229) = 2.59, p = .039, ηp 2 = .022, but not the first 90 ITIs (p = .170). The DAT contrast was significant, F(1,230) = 8.18, p = .003, ηp 2 = .034. The CDR .5 group produced larger SD than the OA group at the 1000-ms (p = .005) and 1500-ms rates (p = .015) and at the 500-ms tap rate when examined across the first 90 ITIs (p = .003). Neither the SEVERITY × tap rate interaction (p = .178) nor the SEVERITY contrast (p = .070) reached significance, indicating fairly stable variability when moving from very mild to mild DAT.
Standard deviation from the timing task for all four participant groups across the unpaced intervals (A) for the first 90 ITIs and (B) for the full trial at each tapping rate. Error bars are ±1 standard error. The asterisk (*) indicates that in the comparison between just the healthy older adult group and the CDR .5 group, the groups differed at the indicated tap rate. YA = younger adults; OA = healthy older adults; CDR = Clinical Dementia Rating.
Coefficient of variation
YA showed similar COVs while OA showed increasing COVs as tap rate decreased, as revealed in Figure 6. This was supported by a significant AGE contrast × tap rate interaction, F(2,230) = 2.95, p = .028, ηp 2 = .025, as well as a significant AGE contrast effect, F(1,231) = 5.90, p = .008, ηp 2 = .025. The DAT contrast revealed a significant effect, F(1,231) = 11.79, p = .001, ηp 2 = .049, because the CDR .5 group consistently showed higher COVs than healthy OA. While COV increased for both groups at slower tap rates, the magnitude of this change was larger for the CDR .5 group across the full trial, F(2,230) = 3.42, p = .018, ηp 2 = .029, but not across the first 90 ITIs (p = .079). Post hoc tests confirmed reliable differences at 1000 ms (p = .006), 1500 ms (p < .001), and at 500 ms determined across the first 90 ITIs (p = .008). There was a SEVERITY contrast × tap rate interaction, F(2,230) = 4.13, p = .009, ηp 2 = .035, and a significant SEVERITY contrast across the full trial, F(1,231) = 4.59, p = .017, ηp 2 = .019, that was simply a trend across the first 90 ITIs (p = .060). Therefore, COV appears sensitive to increasing DAT severity.
COV from the timing task for all four participant groups across the unpaced intervals (A) for the first 90 ITIs and (B) for the full trial at each tapping rate. Error bars are ±1 standard error. The asterisk (*) indicates that in the comparison between just the healthy older adult group and the CDR .5 group, the groups differed at the indicated tap rate. YA = younger adults; OA = healthy older adults; CDR = Clinical Dementia Rating.
Discussion
The purpose of the current study was to examine whether aging and the earliest detectable stages of Alzheimer's disease lead to declines on a simple repetitive tapping task, particularly at slower rates where extant literature suggests increasing reliance on attention. We also lengthened time on task to increase the likelihood of performance drift and tap into vigilance processes. Our major focus, however, was to determine what conditions lead to DAT-specific declines above and beyond those associated with normal aging.
Age-Related Differences in Timing
As expected, YA were more accurate and resistant to drift across all tap rates than OA. All groups were fairly accurate at the fastest rate even when evaluated across the full trial (the condition which required production of the most ITIs). Thus, peripheral (motor) systems do not appear to be major contributors to accuracy declines. In general, OA showed an expected increase in tapping speed, especially at the 1500-ms rate. Surprisingly, aging did not lead to increases in variability as measured by SD. However, increases in COV did emerge. These patterns suggest that while healthy OAs’ overall level of variability was equivalent to YA, it increased disproportionately with their observed decreasing tap rate. Thus, aging is associated with reduced timing sensitivity at slower tap rates.
