It is generally assumed that recurrent deviations from ordinary language in poetry have a non-complex relationship to ordinary grammar. This assumption has been formulated by Fabb (2010) as the Development Hypothesis (DH). In this paper, DH is elaborated within a generative framework and tested upon deviant word order in a sample of 19th century Swedish poetry. The result is that the hypothesis is fairly well corroborated, although not totally. In closing, an alternative hypothesis by Thoms (2010), the Non-Uniformity Hypothesis, is tested. It claims that poetry has more of its own syntax. This hypothesis is shown to yield, on the whole, predictions just as good as OH. Neither of the hypotheses, however, lasts the entire course.