Lecture 4: Sentence Processing I

Question 1

structural ambiguity (2)

Answer 1

• When a sentence has two possible interpretations depending on the groupings of words.
• If sentences were simply processed linearly, each sentence would only be able to have one interpretation.

Question 2

global ambiguity (1)

Answer 2

• When the ambiguity never goes away; i.e. it’s not grammatically ambiguous, meaning you need context to disambiguate.
• e.g. Eleven men needed to feed python.
• e.g. Owners cannot be required to stop barking.

Question 3

local ambiguity (1)

Answer 3

• A temporary ambiguity at a point in the sentence, which goes away by the end of the sentence (there’s a disambiguation point).

Question 4

parsing (3)

Answer 4

• The process of assigning syntactic structure to the incoming words of a sentence during language comprehension. The structure-building mechanisms and procedures collectively are often referred to as “the parser.”
• There are structured relationships between words: representation and hierarchical.
• Language is phrasal; processing involves building those structures.

Question 5

incrementality (1)

Answer 5

• The property of synthesizing and building meaning “on the fly” as individual units of speech come in, rather than delaying processing until some amount of linguistic material has accumulated.

Question 6

garden path sentences (2)

Answer 6

• Sentences that are difficult to understand because they contain a temporary ambiguity.
• Tendency for hearers or readers to initially interpret the ambiguous structure incorrectly, and then experience confusion when that initial interpretation turns out to be grammatically incompatible with later material in the sentence.

Question 7

reduced relative clause (1)

Answer 7

• A grammatical structure in English involving a relative clause in which certain function words have been omitted (e.g. “that was”). This structure often leads to ambiguity.

Question 8

self-paced reading task (3)

Answer 8

• A behavioural task intended to measure processing difficulty at various points in a sentence.
• Subjects read through sentences on a computer screen, one word or phrase at a time, pressing a button to advance through the sentence.
• A program records the amount of time each subject spends reading each segment.

Question 9

garden path theory (4)

Answer 9

• A theory of parsing; an initial “first-pass” structure is built during comprehension using a restricted amount of grammatical information and guided by certain parsing principles or tendencies, such as the tendency to build the simplest structure possible.
• Evaluations of plausible meanings or consideration of the context only come into play at a later stage of parsing.
• Decisions aren’t affected by: lexical characteristics (e.g. frequency of structure), plausibility, linguistic and non-linguistic context.
• Suggests that parsing is a modular system: structures are initially built by a lean, fast parser on the basis of limited syntactic info, and without the benefit of knowledge about semantics or context, in order to resolve ambiguity.

Question 10

Fodor (1983) or modularity view (5)

Answer 10

• A module is domain-specific and only operates on specific inputs.
• The main module is language, and within that are other modules such as syntax, phonology, semantics, etc.
• Modules are informationally encapsulated, and info from other modules can’t “reach into” the module.
• Each module can only access the outputs of the other modules.
• The garden path model suggests that parsing is a modular system: structures are initially built by a lean, fast parser on the basis of limited syntactic info, and without the benefit of knowledge about semantics or context.

Question 11

Frazier et al. (1987) (6)

Answer 11

• Suggested that, rather than activating multiple meanings at once (as the case in word recognition), the parser computes only one structure and its associated meaning.
• The parser chooses the simplest syntactic structure (avoids syntactic complexity), guided by two main principles.
• minimal attachment: Principle that says when more than one structure is licensed and consistent with the input, build the structure with fewest nodes.
  
  • If the simplest structure turns out to be wrong, the structure is reanalyzed. This imposes demands on processing.
  • Relative clauses and sentential complements violate this principle.
• late closure: Principle that says we don’t postulate unnecessary structure; if possible, continue to work on the same phrase or clause for as long as possible.

Question 12

constraint-based approach (3)

Answer 12

• Main competitor to the garden path theory; claims that multiple interpretations of an ambiguous structure are simultaneously evaluated against a broad range of info sources (or constraints) that can affect the parser’s early decisions.
• Constraints can be drawn from different sources: thematic relations, frequency (syntactic frames, lexical categories), context, etc.
• i.e. Isn’t modular, and doesn’t rely on just syntax.

Question 13

thematic relations (4)

Answer 13

• Knowledge about verbs that captures info about the events they describe, including how many and what kinds of participants are involved in the events, and the roles the various participants play.
• e.g. The landmine buried in the sand exploded → unlikely that the landmine is doing the burying, so there’s semantic pressure to shift to the alternative reduced clause reading, in which the treasure is the entity being buried.
  
  • However, in sentences such as, The horse raced past the barn fell, there are two participants involved—the horse, and the implicit, unmentioned horse racer. The noun horse fits nicely with the main clause race, in which there’s only one participant racing, but unfortunately this interpretation is wrong.
• Problem for the garden path theory → people have no problem building a relative clause with the landmine sentence, which violates the minimal attachment principle.

Question 14

Trueswell et al. (1994) (5)

Answer 14

• Participants read sentences with animate/inanimate subjects in either reduced/unreduced structures, and reading times were measured using eye tracking.
• The defendant/evidence (that was) examined by the lawyer turned out to be unreliable.
• Found that the reading times in all four conditions were similar.
• Animate reduced (AR) condition (the defendant examined): people had a significantly longer reading time at the disambiguation point (by the lawyer).
• This is because we’re less likely to assign a reduced relative clause to the animate subject (who can technically perform the act of examining) → garden path effect.

Question 15

effects of frequency of syntactic frames on ambiguous sentences (3)

Answer 15

• Garden path model doesn’t predict differences between words; i.e. parser doesn’t access frequency information.
• However, the most frequent readings and structures for words and sentences are activated most strongly.
• Suggests that parser collects info about likelihood of different structures, and they’re attached to different verbs in the lexicon.

Question 16

effects of frequency of lexical category on ambiguous sentences (4)

Answer 16

• Lexical information: Does something most frequently appear as a noun or verb?
  
  • e.g. The government plans to raise taxes vs. The government plans were kept on hold. It depends on the interpretation of “plans”—frequency will help decide (in this case, it’s more frequent as a verb).
• Many syntactic ambiguities rooted in lexical ambiguity.
  
  • Suggests that the two are interconnected.

Question 17

Tanenhaus et al. (1995) (9)

Answer 17

• Used the visual world eye-tracking paradigm to measure proportions of looks to a redundant towel after hearing Put the apple on the towel in the box.
• Condition 1: One apple on a towel.
  
  • If you interpret on the towel as where the apple should be placed (i.e. the goal of movement), you should look at the other towel a lot.
  • If you interpet on the towel as just a descriptor for the apple (i.e. as a reduced relative), you shouldn’t look at the redundant towel as much.
  • In the unambiguous condition (Put the apple that’s on the towel in the box), the latter was found. In the ambiguous condition, people looked at the redundant towel a lot.
  • This interpretation obeys the garden path model, according to the minimal attachment principle.
• Condition 2: Two apples, one on a towel and one on a napkin.
  
  • According to the garden path theory, which is modular, this second apple shouldn’t make a difference—the parser doesn’t access visual context when computing the syntactic structure.
  • Participants: a) looked at the redundant towel much less in general, and b) looked at the redundant towel more in the unambiguous condition.
• This study provided evidence in favour of the constraint-based model, as visual information is affecting how people interpret the sentence.