1 of 67

Language Processing

Language & Mind 11/19/19

2 of 67

Linguistics as part of Cognitive Science

  • What do native speakers know about their languages?
  • How is knowledge of language acquired (primarily by children)?
  • How is this knowledge used to produce and comprehend language?

3 of 67

Linguistics as part of Cognitive Science

  • What do native speakers know about their languages?
  • How is knowledge of language acquired (primarily by children)?
  • How is this knowledge used to produce and comprehend language?

4 of 67

Competence vs Performance

What we know about language

vs

How we use that knowledge to process language

5 of 67

Competence vs Performance

What we know about language

vs

How we use that knowledge to process language

...produce language.

...comprehend language.

6 of 67

Competence vs Performance

  • How does your knowledge of the lexicon allow you to comprehend or produce words?
  • How do you use your phrase structure rules to assign syntactic structure to sentences?
  • How are literal meanings and implicatures computed by discourse participants?

7 of 67

Lexical knowledge

What types of information does does the Mental Lexicon contain?

  • Form (pronunciation, spelling)
  • Syntactic category (part-of-speech)
  • Syntactic subcategorization
  • Lexical semantics

8 of 67

Lexical processing

  • What is the organization of mental knowledge of lexical items?
  • How is information in the mental lexicon accessed during listening/speaking?
  • What are some processing steps required to recognize or produce a spoken word?
  • Is lexical access an all-or-none process?

9 of 67

When is processing hard (or easy). Why?

10 of 67

When is processing slow (or fast). Why?

11 of 67

How Fast is Processing (on average)?

  • Average rate of speech production is ~125 wpm
    • Comprehension must be able to keep up!
  • In reading, gaze durations of 200 – 250 ms, interrupted by saccades of 25 – 60 ms
  • Word naming times average ~500 ms

12 of 67

Experimental Methods

  • Reading
  • Word Naming: Given a prompt, name the corresponding word
  • Lexical decision: Decide whether a sequence of letters is a word or nonword
  • Semantic categorization: Place a word in one of two categories (e.g., vegetable vs. tool)

13 of 67

What do we find?

Speed is influenced by...

  • Frequency (rain vs. puddle vs. precipitation)
  • Repetition
    • Stronger effect for low frequency words
    • Can persist over several hours or longer
  • Age-of-acquisition (AOA)
  • Word length
  • stimulus quality, etc.

14 of 67

What do we find?

Semantic (or associate) priming

Lexical decision to a word is faster when it is preceded by a word related in meaning.

15 of 67

+

16 of 67

PRIME

(150 ms)

17 of 67

18 of 67

target

Is this a word of English?

19 of 67

What do we find?

Semantic (or associate) priming

NURSE βž™ doctor

BREAD βž™ doctor

Faster

20 of 67

Automaticity in Processing

BLUE

Name the COLOR of the word

21 of 67

Automaticity in Processing

RED

22 of 67

Automaticity in Processing

GREEN

23 of 67

Automaticity in Processing

PURPLE

24 of 67

Automaticity in Processing: Stroop Effect

BLUE

PURPLE

Faster

We can’t β€œturn off” lexical processing!

Stroop 1935, Jaensch 1929

25 of 67

Processing Polysemy

β€œRumor had it that, for years, the government building had been plagued with problems.They were not surprised when they found several bugs…”

Swinney 1979

26 of 67

Processing Polysemy

ANT

Swinney 1979

27 of 67

Processing Polysemy

SPY

Swinney 1979

28 of 67

Processing Polysemy

SEW

Swinney 1979

29 of 67

Processing Polysemy

β€œRumor had it that, for years, the government building had been plagued with problems.They were not surprised when they found several spiders, roaches, and other bugs in the corner of the room.”

Swinney 1979

30 of 67

Processing Polysemy

β€œRumor had it that, for years, the government building had been plagued with problems.They were not surprised when they found several spiders, roaches, and other bugs in the corner of the room.”

Swinney 1979

31 of 67

Processing Polysemy

Bug

insect

listening device

32 of 67

Processing Polysemy

Bug

insect

listening device

ANT

SPY

SEW

Semantic Priming!

?

33 of 67

Processing Polysemy

SPY ANT SEW

Faster than

Faster than

Semantic priming effect for the meanings that are not part of the final interpretation!

34 of 67

Processing Polysemy

Access ALL meanings for a word (creating priming effects for all entries)

THEN select the meaning that fits contextual information (resulting in a single accessed meaning).

35 of 67

Competence vs Performance

Frequency, repetition, semantic priming, etc. affect the processing (use) of knowledge about words, not word knowledge.

But why does the mind/brain exhibit these effects?

Why are frequent or recently seen words recognized more quickly?

36 of 67

Production?

37 of 67

Tip-of-the-Tongue State

β€œA navigational instrument used in measuring angular distances, especially the altitude of the sun, moon, and stars at sea.”

