How the associative strengths of stimuli combine in compound: Summation and overshadowing.

Abstract

When two conditioned stimuli (CSs) are presented in compound, the response is typically stronger than to the individual CSs, implying that their associative strengths combine. However, to identify exactly how associative strengths combine requires an accurate description of the relationship between associative strength and responding. The authors have used the delta rule (Rescorla & Wagner, 1972) to constrain the predicted growth of associative strength (V) to identify the relationship between V and responding across the course of Pavlovian conditioning of two CSs (one auditory, one visual). Responding to the compound was best predicted as 0.6xV(CS1)+0.6xV(CS2), suggesting that only 60% of the associative strength of each CS generalized to the compound. A second experiment confirmed this result and additionally showed that summation of responding between two same-modality CSs (both auditory or both visual) declined across training. A third experiment applied the procedure to compound conditioning, showing that responding to the compound was equal to the sum of the response rates to the individual CSs. The results are discussed in terms of configural and elemental models of Pavlovian conditioning.

An attention-modulated associative network

“…Most of these failures have occurred in autoshaping experiments with pigeons (e.g., Aydin & Pearce, 1995, 1997Rescorla & Coldwell, 1995). Nonetheless, both successes and failures to observe summation have been reported in other paradigms, such as the conditioned nictitating membrane response in rabbits (Kehoe, Horne, Horne, & Macrae, 1994) and the conditioned magazine approach with rats (Pearce, George, & Aydin, 2002;Rescorla, 1997;Thein, Westbrook, & Harris, 2008). Such mixed evidence is troubling for any model that assumes that most of the associative strength of each individual CS generalizes to the compound.…”

The acquisition of conditioned responding.

“…This paradigm measures explicit goal-tracking behavior (entries into the food magazine), which should represent a relatively unambiguous index of learned anticipation of the US. We have recently shown that groupaveraged acquisition functions in this paradigm conform very closely to a simple exponential function (Thein, Westbrook, & Harris, 2008), though as noted above, this could be consistent with either an incremental learning process or arise as an artifact from averaging over many individual step-functions. In both experiments presented here, rats were trained with discrete CSs that were followed by delivery of a food pellet.…”

Summation of reinforcement rates when conditioned stimuli are presented in compound.

“…There are numerous experiments showing that responding to a compound is greater than responding to the individual CSs that make up the compound (e.g., Kehoe, 1982;Kehoe, 1986;Rescorla, 1997;Thein, Westbrook, & Harris, 2008), consistent with the principle of summation.…”