There was someone back in the 90s interested in forming a SIG, or at least he
said so. But instead he got a 'gig' as guest editor of the Language Teacher
magazine, special issue on 'Pronunciation'. However, he was taken off the job
for bad behaviour (such as soliciting manuscripts and then imposing a two-week
deadlines for submission out of spite apparently!). The usual JALT
shenanigans--basically over-educated, over-degreed people with too little to
actually do related to ELT here in Japan.

At any rate, it appears the Pronuciation SIG zombie last raised its head at the
end of 2002. At that time I was very busy with editing ETJ Journal. Actually got
that to have a fairly good run, from 2000 to 2006--with the collapse coming
after I left (predictably enough).

Some web artifacts related to the last time it looked like there might be a
Pronun SIG:

http://www2.gol.com/users/elin/jps/index.html

JALT Pronunciation SIG Homepage

     * 2002 Officers
     * Statement of Purpose
     * Constitution
     * How to Join
     * Newsletters
           o Newsletter 1
           o Newsletter 2
     * Conference Information (not yet available)
     * What you can do

This site was last updated on December 12, 2002

http://www2.gol.com/users/elin/jps/info.html#2002%20Officers

In iatefl_pronsig@..., "pronsig_mod" <pronsig_mod@...> wrote:
>
> I asked around about a JALT PronSIG after last year's conference and the
interest was rather vague and very non-committal. There was plenty of interest
in pronunciation at the conference yesterday, but I don't see much chance of
getting an active SIG going - unless you're offering to take the helm? <<

For most SIGs, that is all there ever is--a vague non-comittal sense of
interest. I remember trying to get the FL Literacy SIG going, attending chapter
meetings and the national conference, discussing different interests such as
extensive reading, intensive reading, EAP, translation, etc. etc. and most
people in JALT said they didn't know what 'FL Literacy' was. Most people
apparently are not synthesists. Usually a SIG gets started around a hardcore
group of people who share an agenda--often self-promotion around a small cluster
of ideas they share or shared from grad school. Usually the focus is much to
narrow to sustain a healthy SIG, except for the fact that the group is a
hardcore clique of self-promoters.

The bigger issue, though, is getting wide membership from within JALT so as to
sustain the SIG--through its general membership, but equally importantly, so
there are enough people to volunteer to run the SIG and take up the officer
positions.

Now a nascent SIG might do this with one or both of the following: cast out a
set of appealing ideas and publish a magazine/journal in order to attract
members from within in JALT (it says it has around 3000 members, but I would bet
the number of actual paying members is closer to 2500). Or it could attract new
people into the SIG. The catch is they have to join JALT too.

And that was the dilemma with the FL Literacy SIG. We attracted a lot of
attention from outside JALT, but very few were willing to join JALT.

If a 'hardcore' group of at least 5 people said they were interested in REVIVING
the Pronunciation/Phonology/Phonetics SIG, then I would be interested. I know
where the most work is--membership (if your membership chair actually does the
job) and newsletter/publication (it only makes sense to stick to a 'newsletter'
if you have a large enough membership to have the need to send out a newsletter
to three times a year--the JALT requirement). Otherwise, it is better to go for
quality content, instead of the usual newsletter filler.

Those are my thoughts.

CJ

PS: who was attacking Vygotsky such that he needed defended?

I asked around about a JALT PronSIG after last year's conference and the
interest was rather vague and very non-committal. There was plenty of interest
in pronunciation at the conference yesterday, but I don't see much chance of
getting an active SIG going - unless you're offering to take the helm?
I think it's mainly you, me and Chris Madden who make up the Japan-based
activity here. Unfortunately I didn't have much time yesterday to promote the
group.
Conference was, BTW, mind-blowing, wish I could have gone to the three days.
Listening to James Lantolf defending Vygotsky has set my mind racing and got me
inspired again. A simple message to all teachers everywhere - you are doing the
most important job in the world: you are shaping people's lives whether you mean
to or not - take care how you do it.

Alex.



--- In iatefl_pronsig@..., "literacyacrosscultures" <jannuzi@...>
wrote:
>
> There was one for a little bit. As with several other SIGs trying to form back
in the late 90s and early 2000s, if asked, I would reply that, based on my own
experiences with SIGs in JALT, forming a new SIG is not advised. I would argue
it was better to integrate the topic areas into existing ones, if possible.
That's because of the ways JALT makes it difficult to get members into a SIG. I
advised against the formation of the Pronunciation SIG because I knew it
wouldn't get enough leadership/officership activity to sustain it long enough to
get members. And then even if it did, it would be a long haul to get members.
>
> However, I see there is a homepage for this SIG. I also see, however, that the
JALT homepage no longer lists it as an active SIG.
>
> Any chance of re-viving this SIG--against all odds, against JALT's
indifference to SIGs?
>
> The thought occurs to me because I see that, unlike the 90s and early 2000s
with the IATEL Pronunciation SIG, the current version of the IATEFL
Pronunciation SIG has activity and membership that is Japan-based.
>
> Charles Jannuzi
> University of Fukui
> Former officer of CUE SIG and FL Literacy SIG of JALT
> Former JALT Fukui officer
>
> PS: FL Literacy SIG now defunct, even though Rob Waring went ahead and tried
to form some sort of 'Extensive Reading' group within JALT, aggravatingly
ignoring the existence of a much more sound, broader themed FL Literacy SIG.
Etc. etc. JALT was always wearisome.
>

There was one for a little bit. As with several other SIGs trying to form back
in the late 90s and early 2000s, if asked, I would reply that, based on my own
experiences with SIGs in JALT, forming a new SIG is not advised. I would argue
it was better to integrate the topic areas into existing ones, if possible.
That's because of the ways JALT makes it difficult to get members into a SIG. I
advised against the formation of the Pronunciation SIG because I knew it
wouldn't get enough leadership/officership activity to sustain it long enough to
get members. And then even if it did, it would be a long haul to get members.

However, I see there is a homepage for this SIG. I also see, however, that the
JALT homepage no longer lists it as an active SIG.

Any chance of re-viving this SIG--against all odds, against JALT's indifference
to SIGs?

The thought occurs to me because I see that, unlike the 90s and early 2000s with
the IATEL Pronunciation SIG, the current version of the IATEFL Pronunciation SIG
has activity and membership that is Japan-based.

Charles Jannuzi
University of Fukui
Former officer of CUE SIG and FL Literacy SIG of JALT
Former JALT Fukui officer

PS: FL Literacy SIG now defunct, even though Rob Waring went ahead and tried to
form some sort of 'Extensive Reading' group within JALT, aggravatingly ignoring
the existence of a much more sound, broader themed FL Literacy SIG. Etc. etc.
JALT was always wearisome.

I think this view from 'speech processing' (by way of psycholinguistics I
believe) is also convergent on what I'm discussing here. They should make this
required reading at all 'Phonology and Phonetics' courses in applied linguistics
and TEFL.

http://www.psy.cmu.edu/~lholt/publications/PhonemeIllusion.pdf

The Illusion of the Phoneme
Andrew J. Lotto
Loyola University Chicago
Lori L. Holt
Carnegie Mellon University

0. Caveat

A caveat is warranted here. While our title is provocative, our ambitions are
much more prosaic. Obviously the debate on the ontological status of the
phoneme has a long and complicated history. We offer neither a summary of this
debate nor a last word on the question. We seek only to question the role of the
phoneme in the perception of speech and, in doing so, we hope to demonstrate
that the empirical evidence for the causal role of the phoneme in perception is
limited.

1. The Ontological Status of the Phoneme

Classifying speech in terms of discrete abstract labels has been an extremely
successful descriptive system for linguistics. While there are well-known
concerns about the "shoe-horning" of different languages' phonetic systems into
universal phoneme sets, there is little doubt that this descriptive symbol
system
has been instrumental in the progress of the speech sciences. Phonemes offer a
communication system for empirical results. In addition, standardized
descriptions of different linguistic sound systems have made the immense
variation in language manageable. But has this desire for simplicity and
generality blinded us to some of the realities of speech as a communication
system?

