Shortinterval Timing

Timing is a guiding force in behavior. Accurate and flexible perception of timing affords predictions about when events in the environment are likely to occur and how long they are expected to last. In the literature, short-interval timing refers specifically to event durations in the seconds-to-minutes range. Examples of human behaviors that involve short-interval timing range from having expectations about when a traffic light will turn from red to green to dancing to the beat of music (Matell and Meck, 2000).

A wide variety of theories of short-interval timing have been proposed (Church and Broadbent, 1991; Gibbon, 1977; Jones, 1976; Killeen and Fetterman, 1988; Large and Jones, 1999; McAuley, 1996; McAuley and Kidd, 1998; Miall, 1989; Treisman, 1963). These theories generally fall into two classes: approaches from a dynamical system perspective that involve coupled oscillators (Large and Jones, 1998; McAuley and Kidd, 1998) and approaches from an information-processing perspective (Church and Broadbent, 1991; Gibbon, 1977; Treisman, 1963).

An emerging distinction between models concerns the perception of durations shorter than a couple of seconds vs. the perception of durations longer than a couple of seconds. One proposal is that different systems are involved in timing events on these two scales (Ivry, 1996; Ivry and Hazeltine, 1995). For event durations shorter than a couple of seconds (e.g., the time intervals defined by successive beats in a musical performance), it has been suggested that temporal regularity plays more of a role in timing processes than explicit memory because patterns of event durations on this scale form directly perceivable rhythms (e.g., music and speech); hence, timing in this range is predictable by virtue of the exogenous timing cues provided by stimulus markers (Fraisse, 1963; Jones, 1976; Jones and Boltz, 1989; Port, 1995). This issue and related ones pertaining to effects of rhythmic context on time perception have recently generated interest from dynamical systems and information-processing theorists (Barnes and Jones, 2000; Ivry and Hazeltine, 1995; McAuley and Jones, 2002; McAuley and Kidd, 1998; Pashler, 2001). The emphasis of this chapter, however, is on the perception of events that occur in isolation (in the absence of a prevailing temporal context) and that last longer than a couple seconds. On this timescale, perceived rhythm tends to break down, and it is more clear-cut that behaviors rely on explicit memory for duration (e.g., knowing how long the red light lasts at a particular intersection).

14.2.1 Scalar Expectancy Theory

Most theories that incorporate explicit memory for time are information-processing models, which involve three independent components: an internal clock used to estimate duration, a reference memory used to store information about duration, and a comparison mechanism used to make judgments about how much time has elapsed relative to a remembered (expected) standard duration (Church and Broadbent, 1991). Within this framework, scalar expectancy theory (SET) has been particularly influential because it has been successfully applied to both human and animal data (Gibbon, 1977; Gibbon et al., 1984; Rakitin et al., 1998). SET posits a neural pacemaker that emits a continuous stream of pulses. Stimulus events marking the beginning and ending of event durations trigger the closing and opening of a switch that gate pulses into an accumulator. The count of the pulses accumulated over the target event duration represents a subjective duration code that is stored in reference memory. Successive time intervals are estimated independently, with relative duration judgments about time intervals involving a comparison between a working memory representation of the accumulator and a criterion time sampled from reference memory. A schematic of the various components of SET is shown in Figure 14.1.

FIGURE 14.1 Schematic diagram of scalar expectancy theory depicting the relation between important components of the model.

14.2.2 Peak-Interval and Simultaneous Temporal Processing Tasks

SET has been used to model human and animal timing in a variety of behavioral tasks. One task, in particular, that is frequently used to assess animal timing is the peak-interval (PI) procedure (Roberts, 1981). The PI procedure is a variant of a fixed-interval (FI) reinforcement schedule. Participants are trained to associate the onset of a light or tone stimulus with the expectation of reinforcement after a fixed time interval. On PI (probe) trials, the stimulus stays on for twice the duration of the FI and no reinforcement is given. The data of interest are the lever presses that the animal makes in the absence of feedback about duration (i.e., probe trials). Human versions of the PI procedure involve essentially the same task, but differ primarily in the method of feedback (Rakitin et al., 1998). Because human participants can be directly instructed about the task, feedback about the duration on FI trials is simply the offset of the stimulus. Similar to probe trials for animals, the stimulus remains on for at least twice the duration of the fixed interval, and the data of interest are the responses that the participant makes in the absence of feedback about the target duration.

