Does grammar steer who 'him' refers to from the very start?
When reading a pronoun like 'him', people consider only the names that grammar allows as its referent from the earliest moment, rather than first grabbing any name of the right gender and filtering later.
Source
Immediate sensitivity to structural constraints in pronoun resolution
Study at a glance
- Design
- Human experiment — Five within-participant reading experiments (four word-by-word self-paced reading, one eye-tracking) crossing the gender match of grammatically allowed and disallowed antecedents with a pronoun; Experiment 2 is a direct replication attempt
- N
- Separate samples of University of Maryland students per experiment: 36 (Exp 1), 26 (Exp 2), 26 (Exp 3), 38 (Exp 4), 24 (Exp 5, eye-tracking); plus 20 online raters who normed noun gender bias
- Population
- Native English-speaking university students, mostly aged 18-28
- Outcome
- Reading times at and just after the pronoun (self-paced reading) and first-pass, regression-path and total reading times (eye-tracking)
Structured fields used in claim comparison tables when every cited study has a complete layer.
What they did
Across five experiments, university students read sentences such as 'John thought that Bill liked him', where grammar (Binding Principle B) allows only the main-clause name to be the referent of 'him'. The researchers varied whether the allowed name and the forbidden, nearer name matched the pronoun's gender, and measured reading slowdowns just after the pronoun. Experiment 1 also varied how similar and how distant the names were, Experiment 2 exactly reproduced an earlier study's materials and procedure, and Experiments 3-5 (including one with eye-tracking) looked for costs when two names matched.
What they found
Readers slowed down right after the pronoun whenever the grammatically allowed name mismatched in gender, a robust 'grammaticality effect' in Experiments 1 and 2. Critically, a gender-matching but grammatically forbidden name never made this slowdown smaller, which is what an agreement-first account predicts. The previously reported cost of having two matching names was never replicated in any of the five experiments, even though each had more data than the original, and a late slowdown from a forbidden matching name appeared only in a couple of conditions, after the early grammar effect.
The limits
What it doesn't show
The conclusions rest partly on null results (no facilitative interference), so a very small interference effect could still exist; the authors argue against this by comparison with verb agreement but did not formally test power. Only the masculine pronoun 'him' and English sentences with names or gender-stereotyped nouns were tested, so the findings may not extend to other pronouns, languages or discourse-level reference. Reading times are indirect measures of memory retrieval, and some reported effects were significant only by participants or only by items.
Key terms
- Binding Principle B
- A grammatical rule stating that a pronoun such as 'him' cannot refer to a noun phrase within its own local clause.
- Antecedent
- The earlier word or phrase that a pronoun refers back to.
- Cue-based retrieval
- A memory model in which features of the current word (such as gender or number) are used as cues to pull matching items back from memory.
- Facilitative interference
- Faster processing caused by a feature-matching but grammatically illegal candidate, taken as evidence that grammar was not used in the initial retrieval.
- Self-paced reading
- A task in which readers press a key to reveal each word in turn, so the time spent on each word indexes processing difficulty.
- Multiple match effect
- A slowdown reported when two candidate antecedents both match the pronoun, interpreted as competition between them.
Flashcards
0 of 11 answers reviewed
Research intelligence for this paper
See its role on concept claims, tensions it is part of, placement history, and related discoveries.
Quiz yourself
What was the core finding across the five experiments?
Common questions
Why does a slowdown after the pronoun matter?
If the only grammatically allowed name has the wrong gender, the reader cannot find a valid antecedent, which costs time. Seeing this slowdown shows readers are checking grammar and gender together.
What result would have supported the Agreement First view?
Faster reading when the forbidden nearby name matched the pronoun's gender in sentences where the allowed name did not, because an agreement-only search would have found a candidate. This never happened.
Why include a direct replication of an earlier study?
A previous study had reported a cost when two names matched, which conflicted with Experiment 1. Using that study's exact materials and procedure in Experiment 2 tested whether the difference came from methods; the effect still did not appear.
More on Language