Why do L1 and L2 learners usually achieve different levels of language mastery? Previous accounts have highlighted the role of biological [1], neural [2] or cognitive [3] differences between children and adults. Here, we focus on another possible difference, the linguistic units adults learn from and how these might differ from the ones children use. Specifically, we suggest that adults might learn from more segmented representations - with word boundaries more clearly marked - while children may begin with larger, less segmented ones that cross word boundaries. We hypothesize that learning from more segmented representations may hinder learning about grammatical relations between units.

We focus on learning grammatical gender, an aspect of language L2 learners have difficulty with [4,5]. Researchers have suggested that children initially treat the determiner and noun as a single unit [6,7]. Adults, in contrast, may treat them as separate units from early on (given previous knowledge and their learning environment). If some of adults' difficulty is related to the units they employ, then manipulating these units should change learning patterns. We show that adult learning of grammatical gender in an artificial language was facilitated when learners were first exposed to sequences of language where the determiner and the noun were less differentiated.

We exposed 32 native English-speakers to a verbally presented artificial language consisting of 14 novel labels for familiar objects (e.g. piano), a carrier-phrase and two determiners. Each noun appeared with only one determiner (there were no phonological or semantic cues for class membership). We collected two learning measures: accuracy of determiner production when prompted by a picture and accuracy of choice between two alternative descriptions of a given picture (one with the correct determiner and one with the incorrect one). We compared learning in two conditions. In the sequence-first condition, participants first heard a block of full sentences (carrier-phase+determiner+noun) and then a block of noun-labels (bare-nouns without determiners). In the label-first condition, participants heard a block of noun-labels and only then a block of full sentences. By the end of the experiment, both groups had received exactly the same input, but in different orders. If starting from larger units facilitates learning, learning should be better in the sequence-first condition.

As predicted, participants in the sequence-first condition showed better learning. They were more likely to produce the correct determiner-noun sequence for a given noun (40% vs. 29%, t(31) = 3.10, p < .001), and more likely to detect a mismatch between the determiner and noun (62% correct, above chance, t(15) = 3.53, p < .001, vs. 54%, at chance, t(15) = 1.47, p > .1).

Our results suggest that at least part of L2 learners' difficulty is related to the units they learn from. We propose a learning theory account that underlines the way learning units individually may block later learning about the relations between them [8], and discuss ways of extending these ideas to other linguistic domains L2 learners struggle with (e.g. verb-preposition pairing, idioms). We discuss implications for models of first and second language learning.

References:
1.Lenneberg, E. H. (1967). Biological Foundations of Language. Wiley.
2.Kuhl, P.K. (2000). A new view of language acquisition. Proceedings of the National Academy of Science. 97, 11850-11857.
3.Newport, E.L. (1990). 'Maturational constraints on language learning. Cognitive Science, 14, 11-28.
4.Scherag, A., Demuth, L. Roesler, F., Neville, H.J. and Roeder, B. (2004). The effects of late acquisition of L2 and the consequences of immigration on L1 for semantic and morpho-syntactic language aspects. Cognition, 93, B97-B108.
5.Lew-Williams, C., & Fernald, A. (2007). Young children learning Spanish make rapid use of grammatical gender in spoken word recognition. Psychological Science, 18, 193-198.
6.Carroll, S. (1989). Second-language acquisition and the computational paradigm. Language Learning, 39, 535-594.
7.Chevrot, J. P., Dugua, C., & Fayol, M. (2008). Liaison acquisition, word segmentation and construction in French: a usage-based account. Journal of Child Language, 1-40.
8.Kamin L.J. (1969). Predictability, surprise, attention, and conditioning. In: Campbell B, Church R (Eds). Punishment and Aversive Behavior. Appleton-Century-Crofts: New York.