Info

Gap of 12 sec

Gap at 15 sec

FIGURE 2.21 Peak-interval gap procedure, fixed duration simulations. These data present the amount of peak shift resulting from gaps of the same durations but with different onset times. The lower panel graphically shows the six locations of the gaps within the interval. The animal data presented are taken from Cabeza de Vaca et al. (1994).

animals would learn that a short break in the timing signal predicts that this trial will not produce any reinforcement. If they did learn this, subjects would stop responding entirely after the gap. But even after weeks of training with gap trials, rats and pigeons do not stop responding. As with most experiments involving nonfood trials, overall response rates do go down in gap experiments, but they do not extinguish as they would if subjects were capable of using the gap as a discriminative stimulus.

Roberts (1981) explored what would happen if the timing signal was not removed, but instead a new stimulus was introduced during an equivalent portion of the interval. On a filled-gap trial, the timing signal (a light) is turned on at the beginning of the trial and remains on for three times the reinforced interval length. Ten seconds after the trial begins, a 10-sec tone is presented. The presentation is

Trials

FIGURE 2.22 Filled-gap simulation. The simulation responded differently to a gap filled with a novel stimulus (squares) than to a standard gap (triangles). The baseline data (diamonds) show the normal peak time.

Trials

FIGURE 2.22 Filled-gap simulation. The simulation responded differently to a gap filled with a novel stimulus (squares) than to a standard gap (triangles). The baseline data (diamonds) show the normal peak time.

equivalent to a normal gap, but the sound can be learned as a discriminative stimulus. In this experiment, subjects' peak response times were initially the same as in a normal PI trial, but over the course of 2 weeks, their peak times gradually became longer and longer, indicating that the subjects had learned that the sound signaled a nonfood trial. Response rates postgap declined as well, approaching zero by the end of the experiment.

A simulation of this experiment using the general timing model was performed, and the results can be seen in Figure 2.22. The peak times become later and later over successive trials. In the model, this shift takes place within 20 gap trials, unlike the rat experiment, where it occurs over 2 weeks. This can be attributed to the fact that the model simply learns faster than its animal counterpart. One might propose that the model "lives" in a much simplified world, with fewer extraneous stimuli to confuse the issue. Even ignoring the world outside the operant chamber, the model never gets satiated, is never distracted by an odd smell in one corner of the chamber, never misses a signal due to looking the wrong way, or any of ten thousand other variables. The model's learning rate is controlled by a parameter and could be reduced to the point where this learning takes the same number of trials as the animals.

2.4.3.3 Scalar Timing

One of the most important attributes of animal timing is what has been termed the scalar property. Also referred to as timescale invariance, it has been defined as the fact that one should not be able to tell the length of the interval from the shape of the timing function (Gallistel, 1999). In practical terms, this means that the PI functions for a 30-sec schedule and a 300-sec schedule should be of similar proportions. This is not an absolute law of behavior. Studies with very long schedules (over

0 0

Post a comment