1.1 The "Categorization Error"

The early Gestalt Psychologists (e.g. K�hler, 1930; Koffka, 1935) warned
against making the "experience error" when theorizing about human perception.
This error arises when "we mistake the result of organization for the cause of
organization" (Koffka, 1935). For the Gestalt theorists, structure and
organization
in perception did not necessarily imply that this structure and organization
must
be present in the input. We believe that this cautionary note applies beyond the
scope of "illusory contours" and perceptual grouping rules. Structure and
organization in behavior need not imply that this structure and organization is
present in mental representation. It is easy enough in many areas of study to
mistake efficient descriptors of the structure of a system's behavior as causal
entities responsible for that structure. For example, when subjects are asked to
rate members of a category for their typicality (e.g. `birds' or `furniture' or
`members of /i/') there is a clear structure in their responses. In particular,
there is
usually a prominence in the ratings across a subset of the members. This
prominence in responses is often described as a prototype. Similar prominences
in
response structure can be observed for a variety of tasks and measures (e.g.
reaction time for identification). The prototype is a very useful descriptor of
the
response structure for these types of tasks. After all, a prominence in
responding
is a very salient attribute and it may have functional significance..
Beyond being an important descriptive entity, prototypes in the response
structure are often taken as evidence for mental prototypes internal to the
subject
which are responsible for the structure of the output (e.g. Posner & Keele,
1970;
Rosch, 1975; 1978). This theoretical approach appears rather compelling: the
structure of mental representations is mirrored in the structure of responses.
However, we now have a variety of models of categorization that lead to response
structures that include a prominence or "prototype". Exemplar models and
connectionist net models deliver similar response structure with no explicit
representation of a mental or internal "prototype" (see, e.g., Reed, 1972;
Nelson,
1974; Brooks, 1978; Medin & Schaffer, 1978; Knapp & Anderson, 1984;
Kluender, Lotto, Holt, & Bloedel, 1998). It is clear, then, that the presence of
a
"prototype" in behavioral data does not necessitate a "prototype" in the mental
representation of the stimuli. To make this sort of presumption about mental
representation structure from response structure is to commit something akin to
the "experience error".i Whereas, the Gestalt psychologists reprimanded the
presumption that subjective organization is determined by organization in the
stimulus, we are suggesting that it is wrong to presume that structure in
response
follows in a straightforward manner from similar structure in mental
representations. The Gestalt version concerned the "experience" of subjective
perception. We are discussing the "experience" of objective data. This is
stretching the metaphor a bit and, thus, we prefer to refer to this as the
categorization error. That is, sometimes we mistake the description of response
structures (i.e. we categorize the structure for efficiency) as causes of
response
structures (i.e. we presume the category to exist as a mental entity).
Another example of a "categorization error" would result from confusing
"rule-described" behavior and "rule-following" behavior (Heil, 1983; Ben-Zeev,
1987). The fact that behavior can be described efficiently by rules does not
entail
that the behavior is the result of the following of explicit rules. This
distinction
was made clear by Wittgenstein's (1953) "skeptical paradox". A contemporary
example is the production of the past tense in English. The morphological
changes of regular verbs from present to past tense can be described quite
efficiently by a rule. This is certainly an example of rule-described behavior.
However, it would be an error to presume that this is equivalent to saying that
the
behavior is a result of explicit rule following. Connectionist networks can map
the morphological change for regular verbs and even novel exemplars without the
representation of explicit rules (Rumelhart & McClelland, 1986). These networks
even appear to model the trajectory of acquisition of correct production of the
past
tense by children, again without reference to explicit rules (though, see Pinker
&
Prince, 1988 for a critical review of these results).ii
Are we committing the "categorization error" when we suggest that the
phoneme is the fundamental representation or unit of speech? It is an efficient
unit for describing speech behavior, but does that give us license to infer that
it is
a causal entity? And worse, could it be that because we give a prominent role to
the phoneme that we start to see more regularity in speech perception and
production than is actually there?

1.2 Phoneme as the Fundamental Unit of Speech
So, is the phoneme merely an efficient descriptor of language behavior or is it
a
fundamental functional unit of speech perception (and production)? It appears
that across distinct theoretical divisions in speech research there is a
pretheoretical
presumption that the phoneme is indeed the fundamental
representation of speech perception.iii For example, here is a quote from some
of
the most respected researchers in the field:
"How is it that on hearing the sounds of speech a listener perceives
phonemes? Since the question is reasonable only if we assume
that phonemes are perceived�we accept it and go on to ask how
such perception might occur." (Liberman, Cooper, Studdert-
Kennedy, Harris, & Shankweiler, 1966, as quoted in Walsh, 1989).
In the decades since this quote there have been some concerns raised about the
accepted role of phonemes (e.g., Studdert-Kennedy, 1976, 1980). The phoneme
representation has also had several explicit defenders (e.g., Nearey, 1990).
However, speech perception researchers (the current authors included) too often
simply proceed as if the issue has been resolved and conduct experiments to
uncover how "a listener perceives phonemes". Most of the disparate theories of
speech perception (excepting some forms of Direct Realism) simply presume that
the end product of perception is the assignment of a discrete phonemic symbol.
This presumption is codified in typical experimental paradigms that demonstrate
effects of acoustic parameter manipulation on the forced-choice phonemic
labeling of speech sounds. In order to avoid the categorization error, empirical
evidence concerning the role of the phoneme in perception has to be brought to
bear on the issue. To accomplish this we need to set down some defining
characteristics of the phoneme that will lead to testable predictions about
perceptual data.

