144
John B. Haviland
3. Metalanguages for meanings and units of lexical knowledge
A second major set of issues for lexical databases is how to represent the
meanings of lexical items, and how to delimit such items in the first place.
Bilingual definitional equivalents are often manifestly inadequate, for the
reasons that have always worried translators: mismatches in grammatical
class, inexactness or lack of equivalence between target and source lan-
guage terms, incompatible ranges of meaning, infinite regress or vicious
circles, and so forth. Much depends on the available metalanguages.
My colleague Matt Pearson, trying to illustrate the interdependence of
different expressive modalities in language, challenges beginning linguistics
students as follows: “Can you define ‘spiral’ without using your hands?”
(You might try it yourself before reading on.)
To repeat, everything depends on the available metalanguages. Even a
novice mathematician can respond by giving a formula for a 3-dimensional
graph, i.e., by defining a series of values for the (x,y,z) axes. Here are some
sample formulas.
(7)
spiral
(cos(t), sin(t), t) [for a spring-like spiral]
(c
t
cos(t), c
t
sin(t), c
t) (where c is some constant)
[for a cone-like one]
Just to see how these formulas work, on the following page are two graphs
of their results, plotted by my statistician colleague Albyn Jones.
The beauty of the mathematical metalanguage involved is its precision,
parsimony, and presumed universality.
11
The drawback is its potential ar-
cane incomprehensibility.
12
Moreover, though the formulas may describe
quite precisely a class of geometric forms, and perhaps even would help
define ‘spiral,’ we might still need recourse to some further (though per-
haps equally general) non-mathematical devices to capture the meaning of
the word in expressions like “Prices are spiraling out of control,” or “We
must control the insane spiral of nuclear proliferation.”
One difficulty with presuming a language-independent semantic meta-
language (aside from prejudging the semiosis of
words and limiting it to
referential information – a worry of the previous section) is that it may do
violence to the conceptual organization of particular languages. Here is the
emic-etic dichotomy of classical anthropological linguistics: do we give
priority to language-specific organization of forms and meanings, or to de-
Chapter 6 – Documenting lexical knowledge
145
Figure 2. (cos(6t),sin(6t),t) for t in (0, pi)
Figure 3. (t
cos(t),t
sin(t),t) for t in the same range
146
John B. Haviland
scriptive categories derived from language-external conceptualizations. An
early and instructive demonstration of the dilemma is Conklin’s treatment
of Hanunoo pronouns.
(8) Hanunoo pronouns (Conklin 1962)
kuh ‘I’ 1s
muh ‘you’ 2s
yah ‘s/he’ 3s
tah ‘we two’ 1du
tam ‘we all’ 1pl
INCL
yuh ‘you all’ 2pl
dah ‘they’ 3pl
mih ‘we (but not you)’ 1pl
EXCL
If we adopt the standard pronominal metalanguage, kuh will be glossed as
“first person singular” or
tam as “first person plural inclusive”. The meta-
language thus involves a ‘person’ component (with possible values 1, 2, or
3), a ‘number’ component (with possible values, for Hanunoo, of singular,
dual, or plural), and an ‘inclusivity’ component (with possible values inclu-
sive or exclusive, and perhaps an unmarked value) which is defective in
that it can by definition apply only to non-singular first person pronouns.
Using such meaning components it should be possible to distinguish be-
tween 11–13 different pronominal forms (three different persons, with three
different numbers, and an inclusive/exclusive distinction on all non-singular
first-person forms). The paradigm has only eight pronouns, however. Worse,
the primitive terms in the descriptive metalanguage (the number and person
categories, plus the terms ‘inclusive’ and ‘exclusive’) themselves total eight,
suggesting that there is little to recommend this particular metalanguage
over just using the raw Hanunoo terms themselves as “primitive” or “un-
definable” elements.
Conklin observed that a better analysis is possible, taking as metrics of
evaluation efficiency (so that exactly three binary distinctions should be
able to distinguish eight [=2
3
] terms), and “faithfulness” to the native Ha-
nunoo logic. His proposed three binary features are ±Speaker, ±Hearer, and
±Minimal, giving a table like Table 1, whose aesthetic symmetry inspires
hope that one is discovering rather than imposing the underlying system.