Eddy J. Davelaar

The Demise of Short-Term Memory Revisited: Empirical and Computational Investigations of Recency Effects.

In the single-store model of memory, the enhanced recall for the last items in a free-recall task (i.e., the recency effect) is understood to reflect a general property of memory rather than a separate short-term store. This interpretation is supported by the finding of a long-term recency effect under conditions that eliminate the contribution from the short-term store. In this article, evidence is reviewed showing that recency effects in the short and long terms have different properties, and it is suggested that 2 memory components are needed to account for the recency effects: an episodic contextual system with changing context and an activation-based short-term memory buffer that drives the encoding of item-context associations. A neurocomputational model based on these 2 components is shown to account for previously observed dissociations and to make novel predictions, which are confirmed in a set of experiments.

Individual differences in semantic short-term memory capacity and reading comprehension

We report three correlation studies, which investigate the hypothesis that individual differences in the capacity of a semantic short-term memory (STM) component in working memory (WM) predict performance on complex language tasks. To measure the capacity of semantic STM, we devised a storage-only measure, the conceptual span, which makes use of a category-cued recall procedure. In the first two studies, where the conceptual span was administered with randomized words (not blocked by categories), we found that conceptual span predicted single-sentence and text comprehension, semantic anomaly detection and verbal problem solving, explaining unique variance beyond non-word and word span. In some cases, the conceptual span explained unique variance beyond the reading span. Conceptual span correlated better with verbal problem solving than reading span, suggesting that a storage-only measure can outperform a storage-plus-processing measure. In Study 3, the conceptual span was administered with semantically clustered lists. The clustered span correlated with the comprehension measures as well as the non-clustered span, indicating that the critical process is memory maintenance and not semantic clustering. Moreover, we found an interaction between subjects’ performance on the conceptual span and the effect of the distance between critical words in anomaly detection, supporting the proposal that semantic STM maintains unintegrated word meanings.

An interference account of cue-independent forgetting in the no-think paradigm

Memory suppression is investigated with the no-think paradigm, which produces forgetting following repeated practice of not thinking about a memory [Anderson MC, Green C (2001) Nature 410:366–369]. Because the forgotten item is not retrieved even when tested with an independent, semantically related cue, it has been assumed that this forgetting is due to an inhibition process. However, this conclusion is based on a single stage to recall, whereas global memory models, which produce forgetting through a process of interference, include both a sampling and a recovery stage to recall. By assuming that interference exists during recovery, these models can explain cue-independent forgetting. We tested several predictions of this interference explanation of cue-independent forgetting by modifying the think/no-think paradigm. We added a condition where participants quickly pressed enter rather than not thinking. We also manipulated initial memory strength and tested recognition memory. Most importantly, learning to quickly press enter produced as much cue-independent forgetting as no-think instructions. Demonstrating the adequacy of two-stage recall, a simple computational model (SAM-RI) simultaneously captured the original cue, independent cue, and recognition results.

Sequential dependencies in the Eriksen flanker task: A direct comparison of two competing accounts

In the conflict/control loop theory proposed by Botvinick, Braver, Barch, Carter, and Cohen (2001), conflict monitored in a trial leads to an increase in cognitive control on the subsequent trial. The critical data pattern supporting this assertion is the so-called Gratton effect—the decrease in flanker interference following incongruent trials—which was initially observed in the Eriksen flanker task. Recently, however, the validity of the idea that this pattern supports a general conflict/control mechanism has been questioned on the grounds that the Gratton effect is only observed with stimulus repetition. We present an experiment testing whether the Gratton effect reflects a stimulus-independent increase in cognitive control or stimulus-specific repetition priming. Although our results support the latter hypothesis, the priming effect is modulated by the congruency of the previous trial. We discuss a new mechanism through which monitored conflict is used to exert executive control by modulating stimulus-response associations.

On the lawfulness of the decision to terminate memory search

Nearly every memory retrieval episode ends with a decision to terminate memory search. Yet, no research has investigated whether these search termination decisions are systematic, let alone whether they are made consistent with a particular rule. In the present paper, we used a modified free-recall paradigm to examine the decision to terminate search. Data from two experiments revealed that the total time engaged in retrieval was a monotonically increasing function of the total number of items retrieved, whereas the time from final retrieval to search termination (exit latency) was a monotonically decreasing function of the total number of items retrieved. These findings were compared to the predictions of previously proposed stopping rules, using the Search of Associative Memory framework. Of the four rules examined, only one predicts the obtained data pattern.

