The issue of young children's creativity in syntax has been complicated by the lack of generally accepted criteria for productivity. Some researchers interpret child usage of particular word order patterns, for example, as evidence that young children have adult-like syntactic competence (Poeppel & Wexler, 1993). Other researchers, who have set different criteria for productivity (e.g., contrastive usage), suggest that child grammar is initially heavily based around particular lexical items (Tomasello, 2000). This paper describes a new methodology measuring early syntactic productivity in observational data from a German corpus.
The subject of this study was one monolingual German-speaking boy. Audio recordings started when the child was 1;11.14. Five one-hour recordings were made each week until age 3, supplemented by daily parental diary entries. We assume that this 'dense' sampling method provides us with 10 % of the child's actual output.
In order to assess the issue of productivity in this child's early complex constructions we collected the first 1,000 utterances with at least three different lexical items and tried to trace them back. That is, we took these most complex constructions (occurring between age 1;11.16 and 2;2.10) and analyzed if and how they are related to previously produced utterances. For example, given the complex construction 'auch raus Garten' [too - out - garden], we would i) search for all earlier occurrences of all three items, and ii) try to find the closest match - defined by morphemic similarity - to this target utterance (in our case, 'auch raus' had been said before three times). Having determined the closest match, we measured the similarity between the target utterance and the closest match with the help of three 'operations': 'add on', 'insert', and 'substitute'. For instance, in this example, to get from 'auch raus' to 'auch raus Garten' requires one 'add on' operation.
Using this procedure, we found that 40 % of the first 1,000 3+word utterances require exactly one such operation. 50 % of the complex constructions can be accounted for by the application of two operations (mostly 'add on' plus 'substitute'). The rest consists of utterances that the child had recently already said (= no modification; 5 %), or word concatenations in which none of the items had occurred together before (5 %). The results suggest that children rely to a large degree on previously produced utterances in their creative language use. Furthermore, early creativity can be described in terms of very few simple operations, which indicates early syntactic usage to be more conservative than previously assumed.