2.4.1 PP- ergative and DP- ergative languages: transitive clauses
The derivation of the transitive verb complex in a PP- ergative language proceeds as follows. First, the lexical verb combines with its complement, and the resulting VP merges with the functional head v. This functional head has the absolutive case as its only case feature. In principle, case features are expected to be checked as soon as possible (Chomsky 1995), so the v head licenses the absolutive on the complement of the verb. Likewise, this v head may also include agreement feature(s) valued by the feature(s) on the internal argument; however, these phi- type features may also be located higher, on the inflectional head. I do not consider the details of phi- feature checking in the derivations below.
At the next stage of the derivation, the resulting vP combines with another func- tional head, v or Voice, that bears the feature [+Active]. This next functional head is responsible for the introduction of the external argument, to which it assigns case.
For the discussion below, it is important to note that this second functional head may not be present in all derivations. In particular, its presence in unergatives is unclear; as I show below, unergatives appear to be the most cross- linguistically het- erogeneous class with respect to derivation. In particular, the status of unergatives with respect to extra projections is far from clear and may be less uniform than the status of transitives. In PP- ergative languages, this functional head has a PP in its specifier; this is where the PP- ergative is licensed and receives case from the P head.
The PP in spec,vP has standard specifier properties: semantically, it serves to make
the meaning of the head more precise, and, syntactically, it marks the boundary of the higher vP.9
(34) vP
PP
PERG DP vP v [+active]
v’
VP v [ABS]
V DP
[uABS]
The idea of a PP specifier is not novel; non- DP specifiers in general, and PP specifiers in particular, have been proposed and motivated by a number of researchers, notably in the work on passives and experiencers (cf. Williams 1981, 1985; Hasegawa 1988;
Belletti and Rizzi 1988; Watanabe 1996; Goodall 1997; Kayne 2002, 2005; Landau 2010; Fukuda 2009, 2012). What I propose is to expand the range of acceptable non- DP specifiers to include ergative expressions. This is very much in keeping with the notion that the ergative is an inherent case in a subset of ergative languages. While some researchers argue that the inherent ergative case can be assigned to a DP directly by a verbal head (Laka 2006; Legate 2008; Aldridge 2004, 2008; among others), I con- tend instead that this case is assigned by a morphologically dependent adposition—
the ergative PP, which is located in its entirety in the specifier of the highest vP. This analysis, which accounts for the subject properties of the ergative expression, is the essence of the ergative- as- inherent- case approach adopted in this work.
The derivation presented in (34) assumes the layered- structure approach that I briefly discussed in the beginning of this chapter. Crucially, the structural absolu- tive and the inherent ergative are associated with two different functional heads in the verb phrase; the absolutive is licensed by the lower functional head v, and the ergative is a PP in the specifier of the higher vP, with the v providing its theta role. It is, however, possible to construct a similar derivation that makes use of only a single functional head v; on such an analysis, that single functional head would license the absolutive and have a PP in its specifier. Thus:
(35) vP
PP v’
PERG DP VP v [+active]
[ABS]
V DP
[uABS]
The distinction between these two approaches may be obscured by the bundling that I mentioned in the beginning of this chapter: two functional heads can be bundled
9. Of course, we must make sure that such an ergative argument is a true PP, not a PP wrapped in a covert DP; I will discuss ways to distinguish between these two possibilities in chapter 3.
10. A variant of this approach assumes that the absolutive of the internal argument is a default case that does not require licensing or selection of any sort (cf. Aldridge 2008;
Legate 2008; Arregi and Nevins 2012: ch. 1). As I show in chapter 11, such an approach does not work for Tsez. I am not disputing its general availability, but it is far from universal.
into a single projection that does all the work (Pylkkọnen 2008; Harley 2013). As I have already mentioned, I remain agnostic as to which approach is correct, and I am treating both structures— (34) and (35)— as viable.
In contrast to the PP- ergative, I assume that the DP- ergative is a structural case.
