Personal tools

The Science and Technology of Music

Home Activities Workshops and Research Meetings Research Workshops Workshop on continuous response data analysis

Skip to content. | Skip to navigation

Workshop on continuous response data analysis

— filed under:

This workshop is organized by CIRMMT Research Axis 5 (Music Perception and Cognition). It will take place on June 14, 2011, from 9:00am-6:00pm in A832 (New Music Building).

  • Research Workshop
When Jun 14, 2011
from 08:30 AM to 06:00 PM
Where A832, New Music Building, 527 Sherbrooke St. West.
Add event to calendar vCal


Registration is mandatory as seating is limited, and is on a first-come first-served basis (30-35 seats).  Reservations for this workshop are now full.  If you have a specific enquiry, please contact Jacqui Bednar (jacqueline.bednar [AT]


This workshop will address issues in the collection, analysis and interpretation of continuous responses while listening to music. Both behavioural and physiological time-series data will be considered. The panel of international experts in time series analysis, functional data analysis and activity analysis will present the different analysis methods and will discuss their advantages and disadvantages. One aim is to determine future directions development of these techniques will need to pursue in order to be relevant for understanding the cognitive and affective dynamics of music listening. 


  • Roger Dean, MARCS Auditory Laboratories, University of Western Sydney and austraLYSIS
  • Hauke Egermann, CIRMMT, Schulich School of Music, McGill University
  • Theodoro Koulis, Department of Statistics, University of Manitoba
  • Daniel Levitin, CIRMMT, Psychology Department, McGill University
  • Clifford Madsen, Center for Music Research, Florida State University
  • James Ramsay, Departments of Psychology and Mathematics & Statistics, McGill University
  • Emery Schubert, Empirical Musicology Lab, School of English, Media and Performing Arts, University of New South Wales
  • Finn Upham, CIRMMT, Schulich School of Music, McGill University


Moderator: Stephen McAdams, CIRMMT, Schulich School of Music, McGill University

  • 8:30 - Coffee/tea
  • 9:00-9:05 - Introduction
  • 9:05-9:40 - Clifford MadsenThe history and development of the Continuous Response Digital Interface (CRDI)
In 1989 a device was developed (CRDI) which made it possible for responses to be registered during music listening. The first two models consisted of a dial that could be moved across a circle and a box with a lever that could be moved horizontally. The next development was a two-dimensional CRDI making it possible to register both arousal and valence via a computer screen. The latest CRDI device allows a person to change various aspects of an actual performance. Many published studies have used the CRDI to measure aesthetic response, emotion, tension, rubato, pitch/rhythm preferences, focus of attention, as well as other aspects deemed important to listeners and performers.
  • 9:40-10:15James Ramsay, Music as a buffered input/output system
In this workshop on functional responses to functional inputs, we have a look at a dynamic model as a relatively minor modification of functional regression, where at least one of the covariates is functional (for example, a musical score or an audio signal) and the output or response variable is also a function of time (such as a rating of tension or some other subjective attribute of music).
  • 10:15-10:50Theodoro Koulis & Daniel Levitin, Input-output systems in psychoacoustics
Continuous response digital interfaces are becoming popular in psychoacoustics for measuring in real time psychological responses to episodes such as a musical performance. Many psychoacoustic experiments of this nature may be viewed as collections of input-output systems. The statistical problem is linking time-varying covariates to the continuous response variate. Using online data obtained from an experiment in psychoacoustics, we showcase statistical tools that incorporate dynamical elements of the response. We outline the issues involved in analyzing input-output systems when the exact form of the underlying mathematical model is not known, and present a calibration method to facilitate inter-subject and intra-subject comparisons.
  • 10:50-11:10 - Coffee/tea break
  • 11:10-11:45Finn Upham, Variability and coordination in collections of continuous responses
Are continuous responses to the same stimulus measurably coordinated in time? Using activity analysis, it is possible to test whether the distribution of response events, such as changes in ratings, differs from that of a collection of random responses. Having applied these tests to dozens of collections, we can conclude that significant coordination cannot be presumed. Activity analysis exposes the disagreement between individual responses; however, a collection may still be coordinated despite never having instances of the majority of responses showing the same activity. When two collections of responses to the same stimulus are each coordinated, they also appear to be coordinated with each other; in other words, they show similar proportions of participants responding to the same moments in the music. While analyses using the average of continuous responses do not represent the "average" individual response, results suggest that, as long as collections are coordinated, these summary time series may still be effective descriptions of the collective listening experience. This talk will be preceded by a five minute introduction to CRAW,  the new Continuous Response Analysis Wiki.
  • 11:45-12:20Roger Dean, Relating perception and performance data to acoustic and structural features, and to agency, in electroacoustic and keyboard music
We are using Time Series Analysis techniques to demonstrate statistical relationships between acoustic variables, such as intensity and timbral parameters, and listener perception of structure and affect in electroacoustic music. In addition, sounds which derive from human agency have been investigated as potential initiators of affective response, particularly in Trevor Wishart's Red Bird. We follow up such correlative analyses with experimental intervention studies, in which the statistically powerful relationships (such as between acoustic intensity and perceived arousal affect) are tested as potentially causal relations by systematic manipulation of acoustic parameters. We have obtained clear evidence of the importance of acoustic intensity from this approach. I will also illustrate more briefly some collaborative studies on the temporal relations between information content flow and performance parameters in classical music, on error correction in anti-synchronic tapping, and on keyboard improvisation, in which we also study skin conductance responses and their psychophysiological correlates.
  • 12:20-1:30 - Lunch
  • 1:30-2:05 Emery Schubert, Initial orientation time in the music-emotion system. Some puzzles.
Initial orientation time (IOT) refers to the amount of time it takes to reach 'operating' position upon commencing a continuous response task.  Under the assumption that a response can be understood as a simple linear model that also exhibits serial correlation and lag, IOT can be understood as an additional non-linearity in the (in this case) music-emotion system.  This non-linearity appears to come about as a result of the listener adjusting to the emotion-tracking task.  Two recent, independent studies have suggested that this IOT can be quantified as occupying about the first 8 seconds of response from the commencement of the music being tracked.  After those 8 seconds, participant responses are more in line with a simple time-series models.  Some questions that arise from this approach are: Which variables affect the IOT estimate?  To what extent do the musical parameters themselves affect this estimate? What is the difference between IOT and (initial) reaction time, and how can the latter be measured in a continuous response scenario?  Answering these questions will help better understand underlying psychological processing of responses to music, and help improve estimates of simple time-series model parameters that are otherwise compromised by IOT.
  • 2:05-2:40Hauke Egermann, Analyzing continuous response data to test a theory on expectation and emotion in a live concert experiment
The talk will present an event-based analytic approach to continuous response data for testing a theory about musical expectations and listener’s induced emotions. Musical structures leading to expectation reactions were thought to be manifested in emotional reactions at different emotion component levels. We continuously measured subjective experience and peripheral psychophysiological changes of participants in a live flute concert using the CIRMMT Audience Response System. Additionally, musical expectancies were predicted by analyzing the musical stimuli used based on an information-theoretic model, and measured via a continuous expectation rating scale. An event-based analysis will show how to investigate relationships between peak information content musical events (high and low), the violation/confirmation of musical expectations (in corresponding ratings), and continuously measured emotional reactions.
  • 2:40-3:00 - Coffee/tea break
  • 3:00-5:00 - Structured discussion
  • 5:00-6:00 - Reception
Document Actions