Thirty-two kindergarten children, nine parents, and two teachers participated in a study of classroom learning environments. Participants included lower and middle income White and Hispanic families; the children were enrolled in three schools in the southwestern United States. Both interviews and pictures of classroom learning events were used to elicit beliefs about learning in school. A learning events taxonomy was developed and became the basis for two sets of pictorial stimuli--one used to elicit personal constructs and the other for preference rank order. Participants were asked to sort the drawings, describe the groups, discuss classroom learning descriptors, and rank favorite learning activities. Three analyses of the data were performed; the first two analyses used multidimensional scaling models: INDSCAL (Individual Differences Scaling model); and EUCLID; both are available in the Alternating Least Scales (ALSCAL). The third analysis used interrater reliability. INDSCAL revealed two dimensions, participant configuration (student participation) and learning activities structured by teachers (work) versus activities engaged in by children (play). The EUCLID model was used to analyze the descriptions of learning environments; a degenerate solution was obtained. Kendall's tau, which was used to analyze the preference rank ordering, indicated no predictable differences in pattern between parents and children. However, the two teachers had very different tau values from each other. Several tables and many figures are provided, including illustrations used in the research. A six-page reference list is offered. (GDC)