Info

Time (100 msec bins) Time (100 msec bins) Time (100 msec bins)

i=i Striatal Neuron

FIGURE 15.4 Striatal neuron activity in relation to lever presses occurring at 10 sec (a) or 40 sec (b). Increases in spike rate immediately before and after a lever press indicate that neural activity is time-locked to this behavior. Note the large difference in the mean spike rate across early and late presses, indicating the temporal specificity of this neuron. The extent of lever press-induced variation in firing rate is masked by the high frequency of pressing during the early and late press periods. Panel (c) shows a more extensive firing rate variation by plotting the spikes resulting from low-frequency presses (1-sec interpress interval). This neuron is the same as that shown in Figure 15.2. Lever presses occur at 0 sec, and data have been smoothed by a 3-bin running mean for presentation.

pressing. Therefore, these data suggest that this neuron does not fall within a pure interval timing region of the striatum. Additional evidence suggesting that this temporally modulated striatal neuron is embedded within a striatal area that is involved in lever pressing is demonstrated by plotting the average lever press-related firing rate on each trial as a function of the average interpress interval on each trial (Figure 15.5). As can be seen, the firing rate of this neuron increased as a function of the interpress interval, indicating that this neuron is encoding detailed features of the behavioral response. Although the correlation was only significant for the late presses (early, r2 = .05, P > .05; late, r2 = .25, P < .01), the slopes of the regression line are identical in both cases (the differing correlation could be due to the extensive duration modulation seen during the early press period). In addition to this motor modulation, this neuron is clearly providing temporally specific information in its firing rate as well. As seen in Figure 15.4, the background rate is much higher during early presses than late presses. This effect is also evident in Figure 15.5, in which the average early press-related firing rate is generally greater than the average late press-related firing rate.

15.5.5 Competitive Interactions

If timing in the striatum is a distributed process, what predictions does it make regarding the other computations that occur in an interval timing system? We

Press-Related Firing Rate as a Function of Inter-press Interval

Average Interpress-Interval (sec) • Early Presses v Late Presses

FIGURE 15.5 Average press-related firing rate in a striatal neuron as a function of average interpress interval. The spike rate and interpress interval were computed during press bursts occurring during the 10- or 40-sec peak. The positive slopes for the regression lines indicate that this neuron encodes aspects of the press topography. Only the regression for the late presses was significant (r2 = .25, P < .01). This is the same neuron that is shown in Figure 15.2.

discussed earlier the likelihood that the striatal decision stage output influences its subsequent input, thereby creating a dynamic loop. The cortico-striato-thalamo-cortical circuit is thought to be composed of anatomically separate loops (e.g., sensory loop, motor loop, association loop, limbic loop, etc.) (Alexander et al., 1990). However, there is undoubtedly some cross talk occurring at many junctions of the circuit. One dramatic source of cross talk occurs at the input from the cortex to the striatum. Although the predominant source of cortical input to any particular striatal region comes from the matched functional cortical region (and eventually returns to that region), input from multiple cortical areas impinge on single striatal neurons (Finch, 1996). This cross talk leads to the ability of different components of the interval timing systems to modulate not only their own input, but that of the other systems as well, at least to some degree. In other words, a cortico-striatal interval timing system, although distributed in terms of its immediate timing processes, may be integrated in terms of the structuring of behavior and cognition. Such integration across systems might thereby induce the formation of behavioral and cognitive chains — chains that are so extensively a part of an organism's temporal behavior that they have been hypothesized to serve as the foundation of timing in the behavioral theory of timing (BET) (Killeen and Fetterman, 1988).

In contrast to BET, which utilizes a single pacemaker for transition from behavioral state to behavioral state and thereby allows only a single state to occur at any one time, a distributed striatal timing process would allow multiple temporally modulated processes to be activated simultaneously (e.g., the delayed match-to-sample processes of attention, gaze, motor preparation, reward expectation, etc.). Although the temporal processing for each of these behavioral and cognitive processes will dynamically influence each other through the mechanisms described above, each circuit would continue to support timing even if one of the other processes is prevented or disrupted. However, if such disruption occurs, the decision stage function for the remaining processes would likely be shifted or altered. The extent of this decision stage alteration would depend on the extent of cross talk between the systems in question. Presumably, preventing the expression of an early motor behavior (e.g., gaze direction toward the instruction cue) will have a greater effect on a subsequent motor behavior (e.g., gaze toward action cue) than on the behavioral expression of reward expectation (e.g., salivation). This proposed influence, rather than dependence, of one set of processes would predict that a subject's temporally informative behavior (i.e., its terminal behavior) would be altered, but not eliminated, by blocking its expression of an early component of its behavior chain. Indeed, Frank and Staddon (1974) have shown that restraining pigeons from producing adjunctive behaviors causes a change in the form of its fixed-interval scallop, but does not dramatically disrupt this behavior in the steady state. Interestingly, their interpretation of this result was that behavioral chains were not necessary for timing, and that adjunctive and terminal behaviors competitively influenced each other.

