child language

Brian MacWhinney macw at
Wed Feb 16 21:49:33 UTC 2011


    This examination promises to be better than the Super Bowl.  And, unlike the Super Bowl, where my home team lost, I don't have strong
bets on either side.  

Seriously, this is an important issue and the more we can articulate the relevant factors involved, the better for theoretical development.  I am happy to see no reference here to Broca's aphasia, with the focus instead on comparing diachrony, SLA, and child language.  My understanding of Dan's earlier emphasis is that he denied the link or correspondences (?) between diachrony and child language. Perhaps he can clarify his position vis a vis SLA.  In regards to SLA, it is important to nominate corpora concretely.  One classic corpus, which is available in TalkBank on the web, is the Klein-Perdue ESF corpus.  But maybe Tom has something else in mind.  Within the ESF, there are lots of subtypes.  The most revealing would be those that involve learning of a Indo-European language from speakers who have non-Indo-European sources (such as Turkish-German or Arabic-French).  Regarding child language corpora, I assume we can dispense with interlinear glosses for English.  For Hebrew, we should soon have a full morphemic line courtesy Shuly Wintner and Bracha Nir.  Instead of Swahili, I would recommend the Demuth Sesotho corpus.  For Spanish, nearly all of the corpora have already been morphemicized.  There is also a grammatical relations tier for the syntactic structure.
On the substantive front, it seems to me that some of these predicted parallels amount to foregone conclusions.  One can take as a general developmental principle for both biology and mind the fact that simple things precede complex things, or that combinations do not arise before their components.  Manfred Pienemann's Processibility Theory or Kim Oller's phonological development theory are cases in point.  One hardly needs more than natural compositionality to predict some of these parallels.  However, some of these predicted parallels cannot be so simply reduced.  More importantly,  do these parallels also work in the same way vis a vis diachrony, where the starting point is already complex?

Good luck in this analysis,

-- Brian MacWhinney, 

On Feb 15, 2011, at 10:13 PM, Tom Givon wrote:

> Dead FUNK people,
> In 2002 I accepted for publication, in a TSL volume I edited, an article by my good friend Dan Slobin. I disagreed strongly with what he had to say, but felt that one might as well have a vigorous discussion. In that article, Dan challenged the meaningfulness of a whole research programme, that of seeking similarities between early child language (1-2 yrs) and other developmental ('emergent') processes such as language diachrony, SLA-pidginization and language evolution. A week or so ago, Dan cited that article, once again implying that it was the last word on the subject, and thus that the discussion should be closed.
> In 2009 I published a whole book ("The Genesis of Syntactic Complexity", Amsterdam: Benjamins) suggesting, and trying to show, that the discussion was far from closed, and that seeking similarities between the four developmental trends of language is both theoretically sound and empirically feasible. At the time, somewhat tongue-in-cheek, I said I was sorely tempted to dedicate the book to my old teacher Noam Chomsky, given his infuriating treatment of the subject in a 2002 article (Hauser et al. 2002). In retrospect, I see that I should have dedicated the book to Dan Slobin, since both he and Chomsky were the proximate catalysts for writing the book; and unlike Chomsky Dan is still my good friend.
> In his recent FUNK contribution, Dan leveled his guns most specifically at the pernicious suggestion (Givon 1979; Bickerton 1981) that the similarities between early child language and SLA pidgin could be legitimate and revealing. I would like to briefly sketch out the beginning of an answer, and also give a promisory note for a more extensive, quantified, text-based study that will back up my claims--tho of course I don't intend to close the discussion once and for all even then. My investigation of this topic started in 1979, in ch. 5, 7 of "On Understanding Grammar" ( NY:Academic Press). I set up there a comparison between some of the main structural and communicative properties of pre-gram,matical vs. grammatical communication. After 30 years of tinkering, that comparison may be summarized as follows:
> 1. Lexicon over grammar: Pre-grammatical communication relies maximally on the lexicon, leaving communicative intent largely un-coded and thus heavily dependent on contextual inferences. In contrast, grammatical communication codes much more of the communicative intent via grammar.
> 2. Coherence scope: The thematic and/or topical coherence units of pre-grammatical communication are much shorter. In early child language, for example (ca. Lois Bloom's 1970/1973 1-word stage, ca. 1 yr old), it approximates 1-clause, thus "mono-propositional coherence". It gradually gets longer, and is clearly multi-propositional by the so-called 2-word stage (Bowerman 1973, ca. 2-yr).
> 3. Dependence on care-taker turns: In early child language (Ochs et al. 1979) the one-word contributions of the child are complemented/supplemented, both syntactically and communicatively, by the adult interlocutor's contributions.
> 4. Context dependency: Early child communication is, therefore, much more heavily context-dependent.
> 6. Speech acts: The ratio of manipulative vs. informative (declarative, question) speech acts will be highest in early child language, and will decrease over acquisition.
> 6. Noun/verb coding ratio:  The ration of N/V in early child comminication (e.g. Lois Blooms 1970/1973 transcipts) is highest, and decreases during acquisition.
> 7. Grammatical morphology: The use of productive (as against frozen) grammatical morphology is minimal in early child language, and increases during acquisition.
> 8. Grammatical constructions: In early child language, in part due the 7. above, syntactic constructions (speech acts, voice, subordination, etc.) show low differentiation. They emerge and differentiate gradually during acquisition.
> 9. Syntactic complexity: Overall, early child language shows minimal syntactic complexity, which then increases gradually during acquisition.
> All these properties can be studied empirically in a quantified manner--in transcripts of natural oral communication. The relevant comparison with the adult grammatical register should obviously involve informal oral communication. None of these properties are absolute either/or; rather, they are all a matter of degree. They must be studied language by language, and one would predict that in languages with more-regular morphology the acquisition of productive (as against frozen) morphology would come earlier. But one would still predict the gradient suggested in 7 above. Further, one could predict that in languages where the adult input of verbs obligatorily comes with some morphology (Hebrew, Spanish, Navajo), lexical verbs will be acquired initially with some frozen morphology; but that morphology would not be productive at the earliest stages. Likewise, one could predict that in languages where the adult input of nouns obligatorily comes with some morphology (Bantu), nouns will be acquired early with frozen morphology; but that the richness of Bantu nominal-class morphology, and in particular its extensive ramifications into the grammar of both NPs and VPs, will be acquired later and gradually.
> So, I hereby promise to find the time to do this kind of  quantified comparative study, and present the result for y'all to inspect & decide. I promise to include at least 4 languages (Hebrew, Spanish, English, Swahili), as well as English-based SLA-pidgin. While this is obviously an impoverished typological sample, it does include some of the major relevant typological parameters. And one needs to know the languages in order to do the study (the CHILDES transcripts, alas, come without inter-lineal glosses). I can only hope that other people may want to pick up the gauntlet and extend this study to other languages.  As in other sciences, theoretical arguments can only be resolved by well-design empirical investigation. The worst thing one would want to see in child language studies is a premature closure.
> Best,  TG

More information about the Funknet mailing list