Psycholinguistics part 2 Flashcards
Sentence processing
Incremental parsing
Refers to how we build the sentential structure and meaning as the
words in a sentence unfold
Temporary vs. global sentence ambiguity
Temporary ambiguity: When sentences are ambiguous (i.e. can be represented by multiple structures) up to a certain point
Ex. The horse rode past the barn fell
Global ambiguity: When sentences are entirely ambiguous
Ex. I saw the dog and the cat on the table
Reduced relative clauses
Relative clauses (clauses modifying a noun) without a pronoun/complementizer/other function words
Ex. The florist [who was] sent the flowers was pleased
Garden path sentences
Sentences which are exhibit temporary ambiguity up until a point of disambiguation
Processing difficulty is experienced at the disambiguating region yielding slower read times
Ex. In ‘The florist sent the flowers was pleased’, slower read times would likely be recorded at the word ‘was’
Self-paced reading
A task used to assess processing difficulty through reading times –participants must press a key to advance through a sentence word by word or phrase by phrase
Garden path theory vs. constraint-based models
Garden path theory: parsers initially build one structure using minimal attachment (the fewest nodes possible) and only later revise their analysis using other information (plausibility,
context, etc.) if necessary
Constraint-based models: different information sources (constraints) like semantic/contextual information or word frequency can simultaneously affect the parser’s early decisions, sometimes negating the garden path effect
Factors affecting processing difficulty
- Thematic relations, ex. inanimate objects are not expected to be the subject of the main verb
- Syntactic frames, ex. transitivity / argument structure (NP-bias verb/S-bias verbs)
- Frequency-based information: more common constructions are preferred – ex.late closure (attaching a newly encountered phrase to the most recent NP) is more likely in English
- Context, ex. a preamble priming the parser for the correct interpretation, a visual aid
Tanenhaus et al. (1995)
Visual world eye-tracking
Participants hear either an ambiguous (‘Put the apple on the towel in the box’) or unambiguous sentence (‘Put the apple on the towel in the box’) as they view 1 of 2 displays:
1. One-referent display: an apple on a towel, a towel, a box, and an unrelated object
2. Two-referent display: an apple on a towel, an apple on its own, a box, and an unrelated object
Eye movements are tracked over time.
Looks to the empty towel indicate a garden path effect.
Results: More looks to the empty towel in one-referent
ambiguous condition than two-referent ambiguous condition.
This supports the constraint-based model of ambiguity resolution as it indicates visual context is used during parsing.
Predictions in sentence processing
As we incrementally process what we hear, we also anticipate the upcoming input using different types of linguistic information including:
* Verb semantics
* Subject + verb semantics
* Case-marking information
Altman & Kamide (1999)
Visual world eye-tracking
Participants hear ‘The boy will [verb]…’ as they view an image of a boy with various objects including a cake, a toy train, and a ball.
Verbs with different semantic characteristics matching the objects are alternately used (i.e. eat, move, bounce).
Results: Participants look to the object that is appropriate to the verb even before an object is named (i.e. eat -> the cake, bounce -> the ball). This indicates that the semantic characteristics of verbs shape our predictions about the objects they will take.
Kamide et al. (2003)
Visual world eye-tracking
Participants hear either ‘The girl will [verb]…’ or ‘The man will [verb]…’ as they view an image of a little girl and a man along with various objects including a glass of beer, some candy, a merry-go-round, and a motorcycle. Verbs with different semantic characteristics matching the objects are alternately used (i.e. taste, ride).
Results: Participants look to the object that is appropriate to the verb AND the subject (i.e. ,man + taste -> beer, girl + taste -> candy, man + ride -> motorcycle, girl + ride -> merry-go-round). This indicates that the semantic characteristics of verbs AND subjects shape our predictions about the objects the verbs will take.
Surprisal as a measurement of processing difficulty
Expectations are met = lower surprisal = shorter reading times
Expectations are not met = higher surprisal = longer reading times
Memory effects in sentence processing, Gibson (1998)
Posits that we hold related parts of sentences in our memory as we read them –The greater the distance between the related parts, the more our working memory is taxed resulting in processing difficulty
Ex. The distance between relative clause gaps and fillers
Interference effects
Similar nouns are harder to hold in our memory than nouns of different types, i.e. in the distance between a filler noun and a gap, if there are more nouns of the same type as the filler, more processing difficulty is expected
Individual effects that can
affect sentence processing
- Working memory, ex. King & Just’s (1991) reading span test showing lower memory span individuals have more difficulty with object RCs
- Cognitive control, ex. Trueswell et al.’s (1999) visual world eye-tracking showing that lower cognitive control = lower abilities to suppress incorrect analyses
