This article present a dual-process model of free and associative recall based on the representation principles of memory outlined in Fuzzy Trace Theory (Fuzzy). Generally, Fuzzy is a theory of decision making and memory processing. This theory assumes that generally we are predisposed to rely on a fast and loose gist processing when making decisions and performing recognition memory tasks. Thus, when given the option we typically prefer to minimize the amount of effort we put forth when making a decision.

In regards to memory, Fuzzy assumes there to be two types of memory traces, gist and verbatim, encoded for a given memory experience. A verbatim memory trace maintains the surface qualities and contextual details of a memory event. For example, if a person is presented with a Guinness beer they would remember they would encode a detailed representation for this particular Guinness. In contrast a gist memory trace only maintains the gist or the meaning present in the memory event. Also, gist is thought to vary along a continuum of precision. For example a person might remember a more precise gist representation of imported stout beer, or a more general gist representation, such as type of beer. Both of these representations capture the gist of the Guinness but one is far more descriptive than the other. Yet even the most precise gist trace still does not directly represent surface details. Although, such surface details can be inferred from gist memory traces.

In regards to recognition memory performance there is a order of operations. There is evidence that suggests that people first make a graded familiarity judgment as to whether or not they have seen a recognition test item previously based on gist. After this initial, fast process is completed a more detailed search for more specific verbatim information is instigated. Thus, a person first makes an overall familiarity judgment and then conducts a search for confirming or disconfirming verbatim traces. This is how Fuzzy assumes the order of operations occurs in recognition memory. However, the order is revered in Free Recall. We first attempt to directly access verbatim information and output events that we explicitly recall being on the test. After this direct retrieval process has failed, we utilize a reconstructive process. This reconstructive process accesses any available gist information and uses it to construct possible candidates events that occurred previously. So a person remembers a type of beer was presented. Maybe, Guinness was presented. Then, a metacognitive process judges the candidate item. If the judgment process ascertains that it is likely that Guinness was presented previously, Guinness will be outputted. But if the judgment process ascertains that it is unlikely that Guinness was presented previously Guinness will not be outputted.

One important assumption of this model is that participants will deplete the available verbatim information before they engage the reconstructive process. As a consequence words that are more likely to have a stable verbatim memory trace will be outputted prior to words that are more unlikely to have a stable verbatim memory trace. For instance, words that were presented on a DRM list are outputted prior to Critical Lures. This is consistent with the notion that direct access to verbatim memory traces is carried out prior to reconstructive gist based processes.

In the present model there are three process which are reliant upon two different memory representations. Thus, this is a dual representation model of free recall.

Direct access = D (reliant upon Verbatim memory traces)

Reconstruction = R (reliant upon Gist memory traces)

Metacognitive Judgment = J (judges constructed memory candidates)

The purpose of the first experiment was to introduce an experimental maipulation that would influence the availability of gist information and not the avialbility of verbatim information. As such this manipulation was predicted to influence the reconstructive process and not the ability of participants to directly access verbatim memory traces.

Participants were informed that they would be presented with four lists of related words. The testing environment was manipulated between participants. Some participants were either presented with a normal testing environment, in which they were prompted to recall as many of the previously presented words as possible (Control group). Some participants were prompted to recall as many words as possible. When they had recalled as many words as possible, to produce words that were probably presented on the list because they were meaningfully related to word sthat were presented, even if they were not able to specifically remember their presentation (Meaning Instructions). While still others, were asked to recall as many words as possible, but additionally were provided with the names of the categories they had been presented with during study (Category Cues).

Results

Not surprisingly, the following pattern was observed in participants’ ability to correctly recall presented items: control < meaning < category label. Estimates of the three memory parameters were found to differ between conditions. More importantly, parameter estimates of the reconstructive process was found to follow the same pattern as to that of the observed data: control < meaning < category. Also, the estimate of the metacognitive judgment was found to be significantly greater in the category label condition in comparison to both of the other conditions: control = meaning < category.

Direct retrieval is dependent upon the ability to access verbatim memory traces, which are composed of surface details. Thus, in an attempt to influence the availability of verbatim memory traces the processing of surface detail of the presented words was manipulated between participants. Thus, participants were in one of the following conditions: Control, unusual fonts, or Consonant Counting. It was assumed that displaying words in unusual fonts and requiring participants to count the number of consonants in a word would disrupt the processing of the orthography of the word.

Again, participants were presented with four lists of category exemplars. And again participants’ memories were tested in three consecutive free recall memory tests.

Results

As predicted, the parameter estimate of D was influenced by the manipulation. However, there were no differences between estimates of R between the three conditions.

Control Font Counting

D .44 .32 .17

R .32 .31 .27

In the present design participants were presented with paired associates. These paired associates are not related to one another in any meaningful manner. To investiate the influence of concreteness and abstractness on D, R and J, four different types of pairs of associates were formed: concrete – concrete, abstract – abstract, concrete – abstract, abstract – concrete. Unlike, the procedures used in the previous experiments, participants in the current experiment studied these associates between testing sessions (i.e., study : test : study : test : study : test). Also, the test was a cued recall opposed to free recall test.

Results

Recall was greater when the pairs included concrete words (either a target or a cue).

The model used to fit the data obtained in the present experiment was different from that used in the first two experiments. This model allowed for estimates of D, R and J to increase across testing conditions. This is change was made because of the addition of multiple study session to the procedure. But what is important is that the parameter estimates for direct access in the third session was greater for concrete opposed to abstract targets. Additionally, the parameter estimate for the reconstructive process obtained from data generated on the second and third cured recall tests were larger for abstract opposed to concrete words. Thus, level manipulating the level of abstraction produced a double dissociation.

The R parameter acted in a different manner to the manipulation than did the D parameter. Oddly, the parameter that appears to be most influenced by this manipulation is J. Concrete words were less likely to be associated with metacognitive judgments (M = .33) than abstract words (M = .63).

Tested the assumption that cued recall is more reliant upon direct retrieval whereas free recall is more reliant upon reconstructive processing.

Results

See table 8.