Transitions + Perception
March 25, 2013
The Home Stretch
• Guidelines for the final course project term paper and/or
presentation
• Papers will be due on Friday, April 26th
• Our one and only presentation will happen on
Monday, April 15th
• Also: I have a perception homework for you
• But we need to learn a few more things before we
can get to it…
• However, you can have your response sheets back!
Laterals
• Laterals are
produced by
constricting the sides
of the tongue towards
the center of the
mouth.
• Air may pass
through the mouth on
either both sides of
the tongue…
• or on just one
side of the tongue.
Lateral Palatography
Lateral Acoustics
• The central constriction traps the flow of air in a “side
branch” of the vocal tract.
• This side branch makes the acoustics of laterals similar to
the acoustics of nasals.
• In particular: acoustic energy trapped in the side
branch sets up “anti-formants”
• Also: some damping
• …but not as much as in nasals.
17.5 cm
4 cm
• Primary resonances of lateral approximants are the
same as those of
for vocal tract length of 17.5 cm
• 500 Hz, 1500 Hz, 2500 Hz...
• However, F1 is consistently low (300 - 400 Hz)
• Anti-formant arises from a side tube of length  4cm
• AF1 = 2125 Hz
Laterals in Reality
• Check out the Mid-Waghi and Zulu laterals in Praat
Mid-Waghi:
[alala]
Velarization of [l]
• [l] often has low F2 in English because it is velarized
• = produced with the back of the tongue raised
• = “dark” [l]
• symbolized
• Perturbation Theory flashback:
• There is an anti-node for F2 in the velar region
• constrictions there lower F2
Dark vs. Clear /l/
•/l/ often has low F2 in English because it is velarized.
[alala]
[l] vs. [n]
• Laterals are usually more intense than nasals
• less volume, less surface area = less damping
•  break between vowels and laterals is less clear
[
]
[
n
]
[l] vs.
• [l] and
are primarily distinguished by F3
• much lower in
• Also: [l] usually has lower F2 in English
[
]
[
]
Glides
• Glides are vowel-like sonorants which are produced…
• with slightly more constriction than a vowel at the
same place of articulation.
• Each glide corresponds to a different high vowel.
Vowel
Glide
Place
[i]
[j]
palatal
(front, unrounded)
[u]
[w]
labio-velar
(back, rounded)
[y]
labial-palatal (front, rounded)
velar
(back, unrounded)
• Each glide’s acoustics will be similar to those of the vowel
they correspond to.
Glide Acoustics
• Glides look like high vowels, but…
• are shorter than vowels
• They also tend to lack “steady states”
• and exhibit rapid transitions into (or from) vowels
• hence: “glides”
• Also: lower in intensity
• especially in the higher formants
[j] vs. [i]
[w] vs. [u]
Vowel-Glide-Vowel
[iji]
[uwu]
More Glides
[wi:]
[ju:]
Transitions
•
When stops are released, they go through a
transition phase in between the stop and the vowel.
•
From stop to vowel:
1. Stop closure
2. Release burst
3. (glide-like) transition
4. “steady-state” vowel
•
Vowel-to-stop works the same way, in reverse,
except:
•
Release burst (if any) comes after the stop
closure.
Stop Components
closure
voicing
vowel
formant
transitions
stop
release
burst
• From Armenian:
[bag]
another
closure
Confusions
• When the spectrogram was first invented…
• phoneticians figured out quite quickly how to identify
vowels from their spectral characteristics…
• but they had a much harder time learning how to
identify stops by their place of articulation.
• Eventually they realized:
• the formant transitions between vowels and stops
provided a reliable cue to place of articulation.
• Why?
Formant Transitions
• A: the resonant frequencies of the vocal tract change as
stop gestures enter or exit the closure phase.
• Simplest case: formant frequencies usually decrease near
bilabial stops
Stops vs. Glides
“baby”
• Note: formant
transitions are
more rapid for
stops than they
are for glides.
“wave”
Formant Transitions: alveolars
• For other places of articulation, the formant transition
that appears is more complex.
• From front vowels into alveolars, F2 tends to slope
downward.
• From back vowels into alveolars, F2 tends to slope
upwards.
• In Perturbation Theory terms:
• alveolars constrict somewhat closer to an F2 node
(the palate) than to an F2 anti-node (the lips)
[hid]
[hæd]
Formant Locus
• Whether in a front vowel or back vowel context...
• The formant transitions for alveolars tend to point to
the same frequency value. ( 1650-1700 Hz)
• This (apparent) frequency value is known as the locus of
the formant transition.
• In the ‘50s, researchers theorized:
• the locus frequency can be used by listeners to
reliably identify place of articulation.
• However, velars posed a problem…
Velar Transitions
• Velar formant transitions do not always have a reliable
locus frequency for F2.
• Velars exhibit a lot of coarticulation with neighboring
vowels.
• Fronter (more palatal) next to front vowels
• Locus is high: 1950-2000 Hz
• Backer (more velar) next to back vowels
• Locus is lower: < 1500 Hz
• F2 and F3 often come together in velar transitions
• “Velar Pinch”
The Velar Pinch
[bag]
[bak]
“Velar” Co-articulations
Testing the Theory
• The earliest experiments on place perception were
conducted in the 1950s, using a speech synthesizer known
as the pattern playback.
Pattern Playback Picture
Haskins Formant Transitions
• Testing the perception of two-formant stimuli, with
varying F2 transitions, led to a phenomenon known as
categorical perception.
Categorical Perception
• Categorical perception =
• continuous physical distinctions are perceived in
discrete categories.
• In the in-class experiment from last time:
• There were 11 different syllable stimuli
• They only differed in the locus of their F2 transition
• F2 Locus range = 726 - 2217 Hz
Source: http://www.ling.gu.se/~anders/KatPer/Applet/index.eng.html
Stimulus #1
Stimulus #6
Stimulus #11
Example
stimuli from
the in-class
experiment.
Identification
• In Categorical Perception:
• All stimuli within a category boundary should be
labeled the same.
Discrimination
• Original task: ABX
discrimination
• Stimuli across
category boundaries
should be 100%
discriminable.
• Stimuli within
category boundaries
should not be
discriminable at all.
In practice, categorical perception means: the
discrimination function can be determined from the
identification function.