Upload
rodger-holt
View
215
Download
1
Tags:
Embed Size (px)
Citation preview
August 16, 1999 SMPC Conference '99 1
Cross-modal integration:Alignment of auditory and visual
accent structures in motion pictures
Dr. Scott D. Lipscomb
Institute for Music Research
University of Texas at San Antonio
August 16, 1999 SMPC Conference '99 2
Acknowledgments• University of California, Los Angeles
– Dr. Roger A. Kendall & Dr. Ed Carterette
• University of Texas at San Antonio– Institute for Music Research
• Dr. Don Hodges, Director
• Fellow Music Researchers
August 16, 1999 SMPC Conference '99 3
Film Music Literature
• past studies deal almost exclusively with the referential aspect of musical sound– “cognitive congruency” (Marshall & Cohen, 1988)
– selected others: Tannenbaum (1956), Thayer & Levenson (1984)
– special issue of Psychomusicology (vol. 13, 1994)
• vs. accent structure alignment– i.e., how often important events in the music coincide
with important events in the visual scene
August 16, 1999 SMPC Conference '99 4
Film Music Perception Paradigm(Lipscomb & Kendall, 1995)
AssociationJudgment
Accent StructureRelationship
Perception
Aural stimulus Visual stimulus
YES
YES
NO
NO
IMPLICITPROCESSES
No Shift ofAttentional
Focus
Shift ofAttentional
Focus
AuralStimulus
VisualStimulus
August 16, 1999 SMPC Conference '99 5
Alignment Conditionsafter Yeston (1975)
b)
a)
c)
Consonant
Out-of-phase
Dissonant
August 16, 1999 SMPC Conference '99 6
Subject Responses …two VAME ratings
• synchronization: “… how often important events in the music coincide with important events in the visual scene”
• effectiveness: “… simply concerns [the observer’s] subjective evaluation of how well the two go together”
August 16, 1999 SMPC Conference '99 7
Results of Previous Study (1997)presented at Penn State ASA Conference
Synchronization Ratings--All Experiments
20.00
30.00
40.00
50.00
60.00
70.00
80.00
90.00
100.00
Consonant Out-of-Phase Dissonant
Alignment Condition
Mea
n S
core
s
Exp 1
Exp 2
Exp 3
Effectiveness Ratings--All Experiments
20.00
30.00
40.00
50.00
60.00
70.00
80.00
90.00
100.00
Consonant Out-of-Phase Dissonant
Alignment Condition
Mea
n S
core
s
Exp 1
Exp 2
Exp 3--UntrainedExp 3-Mod &Trained
August 16, 1999 SMPC Conference '99 9
Revised Stimulus Preparation
• A-V alignment conditions were created using Media 100 software on a Macintosh G3 computer
• sound files were manipulated in Sonic Foundry’s Sound Forge 4.0, using the Time Compress/Expand Sonic Foundry plug-in
• completed A-V composites were recorded directly from Media 100 onto VHS tape
August 16, 1999 SMPC Conference '99 10
Consonant Alignment Condition
aligned as intended by the composer
(visual & music only)
August 16, 1999 SMPC Conference '99 11
Out-of-phase Alignment Conditionalignment conditions based on Preliminary Study
A-V Scene misalignment of audio
"Dots" -100 ms
"Canon" 532 ms
"Synchromy" 532 ms
"Portrait of Elizabeth" 672 ms
"Flashback" -890 ms
"Reunion" 425 ms
August 16, 1999 SMPC Conference '99 12
Dissonant Alignment Conditions
• audio tracks for the McLaren animations were “time expanded” by 115%
• audio tracks for the “Obsession” excerpts were “time expanded” by 110%– since these excerpts were longer, the gradual
misalignment could occur at a slower pace
August 16, 1999 SMPC Conference '99 13
Subject Info & Method• N = 135 UTSA students taking music classes
– musical training n (high = 17, mod = 31, low = 87)– gender n (female = 82, male = 53)– stimuli presented to groups of subjects
