Short-term recognition memory for serial order and timing

Abstract

1724Numerous contemporary models of memory are dedicated to explaining short-term order memory, and in doing so adopt a variety of assumptions to explain people's ability to maintain and recall the order of items or events in a sequence over the short term. In some models, such as chaining models (e.g., Lewandowsky & Murdock, 1989;Murdock, 1995) and ordinal models (e.g., Farrell & Lewandowsky, 2002;Page & Norris, 1998), ordering is assumed to follow from the relationships between items (associations and relative strengths, respectively). In another class of models, the ordering of items is assumed to follow from the pairing of items with some external representation of the location of an item in a sequence. Of interest here is a subset of these models that assume that the order of items is represented by placing them along a temporal dimension. In temporal context models, such as the connectionist OSCAR model of Brown, Preece, and Hulme (2000), items are ordered by associating each item with a vector representing temporal context; being the output of a bank of sinusoidal oscillators, this temporal context evolves over time and can be used to specify the position of an item in a sequence (see also Burgess & Hitch, 1999). In a similar spirit, temporal distinctiveness models take a more abstract approach to list memory, and treat recall as the discrimination of items along a temporal dimension (e.g., Brown, Neath, & Chater, 2007;Glenberg & Swanson, 1986;Neath, 1993). Although differing in level and details of implementation, these time-based models 1 have in common the assumption that memory for the order of events follows from (temporal context models) or is primarily affected by (temporal distinctiveness models) memory for the timing of those events.A fundamental feature of time-based models is that they predict effects of the temporal isolation of items on ordered recall accuracy. Specifically, the more an item is separated in time from nearby items on a memory list, the more accurately that item is predicted to be recalled. In the case of OSCAR (Brown et al., 2000), cross-talk in the connectionist network that stores the associations between temporal context and items in a sequence means that items that are stored in more similar temporal contexts will interfere more with each other at retrieval when one of those temporal contexts is used as a cue. Temporal distinctiveness models predict this temporal isolation effect (TIE) by virtue of decreased temporal distinctiveness of the items. For example, the recent SIMPLE (scale-invariant memory, perception, and learning) model treats time as a dimension along which items may be discriminated, just like perceptual dimensions such as weight or brightness Neath & Brown, 2006). For a demonstration of the prediction of a TIE in SIMPLE, see Figure 1 of Lewandowsky, Brown, Wright, and Nimmo (2006).A number of experiments have recently been published that systematically call into question these time-based models by demonstrating a lack of TIEs in short-term memory for order....

A dynamic approach to recognition memory.

“…Although we leave it as a topic for future investigation, we have no reason to believe that the kind of process we have described for recognition over short timescales might not apply to these kinds of extended events as well. As a longer event, e.g., a sentence or sequence of items (e.g., Farrell & McLaughlin, 2007), unfolds over time, aspects of the event-the incoming words, prosody, and momentary understanding of the semantics of the sentence-join those features already present in working memory and could be used to probe memory in exactly the way our model describes. The dynamics of familiarity are influenced by what features are present in the probe over time.…”

Temporal isolation does not facilitate forward serial recall--or does it?

“…Accordingly, in an experiment that randomly intermixed unconstrained reconstruction trials with forward serial recall trials, and in which people were unaware of the type of test until after list presentation, isolation effects nonetheless selectively occurred with the former but not with the latter test. (In support of the notion that people can shift attention after encoding, Farrell & McLaughlin, 2007, reported a similarly selective use of time between two variants of a recognition test.) Thus, whenever forward serial retrieval is required, people will rely on the positional dimension at recall and thus do not show a temporal isolation effect (see Table 1), notwithstanding the fact that they have demonstrably encoded information about the temporal properties of the list.…”

“…Likewise, the absence of an isolation effect has been observed with both visual and auditory stimuli Parmentier et al, 2006), and it has been observed with single-item probed recall as well as with whole-report forward serial recall . Even with serial recognition, no isolation effects are observed under conditions that are thought to be most favorable to their emergence (e.g., because on some trials, participants are asked to retain timing information, which they are demonstrably able to do; Farrell & McLaughlin, 2007).…”

Memory retrieval as temporal discrimination

“…An alternative approach is taken by Farrell and McLaughlin (2007), who found that serial recognition of items that were irregularly spaced in time during presentation was not affected by the temporal proximity of items, although temporal information was encoded as evidenced by participants' ability to perform a temporal recognition task. Along with other results (e.g.…”