The Implications of Lexical Exceptions for the Nature of Grammar

Lexical Classes in Phonology

“…Lexical strata result from splitting up individual constraints into stratumspecific indexed versions (Fukazawa 1998, Ito and Mester 1999, Ito and Mester 2003 formal implementation within OT is the 'co-grammar' approach advocated by a number of researchers including Anttila 2002, Inkelas, Orgun and Zoll 1997, and Ito and Mester 1995b. In this view, lexical strata result from a class of slightly different stratum-specific grammars ('co-grammars'), with different constraint rankings of constraints, that collectively form a family of grammars.…”

“…Ito and Mester 1995a, with further development in Fukazawa 1998 andMester 1999, among others, argued that the distinguishing formal property of lexical strata in an optimality-theoretic grammar is stratum-specific faithfulness ranking, within an otherwise invariant hierarchy of markedness constraints (see Inkelas, Orgun and Zoll 1997, Pater 2000, and Anttila 2002 for interesting alternative proposals sharing the same basic approach). This results in a containment hierarchy of inventories of the kind depicted in (2): The more markedness constraints are active on a given stratum (by dominating stratum-specific faithfulness), the more structures are ruled out, and the smaller the inventory associated with the stratum.…”

A model of lexical variation and the grammar with application to Tagalog nasal substitution

“…I will simply assume in the tableaux below that the lexical entry is used if available. 16 The idea that some words' behavior is determined by their lexical entries and others' is free to be determined by the grammar has a precedent in Inkelas et al 1997, where the distinction is between fully specified segments and segments with underspecified features, for which Ident is irrelevant. There is also similarity to the proposal of Becker (2009) that known words can be indexed to a particular faithfulness constraint, whose ranking determines their behavior (see Pater 2006), but novel words must be assigned an indexation.…”

“…When, in the course of learning, a ranking contradiction occurs, a constraint is chosen to be "cloned", or split into two differently-ranked versions that apply to different sets of words. This approach belongs to a family of frameworks in which lexical items are indexed either to particular constraints or to constraint rankings: see Inkelas et al (1997); Inkelas and Zoll (2007); Itô andMester (1995, 1999);andAnttila (1997, 2002), among others. (21) and (22)) could lead to the creation of two *[m constraints, one ranked high and indexed to /búhaj/, one ranked low and indexed to /balík/.…”

“…In the English place-assimilation case, we could similarly mark femtoliter as an exception to place assimilation. Or, we could allow assimilation to apply only to nasals of unspecified place, letting the nasal of antler be underlyingly unspecified for place, but the nasal of femtoliter be underlyingly labial (following Inkelas et al 1997)-or simply give up on capturing static generalizations and let the underlying form determine the surface pronunciation, with the prevalence of place agreement left unaccounted for, as a lexical accident that plays no role in the speaker's grammar.…”