FIGURE 15.3 Distribution of lever presses and cortical activity as a function of trial time. (A) Cortical neuron showing modulatory activity in relation to time in the trial. This classification was given to neurons showing significant differences in their spike rate between 10 and 40 sec, but with the magnitude of change from the baseline to the peak at 40 sec less than twice that from the baseline to the peak at 10 sec. (B) Cortical neuron showing ramp activity in relation to time in the trial. This classification was given to those neurons that had significant differences in firing rate between 10 and 40 sec, and visual inspection showed monotonic changes across the trial. Figure notation and data smoothing are identical to that in Figure 15.2.

as a function of time had similar peak-shaped activity patterns, these data suggest that a subset of striatal neurons either are the decision stage output of an interval timer or are downstream from it.

In addition to being representative of a decision stage output function, the striatal activity pattern is also quite similar to the lever press distribution occurring around the short criterion duration. Such a match between the decision stage and motor output is prescribed in the GTM and would, under many experimental circumstances, prevent one from being able to discern in which information-processing component these striatal neurons should be placed in (i.e., decision stage or output stage). However, as can also be seen in Figure 15.2, this behavioral-neural firing match becomes dissociated as a function of time in the trial, as there is a much smaller secondary increase in firing during the behavioral scallop at 40 sec. Given this clear dissociation between the neural activity pattern and the overt behavioral expression of the rat as time approaches the second reinforcement duration, we can conclude that striatal activity does not directly encode the motor programs involved in pressing, and its role in the timing circuit must therefore be upstream from the neural circuitry producing the behavioral output. Thus, we are led to conclude that this striatal firing encodes the decision stage output of the internal clock.

In contrast to the clear peak observed in Figure 15.2, the cortical neurons showed only modulatory or ramp-like firing patterns (Figure 15.3). As can be seen in Figure 15.3A, this modulatory firing pattern generally matched the lever press distribution across the entire trial, with only a slight, although significant, difference in firing rate across the two criterion times. This sort of firing pattern does not match any prescribed functions of the GTM and, given its similarity to the entire lever press distribution, suggests that this type of cortical activity falls on the behavioral output end of the process.

The cortical ramp pattern is, on the other hand, similar to functions that have been proposed to serve as the integration function of the clock signal, notably the linear signal of scalar expectancy theory (Gibbon, 1977) and the logarithmic decay signal of the multiple timescales model (Staddon and Higa, 1999). Despite this similarity, we are not yet comfortable making any inferences regarding the precise role of these cells, for several reasons. First, we had only a small number of cells that showed ramp-like activity during the trial (two cortical cells and one striatal cell). Second, of these ramp cells, both the striatal cell and one of the two cortical cells showed the ramp pattern beginning at delivery of food reward without any alteration in slope at signal onset, suggesting that these cells were not under stimulus control. Because we utilized variable-length intertrial intervals, the failure to show stimulus control suggests that these cells were not serving as an integration function. However, because all of these ramp patterns provide systematically changing levels as a function of time in the trial (i.e., they are isochronic), they may be a potential clock signal and warrant future investigation.

Given that we found peak firing patterns in a subset of striatal cells, but failed to find these same peak patterns in the cortex, these data suggest that the striatum generates, rather than relays, the decision stage output function for an internal clock. Although we acknowledge the possibility that a peak-shaped input may be passed to the striatum by way of other cortical neurons or areas than those from which we recorded, we chose to investigate the cingulate cortex, as it is primarily afferent to the striatal area from which these peak data are derived (Sesack et al., 1989). Therefore, we currently favor the proposal that these striatal neurons are performing the necessary computations, either individually or as an ensemble, to generate a decision stage output.


The generation of a decision stage output in the firing of striatal neurons is consistent with the predictions of the SBF model. Although we do not yet have conclusive data regarding the anatomical location of the clock and memory processes, the SBF model proposes that clock integration and memory stages also occur within these same peak-generating neurons. Given the possibility that single striatal neurons can function as the majority of an interval timing system, we are suddenly confronted with two possible scenarios for implementation of an internal clock in the brain. One notion is that interval timing striatal neurons may be localized to a specific striatal area. In this scenario, interval timing would be carried out by an anatomically localizable group of striatal neurons, all performing roughly the same operations, with the output of these neurons sent to all motor and cognitive cortical areas requiring temporal information. The alternative idea is that single neurons functioning as interval timers may be distributed throughout the striatum, such that they temporally modulate the output of the functional area in which they are localized. As such, interval timing may not be a separate function occurring in one region of the striatum, but an important component operating within this behavioral and cognitive control system. An evaluation of both the anatomical arrangement of the cortico-striato-thalamo-cortical circuit and the physiological data from behaving primates suggests that the distributed interval timing hypothesis is a more likely scenario for temporal modulation of behavioral and cognitive processes.

