Overview of Psychological Measures

This section conceptually introduces Class A and Class B measures, physiological measures, threshold measures and multi-modal nulling. Applications to the measurement of presence appear in subsequent sections.

A measurement, by definition, requires that one thing be compared with another. Spatial comparisons can be made with respect to a standard unit of distance, such as a meter stick. What are the psychological equivalents of the meter stick?

An important property of a good measure is that it be very direct: that is, that the measure require few transformations of its input. This is because each transformation introduces possible sources of noise. For sensory experiments, Brindley [15] introduced a useful terminology for discussing how direct a psychological measure is. The most direct he termed ``Class A'' measurements, which can be expressed solely in terms of a comparison of sensations. For instance, a Class A measure for ability to distinguish colors might involve asking an observer to indicate which two of three adjacent colored lights matched each other.

In many cases, Class A psychological measures are not possible. For example (to stick with sensory experiments, for which Brindley developed his terminology) one might be interested in measuring how observers perceive colors to be related to each other. It turns out that humans are comfortable naming the color space with combinations of four (but not three) colors. Across cultures, the preferred colors correspond to ``red, yellow, green and blue'' [60,52,11]. The color naming experiments which determined this preference (and which contributed to the now well-established opponent-process theory) are not based on a Class A measure for distances between colors. The experiments require participants to compare a sensory stimulus (color) to a mental model: the experiment can not be expressed solely in terms of sensations. Brindley termed ``Class B'' all measures which can not be expressed solely in terms of sensations.

Class B measures can be quite useful, as in the above example, but because Class B measures must pass through observers' mental models they are less direct and more error-prone. Class B measures risk errors of interpretation and estimation. Further, it is known that a great deal of filtering occurs as sensory data moves to more abstract levels of the nervous system. Consequently, Class A measures, which do not require access to mental models, may be more sensitive than Class B measures. Class A measures may be able to detect trends which would only become apparent with stronger stimuli using Class B measures. These issues are discussed more fully below in reference to presence questionnaires (see Section 2.3.3).

The terms ``objective'' and ``subjective'' are widely used in place of ``Class A'' and ``Class B''. This notation is poor, however, as ``objective'' and ``subjective'' have too many meanings. For instance, administering a questionnaire asking for presence ratings is a Class B measure, in that it requires a participant to evaluate a mental state. But one could argue that the results are ``objective'', from the point-of-view of the experimenter. If the experimenter did not influence the outcome of the experiment, and does not impose a bias on the analysis of the data, is not the outcome ``objective''?

Distinct from the Class A/Class B classification are the physiological measures, such as heart-rate or postural changes induced by sensory stimulus. See Section 2.3.4.

Also distinct from the Class A/Class B classification are threshold and multi-modal nulling measures. Whereas Class A measures require an observer to compare two sensations, threshold and multi-modal nulling measures do not. For threshold measures, one evaluates an observer's ability to barely detect some stimulus under different conditions. For multi-modal nulling, one evaluates the degree to which particular stimuli can perceptually overwhelm conflicting stimuli from a different sensory channel. See Section 2.3.5. A specific case of multi-modal nulling is visual-inertial nulling, as used in the presence measure research of Chapter 4.