Message-Id: <m0pcEoK-0000PsC@xiron.pc.helsinki.fi>
Date:         Wed, 2 Mar 1994 22:25:11 -0500
Reply-To:     Logical Language Group <lojbab@ACCESS.DIGEX.NET>
Sender:       Lojban list <LOJBAN%CUVMB.bitnet@FINHUTC.hut.fi>
From:         Logical Language Group <lojbab@ACCESS.DIGEX.NET>
Subject:      Testing Sapir-Whorf - cross-post from
              sci.lang/soc.culture.scientists
To:           Veijo Vilva <veion@XIRON.PC.HELSINKI.FI>
Content-Length: 10880
Lines: 181

Subject: Testing the Sapir-Whorf Hypothesis (was Re: Boundaries Between Science
 and Non-Science)
Newsgroups: soc.culture.scientists,sci.lang
Organization: The Logical Language Group, Inc.
Keywords: science, linguistics, Sapir-Whorf hypothesis, Loglan, Lojban

bouche2@server.uwindsor.ca (Boucher David) responds to me
>#d. If your test is capable of proving the SWH hypothesis false, it would
>#seem to be proving false only a strongest extreme version of the
>#hypothesis.  Indeed, it doesn't really prove anything false.  To say
>#that language 'constrains' thought doesn't necessarily mean that it
>#renders thought outside of language impossible.  Indeed, all it has to
>#to is render thought more difficult, either in general or along certain
>#lines.  Your test says nothing about difficulty of conception, and
>#indeed works backwards from conception to language.
>
>"Difficulty of conception" cannot be measured, and varies from
>individual to individual.  In order to demonstrate that anyone's thought
>has ever been "constrained" by language, one would have to be able to
>eliminate any alternative potential source of constraint, including the
>nature of the thing itself.  If one finds it easier to think of flying
>birds than flying pigs, that probably has more to do with the fact that
>birds fly and pigs don't, rather than with the etymology of the words
>"bird", "pig", and "flying".

Like most things that cannot be directly measured, "difficulty of
conception" has to be inferred by various effects.  Correlating with
language is likely to be a little more subtle that looking at
etymologies.  One example is categorization.  Tying into another posting
I just made, I can easily envision that Russians who do not distinguish
linguistically between 'scientist' and 'scholar' might have a real
problem seeing what all the fuss is about soc.culture.scientists.  They
obviously categorize things differently than English speakers.  A Sapir-
Whorf effect based on this MIGHT be some different attitudes as to what
constitutes science, which then might be reflected in what research is
funded and published.  Of course, this particular chain might be harder
than others to prove that the language causes the effect rather than any
number of social conditions.

That is one reason why researchers seem to choose color words as a
fertile area of research.  The labelling of color is unlikely to be
determined by social effects within a culture (and this could be tested
by statistically looking at different groups within a culture for
different labellings, in any case).  The actual color being observed (in
terms of spectral frequency) is demonstrably culture independent.  Thus,
if the linguistic labelling of the divisions of the spectrum correlates
with some statistically significant bias in how 'objective' colors are
seen, we have probably detected a Sapir-Whorf effect.

In the 50s, no such effect was detected, but in the 80s, using better
methodology, a significant difference was detected.  Following is one
of the researcher's descriptions of the research on Linguist List:

