Recall from chapter 1 that the ergative can bind anaphoric pronouns in the absolu- tive position. This binding fact has long been taken as an indication of the ergative’s structural dominance (S. Anderson 1976, 1977; Aldridge 2008; among others); how- ever, if the ergative is a PP, we arrive at a paradox, since binding out of a PP should be impossible (or at least extremely constrained), as discussed in chapter 3. The facts of reflexive and reciprocal binding in ergative languages, originally welcomed as con- firmation of the dominance of the ergative subject, suddenly become problematic.
One can imagine two possible solutions to this apparent paradox. The first solu- tion places the burden on the binder: by relaxing certain well- known principles of binding theory, we can allow binding from a PP— for instance, through a feature percolation mechanism in which the phi- features of D get transferred to P and then characterize the entire PP (see Režać 2008a for a detailed account). This solution is certainly possible in principle, and would additionally account for the well- known cases of binding out of PPs that occur in English. However, overhauling the stan- dard binding principles so dramatically might have serious ramifications cross- linguistically— and, more dangerous still, this solution create a strange dissonance between accusative languages, in which PP binding feels like something of an excep- tion, and ergative languages, in which all ergatives (but no other PPs) have to be binders.
The alternative is to place the explanatory burden on the bindee. Under this option, the expectation is that languages with PP- ergatives have pronominal expres- sions that are fully exempt from binding-theory principles. More specifically, these languages should lack dedicated anaphors, making use instead of generic pronouns or some other sort of coreferring expression that is not subject to binding theory (see also Reuland 2011: 92– 94). Since there are no anaphors in a language of this
type, there are no violations. The relationship that appears to be binding is estab- lished at the interpretive rather than the syntactic level. This solution is more con- servative, in that it does not sacrifice the general principles of binding theory but simply establishes a class of languages that do not have the elements to which this theory can apply.
Empirical facts from syntactically ergative languages argue in favor of the “bur- den on the bindee” solution: the languages in question do in fact show an absence of dedicated anaphors; in place of such anaphors, they may employ regular pronouns to express reflexive meanings. Some researchers have explicitly noted the absence of reflexives in particular ergative languages. For instance, Mosel and Hovdhaugen (1992) comment on “the absence of reflexive constructions in Samoan” (1992: 726), noting that, while “typical reflexive verbs are found in Russian and Yidiny,” and some languages, like English, express reflexivity through dedicated self- pronouns, Samoan lacks “derived reflexive verbs or reflexive constructions” (1992: 726– 728). Instead, it uses lexically specified reflexives like tā‘ele “bathe,” “labile causative verbs” such as fa‘apa‘ū “fell” (illustrated (1a, b) below), or periphrastic expressions.
(1) a. ergative construction Samoan
Na fa‘a- pa‘ū e le tamaloa le lā‘au.
pst caus- fall erg det man det tree
‘The man felled the tree.’
b. “non- ergative construction” with reflexive interpretation Na fa‘a- pa‘ū le teine i le moega.
pst caus- fall det girl loc det bed
‘The girl threw herself onto the bed.’ (Mosel and Hovdhaugen 1992: 727)
Another strategy can be seen in the following examples from East Uvean; in this language, the lexical intensifier (to)tonu “true, exact” is used attributively, on the structurally lower expression, in order to encode the reflexive meaning.
(2) a. ‘E maheka ia Petelo ki tona foha totonu. East
nspc be.jealous abs P obl 3sg.poss son true Uvean
‘Peter is jealous of his own son.’ (Moyse- Faurie 2008: 136) b. ‘E ilo‘i e Petelo ia ia totonu.
nspc know erg P abs 3sg true
‘Peter knows himself.’ (Moyse- Faurie 2008: 136)
In addition to (to)tonu, East Uvean has an intensifier/ delimiter pē, which can appear more than once within a single clause; as (3b) shows, this particle has other func- tions beyond restricting coreference.1
1. The use of pē in East Uvean is similar to the use of the cognate restrictive particle in Tongan; see chapter 10 for the Tongan data.
