[Lingtyp] spectrograms in linguistic description and for language comparison

[Adam James Ross Tallman]

Hello all,

I would like to start a conversation about something and I’m taking a shot
at lingtyp as a potential starting point for this discussion (perhaps not
the right venue, because the issue is perhaps specific to phonological
typology).

One thing I’ve been confused and/or frustrated about since I started
investigating tone and stress has been the use of spectrograms and/or pitch
tracks in language description. It seems to me that linguists have very
different views about what spectrograms and/or pitch tracks are for, but it
has never been brought out in the open, to my knowledge.

When I was an MA student, I was basically taught that the main purpose of a
spectrogram was to show how one went about measuring some phenomena in the
acoustic signal. A pitch track could be an expositional device to show
variation in the signal perhaps related to speaker differences or
intonation (Cruz & Woodbury 2014). However, spectrograms and pitch tracks
are not “phonetic evidence” for a phonological claim. Due to the
variability of the phonetic signal, acoustic phonetic data only really
becomes phonetic evidence when it is aggregated for the purpose of
statistical analysis (Tallman 2010).

At least that’s what I thought in 2011, but I realized later that this was
not the view shared by many linguists and, at least among non-phoneticians,
my position is perhaps a minority one. In grammars and descriptive works,
linguists often present individual spectrograms and pitch tracks as one off
data points that support a claim. In the vast majority of the cases (except
perhaps when vastly different intonational contours are being compared), I
often struggle to know what the purpose of these displays or pictures are.
How do we know they are not cherry picked? How do we know that these
displays are representative?

The differences of opinion about the use of spectrograms have emerged for
me in the reviewing process – one reviewer says this spectrogram is
useless, another says it's informative etc. one reviewer demands a pitch
track, another says it does not communicate anything . etc. Opinions are
simultaneously contradictory but aggressive and definitive.

Sometimes the subtlety of the pitch phenomena the linguist is describing is
way out of step with the ability of the pitch track to represent. I look at
the pitch track and I think: “I cannot distinguish between pitch phenomena
associated with tones and microprosody in this example so it is unclear
what the purpose of the pitch track is or what it adds” or “if you were to
tell me what tones the language had and give me this spectrogram / pitch
track, I would not be able to associate them with any of the syllables in
any consistent way”. Or perhaps the algorithm used to draw pitch isn’t
appropriate and it's very difficult to understand what is being
communicated by the display.

I have started to wonder whether there were any guidelines or conventions
for the use of spectrograms and whether others perhaps had any thoughts on
the issue. Specifically I am interested in the idea that a single
spectrogram could serve as “phonetic evidence”. I still find this view
strange in light of the well known “stochastic” and “multivariate”
relationship between phonological categories and phonetic realization
(Pierrehumbert, Beckman, Ladd 2000; Mazaudon 2014, among many others), but
it still seems to be widely held in our field.

Cruz, E. & Woodbury, A. C. 2014. Finding a way into a family of tone
languages: The story and methods of the Chatino Language Documentation
Project. *Language Documentation & Conservation *8:490-524.

Mazaudon, M. 2014. Studying emergent tone-systems in Nepal: Pitch,
phonation and word-tone in Tamang. *Language Documentation & Conversation *
8:587-612.

Pierrehumbert, J., Beckman, M. and Ladd, D. 2000. Conceptual foundations of
phonology as a laboratory science. *Phonological knowledge: Conceptual and
empirical issues. *Oxford: Oxford University Press.

Tallman, Adam. J.R. 2010. Acoustic correlates of Lenis and Fortis Stops in
Manitoba Saulteaux. MA Thesis: University of Manitoba.

-- 
Adam J.R. Tallman
Post-doctoral Researcher
Friedrich Schiller Universität
Department of English Studies
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listserv.linguistlist.org/pipermail/lingtyp/attachments/20221127/7bbb8a3b/attachment.htm>