Academic journal article Memory & Cognition

On the Flexibility and the Fallibility of Associative Memory

Academic journal article Memory & Cognition

On the Flexibility and the Fallibility of Associative Memory

Article excerpt

We report the results of four experiments in which we explored the flexibility and fallibility of associative recognition memory. In each experiment, pairs were studied one or more times, and the task was to discriminate intact from rearranged pairs. The critical findings are that the pattern of false alarm rates was dependent on the nature of the recognition procedure (e.g., ratings vs. yes-no) and the situation in which the task was performed. The specific pattern of findings suggest that subjects adopt different recognition strategies in order to achieve a desired level of performance in the most efficient manner possible by varying the degree to which they base their decisions on familiarity versus recollected information. Implications for theories of recognition memory are discussed.

The limitations of memory are widely known. Laypersons usually condemn their "terrible memories." In doing so, they fail to appreciate the remarkable flexibility of human memory, which represents and retrieves varieties of information obtained from different perceptual modalities in the service of countless daily tasks, each performed in countless situations. Flexibility and fallibility are hallmarks of human memory. The subject of the present investigation is what the fallibility of recognition memory can tell us about its flexibility.

Recognition tasks require the discrimination of studied items (or targets) from unstudied items (or foils). Recognition is successful when targets are endorsed (i.e., hits) and foils are rejected. Likewise recognition is unsuccessful when targets are rejected and foils are endorsed (i.e., false alarms). Thus, recognition memory improves as the difference between hit rates (HRs) and false alarm rates (FARs) increase.

Over the past dozen or so years, a dominant question in episodic memory has been how to characterize retrieval from recognition memory. One theory holds that recognition can be characterized by a single retrieval process (e.g., Donaldson, 1996; Dunn, 2004; Wixted & Stretch, 2004), and another theory holds that at least two retrieval processes are required (e.g., Joordens & Hockley, 2000; Reder et al., 2000; see Yonelinas, 2002, for a review). The former class of models is referred to as single-process models, and the latter is referred to as dual-process models. The debate is often framed so that only one model would be needed to characterize all recognition tasks, and accordingly, the same task has always been performed in the same way.

These assumptions are parsimonious, but they might overlook the inherent flexibility of human memory. For instance, there are a wide variety of recognition tasks (including single-item recognition, associative recognition, source recognition, eye witness identification, and memory scanning) and procedures (yes-no, ratings, forced choice), and they are performed in countless situations. Thus, one might assume that these tasks at least have the potential to be performed in different ways. In the present research, the flexible nature of recognition memory is explored in the context of the single- versus the dual-process debate. One hypothesis is that different recognition tasks can be performed in different ways; perhaps, even the same recognition task can be performed in different ways (e.g., Rotello, Macmillan, & Van Tessel, 2000; Yonelinas, 1997).

MODELS OFASSOCIATIVE RECOGNITION

A major theoretical advance in human memory theory has been in its formalization, and here, we will describe three classes of models that recently have been applied to associative recognition. We will pay particular attention to the predictions that they make concerning the effect of pair repetitions on performance, because this is one way that they can be distinguished (Cleary, Curran, & Greene, 2001; Kelley & Wixted, 2001; Xu & Malmberg, 2007).

Single-Process Recall-Only Models

According to recall-only models, associative recognition is performed in a manner very similar to cued recall (but see Anderson & Watts, 1971 ; Postman & Stark, 1969). …

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

Oops!

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.