While OA's faster tapping could signify an increase in speed of an internal pacemaker when there is no external pacing signal, this seems unlikely, given that OA typically show a slower preferred tapping tempo compared to YA (McAuley et al., Reference McAuley, Jones, Holub, Johnston and Miller2006; Vanneste, Pouthas, & Wearden, Reference Vanneste, Pouthas and Wearden2001). Instead, as McAuley and colleagues (2006) argue, aging may be associated with narrowing of the interval range to which individuals can successfully entrain their motor responses. When repetitively reproducing an interval outside this preferred range, OA experience difficulty sustaining the rate and, therefore, adjust it toward their preferred tempo, thought to harbor around 650 to 750 ms (McCauley et al., 2006; Vanneste et al., Reference Vanneste, Pouthas and Wearden2001; although see Baudouin, Vanneste, & Isingrini, Reference Baudouin, Vanneste and Isingrini2004). Another possibility is that poorer online error correction processes contribute to difficulties sustaining an accurate representation of the target rate, thus increasing drift (Krampe et al., Reference Krampe, Doumas, Lavrysen and Rapp2010) and leading to tempo adjustments. Indeed, we found that OA had generally larger negative slopes than YA across a tapping trial.
Of interest, we found similar age-related changes in timing performance for the full trial and the first 90 ITIs. Therefore, we have expanded upon work by Duchek and colleagues (1994) to show that simply lengthening the target tap rate to tax controlled attention is sufficient to elicit normal age-related changes in timing performance, while extending the tapping bout may be less critical. This is consistent with work showing that vigilance appears stable with normal aging (Tucker et al., Reference Tucker, Stern, Basner and Rakitin2011). However, the role of bout length in eliciting age-related changes merits additional study, because while Krampe et al. (Reference Krampe, Doumas, Lavrysen and Rapp2010) did not find these changes during tapping at a supra-second rate under single task conditions, we did. In their study a single bout at the slowest rate lasted less than a minute while ours required tapping for approximately 3 min.
Differences Between Healthy Aging and CDR .5
We focus now on the behavioral timing differences specifically linked to the earliest stages of DAT. In general the CDR .5 group showed faster rates of tapping, increased variability, and larger negative slopes than healthy OA, indicating they had more difficulty maintaining an accurate representation of the target interval across the tapping trial, even when evaluated only across 90 ITIs. These patterns were similar and, in some cases, more pronounced at increased disease severity as shown by the CDR 1 group. Measures of accuracy for the CDR .5 group only reliably differed from healthy OA at the 1500-ms tap rate, suggesting that this condition was most sensitive to DAT-specific accuracy declines. Differences in SD and COV, however, emerged across even faster tapping rates, supporting the idea that increases in intra-individual variability, even in timing, may be a particularly sensitive marker of DAT (Duchek, Balota, Tse, Holtzman, & Goate, Reference Duchek, Balota, Tse, Holtzman and Goate2009; Hultsch et al., Reference Hultsch, MacDonald, Hunter, Levy-Bencheton and Strauss2000). The emergence of significant variability differences between the OA and very mild DAT group at all tap rates across just the first 90 ITIs suggests that there may be an ideal task length over which healthy OA maintain stability in performance, but individuals with the earliest signs of Alzheimer's disease (CDR.5 individuals) do not. Note that even 90 ITIs produced at the fastest rate in our study constitute a bout approximately nearly 3 times longer than that typically used for sub-second tap rates in this paradigm (Duchek et al., Reference Duchek, Balota and Ferraro1994; Ivry & Hazeltine, Reference Ivry and Hazeltine1995).
