One major controversy in the field of language development concerns the nature of children's early representations of word order. While studies using preferential looking methods suggest that children as early as 20 months represent word order as an abstract, grammatical property, experiments using theWeird Word Order (WWO) paradigm suggest that it is represented as a lexical property until age four. In order to shed light on these contradictions, two types of arguments are developed. First, it is argued that the observations taken to support the lexical hypothesis, based on the WWO paradigm, have been incorrectly interpreted. Second, an experiment is reported using the standard WWO paradigm with minimal changes in the design. Two groups of French children were contrasted (mean ages 2;11 and 3;11). Both groups were found to (i) reproduce WWO at a similar, low rate; (ii) correct WWO at a similar rate, even with pseudo-verbs; (iii) reuse the grammatical, SVO order significantly more often than WWO; and (iv) produce grammatical markers, indicating productive use, in grammatical sentences only. We conclude that empirical evidence converges to support the grammatical hypothesis.