This longitudinal study examined the extent to which dimensions of infant negative temperament in the first year predicted IQ at age 3, and whether these associations depended on the quality of the infant-mother attachment relationship. In a sample of 63 infant-mother dyads, mothers completed Rothbart's (1981) IBQ when infants were 4 and 12 months, mothers and infants participated in Ainsworth and Wittig's (1969) Strange Situation at 12 months, and children completed the Stanford-Binet (Thorndike, Hagen, & Sattler, 1986) when they were 36 months of age. Hierarchical multiple regression analyses indicated that 4- or 12-month distress to limitations was not predictive of later IQ, but infants with greater distress to novelty at 4 months had higher IQs at 36 months. Furthermore, greater distress to novelty at 12 months predicted higher IQs but only for infants whose attachment was insecure. Differential implications of temperamental fear versus anger for social influences on cognitive development are discussed. (Contains 5 tables.)