Word count

[Carson Schutze]

Brian MacWhinney wrote:

> Regarding negative contractions, such as don't, I think most people would
> view these as monomorphemic initially and perhaps forever.  In some of the
> most frequent cases, the phonology alone is an indicator that they are not
> simple combinations of auxiliary and negative.

I think these comments need some elaboration/qualification. First, I'm not
sure what the scope of "most people" is meant to be; certainly the "forever"
part would not apply to most theoretical linguists (as opposed to
(developmental?) psycholinguists, if that's what Brian had in mind).

But more importantly on the substance, pre-theoretically it's not clear what
monomorphemicity is meant to entail. Let's assume that the contrasting
position is one in which "don't" is analyzed as containing an auxiliary "do"
and a negation "n't", where each of those has separately identifiable
phonology and meaning--perhaps Tense, in the case of "do", or perhaps some
unmarked mood.
(If instead one takes the view that "do" is a true dummy, i.e. expresses no
meaning at all, then it's not obvious whether it should fit the traditional
definition of a morpheme, so I put that complication aside.)

Claiming that "don't" is monomorphemic could mean

1) it has two bits of meaning, negation and auxiliary (whatever exactly the
latter is), but they cannot be identified with separate chunks of phonology,
i.e. "don't" is a portmanteau form; or

2) it has only one bit of meaning, presumably negation, and no separately
identifiable phonological subparts.

Option 2) is what was claimed in the early acquisition literature, e.g. by
Bellugi--she explicitly states that for children at the relevant stage,
"don't" is not an auxiliary but a negation. Hyams (1986), Stromswold (1990),
and others adopt this version as well, stating that "don't" is not under
Aux, is not tensed, etc. This claim makes strong predictions about the
distribution of "don't"--given an otherwise adult-like phrase structure, it
should not distribute like its adult counterpart. This was of course the
point of the proposal.

On this view it would make no sense to say that the monomorphemic analysis
holds "forever", because then adults wouldn't talk like adults.

Brian's appeal to phonological evidence seems to be proposed as an argument
for the weaker claim in 1). This claim does not, as far as I can see,
predict that "don't" should distribute any differently than under the
alternative bimorphemic analysis alluded to above. In that respect it could
be maintained "forever", i.e. hold for adults as well as children. For the
same reason, however, it would not do the work that Bellugi et al. wanted
their version of monomorphemicity (2) to do. This might be a good or a bad
thing, depending on one's view of the correct treatment of the developmental
facts they were dealing with.

Quite independently of this issue, one can ask whether phonological
considerations bear on the question of portmanteau vs. separate form chunks.
I assume Brian has in mind things like the fact that the vowel in "don't" is
not the same as the vowel in "do", that "won't" doesn't contain the rhyme of
"will", etc.

If these alternations were to be taken as evidence that the contracted forms
do not consist of a concatenation of two phonological chunks, the second of
which is "n't", it would mean denying the existence of stem allomorphy, for
that is all that's going on here. These changes are completely analogous to
things like 'keep' ~ 'kept': by parity of reasoning, one would have to deny
that the latter is divisible into the stem 'kep' and the past tense suffix
't', even though this is the suffix we would expect by the fully productive
rule, and the change in the stem vowel is part of a widely attested pattern
in English. Excluding stem allomorphy would force us to miss a huge number
of generalizations and would gain us nothing that I can see, but maybe "most
of us" have a different take on this.

One might think that this parallel does not go through because the negative
contractions involve cliticization, whereas allomorphy can only be
conditioned by affixes. Even if the latter were an accurate generalization,
it would not be relevant, because Zwicky and Pullum (Language, 1983) have
argued persuasively that "n't" is in fact an affix in so-called auxiliary
contractions.


--

Carson T. Schutze            Department of Linguistics, UCLA
Email: cschutze at ucla.edu     Box 951543, Los Angeles CA  90095-1543  U.S.A.

Office: Campbell Hall 2224B  Deliveries/Courier: 3125 Campbell Hall
Campus Mail Code: 154302     Web: www.linguistics.ucla.edu/people/cschutze
Phone: (310)995-9887         Fax: (310)206-8595