Séminaire de Recherche en Linguistique

The Final-over-Final Constraint (FOFC) is a syntactic generlisation, probably a universal, formulated as follows (see Biberauer, Holmberg & Roberts 2011, BHR, and the references given there):

(1)       Within a single extended projection, a head-final phrase can only dominate  another head-final phrase.

(1) excludes the schema in (2) inside a single extended projection, where αP is the complement β and γP is the complement of α:

(2)       *[_βP [_αP   α  γP ]  β  ]

BHR give evidence that (2) holds in syntax, some of which I will summarise in this talk. The main goal of the talk, however, is to investigate how and whether FOFC holds in morphology. There are four cases to investigate: (i) the general cross-linguistic suffixing preference, observed by Hawkins & Gilligan (1988); (ii) the incidence of FOFC in inflectional morphology, (iii) the incidence of FOFC in derivation and (iv) in compounding. Here I will largely concentrate on (i).

            Hawkins & Gilligan (1988, henceforth H&G) observe the “suffixing preference”. In particular, picking up the observation in Greenberg’s Universal 27, they observe that “[l]anguages with VO and/or Pr+NP word orders in their syntax regularly have  prefixes and/or suffixes in their morphology. But in a suggestively large number of cases, languages with OV and/or NP+Po have suffixes only” (219). H&G look at a database of 16 morphological categories in approximately 200 languages. The general conclusion shows a clear suffixing preference in general, and a much more marked preference in OV languages as compared to VO languages: “the prefixing/suffixing ratios in VO and Pr+NP languages are 34%/66% and 33%/67% respectively, i.e. roughly 2 to 1 in favor of suffixing; and 13%/87% and 16%/84% in OV and NP+Po languages respectively, i.e. between 5 to 1 and 7 to 1 in favor of suffixing” (230). This can be readily accounted for: if prefixes are heads, then a prefixed word is a head-initial category. If the word-phrase boundary is transparent, then a prefixed word constitutes a FOFC violation in a head-final phrase, i.e. in a system with head-final syntax. This situation does not exactly instantiate the schema in (2) since γ, the complement of α, is not phrasal, but that detail is irrelevant if we take bare phrase structure seriously. The structure in (3), which is the case of a prefixed word in a head-final phrase, violates FOFC:

(3)       [_βP  [_W  Prefix  Root ]  β ]

FOFC thus explains the all-but-categorical nature of the suffixing preference in OV languages. For VO languages, nothing forbids suffixing; in other words, the mirror image of (3), shown in (4), is not a FOFC violation although it is, in a sense, an instance of disharmonic word order:

(4)       [_βP β [_W  Root Suffix ]  ]

Several questions arise from this account: (i) Why is there nonetheless a 2-to-1 ratio in favour of suffixing in VO languages? (ii) what is nature of the residue of the suffixing-preference violations in OV languages? (iii) does this account extend to derivation and compounding? (iv) what lies behind FOFC itself? Regarding (i), I suggest the answer is partially diachronic. Regarding (ii), there is a skewing among affix types, with agreement markers and negation showing almost no preference: this can be accounted for treating these elements as non-heads. I will show that FOFC is relevant for compounding, but not (in general) for derivation. Finally, I will speculate on point (iv).