Background: Previous models of the effects of feedback account for lower-order learning outcomes but do not adequately describe experimental findings for higher-order learning. Aims: Based on a connectionist model of feedback effects, this investigation aims to show that feedback that allows only one learner response facilitates proposition-specific verbatim encoding, while feedback that requires the learner to try again on error facilitates relational fuzzy encoding. Sample and methods: Volunteer high school students were randomly assigned to one of 5 print-based lesson treatments that consisted of four science expository texts with adjunct inference-level questions covering science principles. The five treatments included delayed feedback, single-try immediate feedback, multiple-try immediate feedback, and two control treatments, questions without feedback, and text only. A post-test given 5 days after instruction was designed to measure both verbatim and fuzzy outcomes. Results: Multiple-try immediate feedback was best for paraphrased post-test questions (fuzzy) and worse for verbatim post-test questions. Conclusions: Fuzzy trace theory complements a connectionist model of feedback, and may provide a fruitful approach for describing the effects of feedback on different learning outcomes.