Info

identifying the anatomical and physiological underpinnings of the interval timing system. This transition to the study of the biological substrates of interval timing is well timed to stimulate further model development. Because the various interval timing models are already extremely accurate at predicting the behavioral data (Church and Broadbent, 1991; Gibbon, 1977; Killeen and Fetterman, 1988; Staddon and Higa, 1999), much of their attractiveness is associated with their philosophical approach (i.e., behaviorism vs. cognitivism), rather than their predictive accuracy. Although these models fare quite well at explaining behavioral data, because of their fundamental differences, they do not provide us with an unbiased framework from which to search for the neural mechanisms of interval timing. As such, we believe that a theory-free model of interval timing would be valuable. Such a general timing model is a much needed "place to hang our hats" when searching for the neural processes associated with timing and time perception.

In this chapter, we will discuss the basic components of an interval timer and describe how these components may be realized in the brain. To this end, we will present electrophysiological data showing duration-specific activity in the striatum and discuss the implications these data carry for current and future models of interval timing.

15.2 COMPONENTS OF AN INTERVAL TIMER 15.2.1 Generalized Timing Model

It has previously been proposed that information-processing models of interval timing are composed of clock, memory, and decision stages (Church, 1997) — a framework we will refer to as a generalized timing model (GTM). The clock stage encompasses the production of a temporal percept, with processes ranging from the generation of a utilizable temporal signal to the integration of this signal into a meaningful output. The temporal signal is usually characterized as single or multiple, fast or slow oscillators (Church et al., 1991; Gibbon, 1977; Miall, 1989). However, the temporal signal could conceivably be any type of patterned signal, so long as it can be reliably repeated, albeit with some error, for each opportunity to time. In contrast, the integration of the temporal signal is often characterized as a monoton-ically changing function, e.g., linear (Gibbon and Church, 1981) or log (Staddon and Higa, 1999), so that the percept of time varies in a systematic manner. The clock stage may also incorporate processes that allow temporal integration to start anew upon a biologically relevant signal onset (e.g., starting the pacemaker or resetting the accumulation process). Upon occurrence of an important event, the output of the clock stage is stored in long-term memory, thereby composing the memory stage. Given subsequent opportunities to time a similar event, the component processes associated with the clock stage are initiated, and the current clock reading is compared to previously stored values in long-term memory to produce a similarity function, thereby composing the decision stage. The temporally predictive behaviors of the organism are, by definition, based on the output of this decision stage. Each of these stages is conceived as being relatively orthogonal and serially organized from an information-processing perspective, in that information is passed in a single

FIGURE 15.1 Information-processing diagram for a generalized timing model. The core components specific to the processing of time are the clock, memory, and decision processes, indicated in dark boxes. The arrows indicate direction of information flow. Addition of stimulus input and cognitive and motor output extends this GTM to the processes necessary to produce a properly behaving organism. Although not dictated in terms of information flow, the dotted lines indicate the extensive array of feedback loops that may significantly modulate the processing of information.

FIGURE 15.1 Information-processing diagram for a generalized timing model. The core components specific to the processing of time are the clock, memory, and decision processes, indicated in dark boxes. The arrows indicate direction of information flow. Addition of stimulus input and cognitive and motor output extends this GTM to the processes necessary to produce a properly behaving organism. Although not dictated in terms of information flow, the dotted lines indicate the extensive array of feedback loops that may significantly modulate the processing of information.

direction (e.g., from clock to decision stage). The organization of the GTM is shown in Figure 15.1.

In the timing literature, reference to these processing stages has become common, and while we fully agree with this idea within an information-processing capacity, we are concerned that there may be an underlying tendency to parse these stages into anatomically separate brain regions, with each region processing information in a serial manner, akin to the GTM. The temptation may also exist to separate these timing components from other cognitive and motor systems in which such temporal information is utilized. We argue that assuming anatomical and computational independence for the component processes of interval timing is not valid. Similarly, assuming that the cognitive and motor systems which utilize temporal information are distinct from the processes composing the timing system is also unreasonable. Instead, we propose that all of these systems are highly interrelated.

Unlike the timing signal and the temporal integration function, which cannot be determined a priori, the output of the decision stage must be the basis of the temporally controlled behavior of the organism, and will therefore be highly similar in temporal patterning to these behaviors (i.e., the neural discharge of the decision process will be difficult to dissociate from the motor output in terms of their temporal structure). Temporal generalization procedures, such as the commonly used peak-interval procedure (Church et al., 1991; Roberts, 1981), require the subject to indicate when it expects the occurrence of a temporally predictable reward. The distribution of these behavioral indices (e.g., lever pressing) over time in the trial is generally described by a Gaussian-shaped peak centered very close to the normal time of reinforcement, with a width that is proportional to the peak time. This proportionality in spread has been termed the scalar property (Gibbon, 1977) and is a defining property of interval timing in the seconds-to-minutes range. We suggest that this response peak is isomorphic with the decision stage output function occurring in the organism while it is performing these production procedures. The output function may also form the basis for other interval timing judgments by serving as the basic substrate onto which other computations may be performed (e.g., from a rat's perspective, the duration bisection procedure might be a two-duration temporal generalization procedure, rather than a discrimination procedure).

15.2.2 Interdependence of Processes

In terms of a temporal percept (i.e., the current clock reading or a similarity comparison of now vs. not now), the GTM describes all of the processes necessary for timing. However, to understand how the behavior of an organism unfolds in time, two additional information-processing components are required (see Figure 15.1). These components are the stimulus input and cognitive and motor output functions that are needed to detect the signal to be timed and then to respond in a temporally meaningful manner. Although incorporation of these components may seem like unnecessary baggage for a GTM, their importance becomes clear upon the realization that such external processes are neither independent of the functioning of the timing processes nor in the determination of the actual behavior of the animal. The lack of independence between the timing process itself and the processing of the stimulus to be timed is demonstrated by experiments showing that auditory stimuli tend to drive the clock stage at a faster rate than visual stimuli (Penney, this volume; Penney et al., 2000). In terms of the actual motor output of the organism, it is perhaps unnecessary to state that even in our best experimental designs the behavior of an individual subject has many internal and external influences other than its temporal percept. Although behavioral data are frequently averaged across trials, sessions, and subjects with the hopes of characterizing the underlying decision function, averaged data of this sort cannot be used in real-time analyses such as electrophys-iological or brain imaging studies, in which the data of interest must be interpreted with respect to an individual subject's behavior. Furthermore, assuming independence between the timing processes and subsequent motor output excludes the investigation of feedback loops, which may play an important role in interval timing (see below). Due to these additional sources of influence, a GTM that incorporates these input-output components would be a more useful framework for investigating the neural components of the interval timing system.

0 0

Post a comment