The premise is advanced that a major weakness of the everyday generic instructional systems design model stems from a too modest traditional conception of the purpose and potential of formative evaluation. In the typical ISD (instructional systems design) model formative evaluation is shown not at all or as a single, product evaluation step. Yet in practice, formative process evaluation is also widely accepted as desirable and is more pervasive than the models would indicate because it is done informally. A broader theory of product-plus-process formative evaluation for instructional design is offered. The critical factors of this descriptive theory focus upon the desirability and viability of introducing formative evaluation techniques at the very beginning of the ISD process rather than at the end. A revised purpose of formative evaluation is theorized which envisions an ISD system that provides process feedback as well as product feedback, thus enabling the system to become truly self correcting. In the theoretical structure posited, the output of any observable or definable step in the ISD process is a deliverable. Furthermore, the case is stated that not only are all of these deliverables things that can be evaluated, but also that a system of evaluation which calls for all deliverables to be evaluated is parsimonious of effort in the long run and serves as a systemic quality control program. The term introduced here to mean evaluating all project deliverables as they are completed, is front-to-back formative evaluation. To illustrate the applicability of the proposed concept as a prescriptive theory, a graphic linear ISD model is presented which is an elaboration of the widely used Dick & Carey Model. The expanded model, called the Braden Formative Evaluation Model, clearly depicts the relationship between formation evaluation activities and instructional design and development procedures. Predictive evaluation is defined and distinguished from formative evaluation, with the suggestion that its role in instructional design needs to be fully elaborated. (26 references) (Author/BBM)