(3) a. ‘E mamilo te ipu iāte ia pē. East
nspc turn det top obl 3sg int Uvean
‘The top turns on itself.’ (Moyse- Faurie 2008: 136)
b. Ko te hoki ‘ui mai pē ‘aenī
pred det imm say dir int deic
e Soane ne‘e ‘ufi‘ufi pē ia kiā ia pē.
erg S pst be.ashamed int 3sg obl 3sg int
‘Soane just told me that he was ashamed of himself.’ (Moyse-Faurie 2008: 135) Without the intensifier particle, the sentences above have an obligatory disjoint reading. With the particle, there are two readings available: a predicate- focus read- ing and a reflexive reading. Thus, we can conclude that the particle pē has a meaning close to “only,” with the following semantics:
(4) If A is of type ⟨α, t⟩, only A is of type ⟨α, t⟩, and [only A] is the set of all B of type α such that B has the property [A] (i.e., B ∈ [A]), and no other property that is an alternative to that (i.e., in [A]f); [only A]f = {[only A]}.
(after Büring and Hartmann 2001: 248) The predicate- focus reading of pē restricts the two- place relationship to action upon oneself (as opposed to other relationships). This type of focus operation is well known from English only and similar particles elsewhere (see Büring and Hartmann 2001 for a discussion). The other reading of pē, which forces a reflexive interpreta- tion, relies on the ability of pē to limit the set of referents of the object. On this read- ing, pē again behaves like “only,” but in this case, it takes scope over the relationship between the referent of the subject and the referent of the object.
The interpretive relations described here are not entirely free, however. Non- anaphoric binding still obeys the structural hierarchy in that the presumed binder has to be structurally superior; it does not, however, always obey locality (see Mosel and Hovdhaugen 1992: 727 for non- local “binding” in Samoan, and Besnier 2000: 202– 210 for non- local binding- like dependencies in Tuvaluan).
The use of emphatic particles to induce a reflexive or reciprocal interpretation can occur independent of ergativity. In quite a few languages, generic pronouns co- occur with intensifiers whose main function is to limit the set of alternatives to the refer- ent of the NP, thus forcing an assertion of identity between the two participants, one expressed by the subject, the other by the object (McKay 1991; Gast and Kửnig 2006; Haspelmath 2007; Kửnig and Gast 2008; among others).2 The result is the emergence of so- called reinforced pronouns— that is, pronouns accompanied by a deictic marker. A side effect of the presence of reinforced pronouns in a language is
2. Eventually, such emphatic pronouns can coalesce with the intensifier article, giving rise to a dedicated anaphor. Such a process led to the development of the English reflexives in - self, for instance (van Gelderen 2000).
often that the language has no dedicated anaphors; in the absence of anaphors, the principles of binding theory are not violated.
The use of particular forms of pronouns instead of dedicated anaphors allows a given language to establish “interpretive binding” relations while avoiding violation of binding theory principles. I would like to underscore that interpretive binding without dedicated anaphors is not exclusive to ergative languages— see, for example, Cole et al. (2008) for binding-theory- exempt anaphors in languages that have no signs of syntactic or morphological ergativity. However, while reinforced pronouns can be found cross- linguistically in a variety of languages, it is the regular, system- atic absence of non- exempt anaphors in PP- ergative languages that matters for the purposes of the discussion here. Thus, although reinforced pronouns are quite wide- spread and found in many languages, their presence in syntactically ergative lan- guages is systematic— and, I will argue, non- accidental.
Co- indexation between two positions in the same clause can also be achieved when the first position hosts a DP with a referential index and the second hosts a possessed noun phrase in the position of an anaphor. Typically, one finds nouns meaning “head” or “body” in the possessed- NP position. Abkhaz is one language in which the ergative can be co- indexed with the absolutive- possessed DP; note, how- ever, that coreference in this construction is optional. Hewitt (2005b: 110) writes,
“Where the reflexive element functions as Indir[ect]/ Obl[ique] Obj[ect], Abkh[az]- Aba[za] employs the lexeme for ‘head’ outside the verbal complex”:
(5) proi a- sark’a- ỗ’ə si- x ə z- be- yt’. Abkhaz
det- mirror- loc 1sg.poss-head.abs 1sg.erg- see-asp
‘I saw myself in the mirror.’ (based on Payne 1997: 200)
The use of simple pronouns or referential expressions in lieu of true anaphors is just one strategy languages employ to escape the constraints imposed by binding theory. Another possibility involves suppression of the object argument, thus elimi- nating the nominal reflexive altogether; under this configuration, the verb— which could in principle take an anaphor in the object position— is instead intransitive, and the reflexive interpretation is either established pragmatically or follows from the semantics of a dedicated verbal affix.
