phonetics and universal grammar: an open letter to Paul Smolens from Lise Menn
matt at RICE.EDU
Mon Mar 18 13:54:38 UTC 2002
At the LSA meeting in January, during the question period after
your talk on how the constraints of optimality might hypothetically be
represented in the genome, I tried to formulate an issue for you to
consider: which constraints are good candidates for an innate mental
grammar and which are not? I promised you that I would e-mail you a
fuller statement of what I had in mind by my question, since the issue is a
complex one and raises points that you have probably not considered.
Several colleagues in the audience that day, including Shelley Velleman,
John and Manjari Ohala, and Diana Ohala, were also interested in the issue
and encouraged me to attempt a fairly formal exposition of my point, which
1. that many universals of phonology are physiologically based,
2. so that while these are represented in the mind,
3. they are not good candidates for representation in an innate mental
grammar, should there in fact be such an object.
As always, I remain
your genial and loyal opposition,
with sincere hopes of seeing Optimality develop to a level where I
can't find anything to complain about.
Below is my full statement, with lecture notes by John Ohala to make some
of the aerodynamic arguments explicit. It would be better as an
attachment, since it is so long, but network constraints make that
problematic. Anyone who would like either my statement or the PowerPoint
version of John's lecture notes is welcome to e-mail me for them.
Phonetic manifesto: Cave fish are blind, or, Why many innate universals of
language are not candidates for inclusion in an innate grammar
Lise Menn, with Shelley Velleman & John Ohala
References and acoustic physics argument provided by John Ohala
Assuming (only for the sake of argument!) that some parts of phonology are
determined by a genetically-controlled universal grammar (innate UG), and
that 'grammar' refers to a mental language-data processing mechanism, we
argue that EVEN THOUGH CERTAIN CHARACTERISTICS OF PHONOLOGY ARE UNIVERSAL,
they are poor candidates for inclusion in such an innate UG. We use
optimality theory as a basis for exposition.
Cave fish are blind. If natural selection maintains properties of an
organism, they are either useful to survival in some way, or byproducts of
something else that is useful. Natural selection typically fails to
maintain characteristics that are either useless (eyes in a cave) or
redundant. (Another putative example is the ability to synthesize vitamin C
in primates who have good access to fruit and other sources; see article on
Vitamin C in The Cambridge World History of Food.) Useless or redundant
characteristics of course MAY be maintained because they are accidental
by-products of some other capacity (the appendix in the human
gastro-intestinal tract; 'spandrels', in a popular metaphor).
If a property of language is a consequence of human (vocal tract) anatomy,
auditory processing, or other non-mental circumstances, it does not need
also to be part of a MENTAL grammar. For example, the constraint strongly
disfavoring speech sounds made with ingressive pulmonic airflow (as in
gasping, or sucking air in through the teeth) presumably is due to
something about human breath control, the fact that dry air is bad for the
mucous membranes of the mouth and larynx, etc.
Furthermore, speakers making such sounds quickly discover the discomfort
and the effort that they cause, through ordinary sensory feedback circuits,
so there is no need for a pre-wired mental counterpart to the physical
Many violable constraints are demonstrably based on anatomy, muscle
control, and other physical properties of the vocal tract. Some are rather
strong, like the constraint against nasal fricatives (which demand a lot of
airflow); some are very weak, like those against specific consonants -
*/p/, */g/, */h/, which are very common yet occasionally absent. See the
appended adaptation of an Ohala power-point lecture on airflow requirements
for fricatives and voicing.
These constraints are indeed as innate as having a nose and mouth. But
they are not MENTAL constraints, they are PHYSIOLOGICAL constraints. So
the genes which control them must be genes that structure the body, not the
Speakers subconsciously know the constraints of their grammar, regardless
of the source of the constraint. Consider the Jusczyk et al. experiments
showing that infants recognize native vs. unfamiliar phonotactic patterns,
not to mention the common experiences of second language learners who try
to overcome their native language production patterns.
In phonology, to describe language patterns as well as individual speakers'
knowledge, all constraints must be represented, regardless of whether their
source is transparently physiological, or perceptual, or based on general
human cognitive properties, or indeed based on language-specific mental
Under the assumptions that
some constraints might be genetically encoded as part of a mental grammar; and
no constraint will be directly selected for as part of an innate mental
grammar if it is redundantly a consequence of physiology & physics, and
therefore both internal to and learnable by a physiologically intact child
trying to make sounds,
which constraints, if any, are good candidates for representation in a
innate mental grammar?
According to the argument above, the poor candidates include all
constraints that are demonstrably physiological in origin (and therefore
genetically encoded, but not in the mind). I think this includes at least
the majority of common constraints. Indeed, anything that takes the tongue
away from resting position would be easier not to do, from the speaker's
point of view.
[We also know that many constraints can demonstrably be learned by exposure
to the ambient language (cf. the work of Jusczyk and colleagues, Aslin &
colleagues). They would also be redundant in an innate grammar and
therefore poor candidates for membership in a UG - but I suppose it could
be counter-argued that the reason that children can learn them is that they
are part of UG. I am not sure how to reach a testable hypothesis in this
Ohala Lecture notes (edited by LM from the original PowerPoint version)
1. For a fixed mass of air, pressure varies inversely with volume.
(Boyle-Mariotte's Law)*; e.g., in a hand bicycle pump one pushes a plunger
to reduce the volume of the air and thus increase the pressure.
