2. Overview
• Motivation: Alcohol slows the motor skills, so it is our
intuition that the articulators would become slower, causing
differences in drunk vs. sober speech
• Question: How does alcohol affect /s/, /sh/, and /z/?
• Method: Record subjects reading a passage sober, then get
them drunk and have them read the same passage.
• Results: /s/, /sh/, and /z/ show minor lengthening in
duration in all positions of intoxicated speech. Amplitude
differences not consistent, and /s/ does not appear like /sh/
on spectrograms.
• Conclusions: Intoxication does affect speech, but effects
vary from person to person
3. Background
• Pisoni & Martin (1989): Performed an acoustic and
perceptual experiment with sober and intoxicated speech.
Found that police officers and college students could
identify sober vs. intoxicated speech when presented in an
isolated environment.
• Tanford, Pisoni, & Johnson (1991): Studied several
samples of a ship captain saying “Exxon Valdez” to see if
acoustic cues are suitable evidence for intoxication. They
found that pitch, variability, and duration can be used to
make a strong case for intoxication (when stress and fatigue
can be ruled out).
4. Background: Continued
• Biadsy, Wang, Rosenberg, & Hirschberg (2011):
Researched which speech qualities are impacted by
intoxication. Found that prosody alone cannot determine
whether or not an individual is intoxicated, and more focus
must be put on phone duration and phonotactic features.
• Heigl (2013): Studied the behavior of /s/ in sober and
intoxicated speech, and found that the palatalization of /s/
(or slurred speech) is not random.
5. Hypothesis:
• Alcohol and intoxication impair a variety of things
including:
• Balance
• Coordination
• Judgment
• Vision
• And many more…
• How does intoxication affect our production of
/s/, /sh/, and /z/? Does /s/ palatalize and become
more like /sh/ in drunk speech?
6. Method
• Subjects: 3 women, 5 men
• Ages: 23-31
• Different ethnic backgrounds
• All native speakers of English
•
• Before stimulus was introduced:
• We took note of each subject’s ideal Blood Alcohol Content
(BAC) from brad21.org
• The ideal BAC for drunken speech is .11-.18
7. Method: Continued
• Recording:
• Subjects answered several questions:
• 1. How many drinks have you had (to track consumption)
• 2. Where are you from?
• 3. What is your favorite day of the week?
• 4. How was the weather today?
• (Questions 2-4 were included in hopes of obtaining unscripted
/s/, /sh/, or /z/ for comparison. However, they were not
included in our analysis.)
• After questions, subjects were asked to read the following
aloud, at their own pace…
8. Method: Continued
• Stimuli:
• “Hooray for Monday. Time to kick off the sheets, clear out the
cobwebs, and kick out the jams. We are ready, willing, and able.
Bring us a whole case of the Mondays. We’ll grab a bottle, shake it
up, and dance in the champagne shower of awesomeness. While
the drones dread Monday, we see it as a starting line for dreamers
and doers. There’s a fresh week of days on the clock and countless
possibilities for our boundless potential. We’re
relaxed, recharged, and reloaded. Ready to tackle the work we were
born to do. Ready to create something remarkable. This is not the
day to be back at the grindstone, this is the day to be back at our
passion, breathing new life into our projects. We are conduits of
innovation, imagination, and inspiration. It’s Monday, and we can’t
wait to get started. On your mark…get set…“
9. Method: Continued
• After the first round of recordings, subjects were instructed
to consume one drink every thirty minutes.
• The following constitute one drink:
• 1.5 oz of liquor
• 12 oz of beer
• 6 oz of wine
• As each subject reached his or her “ideal” number of drinks
(which most actually surpassed), we repeated the same
procedure for the 2nd recording: 4 questions followed by
reading the passage at their own pace
• **all subjects had a designated driver or spent the night
– we do not condone drinking and driving!**
10. Results
• Duration: /s/ & /sh/:
• We looked at the average duration for /s/ & /sh/ in the
initial, medial, and final positions. /z/ was only analyzed in
the final position.
• The following words were analyzed for each participant:
• Champagne (/sh/ initial)
• Passion (/sh/ medial)
• Fresh (/sh/ final)
• See (/s/ initial)
• Awesomeness (/s/ medial)
• Projects (/s/ final)
• Days (/z/ final)
14. Results
• The production of /s/, /sh/, and /z/ were slowed in
intoxicated speech, which resulted in a longer duration (as
we had anticipated).
• Next: We analyzed the spectrograms of drunk and sober /s/
& /sh/ for each pair:
• Champagne/See (initial)
• Passion/Awesomeness (medial)
• Fresh/Projects (final)
• We were hoping that drunk /s/ would appear more like
sober /sh/ on spectrograms, but that hypothesis failed. As
you can see…
18. Conclusions
• We were able to confirm our hypothesis for /s/, /sh/, and
/z/ in almost all of the positions analyzed. With the
exception of /s/ in the final position, we did see longer
durations in intoxicated speech; final /z/ and /sh/ showing
the most significant changes
• Although our study was unable to prove that /s/ palatalizes
and behaves more like /sh/ in drunk speech, Barbara Heigl
was able to make some interesting generalizations about the
palatalization of drunk /s/ in her paper “[s] Under the
Influence of Alcohol.”