Portfolio assessment is an integral component of the current movement toward classroom-based performance assessments. A variety of descriptions have been presented in the literature, but little consensus exists regarding the goals or procedures for using portfolio assessment. An empirical basis for using portfolio assessment for a range of instructional decisions is virtually nonexistent. This article presents a conceptual model in which portfolio assessment is viewed as a process of collecting multiple forms of data to support inferences about student performance in a skill or content area that cannot be sampled directly by a single measure. In this respect, portfolio assessment is viewed as analogous to the process of construct validation. The suggested model of portfolio assessment serves as a framework for clarifying the goals of instruction and aligning these goals with classroom-based assessment procedures that support individual referenced decisions. The model also occasions an agenda for research and training.