Ergatives are often restricted with respect to their ability to float quantifiers, although as with depictives, the relevant cross- linguistic data are rather scarce. In
genuine agreement with a DP and do not call for an analysis along the lines shown here.
I will present such cases in chapters 7 and 11.
11. A possible confounding factor is that, in some Mayan languages, depictives are in fact embedding verbs which require that the non- finite embedded transitives appear in the AF form (I am grateful to Jessica Coon for pointing this out to me).
seeking to understand the restriction on ergative quantifier float, we can again build on the parallel between ergatives and dative objects as PPs: like ergatives, dative objects are known to be limited in their ability to launch quantifiers (Dryer 1986;
Landau 2010). The generalization seems to be that ergative- floated quantifiers are either impossible altogether in a given language or are possible only if absolutive- floated quantifiers are possible in that language.
Quantifier float has been an immensely popular phenomenon in the linguistic literature, but it is not as common as its theoretical discussions make it appear. First, understanding quantifier float across languages is a challenging task in and of itself.
The challenge has to do with separating D- quantification (associated with deter- miners and attributive expressions such as every, many, etc.) from A- quantification (associated with adverbial expressions such as always, often)— see Bach et al. (1995) and Keenan and Paperno (2012) for a detailed discussion and a broad cross- linguistic overview. Adverbial quantifiers are not directly associated with a particular nominal constituent; hence no quantifier float can be posited.
To take one example, Tongan offers no evidence of floated quantifiers whatsoever (pace Chung 1978), and the word katoa is not a genuine quantifier; I discuss this issue in chapter 10. The Tongan situation is in fact quite representative; even if a language is PP- ergative, it is sometimes impossible to test the differences between PP- and DP- ergative structures because quantifier float is either unattested in that language or has not been investigated.12
Chukchi may be an example of a language in which the ergative cannot float quan- tifiers but the absolutive can. The relevant examples are shown below; the quantifier is əməlʔo “all,” and it can be separated from the absolutive host but not from the
12. In Samoan, which is genetically close to Tongan and is syntactically ergative as well, it appears that the ergative can float quantifiers, but the preference is for the absolutive.
The examples below are from Mosel and Hovdhaugen (1992: 712– 714), who also discuss apparent quantifier float. In (i), the floated quantifier ‘uma “all” is associated with the abso- lutive subject; in (ii), it is associated with the absolutive object, and in (iii), which may be potentially ambiguous, the only possible association is with the absolutive object, not the ergative.
(i) ‘Ua lātou ō ‘uma ‘i Sāmoa. Samoan
pfv 3pl.abs go.pl all to S
‘They have all gone to Samoa.’
(ii) E iloa ‘uma e Seu pese.
aor know all erg Seu song
‘Seu knows (the) songs all.’
(iii) E iloa ‘uma pese e tamaiti.
aor know all song erg children
‘The children know all (the) songs.’
*‘All the children know songs.’ (cf. Mosel and Hovhaugen 1992: 712– 714) However, as with Tongan katoa, it is not entirely clear that ‘uma is actually a quantifier rather than a pluractional adverbial. If it is the latter, the data presented here are simply irrelevant and do not instantiate quantifier float.
ergative one.13 This quantifier can appear either in the unmarked form əməlʔo, which I gloss as such,14 or in the oblique form əməlʔetə. To follow the examples below, it should be noted that Chukchi is not rigidly head- final, and postverbal objects and subjects are freely allowed (see also M. Dunn 1999: 341– 355).
As the following examples indicate, a quantifier can be separated from its host if the host is an absolutive object or subject:
(25) a. ətlʔa- ta əməlʔo jaraŋə ne- čwi- ɣʔen. Chukchi mother- erg all.direct house.abs IPFV- cut- 3sg.3sg
‘Mother divided (lit.: cut) the entire house.’
b. ? ətlʔa- ta əməlʔo ne- čwi- ɣʔen jaraŋə.
mother- erg all.direct IPFV- cut- 3sg.3sg house.abs
‘Mother divided (lit.: cut) the entire house.’
c. ətlʔa- ta jaraŋə ne- čwi- ɣʔen əməlʔo.
mother- erg house.abs IPFV- cut- 3sg.3sg all.direct
‘Mother divided (lit.: cut) the entire house.’