During a typical simultaneous temporal processing task, participants are asked to perform two concurrent PI procedures involving pairs of fixed intervals (Meck, 1987; Meck and Church, 1984; Olton et al., 1988). The shorter of the pair of fixed intervals is referred to as the short stimulus, and the longer of the pair as the long stimulus. As in the standard PI procedure, half of the trials provide the opportunity for reinforcement, and the other half are unreinforced probe trials. Data are analyzed only for probe trials, which can be simple or compound. On simple trials, participants are presented with a single stimulus and can focus attention on reproducing the expected duration of the stimulus. On compound trials, both stimuli are presented together with the short stimulus generally embedded within the long stimulus, and participants divide their attention between the two stimulus durations.

14.2.3 Basic Findings

In both the standard and STP versions of the PI procedure, peak functions are obtained by pooling probe trial data across trials and constructing composite relative frequency histograms for different target durations (i.e., fixed intervals). The precise location of the peak on probe trials is referred to as peak time and is taken as the expected time of the target. In both human and animal studies using the PI procedure, three robust behavioral findings concern the mean, variability, and shape of peak functions. For normal participants, peak functions in both focused and divided attention conditions (simple and compound trials) are approximately (1) centered on the target duration, (2) scalar in variability, and (3) normal in shape (Church et al., 1994; Meck and Williams, 1997; Olton et al., 1988; Rakitin et al., 1998). An especially surprising aspect of STP performance is that in some cases, there is little decrement in timing performance when participants must divide attention between timing two durations compared with timing one duration (Meck and Williams, 1997; Olton et al., 1988). For example, in previous animal studies, normal rats have been shown to accurately time up to three durations, as evidenced by the good correspondence among peak times on compound trials and single PI trials (Meck, 1987).

Results obtained from human participants in our lab during focused attention conditions illustrate the three basic features of the PI procedure (Figure 14.2). Participants were trained to estimate the duration of two time intervals (5 and 8 sec), which were associated with two tones differentiated in pitch (low vs. high). Peak functions were obtained by pooling probe trial data across participants and sessions and constructing composite relative frequency histograms for the 5- and 8-sec target durations. Peak functions on an absolute timescale (Figure 14.2A) and on a relative timescale (Figure 14.2B) were constructed.

As can be seen in Figure 14.2, the distributions of produced time intervals are approximately centered on the target duration. For the 5-sec target, the peak time was 4.91 sec, whereas for the 8-sec target, the peak time was 7.78 sec. In addition, the mean coefficients of variability for the 5- and 8-sec targets were approximately equal (0.143 vs. 0.138, respectively), demonstrating the scalar property. Finally, plotting response times on a relative timescale yields near-perfect superimposition of the two response distributions (Figure 14.2B).

14.2.4 Effects of Neurobiological Manipulations

Previous research has shown that both transient and permanent shifts in peak time (corresponding to under- or overestimation of the criterion duration) occur following various behavioral and neurobiological manipulations. Transient shifts refer to changes in peak time that disappear following continued training with the criterion duration, whereas permanent shifts refer to changes in peak time that remain following continued training with the criterion duration.

FIGURE 14.2 Performance of adult human subjects in a modified peak-interval procedure during focused attention.

Transient shifts in peak time are observed following manipulations of the dopaminergic system, presumably of the nigrostriatal pathway (Meck, 1986, 1996). Drugs that block dopaminergic receptors, specifically the D2 subtype, produce a rightward shift in peak time (corresponding to overestimates of duration), whereas drugs that stimulate D2 dopaminergic receptors produce a leftward shift (corresponding to underestimates of duration).

One explanation of transient shifts is that they reflect changes in clock speed. According to SET, a change in pacemaker rate under the influence of drugs will alter the rate that clock pulses collect in the accumulator. The result is that it takes more or less time to accumulate the number of pulses that correspond to the previously reinforced count. However, with continued training under the influence of the drug, the number of pulses corresponding to the old criterion duration is gradually updated with a new count corresponding to the expected time of reinforcement based on a clock with the altered speed. These findings have led researchers to posit that dopaminergic neurons in the basal ganglia function as the clock (Meck, 1996). However, see Matell and Meck (2000) and Meck and Benson (2002) for an alternative interpretation.