1.3 Definition of Phoneme

It should be made clear that the "phoneme" that we are discussing here is not
the
orhographic unit for phonemic transcriptions of linguists, but a purported
internal
mental representation that may be presumed to be a functional unit for the
speech
perceiver-producer. This is the "mentalistic" notion of the phoneme as described
by Jones (1967). It has its roots in phoneme theory back to its origins with de
Courtenay in the 1870s, who talked of "psychophonetics". The notion of the
phoneme as a psychological (as opposed) to linguistic entity is also explicit in
the
work of Sapir (1925). Thus, this viewpoint is distinguishable from Trubetzkoy's
functional view (1958) or Jones's (1967) "physicalist" notion.
Clearly, the concept of this mentalistic phoneme has gone through many
revisions since de Courtenay. We see three elements that are typical (though by
no means universal) of mentalistic notions of the phoneme. Phonemes are:
Discrete: Some continuous variation in speech sounds is quantized when a
phonemic label is assigned.
Abstract (Symbolic): The mental representation itself is a symbolic label
similar
in conception (though not necessarily similar in structure) to the phonemic
transcription labels.
Language-specific: A phoneme is meaningful only in relation to a particular
language. It is a meaningful functional unit for the idiolect of the particular
speaker.iv
These properties are in direct contrast to the acoustic signal which carries the
phonemic message from speaker to listener. The speech signal is:
Continuous: Discrete markers for phoneme boundaries have been notoriously
difficult to find. The variation in the acoustic waveform is, for practical
purposes,
continuous.
Physical: The waveform is not an abstract symbol, but a lawful product of the
movement and shape of articulators, the medium of sound travel and any
intervening objects or sources.v
Not Linguistically Marked: Obviously, the acoustic waveform is not explicitly
marked as characteristic of a particular language. In fact, it isn't even
explicitly
marked as language.
Because of this mismatch between the continuous signal and the discrete symbol,
speech perception research has been confronted with several contumacious
problems. The "problems" of signal segmentation, perceptual compensation for
coarticulation, lack of invariance, and speaker normalization are all, to some
extent, created by the desire to map a continuously varying signal on phoneme
quanta. Could these long-standing problems be simply a consequence of our
pretheoretical
assumptions?vi
To justify the large amount of work going into solving these
aforementioned difficulties, it seems incumbent upon us to empirically validate
the existence of phonemes or derive their necessity from first principles. We
understand that there have been efforts to provide empirical evidence about the
ontological status of the phoneme. Much of this previous work is based on
patterns of responses in production and perception data (e.g. speech errors,
Fromkin, 1971). These data can be quite compelling, but still we are left with
the
possibility of a categorization error.
We present below some recent empirical data that we feel are relevant to
the debate. The questions we pose are these: What role does the phoneme play in
speech perception? What are the fingerprints left behind by this purported
fundamental functional unit of speech perception?vii Can we display the "causal
efficacy" of phonemic identity in a non-circular manner?
2. Empirical Evidence
2.1 Causal Efficacy
How do we decide on the existence of a proposed entity? One important notion
may be what is termed "causal efficacy" (e.g., Gasking, 1955). This is the
ability
of an entity to cause measurable effects by its presence or absence. It is the
hallmark of the standard empirical approach that we measure the effects caused
by an entity in order to discover something about the entity (e.g. if it
exists).
An example is the search for the existence of neutrinos. These nearly
massless chargeless particles were hypothesized to exist by Enrico Fermi.
Unfortunately, the particles have very weak interactions with other particles.
Thus, it was hard to detect the effects of the neutrino. That is, it was
difficult to
demonstrate the causal efficacy of neutrinos. The existence of the neutrino was
finally established because antineutrinosviii initiate a particular reaction.
Their
presence or absence determines the reaction.
Can we empirically establish the causal efficacy of phonemes? One may
suggest that we have evidence for phonemes in the fact that human adults can
label speech sounds with appropriate phonemic labels. Unfortunately, besides
making us vulnerable to a categorization error, this demonstration is a bit
circular.
We define phonemes by the perceptual behavior of adult humans. We cannot use
that same data as proof that they exist. Instead we need to show that the
hypothesized presence of a particular phoneme has demonstrable effects on
behavior that do not occur in its absence.
2.2 Context Effects
One place where we may look to see the causal efficacy of the phoneme is in
context effects. That is, can the change of phonemic identity (conceptualized as
the activation of a particular mental representation) affect the response made
to a
nearby speech sound?
We do see such effects. For example, Mann (1980) reported that an
ambiguous consonant-vowel (CV) syllable is labeled as `ga' following /al/ and as
`da' following /ar/. This appears to be a causal effect of a phoneme. The
presence of /l/ appears to cause a change in the response to the subsequent CV.
It
may be proposed that the auditory signal corresponding to the initial VC
activated
the representation for the phoneme /l/. This activated representation in turn
affected the label assigned to the CV. Proposing this from the given data would
be an example of the categorization error, however, since we would be presuming
that the /l/ phoneme was present in the mind because a phonetic structure was
present in the description of the signal. In order to make this explanation
tenable,
we need some independent evidence that it was the presence of the mental
phoneme /l/, per se, that caused the response shift and not the structure of the
input signal itself.
Mann (1986) presented these same VC CV disyllables to native Japanese
speakers. These listeners could not distinguish between English /l/ and /r/ and,
therefore, it is unlikely that they would have the mental representations for
these
phonemes. If we are to ascribe causal efficacy to phonemes based on this context
effect then the effect must disappear in the absence of these particular
phonemes
in Japanese listenersix. Contrary to this prediction, Japanese listeners shifted
their
responses to the CV in the same manner as English speakers. Even the size of the
effect (in terms of identification boundary shift) did not differ between
groups.
This demonstration not only fails to provide supportive evidence for the causal
efficacy of phonemes; it also provides some disquieting information about the
possible role of the phoneme as a fundamental functional unit. In Japanese,
there
is no phonemic distinction (in the descriptive sense) between English [r] and
[l]
and both are subsumed under the Japanese /r/. If there are mental phoneme
representations then the Japanese listeners in Mann's (1986) study most likely
represented [al] and [ar] as /ar/ and /ar/. The fact that these equivalent
labels led
to widely disparate effects leads one to question what fundamental role phonemes
play in speech perception. It appears that the context effects are in no way
affected by phonemic identity.
Additional evidence concerning the impotence of phonemic labels in this
context effect was presented by Lotto, Kluender, and Holt (1997). We trained
Japanese quail (Coturnix coturnix japonica) to peck to a key when presented with
the syllable [ga] and to refrain from pecking when presented with the syllable
[da].x In a sense, they learned to label the syllables by their pecking
responses.
After training, they were presented ambiguous CVs preceded by either [al] or
[ar].
They demonstrated a shift in "labeling" in a similar manner to human English and
Japanese listeners. That is, they pecked more (a "ga" response) to CVs in the
presence of [al] than they did to CVs in the presence of [ar]. Since birds are
unlikely owners of mental phoneme representations for English /l/ and /r/, this
is
another example of the context effect occurring in the absence of phoneme
representations. Taken together with Mann's (1986) data for Japanese listeners,
these data undermine the use of context effects to demonstrate the causal
efficacy
of phonemes.
Even if we could validate the existence of mental phoneme
representations, it would be difficult to claim that phonemes are the
fundamental
functional units of speech perception. As the studies described above
demonstrate, sub-phonemic shifts in acoustic attributes of a stimulus can cause
dramatic context effects on perception of subsequent speech. This suggests that
there is a unit more fundamental to the processes of speech perception than the
phoneme. A natural, and historically popular, candidate for this unit is the
distinctive feature. However, distinctive features will not suffice to explain
the
parallels in human and quail behavior.
It appears that the causal level for context effects in speech perception
may be more general and not language-specific. In a recent study, we looked at
the effects of non-speech context on the labeling of CVs. A series of syllables
varying from [ga] to [da] were synthesized by manipulating the onset frequency
of the third formant. These syllables were preceded by either synthesized
versions of [al] or [ar] or non-speech complexes that matched some of the
acoustic attributes of [al] and [ar]. In particular, the non-speech stimuli were
two
flat sine waves at the offset frequencies of the third and second formants for
[al]
or [ar]. These sine complexes were matched in duration and energy to [al] and
[ar]. A 50-msec silent gap separated the CVs and the context sounds. The
listeners' task was to label the CVs as "da" or "ga" using keys on a response
box.
Figure 1 presents the identification data for the speech and non-speech contexts
as
mean percent `ga' responses.
It is clear from the graph that the non-speech context also elicited a shift in
the labeling behavior for the CV. In fact, the identifications in the non-speech
context were statistically indistinguishable from the speech context. The size
of
the shift in responses did not statistically differ with context condition
(F9,117 <1).
These data suggest that general auditory attributes of the speech signals are
causing the context effects and not any linguistic-specific representation. Over
the last couple of years, we have demonstrated a variety of similar non-speech
context effects on speech labeling (Holt, Lotto, & Kluender, 1996, 1998; Lotto &
Kluender, 1998). We believe that these data require one to consider auditory
information as fundamental to speech perception when discussing functional
0
10
20
30
40
50
60
70
80
Speech Non-Speech
Condition
Mean % "ga" Responses
/al/
/ar/
Figure 1. Context effects for speech and non-speech precursors.
units. The importance of general auditory attributes in any discussion of speech
perception (or production) is also expressed convincingly by Ohala and Perkell
(both from this session).
2.3 Boundaries of Causal Efficacy
Another way of talking about an "object" of speech perception is to define the
boundaries of causal efficacy. That is, we can try to empirically determine what
parts of the speech signal act together as a cohesive unit in inducing
behaviorial
changes. This is basically a problem of functional segmentation. Do these
empirically determined boundaries correspond to phonemes?
An experiment reported by Lotto, Kluender, and Green (1996) was
designed specifically to determine the boundaries of casual efficacy. The
experiment was based on previous research demonstrating that the identification
of a syllable-initial consonant can be shifted on the basis of the duration of
the
subsequent vowel (Miller & Liberman, 1979). This effect has been considered
indicative of perceptual normalization for changes in speaking rate. So, for
example, the identification boundary between /bi/ and /pi/ will shift depending
on
the duration of [i], with a longer vowel leading to more /bi/ responses (Green &
Miller, 1985).
Of course, it would be a categorization error to simply assume that the
object whose duration is important for this effect corresponds to the phoneme
/i/
or to any specific linguistic entity. Simply because it is useful to label the
stimuli
as /bi/ or /pi/ does not mean that the underlying perceptual representation will
respect these labels. Lotto et al. (1996) tried to determine if it was the
duration of
phonemes that mattered in this "rate-normalization" effect or whether it was the
duration of some other perceptual unit. The experiment included four series
containing synthesized CV syllables, where the consonant varied from [b] to [p]
through the manipulation of voice-onset time (or more correctly, first formant
cutback, Kluender, 1991). One series (the Long series) included CVs with a long
[i] vowel. The total syllable duration was 350 msec. A second series' (the Short
series) CVs contained short [i] vowels with a total syllable duration of 110
msec.
The final two series contained CVs that were 350 msec in total duration (the
duration of the Long CVs), but contained a sudden disjunction in the vowel after
110 msec (the duration of Short CVs). One disjunction (ID-Shift) was slight in
spectral change but crossed the identification boundary causing the phonemic
identity of the vowel to shift from /i/ to /&#921;/ (as determined by a separate
vowel
identification task with the same listeners). The second disjunction (Spectral-
Shift) was large in terms of spectral change, but did not result in a shift in
phonemic identity. The vowel shifted from a relatively neutral exemplar of /i/
to
a more extreme (in the F1xF2xF3 frequency space) exemplar of /i/.
The predictions are straightforward (even if the stimuli are not). If
phonemes define the boundaries of a perceptual unit for the calculation of
duration in the rate-normalization effect, then the ID-Shift series should be
identified similarly to the Short series because the effective duration of the
following vowel will terminate at the phonemic shift disjunction. On the other
hand, the Spectral�Shift series identification function should more closely
resemble the Long series function because the spectral disjunction can be
ignored
with no accompanying shift in mental phonemic label. That is, the disjunction in
the ID-Shift CVs should define two perceptual functional units (phonemes) and
the Spectral-Shift disjunction should be integrated into one functional unit.
The results listed as identification boundariesxi in terms of msec of voice
onset time are listed below in Table I.
Table I. Probit boundary values and significance test results from Experiment 2
of Lotto, Kluender and Green (1996).
Long Short ID-Shift Spectral-Shift
39.28 msec
Difference:
Short
Spectral-Shift
No Difference:
ID-Shift
34.63 msec
Difference:
Long
ID-Shift
No Difference
Spectral-Shift
38.99 msec
Difference:
Short
Spectral-Shift
No Difference:
Long
33.27 msec
Difference:
Long
ID-Shift
No Difference:
Short
Note: Statistical differences were determined using Tukey least significant
difference (&#945;=.05).
The data were in direct contrast to the predictions arising from a
hypothesis of the phoneme as the fundamental functional unit. The identification
of the ID-Shift series was statistically indistinguishable from the
identification of
the Long series. On the other hand the identification function for the Spectral-
Shift series was statistically equivalent to the function for the Short series.
This
pattern of results suggests that, for the "rate-normalization" effect, spectral
discontinuities define the extent of the duration effect and that phoneme
boundary
play little demonstrable role. Acoustic information that could be described as
separate phonemes (ID-Shift) was effectively treated as a coherent perceptual
unit.
In addition to these data, Rochelle Newman (this session) presents
compelling data from experiments using the rate-normalization paradigm that also
demonstrate that spectral changes appear to be of paramount importance in
determining perceptual boundaries and that phonemes, per se, do little to
predict
the identification shifts.
2.4 Intra-phonemic Variation is Meaningful
The results from the previous two sections demonstrate a surprising dearth of
response predictability associated with phonemic identity. At the least, these
experiments weaken the claim that the phoneme is "the fundamental functional
unit of speech perception". However, one may argue that the phoneme is not a
