Questions: Garden-Path Effects and Reanalysis During Parsing

Question 1

A reader encounters 'The horse raced past the barn fell.' After reading the word 'raced,' what has the parser done, and why does this create difficulty when 'fell' is reached?

• A. The parser has not yet assigned any structure because it processes whole sentences before building syntactic representations
• B. The parser has flagged 'raced' as ambiguous and is simultaneously maintaining two structural analyses, causing processing overload at 'fell'
• C. The parser has committed to treating 'raced' as the main verb using the minimal-attachment heuristic, leaving no open syntactic slot for 'fell'
• D. The parser recognized the reduced relative clause immediately and was surprised when 'fell' did not arrive sooner

Answer: (C)

The parser works incrementally and commits early. After 'the horse raced past the barn,' it has built a structure where 'raced' is the main verb and the sentence is effectively complete. When 'fell' arrives, there is no open slot for it, triggering a reanalysis. The correct parse — 'the horse [that was] raced past the barn' as a reduced relative clause, with 'fell' as the main verb — was available but was never initially considered because the main-verb analysis is simpler (minimal attachment). Option B would predict parallel processing, which the commitment-based model rejects.

Question 2

Two garden-path sentences are presented: (A) 'The actress the director praised left the stage' and (B) 'While the boy fed the dog the cat scratched its ear.' Based on the principle that global reanalysis (revising main-clause structure) is more costly than local reanalysis, which prediction follows?

• A. Sentence A is harder because 'the actress' followed immediately by 'the director' creates a noun-pile that takes extra time to resolve
• B. Both sentences impose equal processing costs because all garden paths require the same structural revision effort
• C. Sentence B is harder because the comma-less attachment creates a local ambiguity that global reanalysis cannot recover from
• D. Sentence A imposes a more costly reanalysis because the parser must revise the main clause structure to accommodate the center-embedded relative

Answer: (D)

Sentence A contains a center-embedded relative clause ('the actress [whom] the director praised'), which requires the parser to revise the main-clause assignment — 'the actress' is not the subject of 'the director,' it is the head of the relative clause. This is a global revision requiring the parser to undo and rebuild the core sentence structure. Sentence B involves a simpler local attachment ambiguity resolved by the second noun phrase. The cost of reanalysis tracks how much committed structure must be dismantled.

Question 3

Garden-path effects demonstrate that sentences like 'The horse raced past the barn fell' are grammatically ambiguous, since competent readers consistently misparse them.

• T. True
• F. False

Answer: False

Garden-path effects reveal processing strategies, not grammatical ambiguity. 'The horse raced past the barn fell' has exactly one correct grammatical parse (reduced relative clause as subject, 'fell' as main verb). The consistent misparse reveals that the parser uses a commitment-based minimal-attachment heuristic — choosing the simpler structure even when it turns out to be wrong — not that the sentence has two legitimate readings. Grammatical ambiguity (like 'I saw the man with the telescope') means multiple correct parses exist; garden-path sentences typically have only one correct parse.

Question 4

The minimal attachment principle predicts that when two structural analyses are available, the parser initially commits to the one requiring fewer syntactic nodes, even if the alternative is more semantically plausible in context.

• T. True
• F. False

Answer: True

Minimal attachment is a structural heuristic that operates before semantic plausibility is fully computed. Experiments show that even when the reduced-relative reading of a garden-path sentence is more plausible (e.g., 'the horse raced past the barn' — horses are frequently raced), readers still experience the garden-path effect, because the structural preference for fewest nodes overrides semantic expectations at the initial commit point. This is what makes garden paths a probe for parser architecture: they expose a structural bias that operates independently of meaning.

Question 5

Why is the garden-path phenomenon more informative about how the parsing system is built than simply observing that people sometimes misunderstand sentences?

Model answer: Ordinary misunderstanding could reflect distraction, noise, or low familiarity with vocabulary. Garden-path effects are systematic and predictable failures in fully competent speakers encountering grammatical sentences in controlled conditions. Their systematicity reveals the architecture of the parsing system: that it works incrementally (word by word, not whole-sentence), that it commits to one analysis rather than maintaining alternatives in parallel, and that it follows the minimal attachment heuristic. Because the failure mode is patterned and the recovery cost is measurable via reading-time spikes in eye-tracking or self-paced reading, garden paths serve as controlled experiments that expose parser design in ways that successful comprehension never could.

The key methodological point is that errors are often more diagnostic than successes. A parser that never failed would be a black box; garden-path sentences pry it open by producing failures at predictable points, with predictable difficulty profiles, that map onto specific theoretical commitments about how parsing is organized.