Psychometrics

INTRODUCTION Physiological measurement is like all other forms of measurement and entails the replicable assignment of numbers to represent properties (Campbell, 1957). Accordingly, basic psychometric principles are as relevant to psychophysiological assessment as they are to the measurement of intelligence, assessment of job performance, or self-reports of emotion. In this chapter we will describe the core psychometric issues relevant to measurement in any area, but with an emphasis on their psychophysiological application. Guiding our discussion will be a very simple theme with wide-reaching implications: Science is essentially an error-correcting enterprise that relies on the ability to identify, estimate, and remove error from fallible observations. We will show how this theme underlies traditional psychometric principles and why psychophysiologists should be concerned about them. We will emphasize general principles rather than problems specific to particular modes of physiological measurement. The number of physiological systems now measured is so numerous and the technological advances are so rapid that a discussion of particular measurement problems would be highly selective and not particularly educational. Other sources (e.g., Thurston, Hernandez, Del Rio, & De La Torre, 2010; Bush, Alkon, Obradović, Stamperdahl, & Boyce, 2011; Boucsein et al., 2012; Burt & Obradović, 2013; Clayson & Larson, 2013), including chapters in this volume, amply document the measurement nuances and idiosyncrasies for particular problem areas. THE BASIC INFERENTIAL TASK All scientists attempt to make inferences about hypothetical constructs (e.g., stress, compliance, reactivity, attention, and memory) that cannot be observed directly. This basic inferential is illustrated in Figure 28.1 and requires operationalizing constructs so that numbers can be attached to them in a meaningful and replicable way. After operational definitions are created, the logic and power of research design and statistics can be used to establish the causal linkages. The operationalization stage requires careful attention to psychometric principles because the quality of the inferences about constructs and construct relations will depend on the quality of the measurements that will stand for or represent those constructs. Figure 28.2 provides a simple example. Suppose that a researcher posed the hypothesis that control loss is stressful, implying a causal relation between two constructs.