functional unit of the process of speech perception, but that phonemes are the
output of speech perception and are fundamental functional units of language for
lexical access and "higher" linguistic processes. This appears to be the view
espoused by Liberman et al. (1966) quoted above. They assume that "phonemes
are perceived". This also appears to be the underlying assumption of research
using forced-choice identification methodology in which the response of concern
is the phonemic labeling of a speech sound. This methodological assumption has
assured the hegemony of the phoneme in speech perception research. Likewise,
there are many computational models of speech perception that output a single
phonemic label when provided with a speech signal. For example, Anderson,
Silverstein, Ritz and Jones (1977) present the "Brain-State-in-a-Box Model" of
speech categorization which gets rid of any differences of responses within a
phoneme and increases differences between phonemes.
The "Brain-State-in-a-Box Model" is an extreme example of one of the
main difficulties with the phoneme-as-output approach. Mapping the continuous
speech signal to a discrete symbolic unit necessitates some loss of information.
There is non-random variation in the speech signal that does not differentiate
phonemes. Recent studies have made evident that this variation has meaningful
consequences on response behavior -- even for behavior related to "higher"
linguistic functions. For example, different acoustic stimuli that are
identified as
the same phoneme can easily be rated by listeners in terms of "goodness" as
exemplars of the phoneme (e.g. Miller & Volaitis, 1989; Volaitis & Miller,
1992).
This demonstrates the ability of listeners to respond on the basis of
intraphonemic
variation which models of speech perception that output phoneme
labels like the one mentioned above can not readily do. In addition, exemplars
from a single phoneme differ in their effectiveness as adapters or as initiators
of
identification shifts (e.g. Samuel, 1982; Miller, Connine, Schermer, & Kluender,
1983; Lotto, Kluender, & Holt, 1998). Also, there have been many recent
demonstrations of speaker effects on phonemic identification and lexical access
(e.g. Schacter & Church, 1992; Nygaard, Sommers, & Pisoni, 1994). Goldinger
(1996) shows that the view that non-phonemic variation in the auditory signal is
discarded for the more efficient abstract phonemic representation is almost
certainly false.
3. Conclusions and Future Directions
The data presented in the last section are but a small proportion of recent work
showing the causal efficacy of variations within a single phoneme. However,
suggesting that variation within a phoneme is meaningful is not the same as
suggesting that mental phonemes do not exist. One may argue that it is possible
that the phoneme is but one of several representations of the speech signal; all
present in parallel with each other. Yes! But this view of phonemes as but one
of
many possible streams of information in the system forces some changes in our
viewpoint.
3.1 The Phoneme is Not Fundamental
The question of this session was to describe the "fundamental", "basic" or
"primitive" units of speech processing. If we view mental phonemes as but one
of several representations of speech signals then it is hard to think of them as
"fundamental", especially since, as we argue below, some of the other
representations are sub-phonemic.
3.2 Basic Functional Units are Task-Specific
The purpose of multiple streams of information must be to allow flexibility with
changes in task. We would like to suggest that future research must analyze the
experimental tasks to determine what type of information is necessary to perform
the task. In a task such as phonemic transcription, some phoneme representation
may be functionally essential.
However, let us not presume that all language tasks require phoneme
representations, just because people can do phonemic identifications. McMurray
and Spivey (this session) presented data on the time course for recognition of a
syllable as /ba/ or /pa/. This time course is too slow to account for the speed
of
word recognition. That is, if recognition of the phonemes in auditory speech was
necessary for word recognition then reaction times for word recognition would be
much slower. Also note that in the connectionist network model presented by
McMurray and Spivey that accounts for their data, there was an all-or-none final
decision on the identity of the syllable. However, if there were "word nodes"
connected to the output of the phoneme nodes, then there would not need to be
full phoneme recognition to recognize the word. The flow of information could
determine the word spoken without all-or-none decisions at the phonemic level.
Again, the units that are output or are "basic" are determined by the particular
task.
As we continue to study speech through forced-choice identification and
discrimination paradigms, we should be cognizant of the fact that the
representational level we are studying in these tasks may be one of several
parallel representations that, perhaps, are not essential for the next "level"
of
perceptual or linguistic tasks.
3.3 Auditory Level is Important for Explaining Language Behavior
As stated earlier, we believe that the evidence supports the notion that some
general auditory representation of the speech signal is significant for
explaining
language behaviors and the development of language as a communication system.
This view was also presented by Perkell and Ohala (both from this session).
These general auditory representations of the effectively continuousxii auditory
signal may be present in parallel with phonemic representations along with a
number of other possible linguistic entities.
4. Summary
So, what is the basic unit of speech perception? That depends on what your
definition of is is. Our (probably overstated) conclusions are that the answer
is
dependent on the particular task for the subject. Our nervous system probably
retains much of the information in the speech signal until a response is
determined. We suggest that hypothesized representations for language be
determined through empirical work and by the application of rational first
principles. In this way we can avoid the categorization error and determine
exactly which "problems" of speech perception need to be solved.
Notes
i This isn't to propose that there are no mental prototypes or that those who
propose prototype
theories are committing a grave error. In fact, there are many strong reasons to
suppose that some
kind of prototype theory of categorization is true. We are merely asserting that
prototypes in
response structure are not prima facie evidence for the existence of mental
prototypes.
ii Another example that is relevant but not concerned with behavioral data is
the adaptiveness that
biological species show in relation to their ecological niche. The wonderful
orderliness and
harmony of animal and environment belies the randomness and capriciousness of
the process of
natural selection. While the current equilibrium of organism and environment may
appear to be a
clear picture of "God's plan" in nature, it does not necessarily follow that
this seeming order
resulted from processes exemplified by such order.
iii We focus here specifically on the question of speech perception. We believe
that many of the
same arguments apply to speech production, but this is a separate question.
iv This property isn't as obviously true if one presumes that there is a
universal set of innately
prescribed phonemes. In this case, there would exist mental phoneme
representations that were
independent of an infant's future "native" language. Given the difficulty that
linguists have in
defining phonemes in relation to a single speaker's "language", it may be
difficult to give precise
definitions for these kinds of innate representations.
v There are some who argue that mental representations gain meaning only to the
extent that they
are lawfully determined by physical sources (e.g. Dretske, 1981; 1986). The
distinction made here
remains true to the degree that the phoneme is not a representation of the
physical properties of the
acoustic waveform.
vi It should be noted that to the extent that distinctive feature systems are
discrete, symbolic
linguistic systems, they are subject to the same problems discussed here for the
phoneme.
vii Perhaps indicative of a Freudian slip, this sentence was originally written
as "What are the
fingerprints left behind by this purported fundamental fictional unit of speech
perception?"
viii The causal efficacy of antineutrinos is presumed to demonstrate the
existence of neutrinos,
given the surrounding theory of particle physics.
ix This is a rather simplistic notion of determining causality and the
philosophy of science has
much more rigorous standards than those presented here. However, we think that
these arguments
appeal to our common sense notions of causality.
x Some birds were trained to peck to [da] and refrain from pecking to [ga].
These birds pecked
more in the presence of [ar].
xi Probit analysis is a class of nonlinear models of estimation for binary
decision variables. The
numbers represent the 50% identification boundary as determined from a linear
regression taken
through z-transformed scores.
xii Of course, the auditory representation can't be purely continuous. The
signal is transmitted
discretely even at the level of neural transduction. However, at the level of
perception we are
suggesting that the representations can be considered functionally continuous.
References
Anderson, J. A., Silverstein, J. W., Ritz, S. A., & Jones, R. S. 1977.
Distinctive Features,
Categorical Perception, and Probability Learning: Some Applications of a Neural
Model.
Psychological Review, 84: 413-451.
Ben-Zeev, A. 1987. A Critique of the Inferential Paradigm in Perception. Journal
for the Theory of
Social Behavior, 17: 243-263.
Brooks, L. 1978. Non-analytical Concept Formation and Memory for Instances. In
E. Rosch & B.
Lloyd (Eds.), Cognition and Categorization. Hillsdale, NJ: Erlbaum.
Dretske, F. 1981. Knowledge and the Flow of Information. Cambridge, MA: MIT
Press.
______ 1986. Misrepresentation. In R. J. Bogdan (Ed.), Belief: Form, Content and
Function.
Oxford: Oxford University Press.
Fromkin, V. A. 1971. The Nonanomalous Nature of Anomalous Utterances. Language,
47: 27-52.
Gasking, D. 1955. Causation and Recipes. Mind, 64: 479-487.
Goldinger, S. 1996. Words and Voices: Episodic Traces in Spoken Word
Identification and
Recognition Memory. Journal of Experimental Psychology: Learning, Memory, and
Cognition, 22: 1166-1183.
Green, K. P., & Miller, J. L. 1985. On the Role of Visual Rate Information in
Phonetic Perception.
Perception & Psychophysics, 38, 269-276.
Heil, J. 1983. Perception and Cognition. Berkeley: University of California
Press.
Holt, L. L., Lotto, A. J., & Kluender, K. R. 1996. Perceptual Compensation for
Vowel
Undershoot May be Explained by General Perceptual Principles. Journal of the
Acoustical Society of America, 99:2589.
Holt, L. L., Lotto, A. J., & Kluender, K. R. 1998. Spectral Contrast in
Perception of VCV
Syllables. Journal of the Acoustical Society of America, 104, 1759.
Jones, D. J. 1967. The Phoneme. Cambridge: Cambridge University Press.
Kluender, K. R., Lotto, A. J., Holt, L. L., & Bloedel, S. L. 1998. Role of
Experience for
Language-Specific Functional Mappings of Vowel Sounds. Journal of the Acoustical
Society of America, 104: 3568-3582.
Knapp, A. G., & Anderson, J. A. 1984. Theory of Categorization Based on
Distributed Memory
Storage. Journal of Experimental Psychology, 10: 616-637.
Koffka, K. 1935. Principles of Gestalt Psychology. New York: Harcourt Brace.
K�hler, W. 1930. Gestalt Psychology. London: G. Bell & Sons.
Liberman, A. M., Cooper, F. S., Studdert-Kennedy, M., Harris, K. S., &
Shankweiler, D. P. 1966.
Some Observations on the Efficacy of Speech Sounds. Paper presented at the XVIII
International Congress of Psychology, Moscow, USSR.
Lotto, A. J., Kluender, K. R., & Holt, L. L. 1998. The Perceptual Magnet Effect
Depolarized.
Journal of the Acoustical Society of America, 103: 3648-3655.
Lotto, A. J., & Kluender, K. R. 1998. General Contrast Effects in Speech
Perception: Effect of
Preceding Liquid on Stop Consonant Identification. Perception & Psychophysics,
60:
602-619.
Lotto, A. J., Kluender, K. R., & Green, K. R. 1996. Spectral Discontinuities and
the Vowel Length
Effect. Perception & Psychophysics, 58: 1005-1014.
Lotto, A. J., Kluender, K. R., & Holt, L. L. 1997. Perceptual Compensation for
Coarticulation by
Japanese Quail (Coturnix coturnix japonica). Journal of the Acoustical Society
of
America, 102: 1134-1140.
Mann, V. A. 1980. Influence of Preceding Liquid on Stop-Consonant Perception.
Perception &
Psychophysics, 28: 407-412.
______ 1986. Distinguishing Universal and Language-Dependent Levels of Speech
Perception:
Evidence from Japanese Listeners' Perception of English "l" and "r". Cognition,
24: 169-
196.
McMurray, B., & Spivey, M. 1999. Categorical Perception of Consonants: The
Interaction of
Learning and Processing. Presentation at the 35th Regional Meeting of the
Chicago
Linguistic Society, April 23rd, University of Chicago.
Medin, D. L., & Schaffer, M. M. 1978. A Context Theory of Classification
Learning.
Psychological Review, 85: 207-238.
Miller, J. L., Connine, C. M., Schermer, T. M., & Kluender, K. R. 1983. A
Possible Auditory
Basis for Internal Structure of Phonetic Categories. Journal of the Acoustical
Society of
America, 73: 2124-2133.
Miller, J. L., & Liberman, A. M. 1979. Some Effects of Later-Occurring
Information on the
Perception of Stop Consonant and Semivowel. Perception & Psychophysics, 25: 457-
465.
Miller, J. L., & Volaitis, L. E. 1989. Effect of Speaking Rate on the Perceptual
Structure of a
Phonetic Category. Perception & Psychophysics, 46: 505-512.
Nelson, K. 1974. Concept, Word, and Sentence: Interrelations in Acquisition and
Development.
Psychological Review, 81: 248-267.
Newman, R. 1999. The Use of Rate Normalization to Investigate the Basic Units of
Speech.
Presentation at the 35th Regional Meeting of the Chicago Linguistic Society,
April 23rd,
University of Chicago.
Nygaard, L. C., Sommers, M. S., & Pisoni, D. B. 1994. Speech Perception as a
Talker-Contingent
Process. Psychological Science, 5: 42-46.
Ohala, J. J. 1999. Acoustic-Auditory Aspects of Speech. Presentation at the 35th
Regional Meeting
of the Chicago Linguistic Society, April 23rd, University of Chicago.
Perkell, J. S. 1999. A Theory of Speech Motor Control with Supporting Data from
Speakers with
Normal Hearing and with Profound Hearing Loss. Presentation at the 35th Regional
Meeting of the Chicago Linguistic Society, April 23rd, University of Chicago.
Pinker, S., & Prince, A. 1988. On Language and Connectionism: Analysis of a
Parallel Distributed
Processing Model of Language Acquisition. In S. Pinker & J. Mehler (Eds.),
Connections
and Symbols. Cambridge, MA: Bradford Books/ MIT Press.
Nearey, T. M. 1990. The Segment as a Unit of Speech Perception. Journal of
Phonetics, 18: 347-
373.
Posner, M. I., & Keele, S. W. 1970. On the Genesis of Abstract Ideas. Journal of
Experimental
Psychology, 83: 28-38.
Reed, S. K. 1972. Pattern Recognition and Categorization. Cognitive Psychology,
3: 383-407.
Rosch, E. H. 1975. Cognitive Representations of Semantic Categories. Journal of
Experimental
Psychology, 3: 193-233.
______ 1978. Principles of Categorization. In E. Rosch & B. Lloyd (Eds.),
Cognition and
Categorization. Hillsdale, NJ: Erlbaum.
Rumelhart, D. E., & McClelland, J. L. 1986. On Learning the Past Tenses of
English Verbs. In J.
L. McClelland, D. E. Rumelhart, & the PDP Research Group, Parallel Distributed
Processing: Explorations in the Microstructure of Cognition. Volume 2:
Psychological
and Biological Models. Cambridge, MA: Bradford Books/ MIT Press.
Samuel, A. G. 1982. Phonetic Prototypes. Perception & Psychophysics, 31:
307-314.
Sapir, E. 1925. Sound Patterns in Language. Language, 1: 37-51.
Schacter, D., & Church, B. 1992. Auditory Priming: Implicit and Explicit Memory
for Words and
Voices. Journal of Experimental Psychology: Learning, Memory, and Cognition, 18:
915-
930.
Studdert-Kennedy, M. 1976. Speech Perception. In N. J. Lass (Ed.), Contemporary
Issues in
Experimental Phonetics. New York.
______ 1980. Speech Perception. Language and Speech, 23: 45-65.
Trubetzkoy, N. S. 1958. Grundz�ge der Phonologie (3rd Ed.). G�ttingen:
Vandenhoek & Ruprecht.
Volaitis, L. E., & Miller, J. L. 1992. Phonetic Prototypes: Influence of Place
of Articulation and
Speaking Rate on the Internal Structure of Voicing Categories. Journal of the
Acoustical
Society of America, 92: 723-735.
Walsh, M. A. 1989. On the Nature of Functional Linguistic Units: The Influence
of Written
Language on Spoken Language Forms. Unpublished dissertation, University of Texas
at
Austin.
Wittgenstein, L. 1953. Philosophical Investigations. New York: MacMilliam.