• stimuli were presented to subjects using a Samsung VR 5855 video cassette recorder and a ProScan PS80690 80” big screen television
– VAME ratings were provided on a continuous line response anchored by either “not synchronized - synchronized” or “ineffective - effective”
– each subject was assigned to one of three random stimulus presentation orders
– response forms were generated so that order of VAME responses was also randomized into three versions
not synchronized synchronized
August 16, 1999 SMPC Conference '99 15
Statistical Analysis - Between(SPSS’s “General Linear Model - Repeated Measures”)
• repeated measures, fully-factorial ANOVA– three repeated measures: 2 VAME ratings x 6 AV
combinations x 3 alignment conditions
• Tests of Between-Subjects Effects– no significant between-groups variation
• musical training: p = 0.090; f(2,129) = 2.457
• gender: p = 0.549; f(1, 129) = 0.361
– interaction between AV combination & musical training is not statistically significant
• p = 0.622; f(2,129) = 0.477
August 16, 1999 SMPC Conference '99 16
Statistical Analysis - Within(SPSS’s “General Linear Model - Repeated Measures”)
• Tests of Within-Subjects Effects– within-groups, ratings were significantly different depending upon
the AV combinations• p < .0005; f(5) = 20.5
– within-groups, ratings were significantly different depending upon the alignment condition
• p < .0005; f(2) = 173.036
• Within-Subjects Interaction Effects– alignment condition x VAME
• p < .0005; f(2) = 3.906
– AV combination x alignment condition• p < .0005; f(10) = 48.720
August 16, 1999 SMPC Conference '99 17
Statistical Analysis - Within(SPSS’s “General Linear Model - Repeated Measures”)
• Within-Subjects Effects - complex interactions– AV combination x VAME x musical training
• p = .006; f(10) = 2.517
– AV combination x alignment condition x VAME• p < .0005; f(10) = 3.885
August 16, 1999 SMPC Conference '99 18
Experimental Results(r = .997)
McLaren Animations
20.0030.0040.0050.0060.0070.0080.0090.00
100.00
Dots
Dots_p
Dots_x
Canon
Canon_
p
Canon_
x
Synch
rom
y
Synch
rom
y_p
Synch
rom
y_x
A-V Stimuli
Mea
n R
atin
g
Synchronization
Effectiveness
August 16, 1999 SMPC Conference '99 19
Experimental Results(r = .977)
"Obsession" Excerpts
20.0030.0040.0050.0060.0070.0080.0090.00
100.00
Portra
it
Portra
it_p
Portra
it_x
Flashb
ack
Flashb
ack_
p
Flashb
ack_
x
Reunion
Reunion
_p
Reunion
_x
A-V Stimuli
Mea
n R
atin
g
Synchronization
Effectiveness
August 16, 1999 SMPC Conference '99 20
Experimental ResultsSubject Ratings Collapsed (n=3)
McLaren AnimationsSubject Ratings Collapsed
0.00
20.00
40.00
60.00
80.00
100.00
Consonant Out-of-phase Dissonant
Alignment Condition
Mea
n R
atin
g
"Obsession"Subject Ratings Collapsed
0.00
20.00
40.00
60.00
80.00
100.00
Consonant Out-of-phase Dissonant
Alignment Condition
Mea
n R
atin
g
Results of Previous Study
August 16, 1999 SMPC Conference '99 21
Conclusions• accent structure alignment does, in fact, play an
important role in subject ratings of both “synchronization” and “effectiveness”– ANOVA & Pearson correlation coefficient
• based on the results of the present & previous investigations, there appears to be a dynamic relationship between the “association judgment” and “accent structure alignment” aspects of the model of Film Music Perception
Revised Model
August 16, 1999 SMPC Conference '99 22
Contact Info
Dr. Scott D. LipscombUTSA Division of Music6900 N. Loop 1604 WestSan Antonio, TX 78249
(210) 458-4354(210) 458-4381 [email protected]
http://music.utsa.edu/~lipscomb