A GENERALIZABILITY ANALYSIS OF SUBJECTIVE PERSONALITY ASSESSMENTS IN THE STUMPTAIL MACAQUE AND THE ZEBRA FINCH

Psychometric findings are reported from two studies concerning the con struct validity, temporal stability, and interrater reliability of the latent common factors underlying subjective assessments by human rate rs of personality traits in two nonhuman animal species: (a) the Stump tail macaque (Macaca arctoides), a cercopithecine monkey; and (b) the Zebra finch (Poephila guttata), an estrildid songbird. Because most th eories of animal personality have historically implied that certain pe rsonality constructs should be relatively universal across taxa, paral lel analyses of similar data are reported for two phylogenetically dis tant species of subject using the same psychometric methods. Each of t he samples was drawn from a socially-housed colony of the same species : that of macaques consisted of 5 mature adult fem ales and 8 of their adult offspring and that of finches consisted of 5 adult individuals. A modified version of the 1978 Stevenson-Hinde and Zunz (SHZ) list of personality items was applied to the macaques at various times during the eight years from 1980-1988 and to the finches during 1992. This s tudy also used the three SHZ scales - Confident, Excitable, and Sociab le - originally derived from principal components. Generalizability an alyses were used to assess the construct validity, temporal stability, and interrater reliability of the hypothesized factors. Both Stumptai l macaques and Zebra finches manifest measurable personality factors t hat are highly valid across multiple items, stable across multiple yea rs, and reliable across multiple raters. The same model fits both spec ies, as predicted by theory. The construct validity of the factors is slightly higher for the finches than for the macaques, although the in terrater reliability is somewhat lower. This study illustrates how gen eralizability analysis can be used to test prespecified confirmatory f actor models when the number of individual subjects is quite small.