skip to content

Centre for Music and Science

 

Overview

I am University Assistant Professor in the Faculty of Music and Director of the Centre for Music and Science. I specialize in computational approaches to music psychology, including cognitive modelling, massive online experiments, and corpus studies. I am particularly interested in understanding the psychological mechanisms that underlie listeners' appreciation and enjoyment of music, and how musical styles have developed to exploit these mechanisms. My published works cover a variety of topics, including statistical learning, creativity, musical pleasure, consonance, voice leading, harmonic syntax, and individual differences in musical abilities.

Research topics and methodologies

My research topics generally fall into the domain of (music) psychology. I'm interested in understanding the human mind, particularly in the context of music perception and production, but I am also interested more broadly in the psychology of aesthetics, learning, and memory. I study these topics mostly with a combination of behavioural experiments (e.g. playing participants musical stimuli and recording their behavioural responses) and corpus analyses (e.g. tabulating statistical patterns in large datasets of music compositions), but I also regularly collaborate with neuroscientists (e.g. studying neural responses to particular musical stimuli).

Most of my work has some kind of computational basis, which manifests in various areas:

  1. High-throughput behavioural experiments. Data collection is one of the most important yet time-consuming part of the psychologist's job. A high-quality quantitative experiment might involve 50 or so participants; traditionally, the experimenter would have to meet each of these participants individually, explain the experiment to them, gain informed consent, present the stimuli, record the responses on paper, debrief the participant, and so on. This can be a very time-consuming process, and hence constitute a major bottleneck in the scientific research process. A major goal of mine is to enhance the efficiency of this process. Since 2016 I have been developing the psychTestR and PsyNet platforms, which allow researchers to develop highly sophisticated psychological experiments for online deployment, so that hundreds of participants can be recruited and tested in a matter of hours with minimal input from the experimenter.
    • Harrison, P. M. C. (2020). psychTestR: An R package for designing and conducting behavioural psychological experiments. The Journal of Open Source Software. https://doi.org/10.21105/joss.02088
    • Harrison, P. M. C., Marjieh, R., Adolfi, F., van Rijn, P., Anglada-Tort, M., Tchernichovski, O., Larrouy-Maestri, P., & Jacoby, N. (2020). Gibbs Sampling with People. Advances in Neural Information Processing Systems.
       
  2. Cognitive modelling. A big challenge in music psychology is how to test psychological theories in the context of rich perceptual environments such as music listening. Even an apparently simple psychological theory – for example, Berlyne's hypothesis that aesthetic reward is maximised by intermediate levels of stimulus complexity – becomes challenging in the context of 'real' music, where attributes such as 'complexity' suddenly become difficult to define with objectivity and cognitive validity. My work makes heavy use of computational cognitive modelling for these kinds of problems. The idea is that we build an idealised computer simulation of the psychological processes relevant to the problem, to which we can feed arbitrary stimuli and hence elicit predictions for how a human might react for these stimuli. By iterating between model construction, behavioural data collection, and model evaluation, we can refine our computational models to produce better simulations of the human participant, and hence improve our understanding of the underlying psychological processes.
    • Harrison, P. M. C., & Pearce, M. T. (2020). Simultaneous consonance in music perception and composition. Psychological Review, 127(2), 216–244. https://doi.org/10.1037/rev0000169
    • Cheung, V. K. M., Harrison, P. M. C., Meyer, L., Pearce, M. T., Haynes, J.-D., & Koelsch, S. (2019). Uncertainty and surprise jointly predict musical pleasure and amygdala, hippocampus, and auditory cortex activity. Current Biology, 29(23), 4084–4092.e4. https://doi.org/10.1016/j.cub.2019.09.067
    • Harrison, P. M. C., Bianco, R., Chait, M., & Pearce, M. T. (2020). PPM-Decay: A computational model of auditory prediction with memory decay. PLoS Computational Biology, 16(11), e1008304. https://doi.org/10.1371/journal.pcbi.1008304
       
