150
John B. Haviland
Mechanically generated wordlists will inevitably reveal areas requiring
further lexicographic work – phrasal lexical units, syntagmatically defined
paradigms, “functional” vs. “lexical” elements, or particles, for example –
and they ordinarily also expose to view especially elaborated lexical do-
mains worthy of deeper exploration. Such domains may, on the other hand,
emerge not from obvious gaps or hypertrophy in lexical sets revealed in
text collections or elicited wordlists, but in clues from the communicative
practices of a speech community itself: aesthetic judgments about “beauti-
ful” or “eloquent” – if not “ugly” or “awkward” – speech, for example,
especially marked and evaluated kinds of talk, or specialized speech genres
or performances, on the one hand; and, on the other, cultural “preoccupa-
tions” with associated lexical expression: elaborated vocabularies for pro-
fessions, activities, or other kinds of interests, or insistence on “getting the
right word” or on “proper” and “accurate” expression.
Most methods for lexical elicitation are, for better or for worse, “exten-
sional” and “referential” – that is, they are based on presenting exemplars
of things or situations in the world to native speakers and asking for appro-
priate linguistic expressions which can be used to refer to or to characterize
them. Such a method is perhaps inescapable for first-level lexical documen-
tation, but it leaves largely unanswered difficult questions about the inten-
sions of words: what they actually mean, what meaning distinctions they
encode, what sorts of meaning relationships they enter into with other words
and expressions, rather than simply what states of affairs they can be used
truthfully to refer to. Such elicitation techniques are also often helpless to
capture such non-referential aspects of meaning as politeness registers,
specialized uses and contexts, and the like. Such issues can – and perhaps
must – be ignored for the first stages of building lexical databases in lan-
guage documentation, but they cannot be ignored forever.
Here is a single example from my own fieldwork on Guugu Yimithirr.
I quickly learned that the everyday Guugu Yimithirr word
nambal meant
‘stone’ but was also extended to mean ‘money’. My primary teacher (and
social father) in the community, who sometimes had occasion to borrow
money from me, often instead used (or whispered) another word to me when
he wanted to refer to money: wambugan. However, wambugan is really a
polite equivalent for the ordinary word nambal in the respectful vocabulary,
obligatory in speech with avoided affines and referred to in the published
literature as “Brother-in-law language” (Haviland 1979). Its denotative
range is in fact somewhat broader than that of nambal – it includes stones
(including specially named grinding stones, quartz, etc., which are not
Chapter 6 – Documenting lexical knowledge
151
normally called
nambal) AND money. Crucially it is an over-polite word,
no longer used in modern Hopevale with avoided affines nor, indeed, widely
known beyond a few old men, and with them still carrying a euphemistic
tone of respect. Both factors combine to make wambugan a perfect code
word for an embarrassing task like asking one’s courtesy son and pupil for
a loan.
Ignoring such difficulties for the moment, let us consider techniques for
supplementing the lexical information haphazardly collected through me-
chanical reversal of text corpora. The trick, obviously, is systematic but
controlled elicitation, by presenting or simulating aspects of “external” re-
ality so as to stimulate native speakers into using words and expressions to
represent as yet unencountered states of affairs. Somewhat artificially I have
divided sample methods according to what aspects of “reality” they purport
to simulate: ‘natural’ facts, socio-cultural institutions, and in the final sec-
tions pragmatic facts of (inter)action, and ideological constructions on lan-
guage and society.
4.1. ‘Nature’
The tradition in anthropological linguistics, variously labeled “ethnographic
semantics” or “ethnoscience,” purports to display culturally specific knowl-
edge about the natural world by detailing the semantics of lexical domains
related to the corresponding natural phenomena: Hanunoo medicinal plants,
Tseltal categories of firewood, ethnobotany or ethnozoology; parts of houses
or bodies, taxonomies of disease, local technology, and so on. A classic
example of the genre is Berlin’s (1968) detailed study of Tseltal numeral
classifiers, a detailed compendium of the several hundred classifiers once
obligatory in Tenejapa Tseltal numeral expressions. Numeral classifiers
specify countable units of different kinds of substance, often on the basis of
shape. The notable feature of Berlin’s study, for our purposes, is his use of
carefully elaborated photographs both as stimuli (i.e. to elicit Tseltal nu-
meral expressions from speakers) and as a vehicle for metasemantic repre-
sentation: the photos accompany and illustrate his verbal characterization of
the Tseltal forms so elicited. (Berlin also used Kaufman’s mechanical pro-
cedure to generate potential numeral classifier roots, as described earlier.)
To give an idea of both the semantic specificity of the Tseltal forms and the
nature of the photographic stimuli, here are two sample pictures from
Berlin’s study. (Note that in Figure 5, illustrating the classifier hiht’, the