CALCAP® Psychometric Properties

Return to main menu

The CALCAP Reaction Time measures have very high internal consistency reliability (.77 - .96), indicating that the constructs measured are assessed in a uniform manner across the multiple trials of each reaction time task (Miller, 1995).

The choice reaction time measures show 6-month test-retest reliability (.43 - .68) that is comparable to that seen in conventional neuropsychological procedures (.47 - .77), though it is likely that, as with the simple reaction time measures, choice reaction time is somewhat more state dependent than conventional neuropsychological procedures. Internal consistency reliability for the choice reaction time measures is quite high (.81 - .96).In general, the simple reaction time measures have very low test-retest reliability (.20 - .29), but very high internal consistency reliability (.77 - .95), suggesting that the psychomotor skills measured by the simple reaction time tasks vary considerably depending on state variables such as mood, attention, fatigue, time of day, etc. This hypothesis is also supported by the modest intercorrelations observed between the first, second and third iterations of the simple reaction time task (.41 - .68) during the standard CALCAP test battery (Miller, 1995).

Multiple iterations of the same simple reaction time task, administered at four separate times during the standard CalCAP procedures, correlate from .41 to .68 with each other. Choice reaction time measures correlate from .31 to .60. Form Discrimination shows the lowest intercorrelations with the other choice reaction time measures. Intercorrelations between simple and choice reaction time are very small (from .11 to .29) (Miller, 1995).

Intercorrelations of reaction time measures with conventional neuropsychological procedures are small (.02 to .37). The conventional procedures that correlate most highly with reaction time are Symbol Digit Substitution (.19 to .37), Verbal Fluency (.13 to .25), and Trail-Making, Part B (.17 to .32). Surprisingly, the Grooved Pegboard, a relatively pure motor measure, had negligible correlations with the reaction time tasks (.07 to .18). A factor analysis of the measures showed independent clustering of the computerized and conventional neuropsychological measures. Simple reaction time measures and choice reaction time measures form distinct factors (Miller, Satz & Visscher, 1991; Miller, 1995).

The CalCAP has been shown repeatedly (see Bibliography) to discriminate cognitively impaired index cases from matched controls as well as or better than conventional neuropsychological tests. These findings have been established both cross-sectionally (Miller et al., 1989; Miller, Satz & Visscher, 1991; Worth et al., 1993) and longitudinally (Miller et al., 1989; Miller, 1992). These data demonstrate the sensitivity of reaction time measures for detecting changes in motor functioning, and support the use of reaction time procedures for assessment and monitoring of symptoms of dementia and other cognitive slowing.

For more information, contact Eric Miller at

Return to main menu