anyconstituent;theellipses are typed variablesthatstand for anyrando mstring ofconstituents, including a nullstring.)
- Constraints can impose conditions on the relation between different stages in the derivation. As mentioned
above, the early formulations of movement lacked traces, so constraints on movement had to be stated on
stages related by a derivational step. For instance, the equivalent of (15) in this earlier framework (e.g. Ross
1967) would be stated as (16).
Among morecontemporary theories, Optimality Theory (OT) proposes thatderivations involve exactlytwolayers: the
Input (which plays approximately the role of underlying form) and a large collection of candidate Output (or surface)
structures. The basic principle of OT is to impose constraints on the relation between these two layers, choosing the
candidate Output that best meets the constraints. For instance, the“Parse”constraint stipulates that every element in
the Input structure should appear in the Output structure. However, constraints in OT are regarded as violable.
Therefore, under certain conditions when other constraints override Parse, some element of the Input will fail to
appear in the Output; this is how OT treats what would be treated as a deletion in a theory with derivational rules. A
major difference between OT and a derivational theory is that OT applies all constraints simultaneously to the
Input–Output pair, whereas a derivational theory sees the relation between underlying and surface forms as a
potentially unlimited sequence of simple steps.
- Alloftheruleswehavediscussedso far involvesyntacticstructurealone.Anotherkind ofconstraintimposes
conditions between two structures of disparate types. This type of constraint, which I will call acorrespondence
ruleor aninterface