Event Abstract

Tracking learning of change in American popular music using single-trial EEG decoding

  • 1 Columbia University, Department of Biomedical Engineering, United States
  • 2 United States Army Research Laboratory, United States

We present results for a detectable signal in electroencephalography (EEG) measurements in response to forced semitone changes of key. In particular, when the key-change occurs at a constant location in the same song, we find that subjects exhibit discrimination that is different from when the semitone key-change is not at a constant location. Upon further investigation, we found through the neural data that each subject habituates to the key-change at different times in the course of the experiment. The stimulus domain is composed of four key-change profiles and one control superimposed on the American popular song “Eye of the Tiger.”

With N = 6 subjects, two versions of the experiment were performed. Both versions entailed 7 runs of 5 trials, where each trial consisted of a 91-second excerpt of the song. However, one version contained completely unrelated key-change profiles across all trials of all runs (i.e., the random version), whereas the other did not (i.e., the non-random version). In particular, the non-random version contained key-changes at the same location in the song, although they could be of different directions (e.g., the tonal center going up or down by a semitone). With these two versions of the experiment, we were able to gauge whether a learning effect could be developed in the subjects for the non-random version, as opposed to the random one. We collected EEG data with a 64-channel passive-electrode BioSemi cap.

For the analysis of data from these experiments, we extended previous work from our lab on perceptual discrimination by using logistic regression analysis. In particular, we performed a leave-one out (LOO) classification of key-shift (key pitch-up, key pitch-down, and generic key-change) vs. control conditions, using the area under the receiver characteristic curve (ROC) as a measure of classification accuracy. For each subject, there were multiple statistically significant (99%) post-stimulus components of key-shift discrimination for the random version of the experiment. This result is insensitive both to the place in the song as well as the absolute pitcg of the tonal center for both the preceding and succeeding versions of the song. Furthermore, when averaging classification results across subjects, several poststimulus components between 200-600ms were seen, indicating a common response across subjects.

Finally, the lack of clear discrimination between key-change and control conditions in the nonrandom experiment indicates a learning effect in the subjects. Upon further investigation, it was found using a subset of epochs from the non-random version that an optimum and significant pre-stimulus discrimination window exists before the key-change. This implies two things: 1) that subjects are learning the pattern of forced key-change as the experiment progresses and 2) that the pre-stimulus discrimination becomes attenuated after that pattern is discovered (later trials) and before it is noticed (earlier trials).

Keywords: contextual perceptual decision-making, EEG decoding, expectation violation, music cognition

Conference: BC11 : Computational Neuroscience & Neurotechnology Bernstein Conference & Neurex Annual Meeting 2011, Freiburg, Germany, 4 Oct - 6 Oct, 2011.

Presentation Type: Abstract

Topic: neural encoding and decoding (please use "neural coding and decoding" as keyword)

Citation: Sherwin J, Conroy B, Gupta A and Sajda P (2011). Tracking learning of change in American popular music using single-trial EEG decoding. Front. Comput. Neurosci. Conference Abstract: BC11 : Computational Neuroscience & Neurotechnology Bernstein Conference & Neurex Annual Meeting 2011. doi: 10.3389/conf.fncom.2011.53.00003

Received: 25 Aug 2011; Published Online: 04 Oct 2011.

* Correspondence: Dr. Jason Sherwin, Columbia University, Department of Biomedical Engineering, New York, United States, niceboston@gmail.com

© 2007 - 2017 Frontiers Media S.A. All Rights Reserved