Empirical arguments in favor of treating the ergative as a structural case have been advanced for a number of languages. These arguments typically include the lack of connection of the ergative to a particular thematic role, the non- preservation of the ergative under nominalization or raising (a pattern observed in Tsez; see chapter 11 for details), and the uniform use of the ergative as a derived subject in raising con- structions (discussed in the context of the Basque modal verb behar “must”: Arregi and Nevins 2012: 20; Režać et al. 2014).
The licensing of the (structural) DP- ergative case is not immediately straightfor- ward. Here, I present some general considerations that reflect the current approaches available in the field and comment on the main issues that may arise with respect to possible derivations. In discussing the derivation of transitives in PP- ergative lan- guages, I started with a layered- head approach, according to which the absolutive and the ergative are associated with different functional heads in the verbal domain. Let me start the discussion of DP- ergatives by considering the same layered approach.
The layered- head approach to DP- ergatives mirrors the layered- head approach to the PP- ergative structure I described above. This approach involves two functional heads: the lower functional head assigns absolutive case to the internal argument, while the higher functional head selects as its specifier a DP with ergative case fea- tures. The structure of the lower vP is the same as in unaccusatives.10
(36) vP
DP [uERG]
v’
vP v
[ERG]
VP v
[ABS] DP
[uABS] V
Two possible difficulties arise with respect to this structure. First, it is quite differ- ent from the corresponding structure of accusative languages, in which the exter- nal argument generally receives case from the inflectional head (T) and the internal argument is licensed in the vP. This issue may not be significant, if we adopt the view that the ergative is indeed distinct from the nominative. A more difficult issue con- cerns the status of unergatives. The issue of case licensing in unergatives transcends the contrast between PP- ergative and DP- ergative languages; I will address this prob- lem in the next section.
An alternative to the layered- head approach would derive DP- ergatives by licens- ing both structural cases, absolutive and ergative, in a single functional head, which would assign first the absolutive to the internal argument and then the ergative to the external argument.
An idealized vP with a structural ergative is shown below:
(37) vP
DP v’
[uERG]
VP v
[ABS; ERG]
DP V
[uABS]
This is not an impossible configuration; furthermore, it has the flexibility of allowing either a genuine single functional head v or a bundling of two functional heads into a single projection that does all the work (Pylkkọnen 2008; Harley 2013). The possible bundled structure is shown below, revised from (37):
(38) vP
DP
[uERG] v’
VP v+v
[ABS]; [ERG] DP
[uABS] V
Such bundling operates independent of ergative type. In particular, bundling has been argued for in Chol, which is syntactically ergative and is therefore a PP- ergative language (Coon and Preminger 2013; Harley 2013), and also in Niuean, which is not syntactically ergative (Polinsky and Longenbaugh, forthcoming).
Throughout this section, I have discussed the derivation of PP- ergative and DP- ergative structures by associating them with an array of functional heads. Although this is the approach adopted in this book, it is also possible to capture the differ- ence between the two types of ergative languages using a configurational approach to case. Under the standard configurational approach to case licensing, the only dedicated functional heads that can assign case are adpositions and complemen- tizers. Both these categories assign inherent case. Beyond that, noun phrases are assigned case by virtue of their structural position relative to certain lexical heads and, more importantly, to other noun phrases in the clause or within a particular c- command domain. All available cases are organized along the following hierarchy:
(39) independent/ unmarked case > dependent case > inherent case
Implementations of configurational case assignment differ (see Bittner and Hale 1996a, b; Marantz 1991; Yip, Maling, and Jackendoff 1987; Baker 2015; Levin and Preminger 2015), but all approaches share the insight that accusative and ergative can be given a unified treatment as dependent cases. The presence of the ergative is
dependent on the presence of the absolutive, while the presence of the accusative is dependent on the availability of a nominative. On a configurational approach, DP- ergative languages can be conceived of as languages that have a dependent ergative, whereas PP- ergative languages are those that have a structural absolutive and inher- ent ergative. Thus, regardless of the actual implementation, the PP type among erga- tive expressions is expected to be an inherent case.
2.4.2 PP- ergative and DP- ergative languages: unergative clauses
Most of the work on unergatives within the layered- head approach has been done on accusative languages, where the external argument is always nominative. As a result, the contrast between external arguments bearing the (structural) absolu- tive case and external arguments bearing the structural ergative case has yet to be addressed.
