Complementizer‐Trace Effects

Abstract

Abstract Perlmutter observed that Englishwh‐movement obeys a puzzling constraint: an asymmetry between subject extraction and non‐subject extraction that interacts with the complementizer system. Whilewh‐extraction of a non‐subject from a finite embedded clause is compatible with the presence or absence of the wordthatintroducing the clause, extraction of the subject is possible only whenthatis omitted: a. ✓Who do you think that Sue met ____?/✓Who do you think Sue met ____? b. *Who do you think that ____ met Sue?/✓Who do you think ____ met Sue? The effect has come to be known as the “that‐trace effect,” a member of a family of possibly broader sets of phenomena, all called “complementizer‐trace effects.” A strong poverty of the stimulus argument makes it clear that these phenomena are rooted in innate properties of human language. The discovery of these effects in multiple diverse and unrelated languages reinforces this conclusion, especially when coupled with the availability of independently supported explanations for the absence of similar effects in other specific languages. On the other hand, quite distinct accounts have been offered for complementizer‐trace effects, and there is no consensus yet as to which of these approaches is most likely to prove correct.Linear accountshave been proposed that specifically bar extraction from positions right‐adjacent to elements of the complementizer system; these accounts are supported by claimed interactions with the presence/absence of prosodic boundaries between the two. Alternativestructure‐based accountsthat have proved influential include structural locality requirements on extraction sites that occupy particular syntactic positions – as well as proposals that take an opposite approach, banning movement that is too local. Thus, although the importance of complementizer‐trace effects for linguistic theory is clear, the deeper source of these effects remains a matter of controversy.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 57 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.