Academic journal article Canadian Journal of Experimental Psychology

A Memory-Based Account of Retrospective Revaluation

Academic journal article Canadian Journal of Experimental Psychology

A Memory-Based Account of Retrospective Revaluation

Article excerpt

We adapt an instance model of human memory, Minerva 2, to simulate retrospective revaluation. In the account, memory preserves the events of individual trials in separate traces. A probe presented to memory contacts all traces in parallel and causes each to become active. The information retrieved from memory is the sum of the activated traces. Learning is modelled as a process of cued-recall; encoding is modelled as a process of differential encoding of unexpected features in the probe (i.e., expectancy-encoding). The model captures three examples of retrospective revaluation: backward blocking, recovery from blocking, and backward conditioned inhibition. The work integrates an understanding of human memory and complex associative learning.

Keywords: Instance theory, associative learning, retrospective revaluation, Minerva 2

(ProQuest: ... denotes formulae omitted.)

In retrospective revaluation, unpresented but associatively activated cues are associated. One example of retrospective revaluation is backward blocking. A simple demonstration of backward blocking involves two successive training phases followed by a test. In phase one of training, a cue compound, AB, is presented followed by an outcome, X. In phase two, A is presented followed by X. Backward blocking is observed if, following training, B is a weak exciter of X (Shanks, 1985). Of course, the result contradicts common sense. In the training phases, B reliably predicts X. Yet, subjects behave as if the opposite was true (i.e., that B does not predict X). Retrospective revaluation implicates a role of memory in learning. Van Hamme and Wasserman (1994; see also Dickinson & Burke, 1996) argued that in phase one of the backward blocking procedure a within-compound association forms between A and B. In phase two, the within-compound association causes A to retrieve B. Because B is retrieved, it can develop an inhibitory link to X. Melchers, Lachnit, and Shanks (2004) proposed a different albeit related memory-based explanation. In their account, the presentation of A in phase two elicits covert rehearsal of phase one trials, and backward blocking falls out of the covert rehearsal process. Here, we develop a novel explanation of retrospective revaluation using an instance model of human memory (Hintzman, 1984, 1986, 1988).

Instance theories of learning and memory operate from a premise that the individual experience (i.e., the instance) is the primitive unit of knowledge and that learning represents the accumulation and deployment of instances from memory. Brooks (1978, 1987) was amongst the first to champion the view. Medin and Schaffer (1978) were amongst the first to formalize it. Hintzman' s (1984, 1986, 1988) Minerva 2 model and Nosofsky's (1986) Generalized Context Model represent formal firstgeneration accounts of the instance-based view of memory. Kruschke's (1992, 1996, 2001) ALCOVE, ADIT, and EXIT models and Logan's (1988, 2002) ITAM model are modern extensions of the exemplar-based view of learning. Whereas different instance-based theories differ in their details, all agree that the instance is the fundamental unit of knowledge and that a competent theory of learning must include an account of how instances are stored and retrieved from memory.

In this paper, we adapt Hintzman' s (1986, 1988) Minerva 2 instance-based model of human memory to an analysis of retrospective revaluation. In short, we propose that traces of individual trials are stored in memory, that learning is driven by a process of expectancy-encoding, that decisions about associative strength fall out of a process of parallel cued-recall, and that retrospective revaluation follows from a process of trace-inversion at retrieval.

Minerva 2

Minerva 2 is a classic instance-based theory of human memory. The theory was developed to understand episodic-recognition and frequency-judgement (Hintzman, 1984, 1986, 1988). Minerva 2 has since been applied to a wide range of phenomena from the study of human memory (Arndt & Hirschman, 1998; Clark, 1997; Dougherty, Gettys, & Ogden, 1999; Goldinger, 1998; Hintzman, 1987; Jamieson & Mewhort, 2009a, 2009b, 2010; Jamieson, Holmes, & Mewhort, in press; Kwantes, 2005; Kwantes & Mewhort, 1999; Kwantes & Neal, 2006). …

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


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.