An Attempt at Blocking of Position Learning by Training with Reward-Memory Associations
Burns, Richard A., Johnson, Kendra S., The Psychological Record
A simple serial learning task with rats may involve the repetition of series in which each trial concludes with a reward value that is unique to a given location in the series. Performance usually comes to reflect the reward values appearing in each location. Here we consider two contemporary theoretical accounts, not necessarily mutually exclusive, of the learning that manifests itself in such problems. One account (e.g., Capaldi, Alptekin, Miller, & Birmingham, 1997) emphasizes the memories of differing rewards being associated with current and future reward values. Another account (e.g., Burns, Kinney, & Criddle, 2000) stresses cues related to the ordinal position of the rewards, which may be associated with reward values in that location.
A good example of a serial learning problem that does not seem to be explicable unless reference is made to reward-memory associations can be found in an experiment reported by Capaldi et al. (1997). A trio of series was repeated each day in unpredictable order. The series were PN, SPN, and SSPN, where S and P indicate sucrose or plain food pellets and N indicates no reward. In these series, the locations of the rewards varied from series to series, making position associations impossible. A simple memory association, however, makes the problem straightforward: P is associated with N. Rats trained in these series readily learn to approach slowly on the N trials following P.
Series such as those used by Capaldi et al. (1997) are rare, however. Usually, the series employed confound position and memory cues, and it is from experiments with this confounding that much of the evidence for position learning comes. An example of such evidence was reported by Burns and Criddle (2001) who trained rats on an SNP series. Here, S may be associated with N, but N also appears exclusively in the second series position. Thus, memory learning, position learning, or both may occur. A transfer test was administered after stable performance differences developed on rewarded and unrewarded trials. The test was an NNN series. If performance in the transfer had been controlled by memory, approach would have been rapid on all three NNN trials because N would have been associated with P, a reward. Position learning would have been indicated if approach had been slow on the middle trial, and approach was slow on that trial.
There is good evidence that rats learn both position and reward-memory associations in serial learning tasks for which both are relevant sources of information (Burns et al., 2000; Capaldi & Miller, 2001). If position learning and reward-memory learning are conceptualized as a form of discrimination learning (e. g., Capaldi & Molina, 1979), then much of the recent work on position and item learning may better be thought of as using overshadowing methodologies (Pavlov, 1927). In an overshadowing procedure cues are presented in compound and later tests of the associative strength of the individual cues may indicate that the learning of one element of the compound was moderated (overshadowed) by the other element of the compound. The Burns and Criddle (2001) procedure, for instance, can be thought of as involving a compound (position and memory) cue. The NNN transfer test changed the memory component of the compound while holding the position component constant. That the animals continued to approach slowly on the middle trial of NNN suggests overshadowing of reward-memory associations by position associations. Many of the experiments suggesting position learning do so by producing results that indicate position associations overshadow reward-memory associations (e.g., Burns, Dunkman, & Detloff, 1999; Burns et al., 2000; Roitblat, Pologe, & Scopatz, 1983).
Another related, and extensively studied, phenomenon of associative learning is blocking (Kamin, 1968). In a blocking procedure one cue is permitted to build associative strength prior to its presentation in compound with a second. …