Chapter 8: How Stimuli Guide Instrumental Action

Summary

Stimulus control techniques provide a powerful method for studying processes of animal cognition.
Animals can sort stimuli into polymorphous categories—that is, categories that are not defined by a single feature (e.g., people, cats, cars, flowers, and chairs). Such categorization probably depends on the basic learning mechanisms discussed in Chapters 4 and 5. The learning process may (a) find the most predictive features and associate them with each category, (b) construct a prototype of each category, or (c) allow the animal to remember each example and respond to new stimuli according to how similar they are. The Rescorla-Wagner model and its successors accomplish (a) and (b) and the Pearce configural learning model accomplishes (c).
Organisms generalize from one stimulus to another depending on their physical similarity. We generalize between two stimuli depending on the number of common features (elements) they share.
Generalization is also affected by learning:
1. Generalization gradients are sharpened by discrimination training, in part because the training introduces inhibition to S-. The presence of inhibition to S- can produce surprising peaks in responding to other stimuli. For example, in peak shift, the highest level of responding occurs in the presence of a stimulus that has never actually been reinforced.
2. Generalization between two stimuli will increase if they are associated with a common event or stimulus. Such mediated generalization allows new, superordinate categories (e.g., furniture, clothing) to be built without physical similarity between the individual stimuli.
3. Mere exposure to similar stimuli (like various birds or types of wine) can make them easier to discriminate. Such perceptual learning may occur because exposure to similar stimuli latently inhibits their common elements, creates inhibition between their unique elements, and/or creates a unitized representation of each stimulus.
For a stimulus to guide instrumental action, it must be perceived, attended to, and remembered.
Attention requires that the stimulus be informative. Attention can also be boosted by attentional priming.
Working memory allows stimuli or events to guide behavior after they are gone. It has been studied in the delayed matching-to-sample method and in the radial maze. Working memory is influenced by practice, and by retroactive and proactive interference. Animals appear to use it actively and efficiently, as when they switch between retrospective codes (remembering events that have come before) and prospective codes (remembering events that are coming in the future).
There are several different types of human long-term or reference memory. These can be difficult to study in animals, although there is evidence that scrub jays have an episodic-like memory that incorporates what, when, and where information about food items they have stored.
Time can be an important guide to instrumental action. Animals (and humans) are sensitive to time of day cues, which appear to depend on a circadian clock with a period of about a day. Organisms are also good at timing intervals on the order of minutes and seconds. Timing may be accomplished with an “internal clock” that may involve a pacemaker and an accumulator or the readout of a set of oscillators whose states change at different rates. Timing might also be accomplished with mechanisms that don’t require a “clock.”
Spatial cues also guide instrumental action, and organisms use beacons, landmarks, and dead reckoning to get around in space. Organisms might also form a geometric representation of the environment. Although spatial learning might be accomplished by specialized learning mechanisms, recent research suggests that it depends at least partly on familiar learning principles that allow animals to associate beacons and landmarks with goals.