<?xml version="1.0" standalone="yes"?>
<Paper uid="J84-3003">
  <Title>On Two Recent Attempts to Show that English Is Not a CFL 1</Title>
  <Section position="3" start_page="0" end_page="0" type="concl">
    <SectionTitle>
3. Conclusion
</SectionTitle>
    <Paragraph position="0"> I think there are lessons to be learned from this admittedly negative review. It is clear that linguists have not succeeded in developing strong and predictive theories of what belongs to the domain of syntax and what belongs to semantics. Rather than attempting to discern the status of each new fact as it becomes crucial to some dispute, we ought to be developing general theories of language from which the correct conclusions follow in a principled way.</Paragraph>
    <Paragraph position="1"> Notice, in the present context, that both the arguments I have reviewed relate to the topic of anaphora. In both cases, I have presented evidence against the assumption that certain anaphoric elements are syntactically constrained to be identical and clausally adjacent to their antecedents (section 1) or to be present in the string (section 2). I suspect that it is generally true that anaphoric devices that can be controlled across sentence boundaries in discourse are never subject to any intrasentential constraint on identity or overt presence. As things stand, however, this is nothing more than a hunch. One conclusion we can draw from the present discussion is that we are in need of a general and widely accepted theory of the syntax and semantics of anaphoric devices. 4 A second conclusion I would draw is that it is time to start applying to semantically interpreted linguistic systems the kind of mathematical analysis that so far is mostly conducted with regard to stringsets. We know little about what mathematical or computation power is inherent in particular systems that do not merely generate sets of strings but pair strings with representations of meanings that are appropriate to particular situations. 5 It would be useful to have more clear results about combined syntactic and semantic systems, since no one doubts that it is the entire mapping between structure and meaning that linguists are ultimately interested in. Moreover, it has been argued fairly convincingly that some natural  result, observed by Len Schubert: the set of sentences assigned denotations by a semantics associated with a CF-PSG can be non-CF. Partee and Marsh (1984), stimulated by Higginbotham's paper, discuss a problem that has potential relevance here. They note that the set of predicate calculus formulae with no vacuous quantification is not a CFL, and conjecture (but are not able to prove) that it is not even an indexed language.</Paragraph>
    <Paragraph position="2"> Computational Linguistics, Volume 10, Numbers 3-4, July-December 1984 185 Geoffrey K. Pullum On Two Recent Attempts to Show that English Is Not a CFL languages cannot be described by a CF grammar in a manner that allows a suitable syntax-to-semantics mapping to be defined (see Bresnan, Kaplan, Peters, and Zaenen (1982) on Dutch)/' ~'Shieber (in press), which I saw in a preliminary version after completing this paper but have not seen in its final form at the time of going to press, extends the result about Dutch to make a much stronger claim about a related language, Swiss German, namely that it has a syntactic analog of the Dutch pattern and overall is not even CF. Moreover, Culy (in press) also has evidence of a natural language (Bambara, spoken in West Africa) that appears to be other than CF. This changes the background to the current dispute a lot, of course. If Shieber and Culy are right, there are some aspects of the syntax of some natural languages that call for parsing by a device with greater than CF power. What I have said above about English remains true, of course, but although there may be general facts about how anaphora works that could have enabled us to predict this, we cannot predict it simply from the proposition that universal grammar does not allow supra-CF grammars. This makes me much less confident about being able to answer the rejoinder to this article that Langendoen and Postal publish in this issue. Although their facts, which 1 saw just as this article went to press, are not totally convincing (because the repeated-head-noun appositive relative construction they discuss is so awkward and unnatural even at the best of times), 1 now do not see why in principle they might not be right. Perhaps I have been speaking and writing a non-CF language all these years, and simply hadn't realized it, like Moli6re's M. Jourdain, who didn't realize he had native competence in prose.</Paragraph>
  </Section>
class="xml-element"></Paper>