Three studies examined amnesic H.M.’s use of words, phrases, and propositions on the Test of Language Competence (TLC). In Study 1, H.M. used 19 lexical categories (e.g., common nouns, verbs) and one syntactic category (noun phrases) with the same relative frequency as memory-normal controls, he used no lexical or syntactic category with less-than-normal frequency, and he used proper names (e.g., Melanie
) and coordinative conjunctions (e.g., and
) with reliably greater-than-normal
frequency. In Study 2, H.M. overused proper names relative to controls when answering episodic memory questions about childhood experiences in speech and writing, replicating and extending Study 1 results for proper names. Based on detailed analyses of the use (and misuse) of coordinating conjunctions on the TLC, Study 3 developed a syntax-level “compensation hypothesis” for explaining why H.M. overused coordinating conjunctions relative to controls in Study 1. Present results suggested that (a) frontal mechanisms for retrieving word-, phrase-, and propositional-categories are intact in H.M., unlike in category-specific aphasia, (b) using his intact retrieval mechanisms, H.M. has developed a never-previously-observed proposition-level free association strategy to compensate for the hippocampal region damage that has impaired his mechanisms for encoding novel linguistic structures, and (c) H.M.’s overuse of proper names warrants further research.