(26) a. əməlʔo ʔitu- t ekwet- gʔe- t.
all.direct goose- abs.pl break.loose- aor.3sg- pl
‘All the geese set themselves free.’
b. ? əməlʔo ekwet- gʔe- t ʔitu- t.
all.direct break.loose- aor.3sg- pl goose- abs.pl
‘All the geese set themselves free.’
c. ʔitu- t ekwet- gʔe- t əməlʔo.
goose- abs.pl break.loose- aor.3sg- pl all.direct
‘All the geese set themselves free.’
Although both forms, direct and oblique, are acceptable with non- absolutives, nei- ther can float in this context, as indicated by the following data. In presenting these ungrammatical examples, I limit myself to a representative sample of the orders made available by Chukchi word- order principles; the quantifier and the host must be adjacent, which rules out a number of possibilities.
(27) a. əməlʔo/ əməlʔetə ŋewəsq- et- te ʔəttə- čg- ən Chukchi all.direct/ all.obl woman- pl- erg dog- big- abs
lʔu- nin.
see- aor.3.3sg
‘All the women saw a/ the huge dog.’
b. *əməlʔo ʔəttə- čg- ən lʔu- nin ŋewəsq- et- te.
all.direct dog- big- abs see- aor.3.3sg woman- pl- erg (‘All the women saw a/ the huge dog.’)
13. This quantifier can also incorporate (Skorik 1961: 325), but this usage, although very common, is irrelevant to the discussion here.
14. M. Dunn (1999: 174) glosses this form as absolutive, but since it can be used with other case forms, I have decided against such a characterization.
c. *əməlʔetə ʔəttə- čg- ən lʔu- nin ŋewəsq- et- te.
all.obl dog- big- abs see- aor.3.3sg woman- pl- erg (‘All the women saw a/ the huge dog.’)
d. *ŋewəsq- et- te ʔəttə- čg- ən əməlʔo/ əməlʔetə lʔu- nin.
woman- pl- erg dog- big- abs all.direct / all.obl see- aor.3.3sg (‘All the women saw a/ the huge dog.’)
e. *ŋewəsq- et- te lʔu- nin ʔəttə- čg- ən əməlʔo/ əməlʔetə. woman- pl- erg see- aor.3.3sg dog- big- abs all.direct / all.obl (‘All the women saw a/ the huge dog.’)
A note is in order here: I am presenting these Chukchi examples with some trepidation.
On the one hand, they offer the clearest illustration I have observed of the contrast between absolutive and ergative in terms of quantifier float. On the other hand, these data may follow from some other constraints, in which case the contrast between the absolutive and ergative with respect to their ability to float quantifiers would be epi- phenomenal. Here’s why. First, it is possible that əməlʔo is a true D- quantifier, while əməlʔetə is adverbial. Many forms in - etə in Chukchi are in fact adverbial, and M. Dunn (1999) glosses əməlʔetə as adverbial as well, although not consistently. If əməlʔo is D- quantificational and əməlʔetə is A- quantificational, the task shifts, and we must explain why the putative A- quantificational form cannot co- occur with the absolutive.
Adding another layer of difficulty, in languages without ergativity, quantifier float is often associated with internal arguments: objects of transitives and subjects of unaccusatives. This restriction is well known, for example, from Japanese (Miyagawa 1989, 2012; Fukuda and Polinsky 2014). Although this restriction is not without exceptions, the preference for internal arguments as triggers of quantifier float also puts the ergative expression at a disadvantage. Relatedly, at least some instances of quantifier float can be subsumed under the transparency of certain clausal con- stituents to subextraction— an issue I addressed in section 4.2. In fact, that section included some Chukchi data showing that only absolutives are open to subextrac- tion. If əməlʔo is D- quantificational, then its inseparability from the ergative may be yet another sign that the ergative expression is not transparent.
To sum up these caveats, the inability of an argument to float quantifiers may result from several factors and, as such, may be a less salient diagnostic of PP- ergativity than some other properties considered above. This does not mean, how- ever, that we should not continue looking for representative cases of quantifier float;
such an inquiry would enrich our understanding not only of ergativity but also of quantification and of discontinuous constituents.