Modelling unsupervised online-learning of artificial grammars: linking implicit and statistical learning

Western listeners detect boundary hierarchy in Indian music: a segmentation study

How are listeners able to follow and enjoy complex pieces of music? Several theoretical frameworks suggest links between the process of listening and the formal structure of music, involving a division of the musical surface into structural units at multiple hierarchical levels. Whether boundaries between structural units are perceivable to listeners unfamiliar with the style, and are identified congruently between naïve listeners and experts, remains unclear. Here, we focused on the case of Indian music, and asked 65 Western listeners (of mixed levels of musical training; most unfamiliar with Indian music) to intuitively segment into phrases a recording of sitar ālāp of two different rāga-modes. Each recording was also segmented by two experts, who identified boundary regions at section and phrase levels. Participant- and region-wise scores were computed on the basis of "clicks" inside or outside boundary regions (hits/false alarms), inserted earlier or later within those regions (high/low "promptness"). We found substantial agreement-expressed as hit rates and click densities-among participants, and between participants' and experts' segmentations. The agreement and promptness scores differed between participants, levels, and recordings. We found no effect of musical training, but detected real-time awareness of grouping completion and boundary hierarchy. The findings may potentially be explained by underlying general bottom-up processes, implicit learning of structural relationships, cross-cultural musical similarities, or universal cognitive capacities.

Mini Pinyin: A modified miniature language for studying language learning and incremental sentence processing

Artificial grammar learning (AGL) paradigms are used extensively to characterise (neuro)cognitive bases of language learning. However, despite their effectiveness in characterising the capacity to learn complex structured sequences, AGL paradigms lack ecological validity and typically do not account for cross-linguistic differences in sentence comprehension. Here, we describe a new modified miniature language paradigm - Mini Pinyin - that mimics natural language as it is based on an existing language (Mandarin Chinese) and includes both structure and meaning. Mini Pinyin contains a number of cross-linguistic elements, including varying word orders and classifier-noun rules. To evaluate the effectiveness of Mini Pinyin, 76 (mean age = 24.9; 26 female) monolingual native English speakers completed a learning phase followed by a sentence acceptability judgement task. Generalised mixed effects modelling revealed that participants attained a moderate degree of accuracy on the judgement task, with performance scores ranging from 25% to 100% accuracy depending on the word order of the sentence. Further, sentences compatible with the canonical English word order were learned more efficiently than non-canonical word orders. We controlled for inter-individual differences in statistical learning ability, which accounted for ~20% of the variance in performance on the sentence judgement task. We provide stimuli and statistical analysis scripts as open-source resources and discuss how future research can utilise this paradigm to study the neurobiological basis of language learning. Mini Pinyin affords a convenient tool for improving the future of language learning research by building on the parameters of traditional AGL or existing miniature language paradigms.

Modeling test learning and dual-task dissociations

Much of cognitive psychology is premised on the distinction between automatic and intentional processes, but the distinction often remains vague in practice and alternative explanations are often not followed through. For example, Hendricks, Conway and Kellogg (Journal of Experimental Psychology: Learning, Memory, and Cognition, 39, 491–1500, 2013) found that dual tasks at training versus at test dissociated performance in two different artificial grammar learning tasks. This was taken as evidence for underlying automatic and intentional processes. In this article, a different explanation is considered based on test learning and similarity, where participants are assumed to update their knowledge at test. Contrasting formal memory models of test learning are implemented, and it is concluded that the models account for the relevant dissociations without assuming a distinction between automatic and intentional processes.

Auditory but Not Audiovisual Cues Lead to Higher Neural Sensitivity to the Statistical Regularities of an Unfamiliar Musical Style

It is still a matter of debate whether visual aids improve learning of music. In a multisession study, we investigated the neural signatures of novel music sequence learning with or without aids (auditory-only: AO, audiovisual: AV). During three training sessions on three separate days, participants (nonmusicians) reproduced (note by note on a keyboard) melodic sequences generated by an artificial musical grammar. The AV group (n = 20) had each note color-coded on screen, whereas the AO group (n = 20) had no color indication. We evaluated learning of the statistical regularities of the novel music grammar before and after training by presenting melodies ending on correct or incorrect notes and by asking participants to judge the correctness and surprisal of the final note, while EEG was recorded. We found that participants successfully learned the new grammar. Although the AV group, as compared to the AO group, reproduced longer sequences during training, there was no significant difference in learning between groups. At the neural level, after training, the AO group showed a larger N100 response to low-probability compared with high-probability notes, suggesting an increased neural sensitivity to statistical properties of the grammar; this effect was not observed in the AV group. Our findings indicate that visual aids might improve sequence reproduction while not necessarily promoting better learning, indicating a potential dissociation between sequence reproduction and learning. We suggest that the difficulty induced by auditory-only input during music training might enhance cognitive engagement, thereby improving neural sensitivity to the underlying statistical properties of the learned material.

