No CrossRef data available.
Published online by Cambridge University Press: 12 March 2026
Relational Grammar and Invariant Ordering Grammar both take the linear order of constituents to be a superficial property of sentences which can be predicted by a set of so-called word-order or linearization rules. The constraints which these two theories have imposed on their respective linearization rules are examined, and are shown to make wrong predictions about the possible word orders that active and passive sentences exhibit across languages. Furthermore, it is shown that these constraints cannot be amended in any obvious way to give the desired results. To account for the facts in question, a new mechanism is introduced and briefly argued for.