We investigated the psychometric properties of the random number
generation (RNG) task in four studies using a mixed sample of young adults
(n = 306), middle-aged adults (n = 40), and patients
diagnosed with schizophrenia (n = 26). Data in study 1 were best
accounted for by a three-factor solution representing inhibition of
stereotypical schemas (seriation), output inhibition (repetition), and
monitoring of previous output (cycling). Modest test-retest correlations
were found, with the seriation factor showing acceptable stability across
time (study 2). In study 3, RNG task performance was related to scores on
concurrent neurocognitive tasks to establish construct validity. RNG
scores correlated with healthy controls' performance on the Stroop
color-word test and patients diagnosed with schizophrenia with executive
dysfunctions. Patients diagnosed with schizophrenia performed poorer on
the seriation factor of the RNG than healthy control participants (study
4). Our results indicate that the RNG task has modest to acceptable
psychometric properties. It primarily taps executive subfunctions (i.e.,
inhibition, updating, and monitoring), which are affected by
psychopathological or neurological deficits. (JINS, 2007,
13, 626–634.)