Docs / Studies / Studies.Sheehan2017b

The final-over-final condition and processing

Sheehan, M. (2017). In The final-over-final condition: A syntactic universal (pp. 79–96). MIT Press

The final-over-final condition bans a head-final phrase over a head-initial one. Read left to right, Hawkins's Early Immediate Constituents prefers the consistent orders of a head over a phrase to the disharmonic ones, and the initial-over-final order to the final-over-initial one, each exactly when the lower head has a complement (ex6_to_ex9), while the plain ratio of constituents to words ties the two disharmonic orders. Polar particles and complementizers show the skew without the preference, and the model ranks the four orders consistently with their counts only for a lower head without a complement, where they coincide (Hawkins1994.binary_zero).

Implementation notes

  • The trees of (6) to (9) enter through Hawkins's Mother Node Construction (Hawkins1994.ics). The chapter's English words replace those of Hawkins's (5.3) and (5.4).
  • (10) is stated for the scores. The step to cross-linguistic frequency is Hawkins's prediction that orders are as frequent as their scores rank them, which polarityComplementizer_unexplained states as the consistency of the ranking at a complement length with the counts.
  • In Table 5.2 the complementizer is the upper head and the particle the lower one, so type C, an initial complementizer over a final particle, is the inverse of the final-over-final order.
  • Note 1 reports that Hawkins's later account could derive the asymmetry only with a preference for attaching to the mother. That account reports the plain ratio of constituents to words (its (5.11)), which ties the two disharmonic orders.
  • Sheehan's three chapters of the book are lettered here by title (adverbs, processing, the head-final filter).

TODO

  • Cecchetto's account, on which a backward dependency is phrase-bound, and Sheehan's objections to it need a model of Agree and selection.
  • The articles counterexample prints no counts, and Table 5.1 is not reproducible from WALS (14 and 41 languages against the printed 12 and 38).
  • §§5.5–5.6, the linearization parameter set by the parser and the analogy with Onset Maximization, are prose.

References

Examples (6) to (9)

(6)–(9), pp. 81–82: the verb phrases of (6) and (7) are consistent and score 100%; (8), the verb before a postpositional phrase, scores 75%, and (9), a prepositional phrase before the verb, 42%.

theorem ex6_to_ex9L79–94

Built from , , ,

Harmony and the final-over-final condition, (10)

(10), p. 83: Early Immediate Constituents prefers the consistent orders of a head over a phrase to the initial-over-final order, and that order to the final-over-initial one, each exactly when the lower head has a complement; the second preference is m / (2 (m + 1)) for a complement of m words.

theorem harmony_iff_fofcL107–122

Polar question particles and complementizers, Table 5.2

Table 5.2, p. 92: the final-over-initial order of complementizer and particle is rarer than its mirror image, which is commoner than the consistent order with both heads initial. Early Immediate Constituents ranks the four orders consistently with these counts only at a lower head without a complement, where all four orders score alike.

theorem polarityComplementizer_unexplainedL152–168

Built from