Tonal Symmetry Induces Fluency and Sense of Well-Formedness

Abstract

Fluency influences grammaticality judgments of visually presented strings in artificial grammar learning (AGL). Of many potential sources that engender fluency, symmetry is considered to be an important factor. However, symmetry may function differently for visual and auditory stimuli, which present computationally different problems. Thus, the current study aimed to examine whether objectively manipulating fluency by speeding up perception (i.e., manipulating the inter-stimulus interval, ISI, between each syllable of a string) influenced judgments of tonal strings; and thus how symmetry-based fluency might influence judgments. In Experiment 1, with only a test phase, participants were required to give their preference ratings of tonal strings as a measurement of fluency. In experiment 2, participants were instructed to make grammaticality judgments after being incidentally trained on tonal symmetry. Results of Experiment 1 showed that tonal strings with shorter ISI were liked more than those with longer ISI while such difference was not found between symmetric and asymmetric strings without training. Additionally, Experiment 2 found both main effects of symmetry and ISI as well as an interaction. In particular, only asymmetric strings were more likely to be judged as grammatical when they were presented at a shorter ISI. Taken together, participants were sensitive to the fluency induced by the manipulation of ISI and sensitive to symmetry only after training. In sum, we conclude that objective speed influenced grammaticality judgments, implicit learning of tonal symmetry resulted in enhanced fluency, and that fluency may serve as a basis for grammaticality judgments.

How to use and report Bayesian hypothesis tests.

“…Here, B HN(0, x) refers to a Bayes factor in which the predictions of H1 were modeled as a halfnormal distribution with an SD of x, where x scales the size of effect that could be expected (see Dienes, 2014). Using the tonal inversion paradigm, when testing on the same length as training, Jiang et al (2012) found an effect of 7% above a control; Li, Jiang, Guo, Yang, and Dienes (2013) found 6%, and Qiao et al (2018) found 6%. Thus, the rough size of effect expected if there is learning of different lengths by a mechanism that learnt the inversions per se is 6% above baseline.…”

Simulations of implicit learning of symmetries: The importance of prior knowledge and the nature of the memory buffer

“…We will explore a possible computational mechanism for learning inversions by reference to a paradigm introduced by Jiang et al (2012) (and explored further by Li, Jiang, Guo, Yang, & Dienes, 2013, Ling, Li, Qiao, Guo, & Dienes, 2016, and Qiao et al, 2018. By controlling both chunks and repetition structures, Jiang et al showed that Chinese participants could implicitly learn to discriminate lexical tonal inversions from non-inversions.…”