Ahh Gentilucci hooked up with Corballis. So one of the 'mirror neuron'people is
working with Corballis, one of the 'gestural language' people.

Thanks for the link to that article. It looks like excellent reading.

You might find this book interesting. If only we could get some important people
in ELT to READ it--and Corballis and mirror neuron research. You see, the
phoneme was supposed to be some sort of cognitively significant representation,
a unit that created 'mental phonetic strings' that built up into 'mental lexical
strings'. However, research does not support that the structuralist phoneme
(even in its various conceptions, including features) is the right sort of unit
for representing language at a sub-lexical level in language processing and
cognition. It turns out, rather, to be a great idealization for putting a spoken
language into an alphabetic form (although note how English and French, for
example, for the most part blitheley ignore phonemic principles in favor of
lexical ones).


http://www.amazon.co.uk/exec/obidos/ASIN/0198299842/ref=ord_cart_shr?_encoding=U\
TF8&m=A3P5ROKL5A1OLE

Phonetics, Phonology, and Cognition explores the cognitive and biological
systems involved in speech. It offers challenging findings on the cognitive
status of phonological representations and their relationship with phonetic
implementations. The book's authors are leading researchers from linguistics and
cognitive science. They consider, for example, the type of control required for
the implementation of articulatory events and the nature of the loop between
auditory and articulatory mechanisms. They show the advances that have already
been made on these and other central issues in psycho- and neuro-linguistics now
that the operations of the brain can be studied directly and neuroscience is no
longer dependent on post-mortem dissection of speech-impaired patients. In its
exploration of the mental and physical representation of sound systems,
Phonology, Phonetics and Cognition demonstrates the value of phonology in
allowing the integration of phonetics and cognition. Its authors are concerned
with both the realization of representations in physical structures and the way
that linguistic sound structure is linked to language form and mental coding. In
sum, this book provides a revealing cross-disciplinary perspective on language,
speech, and cognition which will be of value and interest to linguists,
cognitive scientists, and speech pathologists.