Pressure times Volume is constant for a fixed mass of air; P1V1 = P2V2.
For a given volume, pressure varies directly with the mass of air inside.
E.g., inflating or stiffening a car tire by pumping air into it.
(We neglect the effects of temperature; we assume adiabatic conditions.)
2. The quantity of air passing through a channel varies proportionally with
the diameter of the channel, and is positively correlated with the
magnitude of the pressure differential across the channel. (Coffee flows
from a large coffee urn in greater volume as (a) the tap is more open and
(b) the greater the amount of coffee in the urn.)
U = A (P at entrance - P at exit) to the a power, times C
U is 'volume velocity', the quantity of air per unit time, typically, cm3/sec;
A is the channel diameter;
a varies between 0.5 and 1.0;
c is a constant.
3. Air speed (particle velocity) varies directly with quantity of flow
(volume velocity) and inversely with channel cross-dimension. (This is the
principle exploited in carburetors.)
This principle is relevant because the degree of turbulence and thus the
noise produced during fricatives is dependent in part on how fast the air
is moving past a constriction. The intensity and center frequency* of
frication noise vary monotonically with particle velocity of the air flow.
(When a strong wind is blowing around one's house, one can get an
impression of its intensity by listening to the loudness.* But these
parameters can also be affected by the resonances of the vocal tract of the
noise and its "pitch".)
The Bernoulli* effect:
The pressure at right angles to flowing air is inversely correlated
with the velocity of the air flow. This principle is important for
How airplanes fly
4. The Aerodynamic Voicing Constraint
Voicing requires: Vocal cords adducted (lightly approximated at the midline)
Air flowing through the vocal cords.
Certain articulations impact on airflow.
Note: similar principles apply to trills
Obstruents block or reduce the flow of air out of the oral cavity (by
aerodynamic principle #3).
Therefore, the air accumulates in the oral cavity and, by aerodynamic
principle #2, the pressure behind the constriction increases.
This reduces the pressure drop (DP = [P at entrance - P at exit]) across
the vocal cords
If DP goes below a certain minimum value (~ 1 or 2 cm H2O), the air flow
falls below the level needed to maintain voicing and thus voicing will be
5. Factors favoring voicing during supraglottal sounds:
i) Shorter duration of the consonantal closure, since there is less time
for Po (the oral cavity air pressure) to build up to a level that reduces
DP below the critical level
ii) Larger oral cavity (= more front place of articulation), since a larger
cavity means more surface area to passively expand and thus accommodate
more of the air accumulating in the oral cavity.
iii) Active expansion of the oral cavity by larynx lowering, jaw lowering,
augmenting velum elevation -- again, to accommodate more air accumulating
in the oral cavity.
iv) Velic leakage
Related to the difficulty of maintaining voicing on back-articulated stops
are the following:
Even in languages that have /g/, the incidence of this phoneme -- both in
the dictionary and in connected speech -- is often statistically much lower
than for /b/ (Gamkrelidze; Wang & Crawford).
Phonetically /g/ is often less voiced than /d/ and /b/. E.g., my English
/g/ is commonly voiceless even intervocalically.
The absolute absence of /g/ in some languages, the statistical infrequency
of it in others, and its phonetic devoicing in others are all
manifestations of the same basic universal factor.
There is a "bias" among obstruents to be voiceless. Incidence of [voice] on
obstruents in 706 segment inventories surveyed by Ruhlen. (For similar
data, see Maddieson 1984.)
Back-articulated voiced stops are more likely to be missing from languages
with a voicing contrast among obstruents.
Incidence of stop gaps by place and ±voice in 87 languages surveyed by
Sherman 1975 (see similar data by Maddieson 1984).
Labial Apical Velar
Voiceless 34 0 0
Voiced 2 21 40
Familiar examples: Thai, Dutch, Czech (in native vocabulary)
Fricatives have a greater bias against voicing than do stops. For optimal
voicing, Po must be as low as possible (to keep DP high).
For optimal voicing, Po must be as low as possible (to keep DP high).
Both of these actions on Po cannot be done simultaneously.
The result is that voiced fricatives with strong frication (e.g., [ z,
'ezh'] have a tendency to devoice; those with strong voicing (e.g., [v
'eth' ] tend to have weak, if any, frication.
Definition: an obstruent is a sound that substantially impedes the flow of
air out of the vocal tract; everything else is a sonorant.*
Sonorants are voiced
Obstruents are voiceless
*Note: these may not always be dichotomous categories; there can be a
continuum according to the degree of obstruction of the air flow.
However, in a great many languages, there may be a contrast in voicing in
obstruents, and in a small number of languages there may be voicing
contrasts on sonorants.
Moreover, in many cases the voicing contrast on obstruents hinges on the
relative timing of the voicing with respect to the timing of the
Also, the perceptual cues for these contrasts (as with many phonetic
contrasts) are multiple and frequently involve phonetic features other than
simple [ ± voice].
Although there are some general tendencies, the facts must be determined
for each individual language.
For an attachment with a power-point version of this Ohala lecture, please
e-mail him <ohala at socrates.berkeley.edu> or me.
Beware Procrustes bearing Occam's razor.
Lise Menn office phone 303-492-1609
Professor home fax 303-413-0017
Department of Linguistics
University of Colorado
Boulder, CO 80309-0295
Lise Menn's home page
"Shirley Says: Living with Aphasia"
Japanese version of "Shirley Says"
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Funknet