LINGUIST List 2.721

Mon 28 Oct 1991

Disc: Whorf

Editor for this issue: <>


  1. , Language/Behavior Correlations
  2. , Whorf
  3. "Bruce E. Nevin", Whorf-Sapir ms excerpt 2

Message 1: Language/Behavior Correlations

Date: Thu, 24 Oct 91 12:29:24 EDT
From: <>
Subject: Language/Behavior Correlations
Setting aside for the moment the question of why so many people continue
to insist on attributing to Whorf and Sapir views they did not hold
(or at least did not express), I would like to say something about
the results which are claimed to support the hypothesis that language
and non-linguistic behavior (behavior, for short) exhibit certain
close connections (which people seem to want to interpret as involving
causality going from language to behavior).
(1) Even if we find certain correlations between language structure
and patterns of behavior, this does NOT (as I think I noted earlier)
indicate the direction of causality (as indeed Whorf himself noted
at one point). The color terminology business shows, if anything,
that the complexity of a color terminology seems to depend on the
complexity of the culture, there being, for example, no industrial
or postindustrial cultures whose languages use two or three color
terms. There has also been speculation about the fact that the
lateness of terms for 'blue' may be connected with the relative
scarcity of blue objects (other than the ubiquitous sky) in nature.
This would suggest very strongly that the linguistic pattern comes
second, as a reflection of a culture's need to make certain distinctions.
(2) All the studies that claim to show a connection between language
and behavior that I have seen mentioned seem to deal with two or
at any rate a small number of languages, e.g., Tarahumara and English.
Likewise, I have seen studies by Alexander Guiora on Hebrew and English
and other such small sets, which I don't think have been cited on
LINGUIST so far. Yet, since the claim being tested is correlation
between linguistic structure and nonlinguistic behavior, the relevant
population is languages (not individual speakers), and you cannot
seriously talk about correlations for populations of two (or three or
whatever small number is involved). What we require is a study involving
a dozen or a hundred languages that have the Tarahumara color system
and a dozen or a hundred that have the English one before we can say
anything at all about correlations and things.
Having said this, I would predict that we will find such correlations
but I would also predict that at least some of them will turn out
to have the opposite causality from that suggested (or a more complex
one than either of the simple unidirectional ones).
Is there anybody out there who would like to collaborate on putting
together such a mass crosslinguistic study?
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue

