Context Effects on Reinforcement

9306736 Williams对动物学习的研究通常分为巴甫洛夫条件反射研究和工具性条件反射研究，前者涉及刺激和强化物之间的关系，后者涉及反应和强化物之间的关系。威廉姆斯博士最近的研究表明，这两种条件作用的偶然性效应有着惊人的相似之处。现在，他将扩展这一分析，以更精确地确定两个领域的偶然性效应是否依赖于相同的潜在联想过程。他特别提出，工具行为强度、强化率和不同反应选择的标准测量是否可以用巴甫洛夫联想理论所预测的刺激值概念来解释。这项研究可能会彻底改变一般学习理论的概念基础。在本世纪的大部分时间里，学习理论的基本假设是，“联想强度”是由个体因出现强化和非强化行为而导致的联想价值的增加和减少组成的。这一假设受到了最近的研究结果的挑战，这些研究表明，动物似乎对独立于自身行为的情况下可获得的总体奖励率很敏感。换句话说，动物使用它的行为来估计这些总体奖励率，而不是反向因果方向的奖励产生的变化的“强度”的行为。这个问题的核心是联想概念的可行性，以提供一个还原论的学习说明。这个理论问题对所有神经科学都有影响，因为最近强调识别学习和认知背后的神经机制，必须以正确的学习过程观点为基础，才能有成功的希望。

9306736 Williams Research on animal learning has generally been segregated into the study of Pavlovian conditioning, which involves the association between stimuli and reinforcers, and instrumental (operant) conditioning, which involves the relation between responses and reinforcers. Dr. Williams' recent research has demonstrated striking parallels in how contingency effects operate for the two types of conditioning. He will now extend the analysis to determine in more precise detail whether contingency effects in the two domains depend on the same underlying associative processes. In particular, he asks whether standard measures of the strength of instrumental behavior, rate of reinforcement and choice between different response alternatives, can be explained in terms of the concept of stimulus value as predicted by Pavlovian association theory. This research may revolutionize the conceptual foundations of general theories of learning. The fundamental assumption of learning theory for most of this century has been that "associative strength" is composed of individual increases and decreases in associative value due to individual occurrences of reinforced and nonreinforced behavior. This assumption is challenged by recent findings showing that the animal seems sensitive to the overall rates of reward available in the situation independent of its own behavior. In other words, the animal uses its behavior to estimate these overall reward rates, rather than the reverse causal direction of the reward producing changes in the "strength" of the behavior. At the core of this issue is the viability of associative concepts to provide a reductionistic account of learning. This theoretical issue has ramifications for all of neuroscience, because recent emphases on identifying the neural mechanisms underlying learning and cognition must be predicated on the correct view of the learning processes in order to have any hope of success.