ERP Investigation of Prosodic and Semantic Focus

1
ERP Investigation of Prosodic and Semantic
Focus
Shawn Johnson Charles Clifton, Jr. Mara
Breen Andrea Eileen Martin Joanna Morris Florack
2
Birch Clifton, JML 1995, 2002

• Effects of pitch accent appropriateness on
  discourse comprehension
• Follow-up to Bock Mazalla 1983
• Evelyn kissed Jeremy. RHONDA kissed Jeremy too.
• ? Ronda kissed Jason. RHONDA kissed Jeremy too.
• Faster auditory sentence comprehension times when
  pitch accent fell on the NEW item.

3
Focus Projection Birch Clifton, 1995

• Question Isnt Kerry pretty smart?
• Answers
• A Yes, she TEACHES MATH
• B Yes, she teaches MATH
• C Yes, she TEACHES math
• A and B more acceptable than C
• Focus projects from argument math

4
Birch Clifton, 2002

• Focus does not project from adjuncts
• Question How did Ted get to Minnesota?
• Answers
• A He DROVE SPEEDILY
• B He drove SPEEDILY
• C He DROVE speedily
• A acceptable, B and C not
• 1995 B was acceptable when the final word was an
  argument rather than an adjunct

5
Experiment 1

• Materials 2-speaker dialogs
• Setting Rhonda kissed Jason. (active)
• Question Who else was kissed by Rhonda?
• always passive half theme question, half agent
  question
• Answer JEREMY was kissed by Rhonda, too.
• Always passive half appropriate pitch accent,
  half inappropriate pitch accent

6
Theme/Theme Rhonda kissed Jason. Who else was
kissed by Rhonda? JEREMY was kissed by Rhonda,
too. Agent/Theme Evelyn kissed Jeremy. Who
else was Jeremy kissed by? JEREMY was kissed by
Rhonda, too. Agent/Agent Evelyn kissed Jeremy.
Who else was Jeremy kissed by? Jeremy was kissed
by RHONDA, too. Theme/Agent Rhonda kissed Jason.
Who else was kissed by Rhonda? Jeremy was kissed
by RHONDA, too. Focused material is underlined,
pitch accented material is in BOLD CAPS,
inappropriate responses are in red, and
appropriate responses are in blue.
Appropriate (agent/agent)
Inappropriate (agent/theme)
7
Details

• EEGs were sampled at 500 Hz using a 32-channel
  Neuroscan system.
• Participants judged whether dialogs sounded
  acceptable.
• ERP's collected for the first and second noun
  phrase of the answer (200 ms before onset, 1200ms
  after onset)
• 2 x 2 x 2 design (Presence/absence of pitch
  accent X Appropriate vs. inappropriate accenting
  X Early vs. late noun phrase)
• Focus vs. Non-Focus in answer appeared as
  interaction between presence/absence of pitch
  accent X appropriateness of accenting

8
(No Transcript)
9
(No Transcript)
10

• Electrodes were combined into 2 groups.
• Parietal electrodes (P3, PZ, P4, CP3, CPZ, CP4,
  C3, CZ, and C4).
• Frontal electrodes (FC3, FCZ, FC4, F3, FZ and
  F4).
• Samples from these two electrode groups were
  averaged into 100 ms bins for statistical
  analysis.

11
(No Transcript)
12
(No Transcript)
13
Experiment 1 Conclusions

• Effects of Semantic Focus A phrase that
  presents queried (focused) information elicits a
  prolonged late positivity.
• Doesnt rely on focused phrase having a pitch
  accent
• Like Cutler Fodor (1979) phoneme monitoring
  Extremely similar to Hruska, Steinhauer, Alter,
  Strube (2001) finding
• Effects of Word Position This positivity was
  larger and appeared earlier when the focused word
  was late in the sentence than when it was early,
  especially for the posterior electrodes.

14
Conclusions about Pitch Accent

• Effects of Pitch Accent - An early negativity
  appeared in the Focus/Inappropriate condition
• Negativity was elicited by a missing pitch
  accent (see Hruska, Alter, Steinhauer Steube,
  2001, for a similar effect).
• Extra pitch accents did not trigger any ERP
  activity
• but linguists have noted that early pitch accents
  can be added quite freely in English.