It is also interesting to note how the phenomenological philosopher,
Merleau-Ponty anticipates theoretically and conceptually these views of
language. This goes back to his masterwork, 'Phenomenology of Perception' and on
to work much less well known in philosophy, such as:


http://books.google.com/books?id=Kuk7LDmnXQEC&dq=merleau+ponty+body+and+language\
&source=gbs_navlinks_s

I think the book from OUP is edited by those two American academics and is a
collection of papers, but I might be wrong. They are certainly not the FIRST to
do what the description says. I can't imagine them not having Corballis, among
others, in their bibliography in a book out in 2007.

At any rate, some interesting material:

http://homepages.nyu.edu/~ag63/DEC-cours-active/gafosNLLT02.pdf

A GRAMMAR OF GESTURAL COORDINATION
ABSTRACT. Linguistic form is expressed in space, as articulators effect
constrictions
at various points in the vocal tract, but also in time, as articulators move. A
rather widespread
assumption in theories of phonology and phonetics is that the temporal dimension
of speech is largely irrelevant to the description and explanation of the
higher-level or more
qualitative aspects of sound patterns. The argument is presented that any theory
of phonology
must include a notion of temporal coordination of gestures. Linguistic grammars
are
constructed in part out of this temporal substance. Language-particular sound
patterns are
in part patterns of temporal coordination among gestures.1
1. INTRODUCTION

http://ukcatalogue.oup.com/product/9780195163483.do

     * The first comprehensive attempt to trace the origin of grammar to gesture
     * Authors make use of evidence from human fossil records to support their
claims

In The Gestural Origin of Language, Wilcox and Armstrong use evidence from and
about sign languages to explore the origins of language as we know it today.
According to their model, it is sign, not spoken languages, that is the original
mode of human communication.

The authors demonstrate that modern language is derived from practical actions
and gestures that were increasingly recognised as having the potential to
represent and hence to communicate. In other words, the fundamental ability that
allows us to use language is our ability to use pictures of icons, rather than
linguistic symbols. Evidence from the human fossil record supports the authors'
claim by showing that we were anatomically able to produce gestures and signs
before we were able to speak fluently. Although speech evolved later as a
secondary linguistic communication device that eventually replaced sign language
as the primary mode of communication, speech has never entirely replaced signs
and gestures.

Readership: This volume will be an invaluable resource to students and
professionals in psychology, linguistics, and philosophy.

1: Grasping Language: Sign and the Evolution of Language
2: Language in the Wild: Paleontological and Primatological Evidence for
Gestural Origins
3: Gesture, Sign, and Speech
4: Gesture, Sign, and Grammar: The Ritualization of Language
5: Conceptual Spaces and Embodied Actions
6: The Gesture-Language Interface
7: Invention of Visual Languages


http://www.americanscientist.org/issues/feature/1999/2/the-gestural-origins-of-l\
anguage/1


FEATURE ARTICLE
The Gestural Origins of Language

Human language may have evolved from manual gestures, which survive today as a
"behavioral fossil" coupled to speech

Michael Corballis

In 1934 the behavioral psychologist B. F. Skinner found himself seated at the
dinner table with the eminent philosopher Alfred North Whitehead, and proceeded
to explain to Whitehead what behaviorism was all about. Obliged to offer a
challenge, Whitehead uttered the sentence "No black scorpion is falling upon
this table" and then asked Skinner to explain why he might have said that.
Skinner attempted a reply more than 20 years later in an appendix to his 1957
book Verbal Behavior. He proposed that Whitehead was unconsciously expressing a
fear of behaviorism, likening it to a black scorpion that he would not allow to
intrude into his philosophy. The skeptical reader may be forgiven for concluding
that the reply owed more to psychoanalysis than to behavioral principles.

Be that as it may, Whitehead had articulated one of the properties of language
that seems to distinguish it from all other forms of communication, its
generativity. Whereas other forms of communication among animals seem to be
limited to a relatively small number of signals, and restricted to limited
contexts, there is essentially no limit to the number of ideas or propositions
that we can convey using sentences. We can immediately understand sentences made
up of words that we have never heard in combination before, as Whitehead's
sentence illustrates. Language also allows us to escape from the immediate
present and to refer to events in other places and at other times. We can use
language to fantasize, to describe events that have never existed and never
will. This remarkable flexibility is achieved at least in part through the human
invention of grammar, a recursive set of rules that allows us to generate
sentences of any desired complexity. The eminent linguist Noam Chomsky has
attributed this to a unique human endowment that he calls universal grammar. All
human languages, he suggests, are variants on this fundamental endowment.

-----

How are we to reconcile these alternative perspectives? At least a partial
answer is that language emerged not from vocalization, but from manual gestures,
and switched to a vocal mode relatively recently in hominid evolution, perhaps
with the emergence of H. sapiens. This idea was suggested by the 17th-century
French philosopher �tienne Condillac and revived in the 1970s by the American
anthropologist Gordon W. Hewes. It has not found general favor among linguists
or anthropologists, perhaps because it lacks parsimony and because there is no
direct evidence that any of our hominid ancestors gestured rather than spoke.
Even so, argument in its favor has continued to grow.

--- In iatefl_pronsig@..., "literacyacrosscultures" <jannuzi@...>
wrote:
>
>>What signals syllables in speech has more to do with 'timing' onsets and clear
passage into the syllable 'nucleus', which is dominated by 'vowel qualities' and
features rise in pitch and greater stress.

----

Oops. Finished the sentence but not the thought. Rise in pitch and greater
stress falls on some syllables. Not all.

Which brings me back to the issues of 'devoicing' and
'neutralizing/unstressing'. English will unstress some vowels so much as to drop
them out of speech. I would bet the phonological issues here can be related to
devoicing in Japanese. See my earlier discussion on neutral sounds in English
and Japanese. Neutrality in English is like a part of the mouth Japanese
speakers of English don't know about. It tends to fall around the 'high front'
vowels of E /i/, /i:/ (I mean at the end of a word, like 'velocity', otherwise
it's paradoxical to think of it being lengthened). And the zone extends over to
the mid-to-high central 'schwa'.

Japanese neutral is found at high-front and then skips over to high-back ( e.g.,
J /u/). These are the vowels that 'devoice'.

CJ

>>Of course syllable breaks are always controversial (what we need for writing
and what we need for speech are two different things--speech doesn't need
clearly defined syllable breaks, syllables overlap somewhat).<<

I should add: the word as I romanized it for the discussion reflects the WRITTEN
form of the word in Japanese. Japanese would use three syllabic characters to
write the word. However, speech doesn't need to break up syllables the way a
syllable-based or alphabet-based writing system does. Speech perception research
indicates that syllables don't segment discretely (nor do 'sounds' or 'sound
segments'). Spoken syllables overlap--indeed, sub-syllabic elements can spread,
in terms of articulation, across whole words and word boundaries even. What
signals syllables in speech has more to do with 'timing' onsets and clear
passage into the syllable 'nucleus', which is dominated by 'vowel qualities' and
features rise in pitch and greater stress.

CJ

>> By that I mean, it gives more weight to the issue of seeing when trying to
articulate new sounds. <<

And bodily experiencing language. In EFL in Japan, the assumption is almost like
students are supposed to be 'bar code scanners' of text, which is why they don't
pick up much depth or feel for spoken language or even written language that has
the potential to be spoken. Yes, I agree that using written text might provide
some 'visual reinforcement' to learning a second or foreign language--to help
replace all that other reinforcement that is lost from the time we acquired a
native language and then moved into school-based and social literacy. Our lives
evolve around forms of literacy, so our efforts at teaching and learning a SL or
FL reflect this (but unfortunately, at the expense of how languages evolved to
communicate and how they evolved to be acquired).

So when we acquire a language (be it 'native' or 'second'), we have to hear, see
and experience it with our own bodies, including our 'vocal tract' and our
gestural abilities (which, I argue, incorporate our vocal tract or at least
significantly overlap with it).

So add kinesthetic to the audible and visual.

CJ

Correcting my recollection. Some very active authors on the topic of 'teaching
EFL to Japanese learners' cited the work of Cutler.

CJ

--- In iatefl_pronsig@..., "literacyacrosscultures"
> I remember Cutler's work from back in the 90s--I think. There were even
special publications about 'teaching EFL to the Japanese learner' with her name
on them, if I'm remembering correctly.
>

I took a quick look at the Cutler papers. I'm not really sure what the authors
mean by 'segmentation'. It seems they are working with syllables but referring
to everything as if the epistemelogical assumption was the 'structuralist
segment'. I'll have to look at this stuff more closely.