Permanent shifts in peak time occur following damage of the hippocampus and frontal cortex (Olton et al., 1988). An interesting, yet unexplained, finding by Olton et al. (1988) is that shifts in peak time following hippocampal and frontal cortex damage are in opposite directions. Hippocampal lesions produce leftward shifts in peak time, whereas frontal cortex lesions produce rightward shifts in peak time. These effects of brain lesions remained despite extended training, in contrast to the effects of dopaminergic drugs. Thus, it has been suggested that different components of the timing system were affected by dopaminergic drugs and lesions of the hip-pocampal and frontal cortex (Meck, 1996). One proposal is that the hippocampus and frontal cortex are involved in maintaining reference memory. SET captures these effects by scaling the output of the accumulator by a proportional factor, referred to as a memory constant (Meck et al., 1986; Meck and Williams, 1997). This account, however, does not identify the specific role that the hippocampus and frontal cortex have in the reference memory system, and it does not explain why peak time should shift in opposite directions when these brain structures are damaged.

14.2.5 The Attentional Switch Hypothesis

There have been several proposals about how attention might influence timing performance within the scalar expectancy framework. One proposal is that attention influences the probability that participants' behavior is controlled by stimulus timing on a given trial (Church and Gibbon, 1982; Meck and Williams, 1997). In this view, divided attention decreases the probability that attention is focused on any single stimulus, resulting in an increased asymmetry of the response distributions for each stimulus (Roberts, 1981). An alternative proposal is the attentional switch hypothesis. We have chosen to focus on this second proposal because, in our view, it provides the most consistent explanation of STP data from both human and animal participants (Buhusi, this volume; Fortin, this volume).

The attentional switch hypothesis proposes that attention operates as a switch at the clock stage. The attentional switch influences timing by altering the efficiency with which pulses from the pacemaker are transferred to the accumulator (Allan, 1992; Lejeune, 1998; Macar et al., 1994; Meck and Benson, 2002). Under focused attention conditions, pulses accumulate as a function of time, and the subjective experience of duration is directly proportional to the count of the number of pulses that occur over the temporal extent of the stimulus. However, when attention is divided between two tasks (e.g., a temporal and a nontemporal task or two timing tasks, as in the STP procedure), the assumption is that some pulses may be "lost," with the proportion of lost pulses inversely related to the amount of attention allocated to the timing task.

The implication for behavior is that if attention is directed away from the timing task or is divided between two or more timing tasks, then the effective rate at which pulses collect in the accumulator is reduced due to lost pulses. If participants respond when a given number of pulses have collected in the accumulator (i.e., the count reaches a criterion value), then systematic errors in time estimation should occur. Indeed, many previous studies with human participants have reported data consistent with this view (Brown et al., 1992; Brown and West, 1990; Fortin and Masse, 1999, 2000; Hicks et al., 1977; Macar et al., 1994; Zakay, 1989).

In multiple timing tasks, similar to the STP procedure, Brown et al. (1992; Brown and West, 1990) have found that time judgments during divided attention conditions are less accurate and more variable than during focused attention. Moreover, increasing the number of timed intervals produces a progressive deterioration in time judgment performance (Brown et al., 1992; Brown and West, 1990). When participants are asked to provide time estimates while concurrently performing a nontemporal task (e.g., mental arithmetic), the accuracy of verbal estimations and reproductions decreases monotonically with increasing processing demands of the concurrent nontemporal task (Brown, 1985; Brown and West, 1990; Hicks et al., 1977; Macar et al., 1999; Zakay, 1989). Similarly, Fortin and Masse (1999, 2000) have shown that dividing attention during a time production task results in longer productions (overestimates of the target duration) than in focused attention conditions.

14.2.6 Summary

In summary, SET has been used to explain timing behavior in rats and humans for intervals greater than a couple of seconds. The three major components of SET are the clock, reference memory, and comparator. One way that attention has been incorporated in SET is as an attentional switch at the clock stage. This view is consistent with both human and animal data. Neurobiological studies have suggested that the dopaminergic nigrostriatal pathway may be part of or interact directly with the clock. Other evidence supports the view that the hippocampus and frontal cortex have integral roles in maintaining reference memory. A number of recent functional imaging and electrophysiological studies in animals and humans also link various areas of the frontal cortex (including the prefrontal, premotor, anterior cingulate, and supplementary motor areas) to aspects of temporal processing other than reference memory (Harrington et al., 1998; Macar et al., 1999; Mangels et al., 1998; Macquet et al., 1996; Monfort et al., 2000; Rao et al., 1997; Rubia et al., 1998; Tracy et al., 2000). The remainder of this chapter focuses on the involvement of the frontal cortex in attention and timing (Olton and Pang, 1992).

0 0

Post a comment