Parallel grammatical encoding in sentence production: Evidence from syntactic blends

Using a large, newly available corpus of spontaneously uttered syntactic blends (e.g., cast into question from the targets call into question and cast into doubt) and a new method of speech error analysis, two hypotheses regarding grammatical encoding are compared: the single-buffer hypothesis, according to which alternative formulations of the message are encoded in the same memory buffer, potentially sharing representations, and the multiple-buffer hypothesis, according to which alternative formulations are independently grammatically encoded in separate buffers. Randomly generated, unattested blends were found to be reliably distinguishable from blends attested in the corpus, based on the degree to which they adhere to syntactic alignment constraints, controlling for other important factors. This main finding suggests that elements in similar syntactic positions across plans compete for the same slot, supporting the single-buffer hypothesis.