15
Why is there a late positivity?

• The focus-elicited waveform could reflect some
  sort of gross integration process.
• Kaan, Harris, Gibson and Holcomb (2000) found a
  similar positive deflection under conditions of
  long distance syntactic integration
• Steinhauer, Alter and Friederici (1999) found
  similar positive deflections at intonational
  phrase boundaries (where integration effects
  might conceivably occur).

16
Why did pitch accent have so little
effect?

• Semantic focus effects dominated our data
• Is it because all our target sentences were
  passives? The listener did not need to hear a
  pitch accent to know when the focused phrase was
  going to occur
• If we make the location of focused information
  more unpredictable, will listeners rely more on
  prosodic information and exhibit ERP effects
  related to prosodic appropriateness?

17
Experiment 2 (preliminary)

• Similar to Experiment 1, except
  Target sentence was active or passive
• 16 conditions
  Active/Passive X Early/Late X
  Pitch Accent/No Pitch Accent X
  Appropriate/Inappropriate
• All questions were passives
• Only 10 subjects so far

Active Appropriate
18
Passive Sentence Data

• E.g. - Who else was kissed by Rhonda? Jeremy was
  kissed by Rhonda.
• Data quite similar to Experiment 1, despite
  presence of active sentences in Experiment 2
• Clear late positivity to semantically focused
  word
• Bigger, faster to second than to first word
• A suggestion of early negativity to missing pitch
  accent

19
Passive, focus vs nonfocus
20
Passive, appropriate vs inappropriate
21
Active Sentences

• E.g. Who else was kissed by Rhonda? Rhonda kissed
  Jeremy.
• Patient question, Jeremy the focus
• E.g., Who else was Jeremy kissed by? Rhonda
  kissed Jeremy.
• Agent question, Rhonda the focus

22
Active Sentence Data

• Some results similar to passives
• E.g., possible positivity to focused word,
  beginning 400 ms (parietal electrodes)
• No strong evidence for appropriateness or PA
  effects
• Anterior electrodes, large persistent positivity
  to last word in sentence
• One disconcertingly different result
• Strong early negativity to focused words
  100-200 ms after start of word

23
Active, focus vs nonfocus
24
Active, appropriate vs inappropriate
25
(No Transcript)
26
Early negativity to focused
words

• Perhaps related to nonparallel question-answer
  structure
• Who else was Jeremy kissed by? Rhonda kissed
  Jeremy.
• Who else was kissed by Rhonda? Rhonda kissed
  Jeremy.
• But shows up early as well as late
• And answer structure is not evident early

27
Conclusions

• Remarkably persistent late positivity
  (widespread, bilateral) to semantically focused
  words in answer to question
• Rather little ERP response to prosody
• Early negativity to missing pitch accent
• Responsiveness to prosody does not increase when
  prosody is made (somewhat) more informative (by
  mixing actives and passives)

28
Conclusions (cont.)

• Large positive shift to last word in sentence
• Not the same kind of suggested integration
  signaled by positivity to focus different scalp
  distribution
• Puzzled by early negativity to focused words in
  active sentences
• Would appreciate suggestions for how to interpret

29
REFERENCES Birch, S., Clifton, C., Jr. (1995).
Focus, accent, and argument structure. Language
and speech, 33, 365-391. Birch, S., Clifton,
C., Jr. (2002). Effects of varying focus and
accenting of adjuncts on the comprehension of
utterances. Journal of Memory and Language, 47,
571-588. Bock, K., Mazella, J. R. (1983).
Intonational marking of given and new
information Some consequences for
comprehension. Memory Cognition, 11,
64-76. Hruska, C., Alter, K., Steinhauer, K.,
Steube, A. (2001, June, 2001). Misleading
dialogues Human brain's reaction to prosodic
information. Paper presented at the Oralite et
Gestualite, Aix en Provence, France. Kaan, E.,
Harris, A., Gibson, E., Holcomb, P. (2000). The
P600 as an index of syntactic integration
difficulty. Language and Cognitive Processes, 15,
159-201. Selkirk, E. (1984). Phonology and
syntax The relation between sound and structure.
Cambridge, MA MIT Press. Steinhauer, K.,
Altern, K., Friederici, A. D. (1999). Brain
potentials indicate immediate use of prosodic
cues in natural speech. Nature Neuroscience, 2,
191-196.
30
(No Transcript)
31
(No Transcript)
32
(No Transcript)
33
(No Transcript)
34
(No Transcript)