The interaction between temporal grouping and phonotactic chunking in short-term serial order memory for novel verbal sequences

ABSTRACT

The current study investigated the ways long-term memory contributes to short-term serial order memory of novel verbal sequences, focusing on long-term knowledge of bi-element frequency, that is, co-occurrence frequency of two consecutive elements in a linguistic environment. Participants performed two types of immediate serial recall of nine-element (nine-mora) sequences: low bi-mora frequency sequences where all eight associations between the nine morae were low frequency, and mixed bi-mora frequency sequences, with high-frequency associations for six of the eight bi-morae. Experiment 1 confirmed the bi-directional bi-mora frequency effect, meaning better recall performance for morae having high-frequency association with either the preceding mora (forward association) or the following mora (backward association). In Experiment 2, two temporal pauses were inserted in each list to disrupt high-frequency associations with the preceding mora or the following mora. The results showed that the bi-element frequency effect diminished when the high-frequency backward association was disrupted but the effect remained when the high-frequency forward association was disrupted. We discuss possible mechanisms underlying the asymmetric influence of temporal pauses on interactions between short-term memory and linguistic long-term memory.

KEYWORDS:

• Serial order memory
• immediate serial recall
• phonotactic frequency
• chunking
• temporal grouping

Acknowledgements

This research was financially supported in part by JSPS KAKENHI Grant Number 25380980 and 16K4424 to SS. MN was supported by JSPS Postdoctoral Fellowships for Research Abroad. We are grateful to Steve Majerus and an anonymous reviewer for helpful comments during the review process. We also appreciate David C. Plaut for his comments on earlier draft and daily discussion about the topic of serial order control in general.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1 A Japanese mora consists of a vocalic nucleus, a vowel with onset consonant, a nasal consonant /N/ (only for syllabic coda), or a geminate consonant /Q/. A long vowel /R/ is counted as two morae.

2 The number of morae in a sequence was matched to Tanida et al. (2015), which identified the bi-mora frequency effect in Japanese university students. An experiment in that study employed three tri-mora nonword sequences for immediate serial recall with auditory presentation and oral recall.

3 The set of mora items employed was consistent across participants, since every participant experienced the same mora set sequence as either a seed sequence, the corresponding dextral sequence, or the corresponding sinistral sequence (e.g., a seed sequence ‘mo-to-hi-zo-wa-bo-ge-pu-ke’ might be employed for participant A and the derived dextral sequence ‘ke-mo-to-hi-zo-wa-bo-ge-pu’ for participant B).

4 Barr et al. (2013) warn that analyses using models with only random intercepts increase the possibility of false positive. Therefore, following their outline, we performed separate analyses to assess each fixed effect while considering the influence of each tested random slope. We established models with all fixed effects, all random intercepts, and each tested random slope (when models did not converge, we dropped random slopes that were highly correlated with the random intercept or whose correlations were not calculated) and compared them with models dropping each tested fixed effect term, using likelihood ratio tests. We obtained almost the same results, except for the fixed effect of interaction between backward association frequency and posterior pause. Regression analysis of the model with the random slope of the interaction showed that the interaction was marginally significant (p = .08). Table A3 in Appendix 3 provides detailed outcomes of the regression analysis and the likelihood ratio test.

Additional information

Funding

This research was financially supported in part by JSPS KAKENHI Grant Number 25380980 and 16K4424 to SS. MN was supported by JSPS Postdoctoral Fellowships for Research Abroad.