Presentation is loading. Please wait.

Presentation is loading. Please wait.

Grammatical Illusions and Selective Fallibility in Real- Time Language Comprehension Collin Phillips, Matthew W. Wagers an Ellen F. Lau April 15, 2015.

Similar presentations


Presentation on theme: "Grammatical Illusions and Selective Fallibility in Real- Time Language Comprehension Collin Phillips, Matthew W. Wagers an Ellen F. Lau April 15, 2015."— Presentation transcript:

1 Grammatical Illusions and Selective Fallibility in Real- Time Language Comprehension Collin Phillips, Matthew W. Wagers an Ellen F. Lau April 15, 2015

2  The parser’s implementation of constraints – “selective fallibility”  Grammatical illusions  A presentation of findings from previous researches to create a clearer picture of the parser’s implementation of constraints Main points that will be discussed:

3 Introduction  Historical view of the working memory: capacity as a main issue  Today’s approach: online findings about how structured information is encoded and accessed in memory

4 Two ways of searching structures:  Retrieval by search advantage: structurally constrained disadvantage: can be slow, sometimes the information desired may not be found at all  Content-addressable retrieval advantage: higher speed disadvantage: grammatically illicit constituents can be activated There is evidence that both mechanisms are active in language comprehension.

5 The grammatical constraints that will be discussed:  Island constraints on filler-gap dependencies  Structural constraints of backward anaphora  Forward anaphora: Locality constraints on reflexives

6 Island constraints on filler-gap dependencies:  Previous research shows that the parser avoids violation of island constraints. Example: “What does the teacher think that the children expect her to assign _____ for homework”  Positing a gap as soon as encountering a verb “My brother wanted to know who Ruth will bring us home to _____ at Christmas” “My brother wanted to know if Ruth will bring us home to Mom at Christmas” Reading time slowdowns

7 Structural constraints on backwards anaphora: Implementation of Principle C: blocks coreference between a pronoun and a referring expression that it c-commands. Example: *“He i perused the classified ads while John i was drinking his coffee.” Cataphoric pronoun

8 Forward anaphora: Locality constraints on reflexives:  Principle A: an anaphor must have a local antecedent. example: “John i hoped that Bill j wouldn’t blame himself * i/j for the accident”. Does the parser consider grammatically inappropriate antecedent upon encountering a reflexive? According to most results, it doesn’t.

9 Grammatical Illusions

10 1. Selective fallibility in agreement comprehension:  Verb agreement: the morphological features of the finite verb must agree with the corresponding features on the subject NP. It seems very frequent that this constraint in violated. Example: *“The key to the cabinets are on the table” Interestingly, often reading times are not affected.

11 Proximity concord: “The key to the cabinets are on the table” This is an example of what is referred to as “proximity concord”: a verb incorrectly agreeing with the most proximate noun, rather than with a more distant subject.

12 Violations of the agreement constraint, regardless of “proximity concord”: Agreement violations may also occur when a verb agrees with an NP which is not necessarily the most proximate NP: *“The runners who the driver see each morning always wave”

13 Agreement constraint – a morphological selectivity: It seems that violations of the agreement constraint appear to be caused by the “marked” plural nouns, and not by the “default” singular nouns: *“The runners who the driver see each morning always wave” “The boy whose shirts were torn has wept all night”

14 A retrospective check: When comprehenders encounter a verb which does not properly agree with the NP in the subject, retrospective check is triggered. This is contrary to cases when the verb agrees properly with the NP in the subject: “They key to the cabinets was on the table”.

15 2. Selective fallibility in case licensing Case in German:  *“…that the mother DAT sent the book” – marked dative case  *“that the mother NOM was sent the book” – unmarked nominative case

16 3. Forwards anaphora: antilocality constraints on pronouns: Principle B: a pronoun can have an antecedent as long as the antecedent is not local or does not c-command the pronoun. Examples: *“John i likes him i ” *“The man i that John j had dinner with suspects that Bill k is stalking him i/j/*k ”

17 When is local coreference acceptable? Context: picking out Bill from an old class photo: “Bill i is probably him i ”

18 Activation of local antecedents: Violations of principle B were found in some research, mostly in child literature.

19 Evidence against activation of local antecedents: A. Nicol and Swinney (1989) claim that violations of principle B do not occur among adults. Example from their research: “The boxer told the skier that the doctor for the team would blame him for the recent injury” Priming was found for “boxer” and “skier”, but not for “doctor”

20 B. Clifton, Kennison, and Albrecht (1997, experiment 2) found no activation of local antecedents upon encountering a direct object pronoun. “The supervisor paid him yesterday to finish typing the manuscript” They only did find activation of local antecedents when it was grammatical, as upon encountering a possessive pronoun: “The supervisor paid his assistant yesterday to finish typing the manuscript”

21 C. Lee and Williams (2008) found that reading time was slowed upon encountering a pronoun which didn’t match the gender of the grammatically acceptable subject (midwife): “The surgeon said that Richard introduced him during the opening reception of the medical conference” – reading time is not slowed down “The midwife said that Jane introduced her during the opening…” – reading time is slowed, due to mismatch of gender with the acceptable nonlocal subject.

22 Evidence to support activation of local antecedent: Badecker and Straub (2002) found the reading times were slowed upon encountering a pronoun which matched both the grammatical subject, and the local antecedent: “John thought that Bill owed him another chance to solve the problem” As opposed to: “John thought that Beth owed him another chance to solve the problem”

23 4. Negative polarity items (NPIs) NPIs are licensed in configurations where they are c- commanded by negative-like elements. Example: “No professor will ever say that”

24 Violations of the above which lead to “grammatical illusions”: *“The bills that the senators voted for will ever become law” *“The bills that no senators voted for will ever become law” Acceptance rates for the second sentence were higher than for the first, despite the fact that the NPI and the negative-like element do not c-command each other.

25 5. Comparative illusions “More people have been to Russia than I have”. What does it mean?

26 Comparable elements: “More Americans have been to Russia than Canadians have”

27 Quantities of individuals conveying assertions about quantities of event: “106 million cars crossed the George Washington Bridge in 2007”

28 Repeatable vs. nonrepeatable predicates:  “More undergrads call their families during the week than I do”  “More New Yorkers began law school this semester than I did”

29 Synthesis

30 What may cause selective fallibility? One option: hard-coded differences between the various grammatical constraint.

31 Why is this not a satisfying answer?  It assumes the success or failure of constraint implementations online is arbitrary.  It does not explain the fact that there is also selective fallibility even within the domain of individual constraints.

32 Suggestions to a more systematic account of selective fallibility

33 Directionality  prospective search processes may be more accurate than retrospective search processes.  Doesn’t seem to be strictly systematical.

34 Locality  Searching for local antecedents, such as in principle A – reflexives.  Involves one position, as opposed to two (as in principle B)

35 Higher order representations  Considerations of illicit NPs may be due to semantics or pragmatics, and cannot be considered as syntactically based.  Such as when fleetingly violating principle B: “John asked Danny to wash him”

36 The authors’ suggestion: Structural Priority “structural information constrains dependency formation when it has temporal priority over other information”. That is, when you have an early cue for what you’re looking for later in the sentence. Such as in filler-gap dependency (a wh- phrase) or forwards anaphora (cataphoric pronoun).

37 Conclusion  It is useful to bring together the finindings from different domains, as they allow us to build a clearer picture of how speakers mentally encode and navigate linguistic representations in real time.


Download ppt "Grammatical Illusions and Selective Fallibility in Real- Time Language Comprehension Collin Phillips, Matthew W. Wagers an Ellen F. Lau April 15, 2015."

Similar presentations


Ads by Google