One major controversy in the field of language development concerns the nature of children's early grammatical knowledge. This paper focuses on the early representation of word order. It questions the validity of the results obtained with the Weird Word Order methodology (Akhtar,1999) inwhichchildrenare presented with un grammatical sentences. These results have previously been considered as major evidence for the constructivist,usage-based approach to word order development according to which young children initially encode word order as a verb-specific lexical property which only slowly develops into abstract knowledge at age 3 or 4 (e.g., Abbot-Smith et al., 2001; Matthews et al., 2005, 2007). The critical review presented here addresses various problems with the results and their interpretation. The discussion questions the relationship between theory and data as well asmethodological issues related to thesm all number of observations and the discarding of data not missing at random. It is argued that the data not only fail to support the constructivist account, but they actually bring evidence for the alternative hypothesis according to which children, from early on, represent word order abstractly.