The investigation examined relationships among scales for observing and rating teacher performance. Beginning teachers with varying levels of professional experience (2, 9, and 16 months) were rated by pairs of observers on two occasions. Intercorrelations across occasions fell between .5 and .8. Interrater agreement ranged between .5 and .9. Factor analyses revealed about 67 percent common variance among the scales. Two rotated factors characterized "direct instruction" and "classroom control" dimensions. The extent of unidimensional variance is discussed in relation to underlying "true" versus "attributional" (halo effects) sources of common variance. (Author)