Input and Glue in OT-LFG

In glue logic (Dalrymple 1999, 2001, Asudeh 2004), LFG has found a reasonably stable but developing method for connecting syntactic structures to formal semantic intepretation. However current formulations of glue logic make the rather unsatisfactory move of divorcing grammatical features from any intrinsic connection with what are usually regarded as their meanings. Rather than directly connecting a grammatical feature attribute-value combination such as [NUM PL] to a meaning-constructor that produces a plurality meaning, current glue simply co-introduces a grammatical feature and any associated meaning constructors in the entries of all lexical items introducing the feature. Not only does this lose any account of important traditional observations about how morphology and semantics are related, it also creates an incompatibility with current OT-LFG, which depends on the idea of f-structures having intrinsic semantic content (Kuhn, 2003, pp. 64-65,74). The lack of a means to connect OT-LFG to a formal semantics leaves it incomplete not only in comparison with its contemporary competitors, such as Categorial Grammar, HPSG, and conventional LFG, but even in its own terms, in that there are significant constraints which can’t be formulated without explicit semantics, most noticeably scope-based ones such as Bresnan’s (2001b) faithneg, or the ‘respect-the-tree’ constraints that one would plausibly use to account for scoping modifier and light verb constructions such as those