This report explores background data collected as part of recent large-scale literacy surveys because these data may contribute to an understanding of the variables associated with demonstrated proficiency. The first section reviews various approaches to large-scale assessments of literacy. This review sets the context for exploring variation in literacy performance as revealed in three databases established using the profile approach. These data derive from the Department of Labor Workplace Literacy Assessment and state assessments conducted by Mississippi and Oregon. Analyses using these databases are presented that identify and evaluate demographic and background characteristics associated with literacy proficiency in diverse populations. The evidence provided indicates that measured literacy skills can be predicted relatively well for large heterogeneous populations by using a small set of background variables that are likely to be readily available from census-type information. Estimates from census-like variables are likely to be less useful with homogeneous populations or when subgroup performance is of interest. The final section discusses further considerations--cost-information trade-offs and the use of direct measures in indicator systems for less well-defined and measured entities such as the ability to participate fully in an increasingly technological society. Appendixes include a workplace literacy background and activity questionnaire, a scheme for the coding of background and attitude variables used in regression analyses, 2 data tables, and 1 figure. Contains 28 references. (YLB)