Continuous Response Measurement

Collecting Continuous Responses

Physiological, self-report, and other modes of data collection are usually referred to as “measurements.” Modern equipment used to record these responses continuously usually does so by digitizing the measuring signal (with the signal often consisting of a rating-scale value, or an electrical voltage from a transducer), meaning that the measurements are sampled repeatedly after small, regular time periods, a number of times per unit of time, and so are not strictly continuous. “Time-course” is a more accurate, encompassing expression for the technique in question. The number of samples is referred to as a sampling rate, reported in hertz (Hz), the number of samples per second. Self-reported ratings are generally thought to occur more slowly than physiological changes and so tend to be associated with slower sampling rates, around 2 Hz for emotional responses to music. However, with the availability of high-speed computers, higher sampling rates are normally recorded, with later “downsampling” by the researcher to find a balance between the fastest expected rate [Page 257]of response (which, for accurate tracking, requires a high sampling rate) and minimizing the data used for ease of analysis and interpretation (more downsampling). The resulting data file from a single variable (e.g., continuous rating of the amount of happiness a piece of music expresses over time) is referred to as a “time series.”

Analyzing Continuous Response Time Series

Analysis of time-series data can adopt one or several of a broad range of approaches from visual inspection of the time charts (the simplest) through to sophisticated time-series and functional data analysis techniques. One of the major problems peculiar to time-series data, which distinguishes them from traditional, “single point” (postperformance, retrospective) measures, is that time-series data are often associated with “serial correlation.” This means that a reading made at one point in time may, on the one hand, be causally related to something that happened in the stimulus (e.g., a loud sound in the music led to an increase in the “excitement scale” rating) per the single-point sampling method; but, on the other hand, the reading will be a function of previous ratings, too. That is, in an obvious sense, a self-report rating scale cannot be moved infinitely fast (independently) from one moment to another, and so the rating at one point in time will necessarily be related to at least the most recent point in time.

...