  3. Corpus analyses. Music perception has a dual relationship with music creation. On the one hand, the way that people create music (as composers, as improvisers, as participants) is driven in large part by perception; for example, a composer might add particular appoggiaturas to a melody because they know that these appoggiaturas will cause the listener to perceive a certain emotion in the music. Conversely, the way that the listener perceives the music is deeply shaped by their own previous experiences of music listening; for example, the listener might only perceive the appoggiaturas as expressive through reference to a learned and culturally specific schema for tonality. Consequently, studies of music perception ought to go hand in hand with studies of music creation. These studies most typically take the form of corpus analyses, where we take large datasets of music (e.g. recordings of piano music, transcriptions of pop songs, or jazz lead sheets) and use computational methods to study and summarise aspects of musical structure (e.g. profiles of psychoacoustic dissonance, transition probabilities between chords, or trajectories of surprise and uncertainty). We can then use these analyses to try and explain why listeners hear music in a certain way, or conversely to understand why composers tend to write music in a certain way.
    • Harrison, P. M. C., & Pearce, M. T. (2020). A computational cognitive model for the analysis and generation of voice leadings. Music Perception, 37(3), 208–224. https://doi.org/10.1525/mp.2020.37.3.208
    • Harrison, P. M. C., & Pearce, M. T. (2020). Simultaneous consonance in music perception and composition. Psychological Review, 127(2), 216–244. https://doi.org/10.1037/rev0000169
    • Harrison, P. M. C., & Pearce, M. T. (2018). An energy-based generative sequence model for testing sensory theories of Western harmony. Proceedings of the 19th International Society for Music Information Retrieval Conference, 160–167.

I am deeply committed to the goal of cumulative science. Alongside any theoretical contributions, I'm always keen for my studies to produce something concrete and practical that facilitates future researchers' work in the area. These practical contributions can take various forms:

  1. Reusable experiment implementations. Different experiments often end up using common components, for example particular procedures for selecting stimuli (e.g. item response theory, Gibbs Sampling with People), particular questionnaires (e.g. personality inventories, musical experience surveys), or particular ability tests (e.g. non-verbal reasoning, pitch discrimination). A big motivator for my development of psychTestR and PsyNet was to help experimenters to build libraries of these components that can be shared among the research community for reuse in future experiments. Already psychTestR has been adopted by many researchers in the music psychology community, with many musical ability tests and questionnaires being now available in open-source repositories (see e.g. https://shiny.gold-msi.org/longgold_demo).
    • Harrison, P. M. C. (2020). psychTestR: An R package for designing and conducting behavioural psychological experiments. The Journal of Open Source Software. https://doi.org/10.21105/joss.02088
       
  2. Reusable methodologies. I am interested in developing psychological methodologies that can be applied to a broad variety of domains. So far I have focused primarily on adaptive procedures, which use some kind of computational logic for stimulus selection in order to increase the efficiency of data collection. Early on I worked on adaptive procedures in the individual-differences domain, where the goal is to measure an individual's ability on a particular task (e.g. memory for melodies) as efficiently as possible. I was particularly interested in developing musical tests that bridged traditional item response theory approaches with cognitive modelling, so that we could simultaneously improve data collection efficiency while also exploring the cognitive foundations of the task. Later on my colleagues and I developed a new paradigm called Gibbs Sampling with People, designed to probe rich perceptual spaces (e.g. melody, prosody, faces) by combining responses from many participants in an adaptive procedure based on the Gibbs Sampling algorithm. Both of these methodological approaches are finding many subsequent applications in different musical and non-musical domains.
    • Harrison, P. M. C., Collins, T., & Müllensiefen, D. (2017). Applying modern psychometric techniques to melodic discrimination testing: Item response theory, computerised adaptive testing, and automatic item generation. Scientific Reports, 7(1), 3618. https://doi.org/10.1038/s41598-017-03586-z
    • Harrison, P. M. C., & Müllensiefen, D. (2018). Development and Validation of the Computerised Adaptive Beat Alignment Test (CA-BAT). Scientific Reports, 8(1), 12395. https://doi.org/10.1038/s41598-018-30318-8
    • Harrison, P. M. C., Marjieh, R., Adolfi, F., van Rijn, P., Anglada-Tort, M., Tchernichovski, O., Larrouy-Maestri, P., & Jacoby, N. (2020). Gibbs Sampling with People. Advances in Neural Information Processing Systems.
       