Not all ergative languages treat unergatives uniformly. In some languages, uner- gatives are assimilated to transitives, with the relevant functional head (high v or Voice) invariably assigning ergative case to the external argument. The result is a lan- guage without intransitive unergatives— something that should be possible in accu- sative as well as ergative languages, but may be less apparent. Note that although unergatives are essentially transitives on this configuration (cf. Hale and Keyser 1993 for the idea that prototypical unergatives are covert transitives), they differ from transitives in lacking a genuine internal argument. Instead, they have either an implicit (silent) argument or a cognate object.
Transitive unergatives are observed in Hindi, where there is variation with respect to the way their objects are expressed. Hindi unergative subjects surface with the ergative marker - ne, and at least some unergatives take overt cognate objects.
Furthermore, unergatives in this language combine with the light verbs le “take,”
de “give,” or daal “do,” but not with the light verb jaa “go”; by contrast, only this last verb is used with unaccusatives. The following examples from Surtani et al. (2011) illustrate these three properties:
(40) a. Raam- ne bahut naach- aa. Hindi
Ram- erg a.lot dance- pfv
‘Ram danced a lot.’
b. *Raam- ne bahut ghabraa- ya.
Ram- erg a.lot panic- pfv (‘Ram panicked a lot.’)
(41) Raavan- ne bhayaanaka hasii has- ii.
Ravan- erg horrifying laugh laugh- pfv
‘Ravan laughed a horrifying laugh.’
(42) Raam- ne pahaaR chaD liy- aa/ *jay- aa.
Ram- erg mountain climb take- pfv/ go- pfv
‘Ram climbed the mountain.’
The same surface outcome could also be produced by a structure with an inherent ergative, as in (34), as long as there was no restriction on the transitivity of the func- tional head licensing the ergative. For example, in Chol, unergatives are two- place predicates that always have an internal argument (Coon and Preminger 2010; Coon 2013a). Thus, the presence of an unergative with an ergative- marked subject cannot determine the status of the ergative as an inherent or structural case.
Finally, some researchers suggest that unergatives which are inherently transitive are typical of Basque (Laka 1993), where, as I just mentioned, there are compelling reasons to view the ergative as a structural case (see Režać et al. 2014, pace Laka, forthcoming). However, there is no consensus about the status of Basque unerga- tives; in particular, some researchers argue against the presence of an implicit object in such structures (see Preminger 2012 for a discussion). As a result of this lack of consensus regarding Basque unergatives, I will not be using them as a representative example of the unergative- equals- transitive type.
If a language uniformly assigns ergative case to all its unergatives, the result could be the type of alignment that is sometimes classified as “active”: all agent- like arguments are marked in one way, and all theme- or patient- like arguments, regard- less of transitivity, are marked differently (Klimov 1977; Baker 2015). This type of alignment is probably available with a subset of verbs in most languages and is inde- pendent of ergativity. Only when a comparable— and large— number of intransitive verbs show a difference in alignment might it make sense to classify a language as active- stative. If such an alignment were to be found and if the ergative assigned by the Voice head were inherent, we would predict the same extraction restrictions for all ergative arguments. In theory, nothing rules out such an alignment; why it is not attested remains to be explained.
In a number of ergative languages, unergatives have absolutive subjects. Again, this choice seems to be independent of the status of the ergative as an inherent or structural case. The derivation of such unergative structures requires a different approach. One option is to assume that the absolutive is assigned in different ways depending on whether it is an external or internal argument. On this approach, the internal- argument absolutive is assigned by the lower functional head, while the external- argument absolutive is assigned by the highest inflectional head— making it essentially similar to the nominative. Such differential licensing of two apparently identical forms has been advocated for in a number of languages (cf. Aldridge 2008;
Legate 2008). To represent this assignment pattern schematically,
(43) TP
T vP
[ABS = NOM] [uABS DP= NOM]
VP v
v’
V
On this approach, unergatives seem to have less structure than transitives, especially as regards the higher, external- argument- licensing functional head. Indeed, this
higher functional head seems to be either missing altogether in unergatives or to be present but somehow deficient, unable to assign case. This is not the most appealing solution, and clearly, to be convincing, it would need to be fleshed out in more detail.
