"Basic" vs. "Non-basic" meanings

[Suzanne E Kemmer]

On the distinction between "primary / underlying / basic" meanings
vs. "secondary / derivative / extended" ones:

I think markedness (i.e. asymmetry in a set of oppositions, in this
case, in the cognitive status of related meanings) is a typical
property of category structure, and not just a descriptive heuristic.
It doesn't have to take the form of a linear hierarchy-- in fact,
given a more radial notion of category structure, you can have
increasing markedness along different dimensions, which might or might
not be linearizable into some overall dimension of degree of deviation
from an unmarked or 'most naturally expected' or 'entrenched' meaning.

Psycholinguistic evidence is once source of evidence for investigating
any given category (e.g., we can try to test whether one sense is
quicker to process than the others, or is first activated out of
context, or whatever other measure might show an asymmetry). Obviously
once one moves beyond lexical categories the degree of difficulty of
applying psycholinguistic methodology increases rapidly.  But other
types of evidence are also relevant.

Frequency is one factor, but given grammaticalization it will not
uncommonly give us results that diverge from other measures.  For
example, I think that prototypical 'possession' (roughly:
socially-sanctioned control of a human over a separate, inanimate
entity) is the most basic sense of English _have_, but it's not the
most frequent use, because of the grammaticalization and semantic
extension of the word. Even if we exclude the highly grammaticalized
uses with participles, we have a huge number of 'extended possessive'
uses comprising a schematic 'reference point' meaning (cf. Langacker).

For a given category, I would try to look at as many sources of
evidence as possible and see whether they converge.  I think
acquisition order is relevant, for one thing.  So is typological
evidence--what are the meanings that languages always or almost always
have units to express? Also, we might look at the category structure
of a given unit and note an asymmetry in semantic relationship.

For the case of the English pronouns, for example, it would be strange
to consider impersonal 'you' the basic meaning of English _you_, given
that (a) second person singular is a more crucial meaning for human
languages to express: no languages lack a pronoun with this meaning;
and (b) impersonal 'you' is a specific kind of impersonal meaning that
evokes an idea of participation of the listener--it essentially asks
for identification of the listener with a generic participant. It
is more reasonably thought of as an abstract extension of a basic
second person singular meaning rather than the other way around,
essentially because we can more easily see how such an extension can
come about in usage than the reverse one. I guess these are the kind
of 'natural unidirectionalities' Martin referred to.

To get investigate this further, we might look at acquisition (I bet
impersonal 'you' is much later) and frequency in a spoken corpus (I
would predict the 2nd person meaning would be more frequent) and
weigh that evidence in.  Also diachrony is relevant, but at the
typological level of generalization. Those extensions that
cross-linguistically often become conventionalized are presumably more
natural ones. (Of course this is tricky to apply, because linguists
tend to infer diachronic directionality from general principles of
concrete->abstract and intuitions about basicness. But distributional
facts can also give us better internal reconstructions and support for
intuitions about 'earlier' and 'later'.)

Converging evidence would give us the most robust results; where
the evidence is radically non-convergent we can then talk about
categories that do not have an 'unmarked' sense. I could also imagine
a case of gradual emergence of two different centers of attraction,
where a unit is undergoing semantic split and for a time is
polysemous but with two unmarked uses.

I am aware that there are many complex issues here (e.g. the problem
of identifying senses) and I don't want to imply that it is always
going to be easy to establish these kinds of asymmetries--or even that
they inevitably exist for every category. I just think that markedness
is an important cognitive organizational principle (in fact, Greenberg
suggested in a 1966 monograph that it might be based in or related to
figure/ground organization in perception). More empirical investigation
of sense relations so we can try to firm up our 'fuzzy intuitions'
about basicness and set them on a theoretical basis, is what is needed.

--Suzanne Kemmer