Academic journal article Memory & Cognition

A Constrained Rasch Model of Trace Redintegration in Serial Recall

Academic journal article Memory & Cognition

A Constrained Rasch Model of Trace Redintegration in Serial Recall

Article excerpt

The notion that verbal short-term memory tasks, such as serial recall, make use of information in long-term as well as in short-term memory is instantiated in many models of these tasks. Such models incorporate a process in which degraded traces retrieved from a short-term store are reconstructed, or redintegrated (Schweickert, 1993), through the use of information in long-term memory. This article presents a conceptual and mathematical model of this process based on a class of item-response theory models. It is demonstrated that this model provides a better fit to three sets of data than does the multinomial processing tree model of redintegration (Schweickert, 1993) and that a number of conceptual accounts of serial recall can be related to the parameters of the model.

The serial recall task, in which participants attempt to recall a series of sequentially presented items in the order in which they were presented, has been the most commonly used task to investigate verbal short-term memory. Many models of short-term memory, or simply of the serial recall task, incorporate a process by which degraded temporary memory traces of the items (words, digits, letters, or nonsense words) are reconstructed, or "redintegrated" (Schweickert, 1993), by accessing traces in long-term memory (see, e.g., Brown & Hulme, 1995; Nairne, 1990; Page & Norris, 1998b). Whether the initial degradation of the trace occurs through passive decay (see, e.g., Baddeley, 1986; Page & Norris, 1998b) or active interference (see, e.g., Nairne, 1990) is irrelevant to the redintegration process, but the existence of such a process seems likely, given experimental findings showing an influence of longterm memory factors on serial recall performance. For example, words, by definition, are distinguishable from nonwords in having lexical representations in long-term memory, and serial recall performance is better for lists of words than for lists of nonwords when they are matched on other factors (see, e.g., Huhne, Maughan, & Brown, 1991). High-frequency words differ from low-frequency words in terms of how often they have been encountered in the past, so any superior memory performance for high-frequency versus low-frequency words presumably reflects the longterm memorial impact of those encounters. Robust frequency effects are reported for the serial recall task when pure lists are used (see, e.g., Gregg, Freedman, & Smith, 1989; Hulme et al., 1997; Tehan & Humphreys, 1988; but see Hulme, Stuart, Brown, & Morin, 2003, and Morin, Poirier, Fortin, & Huhne, 2006, for discussions regarding effects in mixed lists). The aims of the present article are to describe a new measurement model of the redintegration process and compare it with an existing model, the multinomial processing tree model of Schweickert (1993).

This model has been applied successfully to data showing a word frequency effect (Hulme et al., 1997) and to the effects of irrelevant speech on serial recall (Buchner & Erdfelder, 2005) by assuming that i, the likelihood of the short-term trace's being intact, decreases through the list, and that r, the likelihood of correctly redintegrating an item, varies across stimulus sets or conditions. For example, high-frequency words have a larger r than do lowfrequency words, given the assumption that they are more accessible within long-term memory.

Although the MPT model has been shown to fit several sets of data, thus providing a good measurement model of redintegration, it includes a number of theoretical assumptions that might be questioned. The model treats the integrity of the short-term trace in a discrete fashion; that is, the short-term trace is either intact or it is not, and if the trace is not intact, it has the same likelihood of being redintegrated regardless of how degraded it is. In fact, the model does not incorporate any notion of degree of degradation. As a result, the model seems at odds not only with common-sense notions of decay and interference, but also with those notions in more formal models (e. …

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.