Message 2: Whorf

Date: Fri, 25 Oct 91 21:03:00 EDT
From: <>
Subject: Whorf
Since many of the readers of LINGUIST are from Missouri, I thought
I would provide some evidence for my recent assertions that Whorf's
position has been widely misunderstood.
In "The relation of habitual thought and behavior to language",
Whorf says among other things"
 That portion of the whole investigation here to be reported may
be summed up in two questions: (1) Are our concepts of 'time', 'space',
and 'matter' given in substantially the same form by experience to all
men, or are they in part conditioned by the structure of particular
languages? (2) Are there traceable affinities between (a) cultural
and behavioral norms and (b) large-scale linguistic patterns? (I should
be the last to pretend that there is anything so definite as "a
correlation" between culture and language, and especially between
ethnological rubrics such as 'agricultural, hunting', etc., and
linguistic ones like 'inflected', 'synthetic', or 'isolating'.
In a footnote on the same page (p. 139 of the Language, Thought,
and Reality book), he says emphatically that "The idea of "correlation"
between language and culture, in the generally accepted sense of
correlation, is certainly a mistaken one" and he cites some
Thus, I believe that Whorf made a clear distinction between culture
(behavior) and language, but he did not make such a distinction
between language and thought. As I said before, he presupposed as
did almost everyone else at the time that if people speak a certain
way then that reflects the way they think. He took it for granted
for example that if the Hopis pluralize the word for cloud (oomaw)
the way that they normally pluralize animate nouns, then they must
think of the clouds as animate.
Of course, this view is naive, as Joseph Greenberg pointed out
in the fifties, since languages make all sorts of arbitrary
distinctions (or fail arbitrarily to make them in certain
environments) without any apparent conceptual consequences.
Essentially, I think the connection works one way, namely,
if a language makes a distinction which cannot be described
in purely structural terms, then we must ascribe to the
speakers the ability to perceive or imagine or whatever
the corresponding distinction in the world. Thus, when
Greenberg points out that nothing important hinges on
the fact that the French use an ordinal in Napoleon Premier
but a cardinal in Napolean Deux, that's OK, because the
choice here can be made w/o reference to the world. The
rule is purely linguistic. And, of course, this could be
the case with the Hopi word for cloud and its plural.
On the other hand, if we find that speakers of Polish systematically
use a different genitive ending for placenames in Poland (and
other Slavic countries) than they do for other placenames,
and do so PRODUCTIVELY, then it IS reasonable to conclude that
they are capable of a conceptual distinction between Poland
(or Slavdom) and the rest of the world.
The distinction between these two kinds of cases is what
seems not to have been entirely clear to Whorf, and that,
as far as I can see, is where he came to sometimes came
to grief.
It is also quite clear that he was not claiming any originality
about the relation of language and thought per se, rather he
was trying to show just how different the language/thought
of one culture could be from that of another in the case of such
basic ideas as that of time, although he points out (p. 158) that
there is not a comparable difference between Hopi and Standard
Average European regarding space.
As to culture, Whorf was faithfully following Sapir in claiming
that there is no more than an "affinity" between language and
culture, but no "correlations or diagnostic correspondences"
(p. 159). For, as I noted earlier, Sapir was one of the
staunchest critics of the late 19th century and early 20th
century linguists who propounded such theories as the "passivity"
of peoples whose languages use the ergative constructions, and
such like drivel.
Incidentally, much of what I have said about Whorf's intent in
bringing the Hopi vs. the SAE treatment of time and matter can
also be said about Sapir's work on the psychological reality
of phonemes. Today, we emphasize the psychological reality
part, but actually in his time, the novelty was the phoneme.
Claims about psychological reality about in the second half
of the 19th century and later (and we find them in all of
Sapir's as well as Bloomfield's early writings). The idea
of the psychological vs. the grammatical subject after
all originated in that period. And, to take one example our
of thousands, when Platt wrote in the 1870's that the Urdu
speakers perceive certain constructions in their language
as active even though they look passive (these are, of course,
ergatives again!), he was expressing himself in a way which
was quite typical for the time (though not for the 17 or the
18th century).
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue

