Bridging the gap: Transitive associations between items presented in similar temporal contexts.

Abstract

In episodic memory tasks, associations are formed between items presented close together in time. The temporal context model (TCM) hypothesizes that this contiguity effect is a consequence of shared temporal context rather than temporal proximity per se. Using double-function lists of paired associates (e.g., A-B, B-C) presented in a random order, the authors examined associations between items that were not presented close together in time but that were presented in similar temporal contexts. After learning, across-pair associations fell off with distance in the list, as if subjects had integrated the pairs into a coherent memory structure. Within-pair associations (e.g., A-B) were strongly asymmetric favoring forward transitions; across-pair associations (e.g., A-C) showed no evidence of asymmetry. While this pattern of results presented a stern challenge for a heteroassociative mediated chaining model, TCM provided an excellent fit to the data. These findings suggest that contiguity effects in episodic memory do not reflect direct associations between items but rather a process of binding, encoding, and retrieval of a gradually changing representation of temporal context.

Sequential learning using temporal context

“…We assume that the context vector is always normalized such that the sum over all its components is equal to one. This differs from most previous treatments of TCM (but see Howard, Jing, Rao, Provyn, & Datey, 2009). The choice of L 1 norm is important in enabling many of the steps of the proofs here.…”

“…The parameter φ is a free parameter of the model that lies in the range (0, 1); its value determines the ratio of the semantic component to the fixed component in the input pattern |c IN . The presence of the fixed representation avoids a collapse of the context vectors to a point (see Howard et al, 2005Howard et al, , 2009, for a discussion), hence we shall take φ to be strictly less than one. In this paper, we will see that the presence of the semantic representation (φ = 0) enables the model to generalize from the sequences it has observed to sequences it has not yet observed, analogous to the recovery of |t i−1 in previous applications of TCM (Howard et al, 2009;Rao & Howard, 2008).…”

“…So, throughout this paper, we will always take γ to be zero. Our semantic representation |s α is composed of the prediction vectors that were present when item α was presented rather than the prior contexts in which item α was presented (Howard et al, 2009;Rao & Howard, 2008). This is essential for sequentially organized materials, such as language.…”

Evaluating models of recognition memory using first- and second-choice responses

“…In the present case, the more complex models show considerable overlap in the range of data patterns that they can account for (see Fig. 2) which leads to model complexity being the major source of model discriminability (for a similar case in free recall research, see Howard, Jing, Rao, Provyn, & Datey, 2009). The fact that models based on distinct assumptions can predict similar results should be interpreted as a major shortcoming of this specific method.…”

Temporal contexts: Filling the gap between episodic memory and associative learning.

“…Convergent evidence for this proposal can be found in a study by Howard, Jing, Rao, Provyn, and Datey (2009), in which participants were presented with lists of paired associates (e.g., AB DE, BC EF) in a random order; the associations between items that were not presented closely together in time but were presented in similar contexts (i.e., with context in this case being the other member of the study pair such that A and C occur in the same context, namely, B) were subsequently examined. As should be expected from the TCM, participants were able to associate distant events that shared a common context.…”