Internal structure and external validity of 39 multiple-choice visual arts achievement test items were examined. These items were developed to assess grade 3 visual arts achievement for a statewide model of a fine arts curriculum. Item responses were evaluated in terms of: (1) fit to the one-parameter Rasch measurement model; (2) item-total correlations and alpha reliability; (3) total score comparisons between art- and non-art-educated groups in kindergarten, grade 3, grade 7, and high school (over 900 students in all); and (4) comparison of art- and non-art-educated groups on six components of visual learning. Most items generally fit a unidimensional measurement model, with good alpha reliability, although six items showed marginal or poor fit. Art-educated students scored higher in each grade, and when items were grouped into the six components of visual arts achievement, art- and non-art-educated students differed significantly as expected, except for knowledge of tools, where no significant difference was noted. This method of assessment appears reliable and valid for children in grade 3 and may be useful for older children as well. An appendix contains a long table of curriculum objectives and interval structure. (Contains 26 references.) (SLD)