Message 3: Whorf-Sapir ms excerpt 2

Date: Mon, 28 Oct 91 10:37:41 EST
From: "Bruce E. Nevin" <>
Subject: Whorf-Sapir ms excerpt 2
Several people have indicated that the excerpt I submitted from my
back-burner work-in-progress ms relating to the Whorf-Sapir Hypothesis
was useful to them. I might as well include some additional excerpts.
Construe this as continuing from the end of my post yesterday. (That
includes the possible response of deleting it now if your reading of the
prior part so indicates to you.) [For "yesterday" above, substitute
Wednesday 10/16. The present submission apparently was a victim of the
crash experienced by tamu on or about 10/17.--BN 10/28/91]
In formal linguistics, Zellig Harris and his co-workers have come full
circle to the work on information structures in discourse that opened
the whole field of transformational grammar. Harris, Ryckman, Gottfried
et al. _The Form of Information in Science_ (1990) develops a
representation of the information immanent in a body of texts written
over a span of years in the history of a subfield of a science
(immunology). Changes in this structure correlate transparently with
historically well-documented changes and developmental stages of the
science during that period, although the structure was determined by
clearly defined formal means and without reference to any knowledge of
that historical context. In this way, they have demonstrated strongly
that structures found in the sublanguage of that science (and not
imposed a priori on it) correlate on the one hand with aspects of the
social reality of the science and on the other with the structure of the
real-world domain which is the concern of that science.
The latter correlation is reflexive, however, in the sense that, as the
structure changed, it (and the undestanding of the scientists writing
the original research reports on which the analysis was done) over time
came into closer conformity with a reality whose nature was in process
of being discovered. Before that change and that concurrent discovery,
certain characteristics of reality could not be stated or thought;
afterward, they could. But the discovery and the change in structure
were simultaneous (though of course the writing down for publication was
not). No better confirmation of Sapir's intuition of the essential unity
of language and thought could be offered by one of his students.
5. The confirmation is equivocal, however, since the work clearly
demonstrates (as Harris stated at the end of _Mathematical Structures of
Language_ (Wiley, 1968)) that language is not identical with thought but
instead provides a rather rigid channel for thought. This corresponds
precisely to the observation above that the discovery and the language
for talking about it co-evolved. By using this term I refer
specifically to the common misperception regarding biological evolution
that e.g. eohippus evolved into the horse in response to environmental
changes, when one must instead acknowledge eohippus and its
pre-grasslands environment co-evolved into the horse and its grasslands
environment. Synecdoche is fallacious in both cases. The claim, then,
is of the unity, but not identity of language and thought.
To illustrate this point further, I should like to adduce a recent
contribution to the enormous literature in the study of kinship
categories, always a favorite topic in anthropological linguistics.
Wierzbicka, in Semantics and the interpretation of cultures: the meaning
of 'alternate generations' devices in Australian languages, proposes a
new set of metalanguage terms for discussing the alternate sets of
pronouns used in many Australian languages. She urges that the
terminology of "generation harmony" and "disharmony" that has become
traditional in anthropology is arcane and psychologically arbitrary,
does not capture native speakers' meaning and does not make that meaning
accessible to people from other cultures, and claims that her new
terminology provides a better fit. This work illustrates a Whorfian
effect in the sublanguage of a specialization within the science of
anthropology. With the traditional terminology, aspects of aborigine
culture are difficult to come to recognize and understand, and not
possible to communicate; she claims that with the proposed new
terminology it is.<6> Thus, while providing an illustration of Whorfian
6. This is part of Wierzbicka's ongoing work on natural language
semantics based, ultimately, on a proposed set of universal semantic
primitives, including: I, you, this, someone, something, want, don't
want, say, think of, imagine, know, become, part, place, and world
(Wierzbicka, Semantic Primitives (1972), Lingua Mentalis (1980). Be it
noted that Harris denies there can be a lingua mentalis or any
metalanguage external to natural language. For one thing, were there
such one would need to account for the grammar and semantics of that
metalanguage, and off we go in an infinite regress of grammatical and
semantic metalanguages. For another, Harris has demonstrated that the
information structures immanent in texts account precisely for the
information that the texts report, so that, like LaPlace, he has no need
for this additional hypothesis. But Wierzbicka's proposal here, however
it may be guided by her broader theoretical interests, concerns only a
sublanguage of English serving as metalanguage for a subfield of
anthropology, and as such is unobjectionable. The _semantics_ of this
sublanguage inhere in its informational structures, per Harris, rather
than in its use of vocabulary from a supposedly universal lingua
effects within a subfield of a science, she proposes to overcome such
effects by devising a perfect metalanguage for that subfield. Since the
subfield concerns an area that is by nature a matter of social
convention and so in social reality rather than physical reality (to
make that Durckheimian distinction again), she may be able to get away
with it. I do not doubt the creativity of human cultures, however, and
would build in means for the sublanguage to evolve.
An abiding interest of Harris, as of his teacher Sapir, has been the
question of refinements and possibly extensions of natural language that
foster international scientific communication. In his analysis,
language-particular characteristics due to the reduction system
(extended morphophonemics) of one language or another are partitioned
from operator-argument structures that `carry' information, which are
remarkably uniform from one language to another. This uniformity
becomes very close indeed in the grammar of a science sublanguage, where
classifications and selection restrictions are much more closely
constrained than in other domains. But even in nontechnical domains
Harris has a great deal to say about linguistic universals,<7> and about
the distinctions between what is universal in language and culture and
what is idiosyncratic and therefore pertinent to the Whorf-Sapir
7. See e.g. _Language and Information_ (Columbia 1989) and _A Theory of
Language and Information_ (Oxford, 1990), which is a more philosophical
companion volume to _A Grammar of English on Mathematical Principles_
(Wiley 1982).
 c 1991 Bruce Nevin
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue