Purpose: Alcohol expectancies (i.e., positive or negative beliefs about the consequences of using alcohol) are powerful predictors of adolescent alcohol use. The accurate measurement of alcohol expectancies has important implications for successful alcohol prevention and treatment efforts. One of the most widely used instruments for measuring expectancies in youth is the Alcohol Expectancies Questionnaire-Adolescent form (AEQ-A). Despite its broad use, the factor structure of the AEQ-A has not been firmly established. Also, few studies have tested whether the AEQ-A meets criteria for measurement invariance (i.e., extent to which it assesses similar constructs across groups). We contribute to the alcohol prevention and treatment research by (1) replicating previous evidence for a two-factor, positive and negative expectancy model and (2) testing the two-factor model for measurement invariance across gender and age. Methods: Using data from a NIAAA-funded study (AA0755409) of the impact of maternal alcohol involvement on child well-being, we conducted a confirmatory factor analysis (CFA) of the AEQ-A with a sample of 313 youth, ages 10-16, to determine whether the two-factor, positive and negative structure held for this sample. To examine measurement invariance, we evaluated a series of multiple-group CFA models in which we tested for the equivalence of the (a) overall factor structure, (b) factor loadings, (c) intercepts, and (d) residual errors across the background variables. Results: To replicate the previous analyses, 28 of the original 90 scale items were included in the CFA. The latent negative expectancy factor was indicated by four items; the latent positive expectancy factor was indicated by four subscales [social enhancement (5 items), personal power (7 items), cognitive/motor enhancement (5 items), and tension reduction (7items)]. Fit indices indicated that the model adequately fit the data [Chi square=42.6; df=18; Chi square/df ratio=2.4:1; CFI=.95; TLI=.91; RMSEA=.07 (p=.09); SRMR=.05]. All indicators loaded significantly on their respective factors. Each of the four dimensions of measurement invariance was confirmed by gender. However, only the basic factor structure was invariant between younger (ages 10-12) and older (ages 13-16) adolescent groups; factor loadings, intercepts, and errors varied. Implications: Our findings have important implications for alcohol prevention and treatment strategies that focus on altering beliefs among youth about the expected benefits of alcohol use. The results support evidence of a two-factor, positive and negative structure for the AEQ-A. Although the data indicate that the AEQ-A assesses equivalent alcohol expectancy constructs among males and females, the same conclusions can not be drawn for younger and older adolescents. This suggests that the AEQ-A may assess somewhat different constructs for youth of different ages. Further research is needed to confirm this finding and to determine the impact on the efforts of alcohol prevention and treatment specialists to reduce underage drinking.