  3. Reusable models. Much of my work involves developing and evaluating various psychoacoustic and cognitive models for simulating various aspects of music perception and production. I'm particularly interested in developing generalisable models that can be applied to arbitrary new experiments and datasets. I always release my model implementations as open-source code, typically in the form of R packages, which other researchers can then easily use in their own studies.
    • Harrison, P. M. C., & Pearce, M. T. (2020). Simultaneous consonance in music perception and composition. Psychological Review, 127(2), 216–244. https://doi.org/10.1037/rev0000169
    • Harrison, P. M. C., Bianco, R., Chait, M., & Pearce, M. T. (2020). PPM-Decay: A computational model of auditory prediction with memory decay. PLoS Computational Biology, 16(11), e1008304. https://doi.org/10.1371/journal.pcbi.1008304
       
  4. Reusable datasets. Many historic music psychology studies work by conducting experiments designed to probe a very specific hypothesis, typically evaluated using a simple group-difference analysis (e.g. t-test, ANOVA). Such experiments are often rhetorically effective in the context of a particular paper, but they typically leave little scope for reanalysis. I am particularly keen on experiments that deliver richer, more general datasets that can theoretically support many more analyses by future researchers. For example, in Harrison & Pearce (2020) we compiled a large dataset of consonance judgments sourced from many historic studies that could then be used to evaluate many different consonance models; this dataset has now been extended by Imre Lahdelma and Tuomas Eerola in the inconMore package, and reanalysed in a recent Music & Science paper (Eerola & Lahdelma, 2021).
    • Harrison, P. M. C., & Pearce, M. T. (2020). Simultaneous consonance in music perception and composition. Psychological Review, 127(2), 216–244. https://doi.org/10.1037/rev0000169
    • Eerola, T., & Lahdelma, I. (2021). The anatomy of consonance/dissonance: Evaluating acoustic and cultural predictors across multiple datasets with chords. Music & Science, 4. https://doi.org/10.1177/20592043211030471

Future projects and postgraduate supervision

I am interested in supervising all kinds of MPhil and PhD music psychology projects that adopt some or all of these methodologies and philosophies outlined above. While I am particularly receptive to projects with computational aspects, I am also happy to supervise other projects as long as they fit well with the 'cumulative science' principles outlined above.

That said, I’m particularly interested in a couple of lines of research right now, in which I'd be especially happy to supervise postgraduate students:

  • Better understanding musical pleasure. Musical pleasure is driven by a broad range of phenomena, ranging from low-level psychoacoustic attributes (e.g. loudness, roughness) to high-level psychological processes (e.g. memory, prediction, surprise, learning). I've focused mostly on the low-level side of this so far (in particular, consonance perception), but I'm interested in moving towards the high-level side, developing and applying sophisticated cognitive models that use machine learning to simulate the listener's cultural exposure.
  • Developing technologies for music education. Musicianship depends on honing a collection of complex skills, such as playing by ear, practising a piece into one's repertoire, and accompanying other musicians. I'm interested in developing technologies that help musicians to develop and improve these skills, taking advantage of recent innovations in digital signal processing and individualised learning.

Prospective students are very welcome to email me to ask for feedback about potential research projects. It is generally advisable to do so well in advance of the postgraduate funding deadlines, which mostly fall in early January for a start date of October (see How to Apply for details). The postgraduate applications require you to write a short research proposal, and this is an important opportunity to demonstrate that you understand the relevant literature, that you can write well, and that your project is a good fit for my supervision. However, it's not a final commitment; there will be significant opportunities to revise and update your plans once you begin your postgraduate degree.

If you think you would be interested in a computational project but don't yet have any programming experience, don't worry – there are amazing online resources available nowadays that allow you to learn programming skills at home or alongside your studies. I recommend in particular taking classes in either R or Python; previous students have had good success with online classes by DataCamp and Udemy. If you want some advice about where to start, you're very welcome to drop me an email.

CMS Logo

Dr Peter Harrison

Contact Details

Email address: 
Takes PhD students
Not available for consultancy

Affiliations

Classifications: 

Latest news

New paper: A case study investigating the interaction between tuba acoustic radiation patterns and performance spaces

27 October 2022

Recent CMS undergraduate Silas Sanders just presented his dissertation project at the 153rd Convention of the Audio Engineering Society. The project is entitled 'A case study investigating the interaction between tuba acoustic radiation patterns and performance spaces'. Silas put together an impressive pipeline of acoustic...

New PhD and research assistant positions: Score design for music reading

11 October 2022

We are happy to share several new positions at our Centre for Music and Science (Faculty of Music, Cambridge). These positions are part of the Leverhulme Trust-funded project Score designs for the facilitation of music reading: Cognitive and artistic perspectives led by Professor Ian Cross (Principal Investigator) and Dr...