Attention is not the only aspect of cognitive function that changes with age; memory problems are a common complaint of older adults (Craik and Jennings, 1992; Light, 1991; Zacks et al., 2000), and an apparent age-related decline in processing speed is often suggested as an underlying reason for age differences on many cognitive tasks (e.g., Salthouse, 1996). As described above, reference memory is a central component of many information-processing models of interval timing. Processing speed plays a special role in mathematical formulations of SET, as substantiated in the K* parameter, the speed at which accumulator values are encoded into and decoded from reference memory (Gibbon et al., 1984; Meck, 1983).
The contributions of these components to age differences in interval timing have not been as heavily researched as the contribution of attention, in part because they are not as amenable to experimental manipulations that can be easily implemented in humans. Furthermore, at least some interval timing experiments using human participants are designed in a way that minimizes the contribution of reference memory as typically defined in information-processing models of interval timing and tested in animal experiments (see Wearden and Bray, 2001). However, recent attempts have been made to examine the relationship between age differences on interval timing tasks and neuropsychological tests designed to target specific cognitive functions (Perbal et al., 2002), and to ask whether older adults may show memory deficits that are specific for time (Rakitin et al., submitted).
Perbal et al. (2002) examined the relations between young and older adults' scores on neuropsychological tests and their performance on different interval timing tasks. The neuropsychological measures included were designed to differentially emphasize processing speed, passive short-term memory storage, working memory (online storage and processing), and long-term, delayed memory storage. The inter val timing tasks included a production task under conditions of divided attention (read-aloud digits presented at random intervals) and a reproduction task in which attention was divided during encoding of the standard, but not its reproduction.***
Standard age and divided attention effects were found on the timing tasks, with older adults overproducing and underreproducing the target intervals to a greater extent than did young adults. Performance on the neuropsychological measures of working memory and long-term memory was moderately correlated with accurate performance on the timing tasks, especially reproduction. There was some tendency for these correlations, especially for long-term memory, to be greater for longer (14 and 38 sec) than shorter (5 sec) intervals. Moreover, when age differences in working memory and long-term memory were statistically controlled for, age differences for longer durations in the reproduction task were no longer statistically significant. Much of the age-related variance in the production task was accounted for by processing speed.
Perbal et al. (2002) conclude that working memory and processing speed contributed to the ability to divide attention between the digit-reading and timing tasks. (It should be noted that this "executive function" of working memory is also often described as working attention (e.g., Baddeley, 1993; Engle, 2001).) They explain the correlations between long-term memory and interval timing performance in the longer durations by suggesting that for long durations, pulses collected early in the interval eventually pass out of the accumulator and are stored in long-term memory. If this were the case, older adults' overproduction and underreproduction would result from deficits in two aspects of cognition: First, age deficits in working memory (or attention) would cause them to miss more pulses during initial accumulation. Second, deficits in long-term memory would cause older adults to lose more pulses from storage for the longer durations than would young adults.
The Perbal et al. (2002) experiments represent an important step in relating interval timing performance to performance on standard measures of attention and memory, but several caveats should be kept in mind when interpreting their results. First, an earlier experiment by this group (Vanneste and Pouthas, 1999) using a task that divided attention between multiple to-be-timed stimuli and a different measure of working memory did not find any relation between timing performance and working memory for either young or older adults. Second, the function of long-term memory as described by Perbal et al. is more closely related to accumulator functioning than to reference memory as usually conceived in models of interval timing. Finally, it is important to note that the contributions of both working memory and processing speed as described by Perbal et al. act through attention's mediation of the switch that gates pulses through the accumulator, not the rate of pacemaker pulsation or K*.
The reference memory component of the interval timing model has been the focus of several recent investigations. McCormack et al. (2002) trained young and
*** Perbal. et al. (2002) also included a control counting condition, in which participants counted aloud while producing the target interval in the production task and while encoding and reproducing the interval in the reproduction task. Participants were extremely accurate in this condition, and there were no age differences, so it will not be discussed further here.
older adults on six tones of increasing duration. On test trials, participants were asked to identify which of the six durations had just been presented. Young adults were quite accurate, but older adults misclassified the tones as being shorter than they really were. In contrast, when asked to make judgments about the pitch of tones rather than their duration, older adults made more errors than did young adults but did not show the same systematic underestimation as for duration. McCormack et al. suggested that older adults had a duration-specific distortion in memory such that they remembered the standards as being longer than they really were. This suggestion receives support from previous human and animal studies suggesting an age-related distortion in duration memory (McCormack et al., 1999; Meck, 1986; Wearden et al., 1997), but the same results could also occur if older adults showed a greater drift of attention from training to test trials.