Cross-cultural differences in implicit learning of chunks versus symmetries

Three experiments explore whether knowledge of grammars defining global versus local regularities has an advantage in implicit acquisition and whether this advantage is affected by cultural differences. Participants were asked to listen to and memorize a number of strings of 10 syllables instantiating an inversion (i.e. a global pattern); after the training phase, they were required to judge whether new strings were well formed. In Experiment 1, Western people implicitly acquired the inversion rule defined over the Chinese tones in a similar way as Chinese participants when alternative structures (specifically, chunking and repetition structures) were controlled. In Experiments 2 and 3, we directly pitted knowledge of the inversion (global) against chunk (local) knowledge, and found that Chinese participants had a striking global advantage in implicit learning, which was greater than that of Western participants. Taken together, we show for the first time cross-cultural differences in the type of regularities implicitly acquired.

Neurophysiological Markers of Statistical Learning in Music and Language: Hierarchy, Entropy, and Uncertainty

Statistical learning (SL) is a method of learning based on the transitional probabilities embedded in sequential phenomena such as music and language. It has been considered an implicit and domain-general mechanism that is innate in the human brain and that functions independently of intention to learn and awareness of what has been learned. SL is an interdisciplinary notion that incorporates information technology, artificial intelligence, musicology, and linguistics, as well as psychology and neuroscience. A body of recent study has suggested that SL can be reflected in neurophysiological responses based on the framework of information theory. This paper reviews a range of work on SL in adults and children that suggests overlapping and independent neural correlations in music and language, and that indicates disability of SL. Furthermore, this article discusses the relationships between the order of transitional probabilities (TPs) (i.e., hierarchy of local statistics) and entropy (i.e., global statistics) regarding SL strategies in human's brains; claims importance of information-theoretical approaches to understand domain-general, higher-order, and global SL covering both real-world music and language; and proposes promising approaches for the application of therapy and pedagogy from various perspectives of psychology, neuroscience, computational studies, musicology, and linguistics.

Time-course variation of statistics embedded in music: Corpus study on implicit learning and knowledge

Learning and knowledge of transitional probability in sequences like music, called statistical learning and knowledge, are considered implicit processes that occur without intention to learn and awareness of what one knows. This implicit statistical knowledge can be alternatively expressed via abstract medium such as musical melody, which suggests this knowledge is reflected in melodies written by a composer. This study investigates how statistics in music vary over a composer's lifetime. Transitional probabilities of highest-pitch sequences in Ludwig van Beethoven's Piano Sonata were calculated based on different hierarchical Markov models. Each interval pattern was ordered based on the sonata opus number. The transitional probabilities of sequential patterns that are musical universal in music gradually decreased, suggesting that time-course variations of statistics in music reflect time-course variations of a composer's statistical knowledge. This study sheds new light on novel methodologies that may be able to evaluate the time-course variation of composer's implicit knowledge using musical scores.

Fluency Expresses Implicit Knowledge of Tonal Symmetry

The purposes of the present study were twofold. First, we sought to establish whether tonal symmetry produces processing fluency. Second, we sought to explore whether symmetry and chunk strength express themselves differently in fluency, as an indication of different mechanisms being involved for sub- and supra-finite state processing. Across two experiments, participants were asked to listen to and memorize artificial poetry showing a mirror symmetry (an inversion, i.e., a type of cross serial dependency); after this training phase, people completed a four-choice RT task in which they were presented with new artificial poetry. Participants were required to identify the stimulus displayed. We found that symmetry sped up responding to the second half of strings, indicating a fluency effect. Furthermore, there was a dissociation between fluency effects arising from symmetry vs. chunk strength, with stronger fluency effects for symmetry rather than chunks in the second half of strings. Taken together, we conjecture a divide between finite state and supra-finite state mechanisms in learning grammatical sequences.

Under the hood of statistical learning: A statistical MMN reflects the magnitude of transitional probabilities in auditory sequences

Within the framework of statistical learning, many behavioural studies investigated the processing of unpredicted events. However, surprisingly few neurophysiological studies are available on this topic, and no statistical learning experiment has investigated electroencephalographic (EEG) correlates of processing events with different transition probabilities. We carried out an EEG study with a novel variant of the established statistical learning paradigm. Timbres were presented in isochronous sequences of triplets. The first two sounds of all triplets were equiprobable, while the third sound occurred with either low (10%), intermediate (30%), or high (60%) probability. Thus, the occurrence probability of the third item of each triplet (given the first two items) was varied. Compared to high-probability triplet endings, endings with low and intermediate probability elicited an early anterior negativity that had an onset around 100 ms and was maximal at around 180 ms. This effect was larger for events with low than for events with intermediate probability. Our results reveal that, when predictions are based on statistical learning, events that do not match a prediction evoke an early anterior negativity, with the amplitude of this mismatch response being inversely related to the probability of such events. Thus, we report a statistical mismatch negativity (sMMN) that reflects statistical learning of transitional probability distributions that go beyond auditory sensory memory capabilities.