Languages that employ this strategy use verbal markers with valency- changing properties such as antipassivization or other detransitivizing marking (cf. Siegel 1998 on object suppression in Inuktitut). In the absence of a dedicated reflexive, the range of arguments in a given language is narrowed by verbal affixation; as a result, the language will lack an anaphoric inventory. The strategy described here is quite common; I illustrate it with Chukchi examples below. In (6a, b), a transitive verb that undergoes detransitivization with object suppression receives the reflex- ive reading. The suffix - tkV- that is used to detransitivize the verb is an antipassive suffix (one of two such suffixes in Chukchi— see Kozinsky et al. 1988; Nedjalkov 2006). Homophony between reflexive/ reciprocal, antipassive, middle, and general- ized detransitivizing affixes is widely attested (see Polinsky, forthcoming- b for an overview).
Transitive verbs in Chukchi agree with both subject and object, but the verb in (6b) agrees only with the subject, indicating that it is genuinely intransitive:
(6) a. proi proj tewla- nenj/*i. Chukchi
shake- aor.3sg.3sg
‘S/ hei shook him/ her/ itj off.’
b. pro tewla- tko- gʔe.
shake- antip- aor.3sg
‘S/ he shook herself/ himself.’
So far, I have identified two main ways of circumventing the syntactic binding rela- tion: (i) the use of reinforced pronouns, which are non- anaphoric expressions, often accompanied by an emphatic marker, and (ii) the suppression of the object argument (detransitivization), illustrated by the Chukchi examples above.
To reiterate, these two strategies are not unique to ergative languages in gen- eral or to syntactically ergative languages in particular; they are a recognizable and respectable alternative to the standard binding configurations that have been at the core of binding theory for years, linking an antecedent and a dedicated anaphor. The availability of different strategies for expressing the identity relation between two elements— pronounced or not— reflects the observation that such a relation can be encoded in the syntax, in the lexical or sentential semantics, or in the discourse.3
The two main strategies for expressing reflexive- like meaning can each apply con- sistently throughout a given language, and the examples above show that such lan- guages are common. Since binding is only problematic for PPs, not for DPs, however, nothing rules out the possibility that, within a given language, DPs will count as legitimate binders, while PPs will be excluded from the range of binders. The result of this scenario would be not the absence of binding proper, as occurs in languages with verbal reflexivization or reinforced pronouns, but simply the exclusion of PPs from standard binding domains.
Assuming such a possibility, we can expect to find ergative languages where the absolutive DP is a legitimate binder for an anaphoric expression in its c- command domain (regardless of the case of that constituent), but the ergative PP is not.
Such is the situation observed in Inuit (West Greenlandic), which has a genuine reflexive immi- (Fortescue 1984: 155– 167; Sadock 2003: 40– 43; Bittner 1994); the distribution of this reflexive is, however, more limited than in the familiar languages with standard binding. Compare the following examples. Example (7a) illustrates a regular transitive clause in Inuit, with the verb agreeing with both ergative and absolutive; (7b) shows the failure of reflexive binding to apply within that transitive 3. Reuland (2011: 125), building on Reinhart (1983, 2006), Grodzinsky and Reinhart (1993), and others, proposes that the encoding of identity relations follows the hierarchy in (i). The encoding style on the left of the hierarchy is more economical and/ or more con- strained; in theory, more leftward encoding styles should be favored because they are, in some sense, less costly than those on the right:
(i) Narrow Syntax < logical syntax (syntax- semantics interface) < discourse
clause; (7c) shows the absolutive DP successfully binding the dative anaphor in the intransitive counterpart of the original sentence.
(7) a. Suulu- p Kaali aallaa- v- a- a. Inuit
S- erg K.abs shoot- ind- tr- 3sg.3sg
‘Suulut shot Kaali.’
b. *Suulu- p immi aallaa- v- a- a.
S- erg self shoot- ind- tr- 3sg.3sg (‘Suulut shot himself.’)
c. Suulu immi- nut aallaa- v- u- q.
