Forgetting in Working Memory

5.2.1 Time-Based Trace Decay

Models of time-based trace decay (e.g., Brown, 1958; Broadbent, 1957; Peterson and Peterson, 1959; Conrad, 1960; Crowder, 1976; Baddeley and Hitch, 1974;

Baddeley, 1976, 1986, 2007; Page and Norris, 1998; Burgess and Hitch, 1999; for a detailed list of decay models, cf. Lewandowsky et al., 2004) assign an important role to the factor Time.

In the late 1950s, increased experimental investigation in human memory launched doubts about the unitary nature of the human memory system (cf. chap-ter 2). Studies on memory performance over short time periods indicated the exis-tence of separate short-term and long-term memory stores. Those findings finally resulted in the creation of multi-component models (e.g., Waugh and Norman,

5.2. FORGETTING IN WORKING MEMORY 73 1965; Atkinson and Shiffrin, 1968). Findings from those early studies suggest dif-ferent kinds of forgetting mechanisms over long and short terms. Brown (1958), for example, run a series of three experiments to test the hypothesis of trace decay in immediate memory. The results show that after a short period of time item re-trieval was impaired when subvocal rehearsal was prevented. Brown (1958) states that experimental stimuli differed from the rehearsal-preventing items. Thus, ac-cording to Brown (1958), the findings cannot be blamed on interference effects of a possible similarity between stimulus material and items that were used to prevent rehearsal. Brown (1958) rather suggests that the rapid forgetting of exper-imental stimuli has to be tied to the passage of time. Thus, according to Brown (1958), the findings support assumptions of a rapid time-based trace decay.

In two experiments, Peterson and Peterson (1959) demonstrated that sequences of consonants were forgotten over a short period of time. The intensity of the degree of forgetting correlated with the controlled amount of rehearsal of the ex-perimental items. Rehearsal was prevented by counting numbers while the items to be remembered were consonants. Just as Brown (1958), Peterson and Peter-son (1959) opted for time-based trace decay. They argued that the stimulus items to be memorized clearly differed from the material that was used to prevent re-hearsal. Therefore, they concluded that interference-based decrease of memory performance had to be ruled out as a possible underlying source.

The influential phonological loop model (Baddeley and Hitch, 1974; Badde-ley, 2007) also suggests that storage of information is subject to trace decay within seconds. Subvocal or overt rehearsal of items is suggested to refresh memory traces. This hypothesis is backed by findings of variations in memory perfor-mance in word list retrieval. Baddeley et al. (1975) report that recall of lists of unrelated words was more error prone when words were long than when words were short. Following the idea of refreshing traces, it takes longer to refresh a long word than to refresh a short word. Thus, Baddeley et al. (1975) argue that in a limited period of time, fewer long words can be refreshed until traces have faded away than short words. Amongst others, this word-length effect is taken as support for time-based decay hypotheses.

More recently, Page and Norris (1998) proposed a resource-driven working memory model that suggests time-based trace decay in short-term serial recall.

The model suggests an exponential activation decay of list items over a short pe-riod of time. Thus, the slower a list of items is presented, the worse will be the performance of recall. This steady loss of resource strength over time is supposed to explain word length and list length effects.

All findings reported above support hypotheses of a time-based trace decay.

However, all those findings only result from studies with unrelated list items.

Nevertheless, there are also sentence processing models that attribute a decrease in memory performance to a fading of memory traces over time. In some models,

trace decay plays an essential role (e.g. Gibson, 1998, 2000, Vosse and Kempen, 2000, Lewis and Vasishth, 2005), while in others, fluctuation of resources is taken as a possible candidate (e.g. Levy et al., 2007).

5.2.2 Interference-Based Forgetting

Models of interference-based forgetting contrast with time-based models. Whereas the latter models assume that time-based trace decay is the underlying reason for forgetting in working memory, so-called interference-based (or ‘event-based’, cf. Lewandowsky et al., 2004) models do not put any weight on the factor Time as the crucial and underlying source in short-term forgetting. Interference-based models (e.g. Melton, 1963; Lewandowsky and Murdock, 1989; Henson, 1998;

Lewandowsky, 1999; Nairne, 1990; Murdock, 1995; Neath, 1999; Farrell and Lewandowsky, 2002; Lewandowsky et al., 2004; Lewandowsky et al., 2008) ad-mit that forgetting in working memory is observed over time, but time is not the reason for the phenomenon of forgetting. As already stated above: iron rusts when it is left outside for a longer period of time, but it is oxidation and not the time per se which is responsible for the occurrence of rust (cf. McGeoch, 1932). To stay in the picture provided by McGeoch: it is the amount of material that has to be pro-cessed which makes sentence processing harder. The fact that parsing processes take place in a certain amount of time does not make time a causal source of for-getting. Therefore, interference-based hypotheses suggest that working memory is affected by interference of cognitive processes. The passage of time is rather taken as epiphenomenal. Lewandowsky et al. (2004) reports a variety of differ-ent interference-based approaches from studies of learning and recalling lists of unrelated items. For the sake of the current experiment, those approaches will be treated alike. The current work does not differentiate between the different approaches, as the common core assumption of all event-based models is: inter-ference of stimulus items and not time causes an increase in sentence complexity.

5.2.3 Coexisting Factors and The Problem of Collapse

As already stated above, the question if memory traces over the short term are subject to interference-based forgetting or if they are subject to time-based decay is a long lasting debate in the field of cognitive research. A major reason why clarifying the roles of time-based trace decay and interference-based forgetting is so complicated, lies in the very nature of the incoming information: the amount of time and the amount of processing are confounded. In order to investigate the underlying source of forgetting in working memory, the two factors have to be dis-entangled. Previous experimental work on this task can be found in Lewandowsky et al. (2004) and Saito and Miyake (2004).

5.2. FORGETTING IN WORKING MEMORY 75