Academic journal article The Psychological Record

Choice and the Initial Delay to a Reinforcer

Academic journal article The Psychological Record

Choice and the Initial Delay to a Reinforcer

Article excerpt

Grace and Nevin (2000) trained pigeons on a concurrent-chains procedure with equal variable-interval (VI) schedules in the initial links and equal VI schedules in the terminal links. The terminal links differed in that one ended after a single reinforcer, which they called "variable-duration" terminal link, whereas the other ended after a fixed period of exposure equal to the average interreinforcement interval (IRI) of the schedule, which they called "constant-duration" terminal link. As Grace and Nevin identified, and as discussed at some length below, an important feature of the constant-duration terminal link is that it probabilistically yielded 0, 1, or multiple reinforcers per entry, although it provided the same average rate of reinforcement overall as the variable-duration terminal link.

Grace and Nevin (2000) found that three of four pigeons clearly preferred the constant-duration terminal link. In their words, the data of a fourth pigeon "demonstrated a consistent right-key bias" (p. 178), and the present conclusion is that its data are more difficult to interpret. In any case, an important question is what variables caused the preference. Ordinarily, one would have expected the pigeons to be indifferent, since the schedules in effect during the alternatives were identical, and each alternative yielded the same overall rate of reinforcement.

Grace and Nevin (2000) initially pondered the role of multiple reinforcers in the constant-duration terminal link, because research has shown that subjects may well prefer a choice alternative associated with multiple reinforcers rather than a single reinforcer per terminal-link entry (e.g., Fantino & Herrnstein, 1968; Mazur, 1986; McDiarmid & Rilling, 1965; Moore, 1979; Poniewaz, 1984; Shull, Mellon, & Sharp, 1990; Shull, Spear, & Bryson, 1981). In particular, Grace and Nevin discussed their findings from the view of "cumulative delayed reinforcement." This view assumes that the value of each terminal-link reinforcer in a multiple-reinforcer terminal link of concurrent chains is discounted as a function of its delay from terminal-link onset; the discounted values are then summed to yield some composite value of the alternative as a whole. A specific discounting function that currently enjoys widespread acceptance is Mazur's (1984, 1986, 1987, 2001) hyperbolic-decay model. Indeed, Mazur's model provides a very accurate description of the data from studies involving different delays, different amounts of a reinforcer, probabilistic reinforcers, and multiple reinforcers.

The Grace and Nevin (2000) interpretation of the role of multiple reinforcers in their data makes a great deal of sense, even though the multiple reinforcers are only a probabilistic outcome of the constant-duration procedure. However, Grace and Nevin further pointed out that the degree of preference for the constant-duration terminal link in their study somewhat exceeded the predictions of the cumulative delay view. Because of this outcome, they argued that some other factor contributed. Perhaps it was the variability in the number of reinforcers per entry associated with the constant-duration terminal link, which they termed "numerosity" (p. 184).

Relevant to the Grace and Nevin (2000) results is an understanding of the temporal distribution of reinforcers in variable- and constant-duration terminal links that employ aperiodic schedules. The variable-duration terminal link is the conventional way of scheduling terminal links in concurrent chains. The IRI selected to be in effect on a given entry begins timing when the terminal link is entered, finishes timing when the reinforcer is arranged, and resumes timing at the beginning of the succeeding IRI when the terminal link is next entered. Thus, the variable-duration terminal link always yields only one reinforcer per entry. It provides a given rate of reinforcement as a function of the average IRI of the terminal-link schedule. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed

Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.