It is shown that lexical decision times to strong associates with an associative strength of approximately 40% are facilitated relative to targets following a neutral prime, “blank,” whereas very weak associates with an associative strength of less then 3% are neither facilitated nor inhibited. It is also shown that relative to the “blank” baseline time, a row of crosses inhibits processing of the following target. The latter finding has implications for earlier studies that have used crosses as a neutral prime. In these studies, facilitation effects have been overestimated and inhibition effects have been underestimated. Neely 1976 has proposed a predict-and-match strategy according to which subjects are assumed to predict one or more targets from the prime and to match the actual target onto the predicted targets. A part of this theory is not supported by the present data. The results are discussed in terms of the two-process theory of expectancy (Posner & Snyder, 1975). They are also considered in the light of a recent theory by Becker (1980). As an alternative interpretation of part of the reported data, a coherence assumption by the subjects about all reading materials is introduced.