Do language users predict word forms as readily as they predict semantic features? Previous studies are conflicting, possibly because they did not differentiate between two types of word form relationship: Head and rhyme relationships, sharing onset or offset features with predictable words. Here, we investigated prediction of form and meaning by means of a priming lexical decision task. People read constraining sentences that disconfirmed their expectations, and indicated, at sentence offset, whether a letter string was a word. Targets were predictable but not presented nouns, semantically related nouns, as well as head- and rhyme-related nouns. Unrelated control nouns were also presented. Results showed facilitation for predictable and semantically related words, with no difference between the two. While no effects emerged for rhymes, head-related words showed slowing, indicating suppression of lexical neighbors following prediction of word forms. Our findings align with word recognition models and prediction-by-production models of predictive processing.