Willett Kempton <willett@Princeton.EDU>:
>I'm a coauthor of the Kay and Kempton study discussed in several
>earlier messages.  (I don't follow this newsgroup regularly, but a
>colleague passed on the thread.)  As pointed out earlier, from the
>tangled cluster of hypotheses referred to as the Sapir-Whorf hypothesis,
>we tested only one question:  Do the lexical categories of a language
>affect non-linguistic perceptions of its speakers to a non-trivial
>extent?  (P.  Kay & W. Kempton, "What is the Sapir-Whorf Hypothesis?",
>American Anthropologist, vol 86, No. 1, March 1984.)
>   Considering the complexities of prior research efforts, our primary
>experiment was simple:  Present three color chips (call them A, B, C) to
>speakers of two languages, such that colors A and B are slightly more
>different in terms of (universal) human visual discriminability, whereas
>B and C have a linguistic boundary separating them in one language
>(English) but not the other (Tarahumara, a Uto-Aztecan language).  As
>noted earlier, the English speakers chose C as most different, whereas
>the Tarahumara chose A or split evenly (there were actually eight chips
>and four sets of relevant triads).
>    I'll add a couple of points of interest that were either buried in that
>article, or have not appeared in print.  First, as the speaker of a
>language subject to this perceptual effect, I would like to report that
>it is dramatic, even shocking.  I administered the tests to informants
>in Chihuahua.  I was so bewildered by their responses that I had trouble
>continuing the first few tests, and I had no idea whether or not they
>were answering randomly.  In subsequent analysis it was clear that they
>were answering exactly as would be predicted by human visual
>discriminability, but quite unlike the English informants.
>   An informal, and unreported, check of our results was more subjective:
>I showed some of the crucial triads to other English speakers, including
>some who had major commitments in print to not finding Whorfian effects
>for color (several of the latter type of informants were available on
>the Berkeley campus, where Kay and I were).  All reported seeing the
>same effects.  We tried various games with each other and ourselves like
>"We know English calls these two green and that one blue, but just
>looking it them, which one LOOKS most different?"  No way, the blue one
>was REALLY a LOT more different.  Again, the Tarahumara, lacking a
>lexical boundary among these colors, picked "correctly" with ease and in
>overwhelming numbers.  The article includes the Munsell chip numbers, so
>anyone can look them up and try this on themselves.
>    Some of the triads which crossed hue and brightness were truly
>unbelievable, as it was perceptually OBVIOUS to us English speakers
>which one was the most different, yet all the visual discriminability
>data were against us.  (The article did not mention the hue/brightness
>crossovers for the sake of simplifying the argument in print.)
>   Our second experiment, like the original visual discrimination
>experiments, showed only two chips at a time.  We additionally made it
>difficult to use the lexical categories.  And we got visual
>discrimination-based results, even from English speakers.  So there are
>ways to overcome our linguistic blinders.  (Which we knew already, or
>the original visual discriminability work could not have been done in
>the first place.)  I don't feel that the differences across these tasks
>was adequately explored, and represent a golden opportunity for a
>research project or thesis.
>   I didn't expect to find this.  The experiment was a minor piggy-back on
>another project.  I believed the literature and the distinguished
>scientists who told me in advance that we wouldn't find anything
>interesting.  The experiment was going to be dropped from the field
>research, saved by a conversation at a wine party with a "naive"
>sociologist (Paul Attewell) who had read Whorf but not the later
>refutations.
>   A simple experiment, clear data, and seeing the Whorfian effect with our
>own eyes:  It was a powerful conversion experience unlike anything I've
>experienced in my scientific career.  Perhaps this all just goes to
>affirm Seguin's earlier quote, as applying to us as both natives and as
>theorists:  "We have met the natives whose language filters the world -
>and they are us."

Of course, it is hard to ascribe a lot of importance to how people
perceive colors.  The importance of Sapir-Whorf effects would be if the
lexicon or some other property of a language had significant,
scientifically measureable effects on thought processes that might make
a real difference in how human beings go about their everyday lives (in
short to make the science 'useful'.  It isn't scientific, but the
categorization effects that seem evident in the definition of the term
'atheism' in the spin-off threads now going on here, seem to show a
correlation with logic errors of 'oppositeness'.

>#e. Indeed, some interpretations of SWH say that it doesn't really talk
>#about individuals and their thoughts at all.  Rather it relates either
>#to the *process* of thinking, or to the collective though processes of a
>#culture.
>
>You can't test a hypothesis if you can't even decide what it means --
>and if Sapir-Whorf has multiple meanings, then any evidence that might
>validate one particular interpretation says absolutely nothing about
>whether or not any other interpretation might be valid.  In effect you
>have admitted that the hypothesis is hopelessly vague.

Yes, I have.

>I.e., first you have to decide what the hypothesis means, which may be
>rather difficult since apparently Sapir and Whorf didn't know either.

There is a whole tangle of hypotheses and in some cases assumptions that
fall under the label "Sapir-Whorf Hypothesis', which fall in the broader
category of 'linguistic relativism' which says merely that there is an
(important) relationship between culture and language.  Some people
assume a version of the SWH without any evidence for it, or in spite of
evidence against that particular version, witness the question of
whether English pronouns and words with "-man" in them encourage or
perpetuate gender discrimination.

I think it is important to note that Sapir and Whorf never STATED the
hypothesis in any form, per se.  They pretty much ASSUMED it and used it
as the basis for their other work.  The bounds on what they assumed are
determined by the arguments that they based on the assumption.
Especially in Whorf's case (he was a student of Sapir), a lot of
inference was possible since much of his writings are directly on the
question of linguistic relativism.

So what we really have is a Sapir-Whorf paradigm, rather than
hypothesis, but the label was applied before Kuhn %^) (and the fact that
scientists can now criticize labelling it a 'hypothesis' shows that the
concept of 'hypothesis' has significantly evolved in the last 40 years -
thereby showing that 'philosophy' DOES have an effect on science'.  In
fact, I suspect that the debates over the vagueness of SWH may have
contributed to that evolution.)
----
lojbab                                              lojbab@access.digex.net
Bob LeChevalier, President, The Logical Language Group, Inc.
2904 Beau Lane, Fairfax VA 22031-1303 USA                        703-385-0273

For information about the artificial language Loglan/Lojban, please
provide a paper-mail address to me.  We also have information available
electronically via ftp (ftp.cs.yale.edu, in the directory pub/lojban)
and/or email (listserv mailing list lojban@cuvmb.cc.columbia.edu).  The
LLG is funded solely by contributions, and are needed in order to
support electronic and paper distribution.