A fragmentation hypothesis of memory: Cued recall of pictures and of sequential position.

Abstract

It is a common assumption that memories consist of sets of interrelated features or attributes. Rigorous and direct investigation of the functional properties of these sets has been relatively neglected however. Two important questions are posed here. First, are the interrelationships of different components reflexive or are they asymmetric? Second, how do components interact when more than one is used as a cue for recall ?These questions were answered by examining memory for series of items whose objective components were statistically independent of one another. The items were color photographs which each depicted a particular object (e.g., a cup) of a particular color (e.g., yellow) in a particular location (e.g., the top left corner of the backdrop); the photograph also occurred in a particular sequential position of course (e.g., as the second photograph of a series). After a filled 25-sec delay, recall of each item was assessed using one, two, or three of these four components as cue for the remainder.A fragmentation hypothesis was suggested to account for the results concerning the visual attributes. This postulates that a memory trace corresponds to a fragment of a perceived situation. A fragment can be excited by the provision of a single constituent as cue: Recall is therefore reflexive, and the cuing of a memory by several of its constituents is no more effective than that by a single one. Multiple cuing will usually produce rather higher overall recall simply because there is an increased chance that at least one of the cue components is stored in the corresponding fragment. Upon rigorous testing it was found that the hypothesis provided an accurate representation of recall. Two other models, an "independent cue" hypothesis and a model based upon that of Anderson and Bower (1973), were significantly discrepant with the data.Sequential position did not conform to the fragmentation hypothesis, correct recall being more likely when sequential position was used as a cue for another component than when it was itself cued for. In the latter case the median recalled value of sequential position was related to the true value by a power function. An implication is that the lower level of recall usually occurring for items in the middle of a list is due to the increasing inaccuracy of the contextual component being used as the implicit cue.

Memory for events and their spatial context: models and experiments

“…The name of someone one met on a single occasion is a good example of both asymmetries (a good cue, but sometimes hard to retrieve), while the location of the meeting is often both a good cue and relatively easy to retrieve. The sequential position of an item in a list of items is another example of a good cue that is relatively hard to retrieve itself (Jones 1976). Considerations such as these demand at least asymmetrical associations, and have fuelled the theoretical distinctions made between the context and content of an event (e.g.…”

Fragment and schema models for recall

“…That is, for some types of material, the fragment model may, in general, represent recall better than does the schema model. Previous work (Jones, 1976(Jones, , 1978c had suggested that the fragment model is successful in representing the recall of material with unrelated components. Consequently, in the experiments to be reported here, participants attempted to recall clusters of words that had not been selected on the basis of their preexisting relations to each other.…”

“…However, they did not note that, in fact, the first and second are different members of the same family of theories, treated here as the horizontal and general versions, respectively, of the fragment model. The general version is the form in which the fragment model has usually been tested previously (e.g., Jones, 1976Jones, , 1978cJones & Payne, 1982). The four different versions are described next.…”

“…If, however, the general version's advantage in accuracy over one of the constrained versions proved empirically to be slight, then it would be appropriate to prefer the more parsimonious version." It is in the form of the general version that the fragment model has usually been tested previously (e.g., Jones, 1976Jones, , 1978cRoss & Bower, 1981a).…”

“…In recent years, two different theoretical models-one within each of the schools-have both enjoyed some success in quantitatively predicting the results of experiments of this type. These theories are the "fragment" model (Jones, 1976(Jones, , 1978cJones & Payne, 1982) and the "schema" model (Ross & Bower, 1981a; see also Anderson & Bower, 1973). For ease of exposition, these models are outlined in the following sections in the context of the type of task used in both the present experiments and those of Ross and Bower (l981a).…”