S.abs self- dat shoot- ind- intr- 3sg
‘Suulut shot himself.’ (Bittner 1994: 82)
We can capture these facts as an empirical generalization by stating that, in Inuit, reflexives can appear only in intransitive clauses. In her discussion of Inuit reflex- ives, Bittner explicitly states that the obligatory detransitivization of clauses con- taining a reflexive remains a mystery. She writes: “For reasons which at this point remain obscure, reflexive arguments in Inuit … have only oblique case forms, and cannot be overtly realized in any direct case, NOM or ERG” (Bittner 1994: 82).4 On the account I have proposed here, the reason for the ungrammaticality of (7b) is clear. Since Inuit is syntactically ergative, we have independent evidence in support of the idea that its ergative is a PP. As a result of its PP status, the ergative cannot bind an anaphor. A binding relation can thus only be established when the anaphor is dominated by a DP, as in (7c), where the verb is intransitive and the subject (the binder) is in the absolutive case. Note that when the relationship between the erga- tive and the c- commanded reflexive is anaphoric rather than syntactic, co- indexation is permissible:5
(8) Juuna- pi ataata- ni immii- nik uqaluttuup- p- a- a. Inuit J- erg father- 3sg.prox.abs self- ins tell- ind- tr- 3sg.3sg
‘Juunai told his father about himselfi.’ (Bittner 1994: 43)
In discussing what she considers the “mysterious” restriction against binding from the ergative in Inuit, Bittner (1994) correctly notes that in some ergative languages overt reflexives can appear as absolutives and be bound by ergatives. This is also expected on the analysis developed here; I argue that those languages that permit binding of anaphors by ergatives will be those languages with a DP- , rather than a PP- , ergative (I will return to this type in chapter 7). The principled difference between the two ergative types reveals itself in binding.
In sum, there are three main ways of circumventing the impossibility of binding from the ergative: (i) the use of reinforced pronouns (non- anaphoric expressions,
4. Bittner may be trying to explain the pattern by appealing to some property of the reflexive pronouns themselves, rather than to a general constraint on binding.
5. The same pattern holds for similar English examples ( chapter 3).
often with an emphatic or deictic marker), (ii) the suppression of the object argu- ment by detransitivization, and (iii) the restriction of the range of binders to DPs.
In principle, it is possible that other strategies for circumventing the ban on PP- ergative binding also exist and remain to be discovered. For the purposes of our dis- cussion, what matters is that PP- ergatives cannot be binders; how this restriction is dealt with is of ancillary relevance.
Assuming the availability of these main strategies, let’s examine the expression of reflexivity in those ergative languages that show syntactic ergativity and hence, hypothetically, have PP- ergatives. Without exception, these languages either employ simple pronouns as reflexives or rely on the use of verbal detransitivization alone:
(9) Expression of reflexivity in syntactically ergative languages
Language Simple form
of pronoun
Dedicated anaphor
Expression of reflexive
Halkomelem Yes No Reinforced pronoun
Straits Salish Yes No Reinforced pronoun
Lushootseed Yes No Reinforced pronoun
Tongan Yes No Reinforced pronoun
Samoan Yes No Reinforced pronoun
Roviana Yes No Reinforced pronoun
Tuvalu Yes No Reinforced pronoun
Futuna Yes No Reinforced pronoun
Tokelauan Yes No Reinforced pronoun
Pukapukan Yes No Reinforced pronoun
Nias Yes No Verb detransitivization
Chukchi Yes No Verb detransitivization
Kuikuro Yes No Verb detransitivization
Katukina- Kanamari Yes No Verb detransitivization
Yarrawa Yes No Verb detransitivization
Trumai Yes No Verb detransitivization
Inuit (West Greenlandic)
Yes Yes Verb detransitivization;
ergative does not bind
Abkhaz Yes No Noun meaning ‘head’;
ergative does not bind
Abaza Yes No Noun meaning ‘head’;
ergative does not bind
From the above discussion, we can conclude that the solution to the binding problem that removes the bindee from the principles of binding theory is supported by the empirical data. In PP- ergative languages, pronouns and reflexive anaphors do not have different lexical forms and do not express different features, so binding theory is not violated.
This conclusion still leaves us with an outstanding question that goes beyond the confines of binding theory: assuming that the languages in question have bound variables, what mechanism allows for the “binding” by the DP inside the PP? The relevant context is illustrated by the English examples below:
(10) a. [Every supervisor]i signed hisi student’s dissertation.
b. [No girli’s parents] could understand why shei was so sad.
I do not have a solution for this problem, but would like to offer some preliminary considerations. First of all, it is important to keep in mind that the analysis of the rel- evant English sentences is also uncertain. Second, we should not assume that every language has the same inventory of quantified expressions as English; it may well be that the relevant languages lack “every,” “any,” or “no” as quantifiers, in which case the problem simply will not arise.6 Theoretical advances in our understanding of sen- tences such as the English (10) will have to be combined with a better knowledge of empirical facts from other languages before this problem can be properly addressed.