It is quite possible that languages of this type exist, but neither of the languages con- sidered in Part II, Tongan and Tsez, fits this profile, despite being different in terms of their properties and in terms of the way their ergatives are licensed.
For languages like Tsez, in which the absolutive in unergative constructions must be licensed by something other than the inflectional head, there are several possibili- ties. First, we can assume that the layered structure of the vP in unergatives is the same size as it is in transitive clauses. If so, it is possible to imagine the structure shown in (44), with the functional head licensing the external argument but assign- ing it a structural absolutive case, not the ergative case. This solution essentially calls for a distinction between transitive and intransitive functional heads that license an external argument— a distinction that needs to be motivated in its own right.
At this point, the only clear way to introduce a distinction between transitive and intransitive functional heads would be by specifying these heads in the lexicon. Such a division of labor is apparently not needed for accusative languages, and more work on argument structure is necessary to explicate this issue.
(44) vP
DP v’
[uABS]
vP v2
[ABS]
VP v1
V
If we reject the transitive/ intransitive division of external- argument- introducing functional heads, two alternative possibilities arise. First, it is possible that unerga- tives always involve a bundled head, thus:
(45) vP
DP v’
[uABS]
VP v1–v2 [ABS] V
The structure in (45) requires us to relativize the concept of bundling. Until now, scholars have treated bundling as a property of individual languages rather than a property of language structures. If the proposal sketched out here is correct, then bundling and non- bundling should be compatible within a single language.
The structure in (45) also paves the way for another possible derivation, in which the unergative vP is simply smaller than the transitive vP, either because it does not include the extra head licensing the external argument or because the two heads are bundled. This possibility, schematized below, is difficult to distinguish from the one in (45). If the unergative vP contains only one functional head, we can assume
that this functional head can project an external argument and license the absolu- tive in that position. On this approach, there are simply two intransitive functional heads: one with the external argument, as shown below, and the other (unaccusa- tive v) without it.
(46) vP
DP v’
[uABS]
VP v
[ABS] V
To summarize, we have considered the following structures for unergatives:
(47) Possible derivations of unergatives in an ergative language:
a. Unergatives are covertly transitive, with a cognate or silent object; their subject appears in the ergative form.
b. Unergatives project an external argument whose case is licensed by the inflectional head, not by a verbal functional head.
c. Unergatives and transitives both have a functional head that projects an external argument, but these external- argument- licensing heads are lexically specified for transitivity.
d. Unlike transitives, unergatives do not include an extra functional head; the absolutive is assigned by the sole functional head to the DP in spec,vP.
Option (47d) may in fact include two different sub- options: one that invokes a sole functional head and one that invokes a bundled functional head. At present, it is hard to tell how— if at all— these two possibilities may be distinguished.
All four options listed in (47a– d) should be equally available in both types of ergative languages, PP- ergative and DP- ergative. In chapters 10 and 11, I will discuss unergatives in Tongan and Tsez, respectively, showing that they both instantiate option (47d).
The wealth of possible derivations for unergative structures in ergative languages is not accidental. Unergatives are a rather heterogeneous class, and this heterogeneity may even find its reflection within a single language.11 Thus, if the numerous struc- tures proposed in this section are all on the right track, it would theoretically be pos- sible to find more than one such structure represented within a single language, with each derivational type associated with a different subclass of unergatives. In particular, some unergatives, such as manner verbs, do not entail a change of state (cf. Alexiadou et al. 2015), which suggests that they include a higher functional head that licenses the external argument. On the other hand, verbs of light and sound emission, which are unergative according to most diagnostics, might not involve a higher functional head, 11. It is surprising that there has been so little discussion of the ways in which uner- gatives may form a heterogeneous class, especially given the attention lavished on the heterogeneity of unaccusatives (for the latter, cf. Levin and Rappaport Hovav 1995;
Sorace 2000).