Dianne Gregory, Center for Music Research, Florida State University

The Continuous Response Digital Interface (CRDI) developed at the Center for Music Research in 1989 is used as a measurement device by researchers at 25 different sites. It has been used to collect real-time data in many diverse areas of music research, including music education, and most researchers tend to agree that its portability and adaptability are major contributions to measurement instrumentation options.

Many of the early studies, and several recent ones, either address the issue of the CRDI’s reliability as a major focus or include reliability data within the scope of the study. In addition, several studies compare CRDI data with that obtained from other measurement devices. Examples of this information provide documentation of the reliability of the CRDI as a measurement device in music research.

The earliest CRDI reliability measures were immediate test-retest applications. Capperella (1989) compared music and nonmusic majors’ focus of attention to five music elements while listening to brief Western art music excerpts. Mean reliability was reported as .90 with ranges from .73 to .98 for the different elements. Another early study compared overall means from undergraduates’ repeated continuous evaluations of choral performances (Robinson, 1990). The correlation coefficient was .85.

Additional test-retest data were reported in a study across eight different research sites (n = 1299) which compared preference responses of subjects listening to brief excerpts (Gregory, 1994). The correlation coefficients of continuous samples/subject were .64 (6th grade students), .82 (high school students) and .84 (music majors). Byrnes (1994) compared mean test-retest data of elementary students listening to longer excerpts and reported a correlation coefficient of .96. In a study of college students’ aesthetic responses to music, Lychner (1995) compared test-retest group means of continuous samples and reported a correlation coefficient of .95.

A study designed to measure the effect of repeated listening on the discrimination of melodic phrase similarity discrimination by elementary students, although not a test-retest measure per se, provides additional data pertinent to the reliability of the CRDI. DeNardo and Kantorski (1995) found no significant differences between three identical listening trials.

Two studies analyzing the behaviors of ensemble conductors during rehearsals report using the CRDI as an alternative to static traditional observation methods. In both studies, the agreement quotient or correlation coefficient primarily used to determine behavioral definition agreement provide additional CRDI reliability data. Blocher, Greenwood, and Shellahamer (in press) reported an agreement quotient comparing CRDI samples indicative of seven different behaviors as .80. Davis (1993) compared summative data across 15 behaviors and reported a correlation coefficient of .85. Both of these measures were obtained from data obtained from a sampling rate of 3/second for 20 minute observations, providing a very stringent time-based agreement measure, particularly for behaviors with brief durations.

Comparison of CRDI data with that obtained from other methods is another approach to determining reliability. For example, high school band and choral students recorded their "feelingful" responses along a "weak-strong" continuum while listening to a music selection before and after three rehearsals of the same music (Nierman and Creasy, 1994). After the posttest listening, 83% of the subjects reported responding differently to the music compared to the first listening before rehearsals. A significant increase was found between pretest and posttest CRDI means.

Another example of instrumentation comparison is a band adjudication study (Johnson, 1992). Three judges used the CRDI to continuously record their evaluation of live band performances. After the continuous evaluation each judge gave a singular performance rating on a 1 to 5 scale. The agreement quotient between the CRDI mean score and the rating was .92. A more detailed and impressive depiction of instrumentation consistency is documented in a study of perceived tension in music (Madsen and Fredrickson, 1993) in which real-time graphic displays of CRDI data were superimposed over graphs from another researcher’s study using a different continuous device. Responses from both devices were virtually the same across time.

Music preference studies comparing continuous CRDI responses with static Likert measures provide additional information. Mean CRDI responses while listening to brief excerpts and post-listening static ratings were similar for music majors but different for nonmusic majors (Brittin and Sheldon, 1995).

These studies represent a wide variety of perceptual tasks with varying response durations by subjects with varying levels of experience. In addition, correlation coefficients, agreement quotients, graphic comparisons, and statistical evaluations represent a variety of ways to approach the issue of reliability. In any event, it seems apparent that early indications of acceptable reliability have been verified in more recent studies. In fact, reliability measures across a wide variety of applications in many different areas of music research perhaps support a generalization of the CRDI’s reliability.

Note: The complete list of CRDI references can be obtained by sending a request to Dianne Gregory, Center for Music Research, Florida State University, Tallahassee, FL 32306-2098.