How Do Children Develop Syntactic Representations from What They Hear?

Children learn language from what they hear. In dispute is what mechanisms they bring to this task. Clearly some of these mechanisms have evolved to support the human speech capacity but this leaves a wide field of possibilities open. The question I will address in my paper is whether we need to postulate an innate syntactic module that has evolved to make the learning of language structure possible. I will suggest that more general human social and cognitive capacities may be all that is needed to support the learning of syntactic structure. I start by briefly discussing precursors to language development that are developing in the first year of life: some of these are probably primate-wide skills, for instance, the capacity for distributional learning (e.g., [1]), others are probably in large part, human-specific, for instance the highly sophisticated socio-cognitive skills that one-year-olds already show (e.g., [2]). Next, I outline an approach to language development that involves the learning of constructional schemas, both specific and abstract. Children are thought to start out with concrete pieces of language and to gradually develop more schematic constructions. All constructions are mappings between the form of the construction and a meaning, though this may not be either the full meaning or the full construction of the adult grammar. For instance a child may say Whats that? for months, perhaps as a request for the name of an object or perhaps as a way of getting attention without connecting the clitic -s to any representation of the verb, to BE. As their language develops children (1) learn more constructions (2) develop slots in constructions as they notice variation (3) abstract a more schematic meaning for each slot, making the constructions more abstract (4) add more slots to constructions making them more complex and (5) relate constructions to each other through distributional and analogical processes. Many previous studies of language development have argued that children could not learn from the input because there is no surface guide to underlying constituency. In support of this, they claim that there is empirical evidence that childrens grammars are abstract from the outset. There are two major problems with assessing such claims. First, until recently, most empirical studies of language development have been conducted on very thinly sampled data. This makes it difficult to know whether either relatively infrequent utterances or the complete absence of an utterance is due to chance sampling or is really indicative of development [3]. In what follows, I report research that has largely been