Novelty

[N. Chipere]

The issue of linguistic novelty is a key theme in my current research and
I would like to share my thoughts on the issue as well as some of my
experimental findings. I hope I will be forgiven for the somewhat long
message, but I need to get some feedback.

On the face of it, the statement that language users can understand novel
sentences appears obvious and redundant. However, the statement
serves an important function of constraining  theories about the
nature of linguistic knowledge and ultimately, about the nature of the
human mind. The basic argument (Fodor & Pylyshyn, 1988) is as follows:

If knowledge of language is considered to be a list of sentences, then
there is no way to account for the ability of native speakers to produce
and understand sentences they have never heard before. On the other hand,
this ability can be explained if knowledge of language is seen as an
infinitely generative set of grammatical rules. And if  it is accepted
that native users of a language possess generative grammars, then certain
important constraints on theories of cognitive architecture must be
observed. Without going into the details,  observing such constraints
leads to the hypohesis that the mind has the general architecture  of a
digital computer. So when it is said that native speakers of a language
can understand novel sentences, a deeper statement is being made, it seems
to me, about the nature of linguistic knowledge and the about nature of
the human mind.

However, it doesn't follow from the fact that native speakers can
understand novel sentences that knowledge of language takes the
form of a generative grammar. Everyday, human beings do things they have
never done before but this ability does not lead to the conclusion that
their actions are the product of generative rule systems. It's quite
reasonable to suppose that the ability to deal with novel situations
depends on previous experience and that novel situations will become more
difficult to deal with the more they stray from the range of an
individual's experience.

This line of thinking forms the basis of an experiment which I carried out
to test the connection between novelty and generativity. According to the
line of thinking outlined in Fodor & Pylyshyn (1988), all the sentences of
a language belong to a generated set, and they should be equally
comprehensible to native speakers. That is to say, since what a native
speaker knows about his or her language is a set of rules capable of
interpreting and producing any sentence in the language, a native speaker
should be able to understand all possible sentences in his or her
language equally well, provided that performance factors are taken into
account. On the other hand, if linguistic knowledge, like other kinds of
knowledge, depends on experience, then native speakers of a language
should find familiar sentence types easier to understand than unfamiliar
ones.

I compared the ability of three groups of subjects to understand
grammatically unusual sentences under conditions in which
memory load was eliminated. Group 1 consisted of graduate native
speakers of English, Group 2 consisted of graduate non-native speakers
of English and Group 3 consisted of non-graduate native speakers of
English. The subjects were asked to answer comprehension questions about
sentences with highly unusual syntactic structures, such as:

1. The doctor knows that the fact that taking good care of himself is
   essential surprises Peter.

   example question: What does the doctor know?

2. The bank manager will be difficult to get the convict to give a loan
   to.

   example question: Who will find it difficult to do something?

3. The lady who Peter saw after overhearing the servant proposing to
   dismiss had lunch in a cafe.

   example qustion: Who might be dismissed?

(These sentences may strike many as ungrammatical, but in fact they are
simply unfamilar combinations of familiar sentence types (adapted from
Dabrowska, 1997)).

I obtained both comprehension and reading time data from the experiment,
but I will mention only the comprehension data. The key results were that
the non-native graduates obtained the highest scores, followed by the
native graduates, with the least scores coming from the native
non-graduates. The native non-graduates were also most affected by
plausibility and often ignored syntactic constaints whereas the non-native
graduates were least affected by plausibility and showed the greatest
mastery of syntax. The non-native graduates, by the way, had learned
English largely through formal instruction. Most of them were also
speakers of East European languages, which, I have been told is one
possible explanation for their facility with complex syntax. However, all
groups performed equally well on control sentences, which were formed out
of familar sentence types. These results give empirical support to the
logical argument that there is no necessary connection between novelty and
generativity. It is quite possible to generate a very large number of
novel sentences out of a small number of familiar sentence types. The fact
that most native speakers can understand such sentences does not entail
that they can readily understand all possible sentences in the language.
In other words, being a native speaker of a language and being able to
understand novel sentences in that language does not entail possession of
a generative grammar of the language.

More details about the experimental design, materials, procedure and
results are documented in an experimental report. The report also reviews
previous psycholinguistic findings which indicate that native
speakers of English often lack full grammatical productivity, and that
education appears to be an important variable in grammatical skill. The
main argument developed in the report is that linguistic ability shares
many of the key traits of skilled performance and can be accounted for
without recourse to an infinitely generative set of grammatical rules. I
am keen to have feedback on the report, which I can make available to
anyone who is interested.

Ngoni Chipere
Darwin College
University of Cambridge