78 / 412
Annals of Otology, Rhinology & Laryngology 123(4)
Table 3.
Repeated Measures Analysis of Variance for Baseline
Laryngeal Electromyography Across Time.
Source
df
F
Value
P
Value
Participant
6
0.85
.555
Time
2
0.02
.978
Table 4.
Repeated Measures Analysis of Variance for Root
Mean Square Values.
Source
df
F
Value
P
Value
Participant
6
1.91
.160
Time
2
1.38
.289
Intensity
1
5.71
.054
Time*Intensity
2
2.23
.150
intensity targets (65 and 75 dB [± 3 dB]). These intensity
targets were chosen to represent a typical healthy vocaliza-
tion intensity range present in everyday speech. In this con-
dition, participants received visual feedback of their
intensity level by monitoring a dB sound level meter placed
16 inches from the individual’s mouth. The participant was
asked to hold the intensity constant for a minimum of 3 sec-
onds. Trials less than 3 seconds were not accepted for analy-
sis. A total of 10 trials of each task condition were recorded.
The initial and final 2 trials were discarded, leaving 6 trials
for post-hoc data analysis.
Habitual fundamental frequency (F
o
) for the vocalization
tasks was initiated by the participant without prompting
from the investigators. The F
o
chosen by the participant was
recorded and played back via an auto-tuner before data col-
lection so that the participant could remain in an acceptable
modal pitch range. Before all data collection procedures
and to ensure performance consistency, each participant
was trained and given time to practice all tasks before data
collection.
Data Analysis
A 2-factor repeated measures analysis of variance
(RM-ANOVA) was used to compare the effect of the 2
independent variables among participants. The independent
variables included (1) LEMG data recording sessions (3
sessions) and (2) the task-related vocal intensity levels (2
dB levels). To evaluate the reliability of LEMG signals
across testing times, intraclass correlation coefficients
(ICCs) using a 2-way mixed random effects model were
calculated. For between-session measurements, standard
error of the measurement (SEM) was calculated to deter-
mine the minimum detectable change (MDC) in microvolts
with 70% and 95% confidence boundaries. Intraclass cor-
relations were also calculated to measure intra-session reli-
ability. All statistical calculations were performed using
SPSS version 18 (IBM, Armonk, New York, USA).
Data Selection and Calculations
A 1-second window from the mid-portion of each 3-second
task recording window was selected for detailed analysis.
Root mean square amplitude values were automatically
calculated for each of the 1-second windows using a cus-
tom sub-routine in LabChart (ADInstruments, Inc). To be
included for further processing and analysis, digitized sam-
ples had to fall within +/– 1 dB of the desired intensity
levels (65 or 75 dB). In certain cases, the RMS amplitude
was not stable for 1 second at +/– 1 dB. In these cases, a
shorter window was averaged to eliminate portions of the
signal with poor signal quality. The inclusion of smaller
analysis windows for these cases was preferable to averag-
ing poor signal quality. Smaller sampling windows do not
significantly affect analyses, as comparisons of EMG sam-
ple windows using RMS have been shown to produce mod-
erate to excellent ICC reliability data from 1000 ms down
to 100 ms window sizes.
13
Results
Statistical Analysis
Group data, including means and standard deviations for
the dependent variable (RMS) for intensities at 65 dB and
75 dB, are presented in Tables 1 and 2, respectively. Results
of the 2-factor RM-ANOVA during baseline and the task
conditions are presented in Tables 3 and 4. Repeated mea-
sures ANOVA for the baseline data did not reveal a signifi-
cant main or interaction effect for testing time (session)
versus RMS, indicating stable LEMG baseline activity
across recording sessions. Results of the RM-ANOVA for
the feedback condition were not significant for the main
effect of testing time (
df
= 2,
F
= 1.38,
P
= .289). However,
the main effect for intensity level closely approached our a
Table 1.
Means for Intensity Level at 65 dB.
Variable
Session Mean SD Min Max
Root mean square
(µV)
1 80.2 34.8 44.7 151.5
2 91.8 28.4 45.4 117.6
3 73.08 19.64 46.31 98.48
Table 2.
Means for Intensity Level at 75 dB.
Variable
Session Mean SD Min Max
Root mean square
(µV)
1
84.0 37.4 46.0 155.5
2 106.5 29.4 69.2 136.9
3
85.04 16.82 55.42 101.56
58