Extending conditioned stimuli before versus after unconditioned stimuli: Implications for real-time models of conditioning.

Abstract

In four experiments using the conditioned suppression procedure with rats, we compared the effects of extending conditioned stimuli (CSs) before versus after reinforcement (called B vs. A extensions). In Experiments I and 2, Group 0 (no extension) received 2-min noise CS trials (3 per day in Experiment 1, 1 per day in Experiment 2) that terminated with a 1-s grid shock unconditioned stimulus (US). For Group B, the CS began 12 min before the US; for Group A, the CS began 2 min before the US but persisted for 10 min past US termination. In Experiments 3 and 4, similar trials (3 per day in Experiment 3, 1 per day in Experiment 4) included a 2-min light CS that always terminated with the US; thus the noise CS became a systematically manipulated context cue in which lightshock pairings were embedded. In Experiments 1 and 2 we found asymmetrical effects of CS extensions: B extensions weakened conditioning more than did A extensions. In Experiments 3 and 4 we found symmetrical effects: A and B extensions weakened context conditioning equally. A realtime model of conditioning that combined the Frey-Sears (1978) dynamic attention rule with the Rescorla-Wagner (1972) model anticipated the asymmetrical effects preasymptotically and the symmetrical effects at asymptote. John W. Moore for his support throughout the project.

Extending continuous versus discontinuous conditioned stimuli before versus after unconditioned stimuli.

“…First, the earlier demonstrations (Ayres & Albert, 1990;Ayres et al, 1987) used between-groups designs as opposed to the within-group designs used in this research. Second, the earlier demonstrations used intermittent CSs (1 s on, 0.11 s off).…”

“…Relative to the baseline condition in Panel 0, the 0 (zero) condition, the A and B extensions weaken conditioning. However, of special interest is the fi nding that the B extension weakens conditioning more (Ayres & Albert, 1990;Ayres et al, 1987).…”

“…The Rescorla-Wagner-Frey-Sears real-time model's (R-model) predicted associative values (vA) resulting from 0, B, BD, A, and AD procedures, assuming that six 2-min parts of discontinuous conditioned stimuli (CSs) are separated by 6-min gaps (top panel) and the sometimes opponentprocess model's (SOP) predictions (bottom panel). (The top panel is replotted from the top left panel of Figure 8 in Ayres, Albert, and Bombace [1987], and the reader is referred there for details, including parameter values used and the reason for their choice. SOP [e.g., Wagner, 1981] made "moment-to-moment" calculations beginning with CS onset and continuing for 10 min past CS termination.…”

“…To explain this asymmetrical effect of A and B extensions, Ayres et al (1987) made two assumptions. First, they viewed each CS as being composed of parts of equal duration.…”

“…Recently, researchers have begun to develop real-time models of Pavlovian conditioning (e.g., Ayres, Albert, & Bombace, 1987;Klopf, 1988;Moore & Stickney, 1980;Sutton & Barto, 1981Wagner, 1981). Like the previous generation of models (e.g., Frey & Sears, 1978;Mackintosh, 1975;Pearce & Hall, 1980;Rescorla & Wagner, 1972), realtime models try to predict and explain the modulation of conditioning to target elements of a compound conditioned stimulus (CS) by other members of the compound.…”