15.5.1 Multiple Processes

An example of the many separate pieces of temporal information utilized by an organism while interacting with its environment may be useful in arguing this point. One common experimental task utilized in primates is the delayed match-to-sample procedure. This procedure progresses through the following phases:

1. The subject rests its hand on a central button.

2. An instructional cue is briefly presented after some delay (e.g., sample A or sample B).

3. An action cue is presented following another delay.

4. The subject performs the behavior designated by the instruction cue (e.g., press button associated with sample A).

5. The reward is delivered following a third delay.

It is clear that many different temporal relations exist within this task, and that the subject must modulate its cognitions and behaviors to conform to these temporal relations. First, attention shifts in a temporally dependent manner to the onset of the instruction cue, action cue, and reward output. Along with these preparatory atten-tional shifts are changes in gaze that will also have a nonuniform temporal distribution. Second, the subject needs to prepare to make a response upon onset of the action cue. Third, processing related to the upcoming delivery of reward is undoubtedly occurring, such as preparatory increases in salivation. At the same time, inhibition of other competing behaviors and attentions must be heightened in a temporally modulated manner in order to prevent behavioral inefficiency.

Because all of these behaviors and cognitions utilize different sources of temporal information (e.g., the preparation of a motor response occurs at a different time than the expectation of reward), there will be a variety of different temporal decisions produced by those striatal neurons performing temporal computations. Electrophysiological data from these types of studies demonstrate that such varied temporal processing, or at least preparatory- and expectation-related processing, is occurring. Specifically, separate striatal neurons have been found to fire in a preparatory manner for the instructional cues, the action cues, and reward (Schultz and Romo, 1992). Differently timed preparatory motor activity for the different behavioral requirements has also been well documented in the striatum (Jaeger et al., 1993, 1995). Thus, in order to modulate the numerous behavioral and cognitive processes required for successful completion of the task, the appropriate temporal information will need to be utilized by the appropriate control structure (presumably anatomically and functionally separate cortical areas). In other words, this temporal information needs to be sent to the appropriate output region of the GTM.

15.5.2 Localized Proposal

In the localized hypothesis, a variety of temporal criterions would be output by a single striatal area that is specialized for timing. Although this localization hypothesis is consistent with the functional topography that has been shown to exist in the striatum (Alexander et al., 1990; Gerfen and Keefe, 1994; McGeorge and Faull, 1989; Selemon and Goldman-Rakic, 1985; Webster, 1961), in order for these timing neurons to impact a variety of cognitive and motor processes, the temporally informative output would eventually need to project to the variety of cortical areas that will utilize this information (e.g., motor cortex for behavioral modulation, cingulate cortex for attentional modulation). However, the anatomical separation of function found in the striatum is preserved as it moves through the basal ganglia output channels on its way back to the cortex (Hoover and Strick, 1993). Therefore, these anatomical data would force the processes of selection and transmission of the appropriate temporal signal to occur within the cortex (i.e., different temporal signals would need to be sent from the cortical area receiving the timing information to the cortical areas requiring the timing information). Although such a cortical selection-transmission requirement is not unfeasible, we feel that it is a less efficient system and that the process is more vulnerable to disruption than that utilized by a distributed timing system.

15.5.3 Distributed Proposal

In a distributed timing system, the criterion time computed by a particular striatal region is restricted to those times that are utilized by the functional region of the cortex to which the striatal region projects. In other words, because temporal information regarding the expected time of the reward is not output from the forearm motor region of the striatum, time-of-reward information does not reach the region of the cortex involved in controlling forearm movement. Instead, only the temporal information specifying when a forearm movement is to be made is passed to this cortical motor region. The anatomical specificity that is required for selecting and transmitting which durations to associate with which behavioral and cognitive processes is already in place, and thereby eliminates the need for the cortex to perform these processes. As such, processing time in a distributed manner seems more likely in terms of the efficiency of information transfer.

Distributed striatal processing of time is also consistent with the cortico-striatal topography and predicts that anatomically focused striatal and cortical regions would be activated during timing tasks. However, the specific cortical and striatal regions that are activated would be highly dependent on the behavioral and cognitive demands of the task. Moreover, the electrophysiological data from the delayed match-to-sample studies described above came from various dorsal and ventral striatal areas, which suggests that such preparatory activity is indeed distributed, rather than localized to a single striatal region. For these reasons, we propose that the interval timing system provides its information in the form of temporally specific activation of striatal neurons that are primarily involved in processing motor or cognitive information through the basic mechanisms proposed in the SBF model.

15.5.4 Neural Evidence

Our ensemble recording data also support the notion that the processing of time is distributed in the striatum. One consequence of a distributed timer is that those striatal neurons showing decision stage output functions are embedded within functional striatal areas that perform processes other than timing. Indeed, we found that the large majority of striatal neurons had small, secondary activity peaks at the other criterion time (e.g., a large neural peak at 10 sec and a smaller neural peak at 40 sec). These different-sized peaks may be the resultant sum of both lever pressing-related activity and temporal expectation-related activity. Specifically, the smaller neural peak at 40 sec would reflect activity solely related to lever pressing, whereas the larger peak at 10 sec would reflect both lever pressing-related activity and the temporally informative decision stage activity.

In order to evaluate this possibility, we compared the neural activity associated with lever pressing during the 10-sec press period with the neural activity associated with lever pressing during the 40-sec press period. As can be seen in Figure 15.4, there is a phasic increase in the firing rate of this neuron immediately before and after a lever press. This modulation in firing rate occurs during both the early and late presses. The behaviorally locked decrease in firing seen in this figure indicates that this neuron encodes some motor or cognitive process associated with lever

Press-Related Activity as a Function of Duration w ®

Early Presses

Early Presses

0 0

Post a comment