Document

advertisement

The Time Course of Processing Emotional Prosody: Behavioral and Electrophysiological Investigations

Lauren Cornew,

1

Leslie J. Carver,

1

and Tracy Love

1,2

1

University of California, San Diego,

2

San Diego State University

Background

• Emotion interacts with cognition at many levels of processing, from basic perceptual 1 and attentional 2 stages to higher cognitive functions such as decision-making 3 and categorization.

4

• Findings from studies of auditory emotion recognition are mixed and include reports of enhanced processing of negative 5 , positive, 6 and neutral 7 prosody.

Questions

• Is there enhanced processing of negative prosody?

• Or, is there an advantage for emotional (positive or negative) compared to non-emotional prosody?

• Two experiments , Behavioral (Exp. 1a & 1b) and

Electrophysiological (Exp. 2), aim to explore these questions.

Experiment 1a Methods

• 43 UCSD undergraduates (monolingual native English speakers, 27 female, mean age = 21)

•Stimuli: 48 Jabberwocky sentences (mean length = 2.7s), recorded by a

Native English female speaker at a regular rate of speech (approx 5 syll/sec), were spoken in happy, angry, and neutral prosody

• Gating paradigm: 8 Sentences edited into successive clips, with duration increasing in 250ms increments; 5s of silence in between (Figure 1 below)

“The hessups ate pea chup after the sholt.”

250ms

500ms

750ms ...

Entire Sentence

Figure 1. Schematic of a spliced sentence

• Mixed factorial design: all participants received all emotions, but no one participant heard any sentence more than once

• Participants randomly assigned to 1 of 3 groups:

Group 1

Sentences 1-16 Sentences 17-32 Sentences 33-48

HAPPY ANGRY

Group 2

Group 3

ANGRY

NEUTRAL

NEUTRAL

HAPPY

• Every clip judged to be happy, angry, or neutral

NEUTRAL

HAPPY

ANGRY

• Variables of interest:

• Percent correct

• Isolation point (length of the clip at which participants chose the correct emotion and did not subsequently change their decision)

Experiment 1b Methods

• 24 UCSD undergrads (monolingual native English speakers, mean age = 22)

• Stimuli: same as in Experiment 1a

• Similar design as above: participants randomly assigned to 1 of 4 groups:

Group 1

Group 2

Group 3

Group 4

Sentences 1-24

ANGRY

NEUTRAL

HAPPY

NEUTRAL

Sentences 25-48

NEUTRAL

ANGRY

NEUTRAL

HAPPY

Exp. 1a Results

• Greater accuracy for neutral prosody,

F (2,

34) = 7.87, p = .001:

*

100

**

90

80

70

60

50

Happy Angry

Emotion

Figure 2

Neutral

= .001

Neut ral > Angry, p

= .001;

< .05

• Faster correct identification of neutral prosody, F (2, 34) = 24.67, p = .000:

1000 *

***

802.28

**

722.88

750

500 443.71

250

Happy Angry

Emotion

Figure 3

Neutral

* Neutral < Happy,

** Neutral < Angry, p

p

< .001

= .001

*** Angry < Happy, p < .05

Exp. 1b Results

• Greater accuracy for neutral prosody than happy prosody, t (11) = 3.52, p = .003, or angry prosody, t (10) = 4.48, p = .001.

• Faster correct identification of neutral prosody than happy t (11) = 2.23, p = .024, or angry prosody, t (10) = 3.06, p = .006.

Experiment 1 Discussion

• Participants in Experiments 1a and 1b demonstrated a processing advantage for neutral prosody, which was identified more accurately and more rapidly than happy or angry prosody.

• This ‘neutral bias’ was not simply due to a greater ease in discriminating emotional from non-emotional prosody.

• Does the “neutral bias” observed in

Experiments 1a and 1b reflect perception, attention, decision/response, or a language processing or acoustic parameter?

Experiment 2 Methods

• 13 UCSD undergrads (right-handed monolingual native English speakers, mean age = 21; note that this experiment is part of a larger study that is still in progress.)

• Stimuli: Jabberwocky sentences from Experiment 1 (presented in their entirety; i.e., not spliced), alternating with tone sequences of ascending, descending, and constant pitch

• Target detection task:

• 6 blocks of trials; each emotion (happy, angry, neutral) served as a target in 3 blocks; order counterbalanced

• Participants instructed to press one button (right/left counterbalanced across subjects) following a Jabberwocky sentence conveying the target emotion, and the other button following any other type of audio clip

• EEG recording and ERP reduction:

• Quik-Cap (CompuMedics,

Inc.) with 32 sintered Ag/AgCl electrodes arranged according to modified international 10-20 system placement (Figure 5)

Figure 4

• EEG amplified 100x with a SynAmps amplifier; electrodes referenced to linked mastoids

• Incorrect trials excluded, as well as time windows containing artifacts

• Participants whose data yielded >70% trial inclusion were retained for data analysis (n = 11).

• ERPs to Jabberwocky emotional prosody stimuli time-locked to the average isolation point for each emotion, as determined in Experiment

1a, in order to examine the recognition process (reference Fig. 3)

• Data analysis: Repeated measures ANOVAs; Greenhouse-Geisser adjustment applied to correct for sphericity violations

• Time windows of interest: -400 – 0ms, 600 – 2000ms

Exp. 2 Preliminary Results

• -400 – 0 window: significant effect of emotion maximal at Pz, P3/4, Oz, and O1/2

• Amplitudes in neutral condition significantly more negative than in happy or angry conditions

(all p s ≤ .001), which did not differ from one another:

Figure 5

Figure 5

Results (cont’d)

• 600 – 2000 window: significant effect of emotion maximal at FT7/8

• Sustained positivity in neutral condition significantly greater than in happy or angry conditions

(all p s ≤ .001), which did not differ from one another:

Figure 6

Experiment 2 Discussion

• ERP data indicate a negative deflection in the 400ms preceding the isolation point, which likely reflects emotion recognition processes.

• Neutral prosody elicited a greater amplitude positivity between 600-

2000ms than emotional prosody.

General Discussion

• Behavioral and ERP data support a bias for processing neutral prosody over happy and angry prosody.

• Future directions include comparing task-relevant vs. task-irrelevant processing of emotional prosody.

References

1 Phelps et al (2006). Psych. Science , 17, 292-299.

2 Carretié et al (2003).

Psychophys., 40, 381-288.

3 Bechara, Damasio, & Damasio (2003). Ann. NY Academy

Sciences, 985 , 356-369.

4 Ito et al. (1998). J Personality and Social Psych., 75, 887-

900.

5 Wambacq, Shea-Miller, & Abubakr (2004). NeuroReport,

15 , 555-559

6 Alter et al (2003). Speech Communication, 40 , 61-70.

7 Schirmer & Kotz (2003). JCN, 15 , 1135-1148.

8 Grosjean, F. (1980). Perception & Psychophys., 28, 267-

283.

Acknowledgements

• This research was supported in part by an NSF Graduate

Research Fellowship to the first author and NIH grants

(DC00494 and DC03885) to the last author.

• Special thanks to Sarah Callahan, Hong Duong, Maxwell

Moholy, Negin Khalifian, Hoa Nguyen, Uyen Pham, Sara

Rust, Danny Sanchez, and Matt Walenski!

Download