The present experiment investigated memory attribute dominance in young children by measuring false recognition responses to associatively and acoustically related words. Second- and sixth-grade children, half of whom were high SES and half low SES, served as subjects. Following Underwood (1969), a shift from acoustic to associative memory attribute dominance was predicted for both the age and social class variables. Contrary to expectations, more associative false recognitions were found at both grade and social class levels. Negative correlations between the associative strength and false positives to associatively related words were also found. Implications for the developmental shift hypothesis are discussed. (Author)