Rakitin et al. (submitted) found age differences on a reproduction task that they attributed to age differences in reference memory for the target durations. They used a peak-interval reproduction procedure in which participants attempted to reproduce a duration learned in an earlier training session. When a single target duration is used, older adults will overreproduce the target more than will young adults (Mal-apani et al., 1998). As described previously, this result likely occurs because older adults' deficits in controlled attention cause them to miss pulses during test trials, requiring more physical time before current accumulator values match those associated with the target duration in reference memory. Rakitin et al. (submitted) found a very different result when they used two durations in the same session. Under these conditions, older adults overreproduced the shorter of the two intervals, but underreproduced the longer interval. Interestingly, this "migration effect" appeared to be specific to memory for durations; no such effect was found for a line-length reproduction task that also used a short and long standard.
Rakitin et al. (submitted) suggested that older adults' vulnerability to the migration effect for durations may be attributed to age-related declines in dopamine function. A very similar effect is found for Parkinson's patients, whose disease stems from a dopamine depletion (Malapani et al., 1998, 2002a). However, the migration effect in non-Parkinson's older adults was not proportionate: the overestimation of the short duration was greater than the underestimation of the long duration. In fact, underestimation of the long duration was only significant for a subgroup of the older adults. Rakitin et al. (submitted) suggested that the migration effect may be a marker for older adults who have especially pronounced declines in dopamine function and are thus particularly vulnerable to the memory effects.
The idea of a vulnerable subgroup of older adults can explain why only some older adults showed a significant underestimation of the long interval, but it does not account for the uneven effects for the short and long targets. One possibility is that age effects on attention and memory compound each other for the short duration and counteract each other for the long duration. That is, for the short duration, attention problems would lead to a loss of pulses and a tendency to overproduce the interval to match the memory representation of the target, a representation that would be inappropriately large because of migration toward the longer duration. For the long duration, attention problems would likewise lead to a tendency to overproduce the interval to match the memory representation of the target, but in this case, migration would shorten that representation, causing the attention and memory effects to partially cancel each other out.
Patterns of scalar and nonscalar variability for the Parkinson's and non-Parkinson's older adults support the idea of a dopaminergic influence. In addition to its role in cognition, dopamine plays an important role in motor functioning; the motor problems caused by dopamine deficits are the primary reason that Parkinson's patients medicate the disease. As described in the introduction, manipulations that influence clock and memory components of the SET model maintain the scalar property of increasing variability with increasing duration (Gibbon et al., 1984, 1997). Manipulations that lead to nonscalar variability are thought to have their effect through nontemporal avenues, such as response criterion or motor output. In the experiments conducted by Rakitin et al. (submitted; Malapani et al., 1998, 2002a), older adults were accurate in training trials, which were conducted with feedback, and inaccurate on delayed test trials without feedback, but showed nonscalar variability under both conditions. Parkinson's patients tested off the medication that remediates their dopamine deficit were both inaccurate and nonscalar for both training and testing trials, but both the migration effect and nonscalar variability were remediated by medication. The conclusion suggested by this pattern is that nonscalar production results from problems with motor processes, whereas the migration effect is located at attention and memory processes, and that both are influenced by dopamine.
To summarize, most of the research on age differences in interval timing has focused on the role of attention, and much less is known about the influence of memory, processing speed, or other cognitive characteristics that change with age. The experiments described above exemplify methods that may be useful for examining these influences. Correlations between interval timing performance and performance on neuropsychological tests designed to tap specific cognitive abilities can help establish connections between group and individual differences in both domains. In addition, the analysis procedures provided by SET and other information-processing models of timing can help to separate the relative contributions of attention, memory, and nontemporal processes to performance on an interval timing task.
Was this article helpful?