I think vowel 'devoicing' is one of those things that got some emphasis in the
'Japanese language is unique' school of thought. This has even been used to
justify why Japanese don't learn English, etc.
It doesn't take an empirical study to show that in some cases 'devoicing' simply
gives Japanese consonant clusters (even though the literacy-based linguistics of
Japanese say Japanese lacks consonant clusters). The point came home very much
to me when I was trying to say the Japanese word for 'ragweed' (one of my
allergy banes). The word is 'butakusa'. If you say it, bu-ta-ku-sa no one will
understand. If you say 'bu-taks-a, they will (if they know the weed). Of course
syllable breaks are always controversial (what we need for writing and what we
need for speech are two different things--speech doesn't need clearly defined
syllable breaks, syllables overlap somewhat). Still, either you get a consonant
cluster [-ks] or you get a [-k] coda to the 'tak-' syllable and the [s-] starts
(the onset) of the final syllable. Either way, the inherited rules of Japanese
have been violated, which means the rules, as usual, are nonsense.

I love the Corballis book. He starts with a rather old but usually dismissed
theory of the origin of language--the bow-wow theory--and builds up a case that
the ability to communicate linguistically evolved and developed gesturally and
then migrated to quite an extent over to the 'phonetic' side of things.
Unfortunately, Corballis is weak on phonetics and phonology--or was when he
wrote that book. Phonology has put forward some very much gesture-based theories
and models of controlled speech, and one phonologist out of Europe (will try to
remember names in more discussion) has put forward a gestural basis for all
grammaticality of language, including, of course, phonology.

I would argue that ALL controlled speech used to communicate linguistically is
to quite an extent gestural. Which is why I produced that paper on a nuanced
version of the 'articulatory gesture'. That is, the basic unit is the 'visually
salient articulatory gesture'.

It's also interesting to note that fluent users of sign languages use their
mouths and faces extensively--and in closely coordinated complexes of actions
with their other 'articulators', their arms and hands and upper body.

CJ

PS: I think when mirror neuron research converges on native and SLA, there are
going to be some real bombshells. But are most of the dominant players in the
field of SLA open to such a huge shift in their thinking?

Two more papers you might want to refer to:

http://www.ulb.ac.be/philo/phonolab/r-atics2/abstracts%20pdf/Labrune.pdf

http://ocs.sfu.ca/fedcan/index.php/cla/acl-cla2008/paper/viewFile/155/99

------------

Additional comment: I think that the beauty of a gestural approach would be the
ability to find what is common across all that reported phonetic variation
(which could mean articulatory, acoustic, allophonic-contextual,
allophonic-dialectal, allophonic-intradialectal, etc. etc.). I'm willing to bet
that if we did audio-visual analysis of Japanese /r/ in the most common words
and across connected speech we could find that there still are 'articulatory
projections' that match up and do not vary. Or, alternatively, language is
actually produced and processed and re-encoded in reference to a cognitive
'syllable-type' plan, in which case there is something the same across the
different words which instantiate a particular syllable type.

>> I don't know whether either of you have seen Anne Cutler's work on l  & rs
and the closeness of some sounds vs dissimilar. Not my field exactly but
fascinating reading. Unfortunately haven't got papers' titles here at mo but
googling would bring them up.<<

I remember Cutler's work from back in the 90s--I think. There were even special
publications about 'teaching EFL to the Japanese learner' with her name on them,
if I'm remembering correctly.

This brings up a confusing issue--people in ELT will often say things like 'this
or that sound are phonetically similar (or dissimilar)'. But this is very
confusing. That is because you might have two sounds in real speech that are
achieved in very distinct ways in terms of articulation but that converge
acoustically. And you might have two sounds that are made in very similar ways
in terms of articulation and yet their acoustic signature is very different. And
all this has to be understood with the idea that in face-to-face communication,
the listener interprets the interlocutor's speech based on both visual and
audible information. This sort of two channel information might also be crucial
to ACQUISITION. Which is why parking a child in front of a TV with foreign
language coming out does not prove sufficient for language acquisition to take
place. Also, this is why visually-impaired children experience delays in native
language development. It takes them longer to work out and build up a language
system that lacks the visually salient information you get from watching someone
speak.

CJ

At a 'popular level', the Japanese accent of English is made fun of by
interchanging [r] and [l] sounds. The native English speaker perception is,
then, that the Japanese speaker of English produces [r]s where an [l] is
expected and vice versa.

>>> "Also, the Japanese [􏰌] is more likely to be per- ceived as English /l/
when substituted for /l/ (p. 245). Interestingly, Aoyama et al. stated, “An L2
sound that is similar...may enjoy an advantage in the early stages of L2
acquisition” (p. 245) because of the intelli- gibility of the L1 sound, but
ultimately, learners may achieve greater accuracy with the more dissimilar
sound. However, for the English /r/, Lado’s previously mentioned
assertion�"sounds not in the L1 sound in- ventory are the most difficult to
acquire�" seems to apply."<<<

This is one of the 'learnability' issues that actually does get a lot of press,
emerging from theories and concepts about 'markedness' and in the old days of
'contrastive analysis'. That is, which features of English are difficult to
acquire and which ones of these are accounted for by differences with the
language of the learner. Most of the 'conclusions' are controversial and there
are many contradictions.


>
> To me, this says that there is no natural 'r' sound in Japanese, but maybe
that's my interpretation...

I would say there are plenty of 'r'-like sounds in spoken forms of Japanese, but
their distribution in spoken Japanese is so different that it would be easy to
think they had no [r] in Japanese that they could easily substitute for an
acceptable English [r] or modify easily to create a Japanese English [r] in
'interlanguage'.

You might find this paper interesting. I find it hamstrung by the old
structuralist baggage (those mythic phonemes). But it does cite a lot of various
sources in trying to get down to the nitty-gritty of the English /l/ /r/ issue.
I think the biggest most interesting set of data hasn't been captured yet--I
would like to compare and contrast the mouth movements--the gestural sets--of
English native and Japanese natives speakers saying similar words, like E 'ray'
and 'lay' and 'day' vs. J 'rei'.

At any rate (you probably already have seen it):


http://web1.hpu.edu/images/GraduateStudies/TESL_WPS/05Carruthers_Phono_a17236.pd\
f

excerpt:

Liquids. In addition to fricatives, English
liquids are a notorious challenge for
Japanese learners. The English /l/ and /r/
are often described by teachers and learners
as corresponding to a single Japanese
sound; however, this categorization is not
entirely accurate. Avery and Ehrlich (1992)
described these two sounds not as allophones
of a single Japanese sound but
rather as “one liquid sound which is between
the English /r/ and /l/” (p. 135).
Thompson (2001) characterized the Japanese
/r/ as a flap-like sound, like short /d/
(p. 298). Price (as cited in Aoyama, Flege,
Guion, Akahane-Yamada, & Yamada, 2004)
agreed, describing the sound as an “apicoalveolar
tap []” (p. 234), a flap articulated
with just the tip of the tongue. In this paper,
I have chosen to treat these sounds as allophones,
as Aoyama et al. (2001) proposed:
“Despite the articulatory difference between
Japanese and the English liquids /l/ and
/r/, Japanese speakers seem to perceptually
assimilate both English liquids to Japanese
/r/” (p. 234). Thus, although they are not
technically allophones of Japanese [], these
sounds act as allophones because of perceived
similarities in place and manner of
articulation.
20
However described, both English /l/
and /r/ are a challenge to discriminate and
produce for JSE. Moreover, the native
speaker of English likely misunderstands
the JSE’s production of these sounds. Avery
and Ehrlich (1992) stated that because the
JSE produces [] for both English /l/ and
/r/, the native speaker listener will hear
“the opposite of what the Japanese speaker
intended” (p. 135). Further confusing the
situation, as Aoyama et al. (2001) observed,
although both are assimilated as Japanese [],
English /r/ is “more dissimilar phonetically”
(p. 234), despite being regularly transliterated
as r orthographically (i.e., romaji).
In a study on the intelligibility of /l/ and
/r/ production, Aoyama et al. found that
JSE, particularly children, are better able to
improve /l/ pronunciation (p. 245). Also,
the Japanese [] is more likely to be perceived
as English /l/ when substituted for
/l/ (p. 245). Interestingly, Aoyama et al.
stated, “An L2 sound that is similar…may
enjoy an advantage in the early stages of L2
acquisition” (p. 245) because of the intelligibility
of the L1 sound, but ultimately,
learners may achieve greater accuracy with
the more dissimilar sound. However, for the
English /r/, Lado’s previously mentioned
assertion�"sounds not in the L1 sound inventory
are the most difficult to acquire�"
seems to apply.

The task:

>>Raw   Reef    Roof    Ray   Row

Law   Leaf    Loose  Lay   Low
Dawn Deep  Dupe   Day   Doe

Recorded 7 ppl in QuickTime, then played them for my students and asked 80 of
them to evaluate which set of 5 sounds more like the
Japanese　らりるれろ。<<

The sum of the results:

>>R= 190  L= 776  D=18<<

One big problem here is you are asking basically about English syllable types
(ra, ri, ru, re, rou, la, li, lu, le, lou, da, di, du, de, dou--to borrow
Japanese-style romanization) in reference to a unified Japanese sound, the
mythic Japanese /r/ of ra, ri, ru, re, ro.

