Statistical learning and probabilistic prediction in music cognition: mechanisms of stylistic enculturation

Predictability and Uncertainty in the Pleasure of Music: A Reward for Learning?

Music ranks among the greatest human pleasures. It consistently engages the reward system, and converging evidence implies it exploits predictions to do so. Both prediction confirmations and errors are essential for understanding one's environment, and music offers many of each as it manipulates interacting patterns across multiple timescales. Learning models suggest that a balance of these outcomes (i.e., intermediate complexity) optimizes the reduction of uncertainty to rewarding and pleasurable effect. Yet evidence of a similar pattern in music is mixed, hampered by arbitrary measures of complexity. In the present studies, we applied a well-validated information-theoretic model of auditory expectation to systematically measure two key aspects of musical complexity: predictability (operationalized as information content [IC]), and uncertainty (entropy). In Study 1, we evaluated how these properties affect musical preferences in 43 male and female participants; in Study 2, we replicated Study 1 in an independent sample of 27 people and assessed the contribution of veridical predictability by presenting the same stimuli seven times. Both studies revealed significant quadratic effects of IC and entropy on liking that outperformed linear effects, indicating reliable preferences for music of intermediate complexity. An interaction between IC and entropy further suggested preferences for more predictability during more uncertain contexts, which would facilitate uncertainty reduction. Repeating stimuli decreased liking ratings but did not disrupt the preference for intermediate complexity. Together, these findings support long-hypothesized optimal zones of predictability and uncertainty in musical pleasure with formal modeling, relating the pleasure of music listening to the intrinsic reward of learning.SIGNIFICANCE STATEMENT Abstract pleasures, such as music, claim much of our time, energy, and money despite lacking any clear adaptive benefits like food or shelter. Yet as music manipulates patterns of melody, rhythm, and more, it proficiently exploits our expectations. Given the importance of anticipating and adapting to our ever-changing environments, making and evaluating uncertain predictions can have strong emotional effects. Accordingly, we present evidence that listeners consistently prefer music of intermediate predictive complexity, and that preferences shift toward expected musical outcomes in more uncertain contexts. These results are consistent with theories that emphasize the intrinsic reward of learning, both by updating inaccurate predictions and validating accurate ones, which is optimal in environments that present manageable predictive challenges (i.e., reducible uncertainty).

Music predictability and liking enhance pupil dilation and promote motor learning in non-musicians

Humans can anticipate music and derive pleasure from it. Expectations facilitate the learning of movements associated with anticipated events, and they are also linked with reward, which may further facilitate learning of the anticipated rewarding events. The present study investigates the synergistic effects of predictability and hedonic responses to music on arousal and motor-learning in a naïve population. Novel melodies were manipulated in their overall predictability (predictable/unpredictable) as objectively defined by a model of music expectation, and ranked as high/medium/low liked based on participants' self-reports collected during an initial listening session. During this session, we also recorded ocular pupil size as an implicit measure of listeners' arousal. During the following motor task, participants learned to play target notes of the melodies on a keyboard (notes were of similar motor and musical complexity across melodies). Pupil dilation was greater for liked melodies, particularly when predictable. Motor performance was facilitated in predictable rather than unpredictable melodies, but liked melodies were learned even in the unpredictable condition. Low-liked melodies also showed learning but mostly in participants with higher scores of task perceived competence. Taken together, these results highlight  the effects of stimuli predictability on learning, which can be however overshadowed by the effects of stimulus liking or task-related intrinsic motivation.

Reduced prediction error responses in high-as compared to low-uncertainty musical contexts

Theories of predictive processing propose that prediction error responses are modulated by the certainty of the predictive model or precision. While there is some evidence for this phenomenon in the visual and, to a lesser extent, the auditory modality, little is known about whether it operates in the complex auditory contexts of daily life. Here, we examined how prediction error responses behave in a more complex and ecologically valid auditory context than those typically studied. We created musical tone sequences with different degrees of pitch uncertainty to manipulate the precision of participants' auditory expectations. Magnetoencephalography was used to measure the magnetic counterpart of the mismatch negativity (MMNm) as a neural marker of prediction error in a multi-feature paradigm. Pitch, slide, intensity and timbre deviants were included. We compared high-entropy stimuli, consisting of a set of non-repetitive melodies, with low-entropy stimuli consisting of a simple, repetitive pitch pattern. Pitch entropy was quantitatively assessed with an information-theoretic model of auditory expectation. We found a reduction in pitch and slide MMNm amplitudes in the high-entropy as compared to the low-entropy context. No significant differences were found for intensity and timbre MMNm amplitudes. Furthermore, in a separate behavioral experiment investigating the detection of pitch deviants, similar decreases were found for accuracy measures in response to more fine-grained increases in pitch entropy. Our results are consistent with a precision modulation of auditory prediction error in a musical context, and suggest that this effect is specific to features that depend on the manipulated dimension-pitch information, in this case.

