I apologize if this posting has gone out to the list twice. Jim Kenkel *************************************** I thank Johanna for her helpful reply and for providing the information from Goldberg's site. The description/definition of a construction from Goldberg seems to have weaker criteria than Croft applies. Although Goldberg's notion of "predictability" is easier to follow than Croft's notion of "wholeness," I still have reservations. My biggest difficulty with "go Xing" as a "construction" was that I didn't see it as a "whole." Its interpretation depends crucially on analysis into semantic categories, as the discussion has made clear. I suppose that Croft's goal is to minimize syntactic analysis, but explaining the interpretation and acquisition of these structures does involve hypothesizing distinct semantic categories and then labeling them. Thus, it seems to me that these expressions also have a level of predictability to them. The learner, with little exposure, by applying his semantic analysis of 'activity types' into 'leisure activities' and 'non-leisure activities,' which presumably reflects our cognitive organization, can predict the meaning of a whole class of expressions - which can't be done with "kick the bucket" type expressions. Is it unfair to ask if the "semantics is unpredictable," why does everyone agree so readily on the intepretation of these expressions? Constructions are claimed to be learned as complex lexical items - all of a piece. However, it seems that some kind of formal analysis is necessary for the acquisition of at least some lexical items - for instance, compounds. Pinker points out the children produce both 'mouse catcher' and mice catcher' but don't produce 'rats catcher,' saying only 'rat catcher.' Describing reflexive structures as 'constructions' is much more problematic. I don't think that it is plausible that the interpretation of the reflexive sentences we've been considering can be accounted for without considerable formal analysis. If through frequent exposure to sentences like 1. Mary wants to help herself the learner figures out that 'herself' is coreferential with the first noun phrase/person/"grounded thing" mentioned, how would that learner not misinterpret this sentence without access to some kind of formal analysis: 2. Alice wonders who Mary wants to help herself. To my mind, construction grammar's challenge is to explain how the learner knows that, in spite of surface distributions, 'Mary wants to help herself' is one kind of construction in the first sentence but is a different construction in the second case - and how this is done without some kind of formal analysis. Moreover, in the learning account sketched by Johanna, what kind of input would the learner get which tells her that contraryto surface distributions, 'herself' receives different interpretations in the two sentences? Is this something learners are taught in school? Wouldn't we expect some people to misinterpret the antecedent of 'herself.' Although some may find the sentence to be inelegant, they don't misinterpret the coreference. Johanna asks about the importance of 'learnability' to discussions on this list. I think that important goals of the ATEG list are to try to understand something of the nature of language and to relate that understanding to problems of language use for students. Linguists have observed not only that much of our knowledge of language is both very complex but that it is also not consciously learned. We might say that much of our complex linguistic knowledge comes for free. The accounts of learning attributed to "construction grammar" place a heavy burden on the learner without providing an account of how learners get the needed negative input or corrections from others. I don't know how such an approach would explain the acquisition of reflexives, for instance. Grammatical descriptions which aren't constrained by plausible accounts of language acquisition are more likely to be artifacts of the grammarian's taxonomy than reflections of the grammar which the learner has tacit knowledge of. I think that presenting these kinds of descriptions to learners adds to their burden, a situation a teacher would like to avoid. At this summer's ATEG conference, Bob Yates and I argued this point with regard to "sentence patterns." We claimed that what students "know" is argument structures of verbs. This is what they learned when they acquired the language. Sentence patterns are taxonomical artifacts but are not reflective of the learner's grammar and therefore present the students with an unnecessary learning burden. A second goal of teaching grammar to native speakers is to help them apply their conscious knowledge of grammar to problems of language use. Robert DeBeaugrande, in his 1984 paper, "Forward to the basics," suggests a number of criteria for writing learner grammars. One of these is that grammatical accounts should mesh with what learners know. For the reasons given above, I am not convinced that anyone "knows" that "Mary want to help herself" is a construction in the sense discussed here. I don't know how conceiving of the whole grammar as "constructions" would help learners trying to apply their conscious knowledge of grammar to challenges of language use. I am not claiming that all aspects of language are learned in the same way. I think that the 'go Xing' are not learned the same way as reflexives. I know that issues of acquisition and learnability are controversial, but I believe that we need to keep these issues in mind when we offer grammatical descriptions to our students. Jim Kenkel, Eastern Kentucky University To join or leave this LISTSERV list, please visit the list's web interface at: http://listserv.muohio.edu/archives/ateg.html and select "Join or leave the list" Visit ATEG's web site at http://ateg.org/