At any rate, the biggest limitation is this is a literacy task with phonological
implications rather than a simple phonological task (and after 5 years of
trying, I can tell you--one, it's hard to get rid of literacy in such
experiments and two, there is no such thing as a simple phonological task for an
experiment).

Which brings me to my conclusion. The reason there is no such thing as a simple
phonological tasks includes the following:

1. Literacy has phonological implications (for language development, for
language processing, for language comprehension, for production and
comprehension of spoken lanuage), and yet we don't really have a clear picture
of what they are.

2. Presumably it is the same language control system that governs language
production, language perception, and language comprehension. However, it's hard
to design tasks to get at the particular sub-systems. But look at some of the
amazing things the brain does in face-to-face real-time speaking: it integrates
a visual channel with an audio one so that the visuals seem to be in synch with
the audibles, which means there is some sort of buffer memory that holds the
visual information and then integrates it with the audibles, that are then
processed into linguistically interpretable data, and finally it all gets
re-encoded as 'speech' (like what we are conscious of in working memory).
Meanwhile, it is getting ready to reply--produce language that is relevant to
what is being comprehended, etc. etc.

3. We are locked into interpreting phonetic and phonological systems by our own
language capacities. That both enables us and limits us severely.

CJ

> 1. Was it a pure listening task or did it involve texts?Each student had a
piece of paper with the word lists printed 7 times, and the instructions to
circle which set sounded closer to the Japanese らりるれろ

Then it is a literacy task.

>
> 2. I didnt want the NS to try to pronounce Romanized Japanese words, too
fraught with difficulties. I set up the task to replicate the possible onset
sounds of  らりるれろ only, just to keep it simple.

I'm not sure I understand your comment here. In Japanese, words that begin with
J /r/ are almost all foreign loan words or built up from loan roots--first from
Chinese, and then from European languages like English. This could well point to
outside influences changing the phonology of Japanese (the way Norman French is
supposed to have changed English, for example).

The other point being that you have sounds being invoked in real English words,
and yet you are having them referred to written syllables of Japanese, not
spoken words. This can cause distortions. For example, note how Japanese
read-pronounce the Japanese syllabic /n/ at karaoke compared to how it is
actually realized in speech. Often way different.


>
> 3. Pure listening tasks themselves can be different than audio-visual tasks,
where the listener sees the face of the person while they are saying the word or
syllable.Agreed. Didnt want to ask NS to video their faces...

And yet it is in face-to-face communication that speech in all languages
evolved. Something we tend to forget, first because of centuries of written
texts and reliance on them and now with multi-media, which removes us from face
to face communication but gives us audio-visual use of language (combined with
other audio-visuals, such as what you get in a film that tells a story). This,
however, it should be noted makes real time listening more like reading--the
language is scripted, it isn't immediately interactive, etc.
OTOH, we see with computers and the internet, written interaction is becoming
more like speech while going into new directions as well.



> 4. In short, what were the results? Is there a pattern to Japanese EFL
listening perception of English /l/ and /r/?
>
> Sample size 90 students, times 7 speakers = 360 possible answers:R= 190  L=
776  D=185. Also note that you have given them only initial [r], [l] and
[d]s.Yes, again, wanted to keep it simple.

But how do we interpret that data relative to what we know about English,
Japanese and speech perception?


I'd be interested in seeing a more detailed explanation of how you interpret
this set of data. Is there a paper involved with this presentation?

CJ

>>> Raw   Reef    Roof    Ray   Row
>
> Law
>   Leaf    Loose  Lay   Low    Dawn Deep  Dupe   Day    Doe
>
>
> Recorded 7 ppl in QuickTime, then played them for my students and asked 80 of
them to evaluate which set of 5 sounds more like the
Japanese　らりるれろ。
>
> Interesting data, again.<<<<

1. Was it a pure listening task or did it involve texts?
2. Why not ask them to compare with Japanese words instead of the syllable set,
ra, ri, ru, re, ro (or la, li, lu, le, lo, depending on how you romanize them)?
3. Pure listening tasks themselves can be different than audio-visual tasks,
where the listener sees the face of the person while they are saying the word or
syllable.
4. In short, what were the results? Is there a pattern to Japanese EFL listening
perception of English /l/ and /r/?
5. Also note that you have given them only initial [r], [l] and [d]s.
6. One word that caused l/r confusion in the absence of visual clues (mouth
contours in motion) was row vs. low. That is in my own research. The reason is
that the proceeding nucleus [ou] causes lip-rounding before phonation, and so
the key element is more in the timing of the lip-rounding, which is better
discerned with visual clues.

CJ

Yeah, it's a pain that they're at the same time. Mine was recently re-scheduled
as I originally thought I could attend on the Monday.
I say your pres last year. Do you have new data?

I think I'm still using old Yahoo! so I can't read the katakana in your p.s.

--- In iatefl_pronsig@..., "Chris P. Bluetree" <japaneasynow@...>
wrote:
>
>
> Drat! With over 500 workshops you would figure they have the decency
> NOT to schedule us at the same time! I am presenting on my usage of
> C=S+R, and the results of some data... I can send you some info
> beforehand if you like. But yes, let's connect, for sure. You can
> contact me at chrispsensei@gmail for details.
>
> Cheers!
>
> c.
>
> ps; i will let you in on a little secret: Katakana de asobou: can you read
this?
>
> カフィオンザ　テーボアッ　メダーナズ
>
>
>
>
>       New Email addresses available on Yahoo!
> Get the Email name you&#39;ve always wanted on the new @ymail and @rocketmail.
> Hurry before someone else does!
> http://mail.promotions.yahoo.com/newdomains/aa/
>

http://www.haskins.yale.edu/research/gestural.html

brief introductory excerpt:

Gestural Model

Introduction to Articulatory Phonology and the Gestural Computational Model

Traditionally, human speech has been seen as having two structures, one
considered physical, and the other cognitive, where the relation between the two
structures is generally not an intrinsic part
of either description. From this perspective, a complete picture requires
'translating' between the intrinsically incommensurate domains (as argued by
Fowler, Rubin, Remez, & Turvey, 1980).
Articulatory Phonology ( Browman & Goldstein, 1986; 1989; 1990a,b; 1992) begins
with the very different assumption that these apparently different domains are,
in fact, the low and high dimensional descriptions of a single (complex) system.
Crucial to this approach is identification of phonological units with
dynamically specified units of articulatory action, called gestures. Thus, an
utterance is described as an act that can be decomposed into a small number of
primitive units (a low dimensional description), in a particular spatio-temporal
configuration. The same description also provides an intrinsic specification of
the high dimensional properties of the act (its various mechanical and
bio-mechanical consequences).

>>When i do connected speech runs with the katakana i am always amazed at how
Native English my students are, and i know they are too! It is almost scary! <<

I use this aspect of English (i.e., a model of 'rapid connected speech') a part
of dialogues/Cummins' devices. One, it's a step to get students to use dialogues
as speech practice rather than a labored reading exercise (where they stick
close to the text and read it word by word slowly). It's actually a simple step
to monitor comprehension: if they can't read the text silently chunk by chunk
and then look up and say it chunk by chunk, chances are they don't understand
what they are trying to process.

So back to the drawing board. First we do dialogues as language
practice--vocabulary and pronunciation of words and phrases. I'm all for making
this step bilingual for beginning level students (and that is mostly what I
teach here in Japan). Then we do it as a 'read silently to yourself'
exercise--reading practice. Then we do them as speech practice and memory
training (read and look up and say). Then we go back and really drill
pronunciation points. As I've said before, it's juncture, liaison,
co-articulation and 'sound changes' in rapid speech that are important to sound
more natural.

Also remember fluent users of a language are unconsciously expecting a flow of
120-300 words per minute from their interlocutor, so speakers who are too slow
and labored end up not getting much of a chance to say anything in real
conversations.

Anyway, the point of all this is that I use dialogues as places to practice
pronunciation because when students do actual communication activities, there is
no place to do pronunciation practice during the activity--that is the whole
idea of 'communication'--the learner is supposed to be on their own to some
extent. There is, of course, a natural 'decay' in the achievement of better
pronunciation when communicating--straining to communicate seems to draw on
'general' memory resources and so detracts from achievement of improved
pronunciation. Until one gets more and more fluent.

The structuralists said that foreign language required 'overlearning' and in
that sense they were right.  A different metaphor would be the 'bootstrap'.
Boostrapping up to get to the point where you can use a language to continue to
learn it is laborious and tedious in its initial stages. It's those initial
stages, unfortunately, that so many of our learners never get past.

>> Have you taught in other Asian countries?

I have taught a bit around in other places in Asia, but for all intents and
purposes you can say I have been mostly limited to Japan. However, I have taught
a considerable number of Korean, Chinese, Malaysian, Thai, Vietnamese and
Indonesian students over the past 20 years. Also, I'm getting more and more from
the Arab countries of the ME.

CJ

http://japanheo.blogspot.com/
http://wearechikineko.blogspot.com/