This competitive transition from one behavioral state to another might evolve naturally through a winner-takes-all process in the striatum. In a winner-takes-all system, a single output signal is selected from a variety of output signals based on which one has the greatest signal strength. Such a network property is thought to arise in systems with lateral inhibitory networks, as the fastest firing neuron (or region) produces the greatest inhibition of its neighbors. As the neighboring neurons are inhibited, their degree of lateral inhibition is also dampened, thereby further increasing the signal of the loudest neuron. Such a mechanism has been proposed to operate in the striatum (Kropotov and Etlinger, 1999; Wickens et al., 1996), through its inhibitory interneurons (Kawaguchi et al., 1995), or through axon collaterals of the GABAergic striatal spiny neurons, which innervate neighboring dendritic fields and may inhibit their likelihood of firing (Kawaguchi et al., 1990).

Interestingly, a winner-takes-all mechanism might explain the slight rightward skew frequently found in peak procedure data (Gibbon and Church, 1984), as well as the finding of the duration bisection at a point at or to the right of the geometric mean of the anchor points (Church and Deluty, 1977; Wearden, 1991), by extending the period in which the strongest signal produces the winning output.* To illustrate this point, imagine an experiment in which a subject is required to time two different

* In a bisection procedure the rat is asked to respond on the left lever after presentation of a short anchor duration (e.g., 2 s) and respond on the right lever after presentation of a long anchor duration (e.g., 8 s). Once this discrimination is well learned, intermediate length stimuli (e.g., 3, 4, 5, 6, or 7 s) are presented as probes (i.e., no reinforcement is given), and the rat's classification of the stimuli (i.e., short or long) is determined. The point of subjective equality (i.e., the duration the rat classifies as short 50% of the time and long 50% of the time) is frequently found to occur at the geometric mean of the anchor durations (e.g., 4 s).

durations and respond on two different manipulanda, e.g., bisection procedure (Meck, 1983) or tripeak procedure (Matell and Meck, 1999). One possible algorithm for performing these procedures would entail timing each of the two durations in an independent manner and responding on the appropriate manipulanda in direct proportion to the decision stage output of each timer. If these curves were simply Gaussian in shape, a function that has been used as an approximation of peak data, they would cross at the harmonic mean, a point that occurs to the left of the geometric mean, thereby indicating that additional computations are required or that a Gaussian approximation is incorrect. One possible solution to this problem is to introduce rightward skew into the Gaussian approximation of the decision stage output function. The effect of adding rightward skew to the curves plotted in our bisection example would be to induce a rightward shift in the point at which these curves meet. The precise point at which these curves would cross depends on the amount of skew introduced. Although a rightward skew can be seen in the majority of peak-interval data, and suggests that it is a common feature of timing data, Church et al. (1991) have shown that this skew can be greatly minimized, or eliminated, by varying the length of the probe trials or the intertrial interval. These data therefore indicate that the extent of skew in the temporally informative behavior of an organism is not fixed, but depends on the task parameters. In other words, rightward skew might not be an inherent property of the decision stage output, but may arise through the interactions across multiple timed processes. Such interactions would be expected to occur in a distributed striatal timing model. Specifically, the winner-takes-all mechanism proposed to operate in the striatum might allow such rightward skew to develop naturally simply through the inhibition of competing processes. To reiterate, the combination of a dynamic timing system (i.e., feedback loops from decision stage to clock stage) and a distributed timing system embedded within functional regions of the striatum may lead to the development of competitive interactions between functionally separate striatal outputs, thereby producing the behavioral chains frequently found in timing tasks.

0 0

Post a comment