near universals (and pan-Americanisms)

[Lyle Campbell]

----------------------------Original message----------------------------
Let me attempt to offer some thoughts in reply to the question Roger Wright
addressed to me in connection with Miguel Carrasquer's question about the
so-called "pan-Americanisms".  Roger wrote:
 
At the Historical Linguistics conference in Stanford in 1979, Lyle
Campbell was assuring us that "universals" could be over-ridden by areal
pressure, thereby producing a phenomenon that seems self-contradictory,
a "near-universal". His examples were specifically from Central
America. ("Explaining Universals and Their Exceptions", Papers from
the 4th ICHL [Benjamins, 1980], 17-26). If we accept this possibility,
then areal pressure is given extraordinary power, and hardly anything
can be deduced for sure about any relationship in the past ..... I
wonder, Lyle, if you're out there, whether you would still agree with
what you said then. I was startled when I heard it.
 
        First some general context:  (1) This isn't really connected with
the pan-Americanism question.  (2) The notion of a "near-universal" or
"statistical universal" is not mine, but rather comes from Joseph Greenberg
and has been written about (and accepted) by most who work in typology and
universals, not just me.  (3) The "phenomenon" would seem
"self-contradictory" only if one were to insist that 'universals' (as the
name would seem to imply), must be absolute (which in fact they do not
according to standard definitions of this technical term); Moreover, though
it is a matter of definition, there are solid substantive reasons why
people don't insist that all universals must be absolute (i.e.
exceptionless).  (4) Not an important point, my examples were not Central
American, but mostly from the Northwest Coast of North America.  (5) So,
yes, I still agree with what I said, more so now than ever, precisely
because of the typological findings of recent years.  Let me give some
justification.
        As Greenberg defined them, a statistical universal (near-universal)
is one where with greater than chance frequency languages tend to exhibit
the trait (e.g. nearly all languages have nasals; most SOV language have
postposition).
        So, what does it mean to have "exceptions" to universals?  In this
view, an exception does not invalidate the universal, but only converts it
into a statistical universal (near universal).  All the questions remain
the same:  we still need to explain why the phenomenon is so frequent
(nearly absolute) in the world's languages, and why so few languages have
exceptions.  Such near universals (universals which have exceptions) seem
still to have explanatory value, since the violating languages are often
soon brought back into line with the universal through language changes.
For example, the universal that q > k (where > = implies) (q = uvular
stops, k = velar stops) had exceptions in the NW Coast linguistic area; the
sound change of k > "ch" [palato-alveolar affricate] diffused across many
languages of different families, leaving languages with /q/ and /"ch"/ but
no /k/, but a new /k/ soon entered the systems, from loans and sound shifts
in the languages involved.  That is, it appears that the universal soon
brought these languages back in line.  Most languages have nasals, but in a
small area of the Northwest Coast, languages of several families lack
nasals -- here original m > b, and n > d under areal pressure.  We still
need to know why nearly all languages, these (and perhaps a few in Papua
New Guinea) excepted, have nasals.  There is strong motivation for
languages to deploy nasals, since they are perceptually the most salient of
all consonants, great for carrying the phonological contrasts of your
message clearly and unambiguously.  So, exceptions are not necessarily
counter-examples; a well-motivated (near) universal need not be
abandoned/rejected because some few languages do not conform for whatever
reasons.  What is important is not that some linguist declares some
phenomenon to be universal or not, but rather the reasons/explanations for
why it is so.
        This view correlates the likelihood of something being universal
(absolutely or nearly/statistically) to the role it plans in the function
of language -- the more some linguistic thing facilitates languages to
achieve their communicative ends, the more likely it is to be found in most
languages.  The possibility of exceptions is correlated with the degree of
disruption to efficient functioning/processing such exceptions would cause
in languages not conforming whole-heartedly to the universal.  We are
interested in the general principles of language -- what the pieces are and
how the pieces interrelate with one another.  Some of these principles have
been considered to be absolute (exceptionless) universals, others near- (or
statistical) universals.  Still others have not necessarily been associated
with universals, but rather are talked about in terms of typological
connections.  The more efficient the language (i.e., not permitting
borrowing or language contact, to mention just one reason cited by Roger,
to motivate departures from principles), or the greater the value of the
universal for facilitating language processing, the more likely it is that
absolute conformity will be found.  Thus, since the burden on perception
and production for speakers of a language which lacked vowels would be so
great as to make communication viartually impossible, it is safe to say
that the universal that all languages have vowels will probably never have
any exceptions.  However, the word order universals that say it's great to
have all the heads preceded by their complements throughout a language
(Genitive-Noun, Adjective-Noun, etc.) or all the heads followed by their
complements throughout (Noun-Genitive, Noun-Adjective, etc.) make a
languages very user friendly -- it's easier to parse, to figure out what
the constituents, are if you can rely on the strategy that the same sort of
constituent will show up in the same order across related grammatical
categories; however, a language which happens not to conform does not
become impossible to learn or use; it's just harder on its speakers and
hearers, explaining why most languages conform, but a few are able to get
by without conforming.
        The color universals provide a clear example.  By the human color
universals the foci (centre, truest "red", etc.) is perceived as the same
everywhere; human perception of color is universal and is reflected in
languages in universal ways, in the implicational universals established by
Berlin and Kay (e.g. the presence of a basic color term for 'yellow'
implies the language also has one for 'red', and 'red' in turn implies
terms for' 'black' and 'white').  Nevertheless, there are some instances
where a language's basic color terms do not correspond to the universal
foci (according to human perception) but rather to the color of culturally
salient objects.  For example the basic color terms in Pukapukan (Samoan
outlier, Cook Islands) correspond to colors of parts of tubers, extremely
salient objects in the culture, e.g. 'red' matches the color of the 'inner
layer of a variety of tuber', not the universal 'red'.  This, however, does
not change the fact that the vast majority of languages have their term for
red matching the one expected from the universal; it just means that in
some cases culture is able to mediate the universal -- how could we even
talk about the handful of exceptions if we didn't recognize the
generalization, the "near" universal, and acknowledge that it has
explanations underlying it?  Just so, linguistic universals have
explanations underlying them and absolutely every language may or may not
conform depending on the strength of the explanation/motivation behind the
particular universal at hand.
        It may not be that all are sympathetic to this relativization of
universals, but most textbooks and programmatic papers do readily
acknowledge near or statistical universals as part of the overall research
program aimed at universals of language.
 
Lyle Campbell (Professor)
Dept. of Linguistics
University of Canterbury
Private Bag 4800
Christchurch, New Zealand
Fax:   64-3-364-2969
Phone: 64-3-364-2242