7. The idea that verbal exponents indexing subjects and objects in Basque are all agree- ment markers is not universally accepted, though; see Preminger (2009) and Arregi and Nevins (2008, 2012) for a different analysis.
resumption. This expectation is confirmed, but I wish to caution from the outset that this strategy is probably less common than it may seem. In addition to resump- tion, there are other ways of working around the restriction that ergatives cannot be extracted leaving a gap at the base position; I discussed these other strategies in the appendix to chapter 1, where I showed that, in addition to resumption, they include the use of antipassives, anti- agreement, and nominalizations.
Western Polynesian languages probably offer the best- known instances of resumption in the base position of the A- bar- moved ergative. As I already showed in chapter 1 (and will show again in chapter 11), in Tongan, the ergative has a dedicated resumptive pro- form; this pro- form differs from the resumptive form used with other expressions:8
(18) a. Na‘e langa ‘a e fale ‘e he tokoua. Tongan
pst build abs det house erg det guys
‘The guys built a house.’
b. ‘a e tokoua [na‘a *(ne) langa ‘a e fale]
abs det guys pst rp build abs det house
‘the guys that built the house’ (lit.: the guys that they built the house)
Norcliffe (2009) proposes a resumption- based approach to the alternation of voice forms in Mayan languages, with primary emphasis on Yucatec. In the dis- cussion of Tzotzil in the preceding section, I already mentioned agent focus (AF), a special dependent verb form, which alternates with the normal transitive verb form of the main clause when the subject of a transitive verb is focused, ques- tioned, or relativized. Observe the alternation in Q’anjob’al: (19a) is the baseline sentence, (19b) shows that the absolutive object can extract leaving a gap at the extraction site, and (19c) shows that this option is unavailable for the ergative subject:
(19) a. Max y- il[- a’] naq winaq ix ix. Q’anjob’al asp erg3- see- tr clf man clf woman
‘The man saw the woman.’
b. Maktxeli max y- il[- a’] naq winaq _ _ i ? who asp erg3- see- tr clf man
‘Who did the man see?’
c. *Maktxeli max- ach y- il[- a’] _ _ i ix ix?
who asp- abs erg3- see- tr clf woman (‘Who saw the woman?’)
8. Other arguments in Tongan must be resumed with the pronominal ai, accompanied by a preposition if necessary (see Chapin 1974 on the distribution and origins of ai).
In order to ask a question about the agent, the verb has to appear in the special AF form (for details, see Coon et al. 2014):
(20) Maktxeli max- ach il- on- i _ _ i ? Q’anjob’al who asp- abs2 see- af- intr
‘Who saw you?’
With respect to Yucatec, Norcliffe proposes that wh- questions and focus construc- tions involve (pseudo- )clefts, and thus are based on relative clauses. The fronted ele- ment (a wh- word or a focus expression) serves as a predicate, taking the embedded clause as its argument. That argument is formed by a silent- headed DP with a relative clause attached:
(21) a. T- inw- il- ah [DP le [NP máak [CP h- súut]]=o’]. Yucatec asp- erg1- see- suf det man asp- return=deic Mayan
‘I saw the man who returned.’
b. [PredP Maax] [DP [NP NULL [CP h- súut]]=o’]?
who asp- return=deic
‘Who returned?’ (Norcliffe 2009:138)
Norcliffe treats the Yucatec ergative marker as a pronoun rather than an agreement marker. In her analysis, this pronoun functions as a resumptive form in a long- distance dependency. In contrast, when a verb appears in the AF form, there is no resump- tion. Thus, Norcliffe suggests that the alternation between the transitive form with the ergative marker and the AF form without it should be analyzed as an alternation between extraction with resumption and extraction with a gap. On this approach, the difference between Mayan languages with AF and other languages with resumption is simply in the locus of the alternation. Rather than employing a resumptive clitic or agreement marker (as in Polynesian or Northwest Caucasian), the Mayan alternation involves two verb forms: one that carries a morphologically dependent subject pro- noun (the regular transitive form) and one (the AF form) that does not.
Mayan languages are not all alike, and the AF construction has been analyzed in a number of ways (see Coon et al. 2014 for a recent analysis that does not rely on resumption, and see Stiebels 2006 for an extensive overview of the issue). It is possi- ble that the grammar of AF across Mayan is not uniform, so Norcliffe’s approach may ultimately be challenged. However, inasmuch as this resumption- based approach is plausible, it allows us to expand the range of morphosyntactic loci of resumption and confirms that resumption is needed to extract an ergative constituent.
In Salish languages, absolutive subjects and absolutive objects extract directly, leaving a gap at the extraction site. Such direct extraction is impossible for both ergatives and oblique objects. However, both types of expression can be extracted if the embedded clause is nominalized (Gerdts 1988; Galloway 1993; Wiltschko 2006a, b). Although this pattern does not show resumption, it nevertheless rein- forces the notion that ergative expressions and PPs pattern identically with respect to extraction.