Statistical characteristics of tonal harmony: A corpus study of Beethoven's string quartets

Tonal harmony is one of the central organization systems of Western music. This article characterizes the statistical foundations of tonal harmony based on the computational analysis of expert annotations in a large corpus. Using resampling methods, this study shows that 1) the rank-frequency distribution of chords resembles a power law, i.e. few chords govern a large proportion of the data; 2) chord transitions are referential and chord predictability is significantly affected by distinguished chord features; 3) tonal harmony conveys directedness in time; and 4) tonal harmony operates differently at the hierarchical levels of chords and keys. These results serve to characterize tonal harmony on empirical grounds and advance the methodological state-of-the-art in digital musicology.

Learning and Consolidation as Re-representation: Revising the Meaning of Memory

In this Hypothesis and Theory paper, we consider the problem of learning deeply structured knowledge representations in the absence of predefined ontologies, and in the context of long-term learning. In particular, we consider this process as a sequence of re-representation steps, of various kinds. The Information Dynamics of Thinking theory (IDyOT) admits such learning, and provides a hypothetical mechanism for the human-like construction of hierarchical memory, with the provision of symbols constructed by the system that embodies the theory. The combination of long-term learning and meaning construction in terms of symbols grounded in perceptual experience entails that the system, like a human, be capable of memory consolidation, to manage the complex and inconsistent structures that can result from learning of information that becomes more complete over time. Such consolidation changes memory structures, and thus changes their meaning. Therefore, memory consolidation entails re-representation, while re-representation entails changes of meaning. Ultimately, the theory proposes that the processes of learning and consolidation should be considered as repeated re-representation of what is learned.

Intracranial Recordings and Computational Modeling of Music Reveal the Time Course of Prediction Error Signaling in Frontal and Temporal Cortices

Prediction is held to be a fundamental process underpinning perception, action, and cognition. To examine the time course of prediction error signaling, we recorded intracranial EEG activity from nine presurgical epileptic patients while they listened to melodies whose information theoretical predictability had been characterized using a computational model. We examined oscillatory activity in the superior temporal gyrus (STG), the middle temporal gyrus (MTG), and the pars orbitalis of the inferior frontal gyrus, lateral cortical areas previously implicated in auditory predictive processing. We also examined activity in anterior cingulate gyrus (ACG), insula, and amygdala to determine whether signatures of prediction error signaling may also be observable in these subcortical areas. Our results demonstrate that the information content (a measure of unexpectedness) of musical notes modulates the amplitude of low-frequency oscillatory activity (theta to beta power) in bilateral STG and right MTG from within 100 and 200 msec of note onset, respectively. Our results also show this cortical activity to be accompanied by low-frequency oscillatory modulation in ACG and insula-areas previously associated with mediating physiological arousal. Finally, we showed that modulation of low-frequency activity is followed by that of high-frequency (gamma) power from approximately 200 msec in the STG, between 300 and 400 msec in the left insula, and between 400 and 500 msec in the ACG. We discuss these results with respect to models of neural processing that emphasize gamma activity as an index of prediction error signaling and highlight the usefulness of musical stimuli in revealing the wide-reaching neural consequences of predictive processing.

Atonal Music: Can Uncertainty Lead to Pleasure?

In recent years, the field of neuroaesthetics has gained considerable attention with music being a favored object of study. The majority of studies concerning music have, however, focused on the experience of Western tonal music (TM), which is characterized by tonal hierarchical organization, a high degree of consonance, and a tendency to provide the listener with a tonic as a reference point during the listening experience. We argue that a narrow focus on Western TM may have led to a one-sided view regarding the qualities of the aesthetic experience of music since Western art music from the 20th and 21st century like atonal music (AM) – while lacking a tonal hierarchical structure, and while being highly dissonant and hard to predict – is nevertheless enjoyed by a group of avid listeners. We propose a research focus that investigates, in particular, the experience of AM as a novel and compelling way with which to enhance our understanding of both the aesthetic appreciation of music and the role of predictive models in the context of musical pleasure. We use music theoretical analysis and music information retrieval methods to demonstrate how AM presents the listener with a highly uncertain auditory environment. Specifically, an analysis of a corpus of 100 musical segments is used to illustrate how tonal classical music and AM differ quantitatively in terms of both key and pulse clarity values. We then examine person related, extrinsic and intrinsic factors, that point to potential mechanisms underlying the appreciation and pleasure derived from AM. We argue that personality traits like “openness to experience,” the framing of AM as art, and the mere exposure effect are key components of such mechanisms. We further argue that neural correlates of uncertainty estimation could represent a central mechanism for engaging with AM and that such contexts engender a comparatively weak predictive model in the listener. Finally we argue that in such uncertain contexts, correct predictions may be more subjectively rewarding than prediction errors since they signal to the individual that their predictive model is improving.