Questions: Constituency Testing and Phrase Diagnostics

Question 1

A linguist tries to cleft 'old man from the' from 'The old man from the corner store laughed,' producing: 'It was old man from the that laughed the corner store.' The result is ungrammatical. What does this show?

• A. The entire sentence is grammatically ill-formed
• B. 'Old man from the' is not a grammatical constituent — it spans a phrase boundary
• C. Clefting is not a valid constituency test in English
• D. The test is inconclusive; only movement tests can determine constituency

Answer: (B)

The cleft test frames a candidate string in 'It is ___ that...' — genuine constituents yield grammatical clefts; non-constituents produce ungrammatical results. 'Old man from the' fails because it cuts across a phrase boundary: 'from' belongs to the PP 'from the corner store,' not to the NP. This is exactly the diagnostic signal. The source sentence is grammatical; the test targets the candidate substring, not the whole sentence.

Question 2

Which test most directly confirms that 'the woman in the red coat' is a noun phrase constituent?

• A. The string begins with a determiner, proving it is an NP
• B. It can be replaced by the pronoun 'she,' yielding a grammatical sentence
• C. It appears at the beginning of the sentence
• D. It contains multiple words, which all constituents must

Answer: (B)

Pronoun substitution is one of the core constituency diagnostics: if replacing the candidate with a pronoun yields a grammatical sentence, the candidate is a constituent. 'She laughed' is grammatical, supporting the constituency of 'the woman in the red coat.' Option A identifies word class but does not test phrase boundaries. Option C is irrelevant to constituency. Option D is wrong — single words can be constituents, and not all multi-word sequences are.

Question 3

A genuine syntactic constituent should pass multiple constituency tests, even if it fails one.

• T. True
• F. False

Answer: True

This is the principle of converging evidence. No single test is infallible — phonological weight, word order constraints, or language-specific factors can block individual tests even for genuine constituents. A true constituent typically passes most tests while a non-constituent fails most. When a string passes substitution, clefting, and movement tests, the evidence is strong. A single failure should prompt investigation of why the test was blocked, not immediate rejection of constituency.

Question 4

If a word sequence fails the cleft test, it is definitively not a syntactic constituent.

• T. True
• F. False

Answer: False

A single test failure is not definitive. Constituency tests can produce false negatives: a genuine constituent may fail a particular test due to phonological weight (heavy-NP shift is blocked for short phrases), register restrictions, or language-specific constraints (not all languages allow clefting equally). A true constituent that fails one test may pass others. Constituency diagnosis requires converging evidence across multiple tests, not a single result.

Question 5

Why do syntacticians use multiple converging tests rather than a single definitive test for constituency?

Model answer: No single constituency test is both necessary and sufficient for all cases. Individual tests can produce false positives (accidental grammaticality) or false negatives (blocked by phonological weight, register, or language-specific constraints). Using multiple independent tests — substitution, clefting, movement, coordination — exploits the fact that genuine constituents pass most of them while non-constituents fail most. When tests converge, confidence is high; when they diverge, the divergence itself is informative, pointing to factors blocking a specific test rather than undermining constituency.

This reflects the broader empirical methodology of linguistics: syntactic structure is not directly observable, so linguists triangulate from multiple behavioral diagnostics. The approach mirrors scientific practice — a hypothesis supported by multiple independent lines of evidence is stronger than one supported by a single observation. Using converging evidence also guards against building theory around idiosyncratic properties of a single construction.