Sorace (2000): Gradients in auxiliary selection with intransitive verbs #

[Sor00] argues that the perfect auxiliary of a monadic intransitive verb in Italian, French, Dutch and German depends on the verb's place on the Auxiliary Selection Hierarchy (AuxiliarySelectionHierarchy, Table 1): the transitions and states, ordered by decreasing telicity, followed by the processes, ordered by increasing control. Verbs of telic change of location take be and verbs of controlled nonmotional process take have, categorically, in every language and whatever the rest of the sentence contributes; these core types are the two ends of the hierarchy. The types between them allow both auxiliaries, to a degree that depends on their position, and respond to the telicity of the predicate and the agentivity of the subject. Each language draws a cutoff point on the hierarchy between the verbs that take be and those that take have, and the cutoff moves from language to language without reaching the core (§6).

The data are the paper's examples, one row per sentence with the judgment of each auxiliary it prints. The theorems are the generalizations the paper draws from them: the rigid core (changeOfLocation_selects_be, nonmotionalProcess_selects_have), variation confined to the types between (free_alternation_iff), the cutoff point of each language (italian_cutoff, french_cutoff, dutch_cutoff, cutoff_varies), and the sensitivity of verbs of motion to a directional phrase (directional_motionalProcess, italian_directional_split). Italian cuts at the seam of the two halves of the hierarchy, where the states end and the processes begin.

The German examples cross (german_crossing): halten 'last', a verb of continuation of a state, takes haben and rejects sein (17a), while rennen, laufen and schwimmen, controlled motional processes further down the hierarchy, take sein and reject haben under a durative adverbial (38). So no cutoff point separates the German verbs at the granularity of Table 1 (german_no_cutoff). The paper reports the German motion verbs (§4.3) and allows a language to merge types or to divide them more finely (fn. 17, §6); a German cutoff needs the types from continuation of a state to motional process merged into one.

The examples are linked by their verb to the entries of the four fragments. ofVerb places each linked verb on the type the paper gives it, except where the entries draw a finer line than Table 1 (ofVerb_entry). Each fragment records the auxiliaries its reference grammar gives a verb, [MR07] §14.20 for Italian, [BC26b] §§2.1–2.2 for Dutch, [Dur11] §12.3.2 for German and the lists of [GG08] §§811–813 for French, and each linked verb with one auxiliary has the one the paper's judgments prefer (entry_aux_of_prefers). With the entries placed by ofVerb, each grammar cuts the hierarchy exactly where the judgments do (entries_cut_iff_cutoff), and the Italian rule is that cut for every verb (italian_perfect_eq_be_iff).

Implementation notes #

A row's verb type is the type of the section its example illustrates; (36), from §4.2's controlled affecting processes, a subclass of the controlled processes, is a nonmotional process.
The diacritics are read on the Judgment scale: unmarked is acceptable, ? marginal, ?? questionable, ?* and *? unacceptable, * ungrammatical. The paper's ?* lies between ?? and *, which is where unacceptable sits, although the scale glosses that value as a pragmatic failure.
A sentence prefers one auxiliary when the paper prints it with both and judges one better; fn. 3 measures the strength of a preference by the distance between the two. The cutoffs of French and Dutch rest on this: counting a sentence printed with one auxiliary as a choice would add crossings, French rougir 'blush' with avoir (13) below rester 'remain' with être (19a), Dutch duren 'last' with hebben (18b) below blijken 'seem' with zijn (24c).
A row records the departures from the default of its type that the paper points to: a directional or bounding phrase that makes the predicate telic ((35), (39b), (40b), (41b), (41d), (42b), (49), (51)), a durative adverbial that makes it atelic ((4), (11)), an agentive subject ((5a), (15c), (45b), (46b)), or a nonagentive one ((34), (36b), (43), (44b)). The cutoff point is tested on the rows without one.
Omitted examples: (2) and (3), frozen be in Romanian and English, which have no choice of auxiliary; §3.5's anticausatives, dyadic and off the hierarchy, except (30a), the one change-of-location verb printed with have; and (14b), the sense 'be elusive' of échapper, which the paper places on no type.
§5's discussion of the projectionist and constructional models of the lexicon-syntax interface is argument, not formalized here.
A row names its verb by the infinitive, the form of its entry among the monadic entries of its language's fragment, those whose basic frame, the first, has no complement, with the auxiliary their grammar gives them. A verb its grammar gives either auxiliary by meaning is left out, as French paraître (14a) ([GG08] §813a), Italian correre and durare ([MR07] §14.20) and Dutch lopen are.
The grammars are tested on the paper's preferences. (21a) and (21b), bastare 'be enough' and appartenere 'belong' printed with avere alone, express none; the Italian grammar gives bastare essere, as (20b) does, and appartenere either.

TODO #

The entries of tossire 'cough' and squillare 'ring' record no subject entailments, and LevinClass.subjectProfile gives none for Levin's classes of these verbs, so ofVerb places neither. A profile without volition would put (46) and (47) among the uncontrolled processes.

Sorace (2000): Gradients in auxiliary selection with intransitive verbs #

Implementation notes #

TODO #

References #

The core #

The cutoff point #

Directional phrases #

The fragments #