Understanding sentences

Chomsky: study of syntax (arrangement of words in a sentence) is to describe the set of rules or grammer that enables us to produce and understand language
1. Language involves a finite set of rules even if we aren't aware of them
  1. Underlying structure changes in a sentence changes meaning this understanding is unconscious
2. Grammar that is not taught is implicit knowledge that allows you to structure a sentence and interpret it
  1. Put words together to understand, abstract rules or knowledge
  2. Constituent can be replaced by another word ie cat can replaced by it but 'annoyed the' cannot
The goal of sentence interpretation is to assign thematic roles to words in the sentence being processed-who is doing what to whom
1. To do this we must compute the syntactic structure of the sentence known as parsing. First step is to determine syntactic category to which each word belongs. Word order is important for representation
  1. To parse is to find a grammatical analysis, to work out sentence structure
2. Typical thematic roles: Agents - willfully doing something, Patient-thing being done to, Instrument - thing used, Goal- is where thing ends up
Phrase structure shows grammatical relationships between words and sentences, thematic roles are assigned according to grammatical relations
1. Syntactic ambiguity (not obvious who did what to whom in sentence): two different structures two possible meanings
Ambiguity is very pervasive in natural language "time flies like an arrow" usually pick most common interpretation. It is used asn a tool to investigate procedures people use to interpret a sentence
Garden path sentences can tell us about preferences for resolving ambiguity. Ambiguous sentence you get wrong analysis sometimes, you have to go back and re-track sentences. Eventually you end up with sensible interpretation, conscious of finding this (usually unconscious decision). Can tell us what bias and procedures we use to decipher structure
1. Assumed that sentence processing is incremental, word by word grammatical update. Therefore with ambiguous sentences you start working out structure and change it based on proceeding words, that's why we get GPs.
Theories of ambiguity resolution/parsing
1. The garden path model (Frazier, 1979) is an autonomous (modular) 2 stage model. Serial parsing, one analysis at a time
  1. First stage: use only internal info, grammatical/syntactic only. If ambiguous sentence only one structure choosen, dictated by two principles of minimal attachment and late closure
    1. Minimal Attachment: Choose analysis that uses the simplest syntactic structure, constructed using fewest nodes (ie 'the women' is a node) possible ( Frazier & Rayner, 1982)
    2. Late closure: incoming/new material should be incorperated into the phrase being processed
    3. If there is conflict MA takes precedence
  2. Second stage: uses non-grammatical info to evaluate choice, thematic info about semantic roles can only be used (Rayner, Carlson & Frazier 1983)
  3. One sentence is analysed according to MA/LC if sentence continues way which is inconsistent with the choice, there should be some processing difficulty because new analysis has to be found (reanalysis). Tests for this.
    1. Rayner & Frazier (1987) eye-tracking in ambiguous sentences the noun pharses should be attached as object, so there should be processing difficulty at the disambiguating region. In unamgiguous sentences noun phrases attached as subject so less difficulty at disambiguating region.
      1. Fixation times were longer for ambiguous sentences in the disambiguating region, shows that people initially preferred the direct object analysis. Support for MA preference
  4. Create syntactic representation then semantic factors kick in, our analysis was wrong look back
2. Constraint based model (Boland, Tanenhaus & Gainsey 1990) info from one domain may be used to resolve conflicts in another (diff from GP only syntactic info for syntactic analysis)
  1. Initial choice in syntactic ambiguity can be solved using any multple info (syntactic, semantic, discourse) called constraints, if non syntactic info found to influence this then evidence against GP
C-B theory claim v important role for context, it can be main driving force for resolving syntactic ambiguity.
1. 'Towel' as destination would be GP preferred, and as modifier would be C-B preferred as it matches the context better
  1. Tanenhaus rt al (1995) Visual context (seems more 'immediate' than linguistic) on ambiguity resolution, manipulated visual contexts (one or two) . Eye-tracking while listening to sentence, early moment as theories are about initial ambiguity resolution
    1. Can context difference affect syntactic analysis during interpretation of sentence? GP model would predict ambiguous sentences causes you to look at wrong destination but unambiguous won't regardless of context. C-B model same in one apple context but in two apple context the difference bet am and unam should disappear
      1. People show no signs of looking at wrong destination when context has two referents, visual context quickly used to resolve ambiguity. Support for C-B model.
Macdonal et al (1994) syntactic ambiguity could be seen as a form of lexical ambiguity resolution, competition when two meanings simultaneously available and balanced causes difficulty. Similar process. In C-B model syntactic ambigutiy resolved by competition. of constraints when alternative constraints are similar acitvation = difficulty
1. Van Gompel et al (2001) tested this, a sentence with a balanced preference for meanings was read more quickly than 2 other sentences only plausible for one meaning. Evidence against competition
  1. From this Van claimed Syntactic ambiguity resolution acts like a race, one computed first is adopted. If equal preference not going to slow down race , no processing difficulty. If winning meaning is imcompatable with continuing sentence the proccessor has to reanalyse, 50% of trials difficulty. However if sentence is compatible with either meaning theres no need for reanalysis
    1. No competition, no equally possible choices slow you down initially, just wrong choice that slows you down later. Different from lexical ambiguity.

Next up

Understanding sentences

Description

Resource summary

Similar

	Created by issy_hinds almost 11 years ago