35:10-37:00 (and also up to 38:00) of

https://www.youtube.com/watch?v=_HfEKKkaTdk&index=3&list=PLeXljLyrouzdmJ29eYOZgQ3jE5MZcy7jX

is also interesting.

The latter video is reproduced as the Appendix of Ludlow's The Philosophy of Generative Grammar.

I have things to say about the relevant issues from the perspective of language faculty science as I pursue it in my forthcoming book, and I will try to make some postings later.

Chomsky once said, 20+ years ago, as in the "Generative Enterprise" interview, that anaphora is one of the few effective probes into properties of the language faculty.
His current position, as stated in the interviews in "Science of Language" and also in the above interview with Peter Ludlow is that anaphora is right on the border between the inside and the outside the language faculty, being on the outside. I understand that his conceptual reason is that it seems to involve a non-local relation. I also understand that his empirical reason is that the informant judgments can be affected by pragmatic considerations.

Being an internalist, I ultimately do not believe anything unless I can test it in my own Lab (i.e., against my own judgments) and unless I see the proposed hypotheses in fact deduce definite and testable predictions that get supported experimentally (i.e., in a single-researcher-informant experiment with myself being the sole informant). (For the reasons alluded to in the paragraph quoted below, which is the last paragraph of Appendix II of my book draft, the relevant testing can involve a rather complicated network of judgments, along with an interaction of several or more interrelated hypotheses.) I thus naturally look at works that deal with Japanese. Based on what I have been able to determine, there is no compelling evidence in Japanese in support of the manifestation of local operations of the sort that Chomsky has been addressing, if we are engaged in testability-seeking research rather than compatibility-seeking research. The research that deals with Japanese either fails to deduce definite and testable predictions (or fails to articulate how such predictions can be put to rigorous empirical test) or the predictions they make fail, rather miserably, to be supported experimentally.

The only area where we can deduce definite and testable predictions (that have a promise of surviving a (minimally) rigorous empirical test) in Japanese, as a manifestation of universal properties of the language faculty, has to involve dependency interpretation (based on the LF c-command relation), as far as I can determine at this point. The definite prediction is about the complete impossibility of a certain dependency interpretation corresponding to any Example instantiating a certain schema and the absence of such complete impossibility corresponding to some Examples instantiating a schema that is minimally different from the first schema. The schemata are constructed on the basis of the hypotheses under discussion.

I have been able to replicate the results in my single-informant-researcher experiment in multiple-non-researcher-
informant experiments -- crucially making reference to the results of Sub-Experiments that serve as a basis of informant classification. And that will be part of the empirical discussion in my book.

I would like to note that I have also been able to obtain experimental results in accordance with our definite predictions with regard to the Weak Crossover cases in English -- with a language-particular hypothesis that the dependency interpretation in question is an effective probe into properties of the language faculty at least for many speakers of English. None of the informants who have been classified as "reliable informants" with regard to the validity of the hypotheses under discussion, on the basis of the relevant Sub-Experiments, accept the Example sentences that are predicted to be impossible with the relevant dependency interpretations. It is interesting to note that some of those informants accept the coreference that is predicted to be impossible, such as the cases of Condition C/D violations. (It is possible that one might be able to devise more effective Sub-Experiments that would allow us to focus on the informants whose reported judgments are significant with regard to the validity of Condition C/D. But because of the absence of clear effects of Condition C/D in Japanese, I do not think such will turn out to be the case in the end.)

My position at this point is that:
(i) anaphora (and scope dependency as well, as Chomsky pointed out years ago) will in fact be an effective probe into properties of the language faculty although much more rigorous and careful work has to be carried out than we typically see practiced in the field, if we are to obtain experimental results in accordance with our definite predictions, which would involve an articulation of how our predictions are deduced and how our experiment is designed accordingly
(ii) it is only through such rigorous and careful research we will be able to determine whether something that seems to involve non-local relations is indeed part of the language faculty and/or how it can be revealing about properties of the language faculty.

As I state in my book, I understand this to be a consequence of taking the internalist view and the methodological naturalist view seriously.


The last paragraph of Appendix II of my book draft.
" As stressed above, the replication of particular judgments by informants on a set of particular Examples is not our concern. We are concerned ultimately with the replication of our experimental results at a more abstract and general level. We are interested in finding out universal properties of the language faculty. We have chosen to work with a dependency interpretation as a probe for that purpose; see Chapter 3 for a conceptual basis for our choice. What type of dependency interpretation can be a good probe for the purpose may differ among languages, and even among speakers of the "same language." In our experiments dealing with individual speakers of a particular language, we check predicted schematic asymmetries given rise to by universal hypotheses, language-particular hypotheses and bridging hypotheses. It is the universal hypotheses among them that would help us see what universal properties underlie individual informants' judgments on Examples of "different constructions," with "different dependency interpretation," in "different languages." Before we begin to be able to address replicability of our experimental result at such an abstract and general level, however, a great deal of work has to be carried out dealing with particular languages, starting with the establishment and the accumulation of confirmed predicted schematic asymmetries, first in a single-informant experiment and ultimately in multiple-non-researcher-informant experiment."