What to Make Of and How to Interpret Process Data

Maddox (2017) argues that respondents’ talk and gesture during an assessment inform researchers how a response product has evolved. Indeed, how a task is performed represents key information for psychological and educational assessment. In an ancient example: Gideon was required by the Lord to select those men who lap the water with their tongues, but not those who kneel down to drink (Judges 7:5 New International Version). In cognitive ability testing, process data can be defined as empirical information about the cognitive (as well as meta-cognitive, motivational, and affective) states and related behavior that mediate the effect of the measured construct(s) on the task product (i.e., item score). Thus, operationally, process data can be regarded as the empirical data reflecting the course of working on a test item. Recently, process data has increasingly gained attention in cognitive ability testing given the digitalization of measurement and the possibility of exploiting log file data. Other sources of process data are, for instance, concurrent think aloud protocols, screen capturing, eye tracking, facial expression, video-recorded behavior, and physiological sensor data (Azevedo, 2015). As shown by Maddox for large-scale assessments, even talk and gesture can be regarded as useful process data. In this case, the process data is not only video-recorded but also observed by the interviewer in situ; the interviewer interactively uses it to influence the test-taking process and to reduce construct-irrelevant variance. Thus, like product data (e.g., scores), process data is used to draw inferences. We argue in the following that the interpretation and use of process data and derived indicators require validation, just as product data do (Kane, 2013). This theoretical background, including some examples about log file data, sets the ground for our comments on Maddox’s use of “talk and gesture as process data.”