[Online] Effect of Strength on Response Latency in Free Recall


Many memory models treat free recall initiation as a decision-making process in which items compete with each other to be retrieved. During this competition, the list-strength effect (LSE: Ratcliff et al., 1990) suggests that strengthen some of the list items would impair memories of other unstrengthened items. A key question remains is that how exactly do these strengthened items interfere with those unstrengthened items? While traditional linear ballistic accumulator (LBA: Brown & Heathcote, 2008) makes absolute strength prediction that items compete with each other independently; the search of associative memory (SAM: Raaijmakers & Shiffrin, 1981) model and makes relative strength prediction that increasing memory strengths of some list items would decrease strengths of the other items. Therefore, we use drift-rate normalisation to instantiate relative strength in LBA to investigate if inhibition between items is responsible for observation of LSE.

Research Questions / Hypotheses

We investigated whether strengthening memories for some list items would affect memory strengths of other items, and whether additional presentations of items would create copies of their original memory representations or simply strengthen them.


32 participants completed the study.


Participants were to complete the experiments on their own computers and complete in three sessions with 24-hour in between. During the experiments, participants were presented with 10-item study lists, then distractor arithmetic tasks, then post-cued to recall as many words as they could in any order they like.


First, the incremental-strength models that assume additional presentations of items would simply strengthen their original memory representations won over the multiple-copies models that assume additional presentations would create copies of original memory representations.    Secondly, the models without drift-rate normalisation outperformed those without.    Third, LSE was observed only in analysis of complete sequence of recall, but not in first recalls.


Firstly, additional presentations of items would strengthen original memory representations rather than create multiple copies of them.    Secondly, as normalisation models failed to address first recall data while LSE was not observed in such data, it is possible that output interference in complete sequence of recall is responsible for observation of LSE. In particular, a specific mechanism for output interference that is not inhibition-based (e.g., noise) could better account for LSE.    Planned communication of results: honours thesis.