(Brown & McNeill 1966, Bock & Levelt 1994, Aitchison 2003)

38 of 67

Tip-of-the-tongue State

Sextant

Participants often guess semantically or phonologically words

compass, protractor, dividers,

secant, sextet, sexton

(Brown & McNeill 1966, Bock & Levelt 1994, Aitchison 2003)

39 of 67

Tip-of-the-tongue State

Brown & McNeill 1966

40 of 67

Lexical processing

  • What is the organization of mental knowledge of lexical items?
  • How is information in the mental lexicon accessed during listening/speaking?
  • What are some processing steps required to recognize or produce a spoken word?
  • Is lexical access an all-or-none process?

41 of 67

Sentence processing

A Mental grammar allows listeners/speakers to comprehend/produce an unbounded set of sentences

42 of 67

Sentence Processing

How do listeners process sequences of words that are

structurally ambiguous?

  • initially consider all grammatical parses?
  • initially consider only one grammatical parse?
  • are parsing decisions subject to reanalysis?

How do listeners interpret displaced phrases, such as wh-phrases moved to the beginning of the sentence?

43 of 67

Task: assign an abstract syntactic structure to a string of words

How does the parser choose a structure?

44 of 67

Parsing: Hypothesis 1

The robot attacked the aliens from Mars

NP

V

NP

PP

S

VP

NP

The parser waits until all information is available before interpreting

45 of 67

Parsing: Hypothesis 2

S

VP V

NP

NP

The robot attacked the aliens…

The parser begins interpreting as soon as it can…

46 of 67

Parsing: Hypothesis 2

NP

NP

PP

V

NP

The robot attacked the aliens from Mars

S

VP

X

… and must reanalyze when it makes a mistake.

Incremental Parsing

47 of 67

Parsing: Global Ambiguities

V

NP

The robot attacked the aliens from Mars

NP

NP PP

β€œThe robot attacked the aliens from Mars.”

S

VP

The robot attacked the aliens from Mars

VP

S

NP

VP

V

NP PP

48 of 67

Garden-path sentences

VP VP

???

NP PP NP

(e.g., Tanenhaus et al 1995; Trueswell et al. 1999)

Put the frog on the napkin in the box

49 of 67

Garden-path sentences

VP VP

???

NP PP NP

(e.g., Tanenhaus et al 1995; Trueswell et al. 1999)

Put the frog on the napkin in the box

50 of 67

Garden-path sentences

VP VP

???

NP PP NP

(e.g., Tanenhaus et al 1995; Trueswell et al. 1999)

Put the frog on the napkin in the box

51 of 67

Garden-path sentences

VP VP

???

NP PP NP

(e.g., Tanenhaus et al 1995; Trueswell et al. 1999)

Put the frog on the napkin in the box

52 of 67

Garden-path sentences

VP VP

???

NP PP NP

(e.g., Tanenhaus et al 1995; Trueswell et al. 1999)

Put the frog on the napkin in the box

53 of 67

Garden-path sentences

VP VP

???

NP PP NP

(e.g., Tanenhaus et al 1995; Trueswell et al. 1999)

Put the frog on the napkin in the box

54 of 67

Visual world eye-tracking

55 of 67

(Tanenhaus et al 1995)

A B C D

A’ B’

56 of 67

Hypothesis 1:

Hypothesis 2:

The parser begins interpreting as soon as it can.

The parser waits until all information is available before interpreting

57 of 67

Long Distance Dependencies

Lee washed the dishes with the soap.

🠊 What did Lee wash ___ with the soap?

58 of 67

Long Distance Dependencies

What did Lee wash the dress with ?

59 of 67

Long Distance Dependencies

What did Lee wash the dress with ?

60 of 67

Long Distance Dependencies

What did Lee wash the dress with ?

61 of 67

Long Distance Dependencies

What did Lee wash the dress with ?

62 of 67

Long Distance Dependencies

What did Lee wash the dress with ?

63 of 67

Long Distance Dependencies

What did Lee wash the dress with ?

64 of 67

Long Distance Dependencies

What did Lee wash the dress with ?

65 of 67

Long Distance Dependencies

What did Lee wash

_____ ?

the dress with _____ ?

66 of 67

Long Distance Dependencies

When does the parser try to β€˜ο¬ll’ the gap?

– E.g. when does the parser interpret fronted what

  • Hyp. 1: Wait and see where there is a missing complement?
  • Hyp. 2: Actively postulate gap at first grammatically legal position.

67 of 67

Summary

  • Distinction between linguistic competence (knowledge) and performance (use)
  • Psycholinguistic Tasks:
    • Visual Word Recognition: tasks and effects
    • Spoken Word Production: TOT, speech errors
    • Sentence Parsing: structural ambiguity, incremental parsing