435 THE DEVELOPMENT OF TEMPERAMENT IN INFANCY: CONTINUITY AND CHANGE Carmen Gonzales Jose A. Carranm, and Manuel Ato Depto. Psicologia Evolutiva y de la Education. Facultad de Humanidades y CC. Education. Unviversidad de Ahneria. Crta Sacramento, s/n. Cabada de San Urbano. 04120. Almeria (Spain) Temperament has been conceptualized as individual differences appearing early in life and constituting the basis of personality. Many investigators have therefore centered their research on temperament in infancy, assummg they will be able to identify the emergence and establishment of most of the temperamental traits. Nevertheless, normative developmental trajectories have not been well established for most of the temperamental dimensions. Longitudinal designs are especialy suited to demonstrate this stability and change throughout time, but studies often are limited in: a) a theoretical model to guide the research questions, b) measures -in particular, the use of appropriate instruments for measurement on multiple occasions, and c) designs -for example, the use only of two measurement points in the time-. Besides these deficiencies, limitations of conventional multivariate repeated measured models have been pointed out (e.g., Ware, 1985), and Hierarchical Linear Models have been proposed as a modem and powerful set of techniques to study the individual change. In our work, 60- healthy infants (29 girls, 3 1 boys) and their mothers were followed up in a longitudinal design from 3 to 24 months old, taking multiple frequent measures (measurement points: 3,6, 9, 12, 15, 18 and 24 months) using mother’s report. From 3 to 9 moths, mothers filled out the Z@rnt Behavior Questionnaire (‘Be), and f?om 12 to 24 months, the Toddler Behavior Questionnaire (TBAQ,J. On all occasions, except 15 months, mothers and children visited the laboratory, located in a kindergarden. While the children were in the observation room with the experimenter, the mothers filled out the questionnaire in an adjacent room, and staffmembers were available to resolve any questions about the questionnaire. At 15 months, the questionnaire was posted and returned by mail. Laboratory data have not yet been analyzed; our report is based on the parent reports. To trace the developmental course of each temperamental dimension, a data analytic method addressing both stability and change over time was used, the multilevel method applied to longitudinal designs (Goldstein, 1986; Brik t Raudenbush, 1987; Prosser, Rashbash & Goldsmith, 1992). Unconditional two-level models were used to trace the developmental trajectory for Activity Level, Anger, Fear, Duration of orienting and Pleasure. Formally, multiple observations on each individual are seen as nested within the person. For Fixed Parameters, Curvilinear tendencies were found for four of the five dimensions, with significant cubic parameters for Activity Level (t-a.594; pc.001) and Pleasure (t=2.137; pc.05); significant quadratic components for Fear (t=-4.018; pc.001) and Duration of orienting (t=2.137; pc.05). For Anger, a Linear Function appeared more accurate (t=4.422; p<.OOl). For Random Parameters, individual variation around the average curve was not significant for Activity Level, Anger, Duration of Orienting and Pleasure. Our results suggest that temperamental dimensions show different rates of development during infancy, with drastic acceleration of some scores during the fmt year of life. Gur data are in agreement with Rothbart’s (1988) proposal that behavioral tendencies have different developmental rates, which can be explained both by the maturation in neurological systems, and by ambient influences. Finally, for Activity Level, Anger, Duration of orienting and Pleasure, but not for Fear, infants in our sample appeared similar in their pattern of change. Thus, consistency in temperament development throughout change can be found during infancy.