This paper represents an attempt to study the graphemic-phonemic associations that are formed during the acquisition and subsequent retention of beginning reading responses and to evaluate the heuristic value of viewing the formation of these associations as a classical conditioning process. Two experiments--one on first graders and one on first and third graders--are contained in this paper. These replications and extensions of Samuels' (1967) tests in both laboratory and classroom of the distractability of pictures in the initial acquisition of reading responses to printed words resulted in both failure to replicate the original findings and in a major reinterpretation of the design. Application of the classical conditioning model to Samuels' test of attentional processes in beginning reading revealed that it does not provide an adequate test of attention, but rather, constitutes a test of either learning to read using intrinsically-produced, visually mediated responses or learning to read using extrinsically-produced, verbally mediated responses. (Author/RB)