Neuromodulation of decision and response selection

We present a model for the attentional neuromodulation of decision and selection processes. The model assumes that phasic responses in the brain nucleus Locus Coeruleus modulate, via the transmission of norepinephrine, the synaptic efficiency of neural circuits, at specific (stimulus and task dependent) time intervals. The model is applied first, to a task of perceptual choice, simulating attentional fluctuations and accounting for a series of behavioral and neurophysiological data. Second, the flexibility of information processing, whereby the parameters of the local circuits are modified online, is illustrated in the application of the model to a task of selection from short-term memory.

A computational study of conflict-monitoring at two levels of processing: Reaction time distributional analyses and hemodynamic responses

The conflict-monitoring hypothesis of cognitive control proposes that response-conflict is higher in incongruent conditions compared to congruent or neutral conditions and that increases in conflict lead to increased control on subsequent trials. A neurocomputational model is used to address data on reaction time distributions and hemodynamic responses in a flanker task with neutral (N), congruent (CO), stimulus-incongruent (SI), and response-incongruent (RI) trials, allowing investigation of stimulus- and response-conflict. A computational study is presented in which the conflict-signal is (a) computed at every level of processing (response, stimulus) and is (b) used to modulate the input in the same trial. Results show that the models capture (1) the profile of distributional plots seen in the behavioral literature and (2) the patterns of hemodynamic responses seen in the neuroimaging literature. Based on the simulations it is suggested that the prefrontal cortex processes response-conflict and that the parietal cortex processes stimulus-conflict.

Semantic similarity dissociates short- from long-term recency effects: Testing a neurocomputational model of list memory

The finding that recency effects can occur not only in immediate free recall (i.e., short-term recency) but also in the continuous-distractor task (i.e., long-term recency) has led many theorists to reject the distinction between short- and long-term memory stores. Recently, we have argued that long-term recency effects do not undermine the concept of a short-term store, and we have presented a neurocomputational model that accounts for both short- and long-term recency and for a series of dissociations between these two effects. Here, we present a new dissociation between short- and long-term recency based on semantic similarity, which is predicted by our model. This dissociation is due to the mutual support between associated items in the short-term store, which takes place in immediate free recall and delayed free recall but not in continuous-distractor free recall.

Age-related declines in context maintenance and semantic short-term memory

This study reports age-related declines in context maintenance (Braver et al., 2001) and semantic short-term memory (STM) and evidence for a relation between the two. A group of younger and older adults completed a context maintenance task (AX-CPT), a semantically oriented STM task (conceptual span), a phonologically oriented STM task (digit span), and a meaning integration task (semantic anomaly judgement). In the AX-CPT task, a target response is required to the probe letter “X” but only when it is preceded by the letter “A” (the context). Either three (short interference) or six distractor letters (long interference) were presented between the cue and the probe. Results indicated an age-related deficit in context maintenance. Age-related declines were also observed for conceptual span and semantic anomaly judgement but not for digit span. Context maintenance was correlated with conceptual span and semantic anomaly judgement but not with digit span. These correlations were largely mediated by age differences, which also explained variance that was unique to (and not shared among) context maintenance, conceptual span, and semantic anomaly judgement.

Sequential Retrieval and Inhibition of Parallel (Re)Activated Representations: A Neurocomputational Comparison of Competitive Queuing and Resampling Models

Sequential behavior is observed in various domains of cognitive psychology, including free recall paradigms. In this article, within a neurocomputational framework, resampling (RS) mechanisms are compared with competitive queuing (CQ) mechanisms. While both types of implementations select the most active representation, the subsequent inhibition is at the level of selection for RS models and at the level of (re)activation for CQ models. It is shown that despite the overwhelming success of CQ models in serial recall (with regard to types of sequencing error) RS models outperform CQ models with regard to inter-response times in a free recall task. Additional analyses show that decay of response suppression reduces the difference between the models. The RS model is sensitive to the size of the search set and accounts for memory selection performance in patients with Alzheimer’s dementia or Huntington’s disease. Finally, a non-mnemonic clustering behavior is observed, which is related to the dynamical process of the selection mechanism.