Much extant work on timing in DAT has investigated temporal discrimination rather than response timing. The most consistent result from this earlier work points to increased performance variability at early stages of DAT, similar to the current findings (Caselli, Iaboli, & Nichelli, Reference Caselli, Iaboli and Nichelli2009; Nichelli et al., Reference Nichelli, Vernneri, Molinari, Tavani and Grafman1993; Papagno et al., Reference Papagno, Allegra and Cardaci2004; Rueda & Schmitter-Edgecombe Reference Rueda and Schmitter-Edgecombe2009). We have now demonstrated that DAT-related breakdowns emerge even for a simple repetitive tapping task when target tap rates are lengthened. In general, it appears that DAT-related changes in the current paradigm are similar to those found under conditions of divided attention with longer intervals (Krampe et al., Reference Krampe, Doumas, Lavrysen and Rapp2010). McAuley and colleagues (2006) argue that task event structure strongly influences the degree to which attentional resources are coordinated to support performance. With continuous tapping, rate can influence one's ability to focus attentional resources at the right moments in time to facilitate accurate performance. This stems from the dynamic attending approach which argues that attention involves both internal temporal expectancies and rhythms created by external events (e.g., a pacing signal, see Large & Jones, Reference Large and Jones1999). The internal rhythm is consistent with McAuley's view that ideal tempo drives predictions about where and when to focus attention. Successful “attending” to external events occurs when synchrony is achieved via entrainment of internal to external rhythms. In the current task, achieving synchrony after disappearance of the pacing signal requires that individuals generate expectancies about when they should press the button for each subsequent reproduction, engaging attentional selection, inhibition of competing internal expectancies, and error correction processes. Since individuals at the early stages of DAT experience breakdowns in attentional control systems that support these processes as well as the maintenance of current task goals (Balota & Faust, Reference Balota and Faust2001), they experience more difficulty resolving these competing influences. This leads to poorer error-correction processes, regressions toward their ideal tempo, and increased drift and variability, especially at the slower target rates where there are larger deviations between the target ITI and internal expectancies.
We note that the age- and DAT-related changes in our study appear at first glance to be inconsistent with the predictions of the pacemaker-accumulator account of timing involving an attentional gate (Rakitin, Reference Rakitin2005; Zakay & Block, Reference Zakay and Block1997). Namely, full attention to timing during paced ITIs followed by OAs’ difficulty attending to time during the continuation phase should have led them to tap more slowly and show increased variability that was most marked at the fastest tap rate. However, results consistent with this model are typically obtained with the use of discrete, as opposed to repetitive timing tasks. As mentioned earlier, studies using a repetitive tapping paradigm obtained results similar to those found in the present study, even for YA under divided attention (Krampe et al., Reference Krampe, Doumas, Lavrysen and Rapp2010; McAuley et al., Reference McAuley, Jones, Holub, Johnston and Miller2006). This suggests that continuous repetitive tapping engages additional online processes that are not used during discrete timing tasks.
Conclusion
In summary, both DAT and healthy aging are linked to poorer accuracy and sensitivity at the slowest tap rate. Poorer variability in early stage DAT also consistently emerged for the 1000- and 1500-ms rates, and in some cases, for the 500-ms rate. Thus, mechanisms that contribute to timing variability may suffer greater impacts at early stages of DAT than mechanisms associated with overall accuracy. Explorations of these mechanisms and identification of tasks which are sensitive to these increases in intra-individual variability may be advantageous for developing useful tools for discriminating healthy aging from DAT. As shown here, a timing paradigm involving the continuous marking of events in the supra-seconds range over an extended period could prove helpful. Benefits of the current paradigm include its simplicity and the fact that the slowest tapping condition takes less than 5 min. Further work examining performance at longer intervals with even higher disagreement between internal and external temporal expectancies may be useful for identifying the conditions most sensitive to these early declines.
Acknowledgments
This work was supported by grants NIA T32 AG000030-32, NIA PO1 AGO3991, and NIA PO1 AGO26276. We thank Christopher Grant, Amy Heidebreder, Rebecca Howard, and Jeanne Mishkin for their help with data collection; Jeremy Missuk for his invaluable assistance with coding and analyzing data; John Morris and the Clinical Core for their careful description of the participants; Martha Storandt for the psychometric data; and Jan Duchek for her help in coordination of the data analyses. The authors report no conflicts of interest.
