Semantic feature effect in verbal short-term memory

ABSTRACT

The present study explored the influences of semantic features in immediate serial recall in order to further examine the involvement of semantic knowledge in short-term memory. The number of semantic features (NoF) was found to have a positive effect on short-term recall where high NoF words were remembered better than low NoF words (Experiment 1). This effect was replicated in a second experiment and was found to persist even after controlling for a potential confound (number of distinguishing features). It was further found that having more distinctive features facilitated recall performance of words whose representation was semantically poorer (Experiment 3). These results provide additional evidence of semantic influences in short-term memory and demonstrate that the organisation of semantic knowledge is reflected in short-term memory performance.

KEYWORDS:

• Semantic features
• immediate serial recall
• semantic knowledge
• short-term memory

Acknowledgements

This research is based in part on MCL’s PhD thesis. This research has been conducted with the support of the Australian Government Research Training Program Scholarship.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 Based on the effect size observed in Experiment 1, the results of a power analysis showed that the estimated power to observe significant effects with a sample size of 20 was 99%. The effect size was based on the partial eta-squared obtained from the repeated measures ANOVA analysis which has originally motivated the selection of sample size and design in subsequent experiments.

2 That is, as the number of items correctly recalled increases, so does the number of order errors.

3 This model included random intercepts for participants. We further examined 2 other models: (1) Having random slopes for NoF and serial position fitted, in addition to the inclusion of random intercepts for participants; and (2) having random slopes for NoF, in addition to the inclusion of random intercepts for participants. The first model failed to converge, while the second model did not change the pattern of results reported. The second model was compared to the model that included only random intercepts for participants, and the results showed both models did not significantly differ in terms of fit to the data. For these reasons, subsequent experiments will be analysed using the model that included only random intercepts for participants.

4 F-values were obtained from the Analysis of Variance Table with Satterthwaite’s method that accompanied the linear mixed effects analysis output.

5 Similar findings were observed when the BF of the model [NoF + position + presentation + participant] was compared to the BF of the model [position + presentation + participant]; the model containing the NoF term was preferred by a factor of 194.

6 Similar findings were observed when the BF of the model [NoDF + position + presentation + participant] was compared to the BF of the model [position + presentation + participant]; the model excluding the NoDF term was preferred by a factor of 4.

7 After each testing session, participants were asked informally during the debrief session whether they noticed any differences between the word lists. All participants indicated no perceived differences among the words presented.

Additional information

Funding

This research has been conducted with the support of the Australian Government Research Training Program Scholarship.