(33) A comparison of PP and DP in subject position
PP DP
Can extract (A- bar move) leaving a gap at the extraction site
No Yes
Can serve as pivot of cleft No Yes
Can determine agreement Only if agreement with absolutive DPs (subjects and objects) is possible
Yes
Can serve as binder No Yes
Is accessible to A- movement No Yes
In this table, most of the properties associated with PPs in subject position are negative. I would like to argue that this is not accidental. Specifically, I contend that the PP- ergative is the acquisition default for ergativity, the one that a learner starts out with when acquiring a morphologically- ergative language. Only if evi- dence against this default is provided does the child conclude that s/ he is learning a DP- ergative language. On this approach, the negative values of several features associated with a PP- ergative make sense, and are overturned only under sufficient evidence. Assuming a default status for the PP- ergative is logical, since it is further away than the DP- ergative from the nominative- accusative type in the space of pos- sible languages.
If we adopt the hypothesis that PP- ergativity is default ergativity, then syn- tactic ergative behavior may be the triggering evidence in support of that default.
I suggest that, in terms of syntactic ergativity, the use of resumption may be neces- sary and sufficient for a child to determine that s/ he is learning a PP- ergative lan- guage. Resumption is salient in L1 acquisition because it is found in wh- questions and relative clauses, both of which belong to the class of structures that are acquired relatively early (Bloom et al. 1982; Stromswold 1995; Thornton 1990; Kidd 2011 and references therein). Due to the robustness of this cue, languages with PP- ergatives and resumption should be straightforward to learn. However, it is possible that such a cue would be insufficient. After all, resumption is not present in every PP- ergative language; some languages resort to other methods, such as antipassivization, for relative clause and wh- question formation.
The idea of the PP- ergative as the acquisition default is speculative, but I will offer a few considerations in support of it here. These considerations are drawn from the literature on the acquisition of ergative languages. Despite excellent work on several ergative languages, which I will reference below, the acquisition of ergative languages is still largely unchartered territory, so these considerations are only preliminary.
The main observation I would like to offer is that there is no evidence that children learning PP- ergative languages have difficulty with forming relative clauses or wh- questions from the ergative, whereas in the acquisition of DP- ergative languages, structures with an ergative gap are problematic for young learners. This pattern lends support to the notion of a PP- ergative default.
Although, as I mentioned above, the literature on the acquisition of ergative lan- guages is rather limited, there exists enough evidence to discern a pattern. Among PP- ergative languages, there is work on the acquisition of Samoan (Kernan 1969;
Ochs 1982, 1988), West Greenlandic/ Inuktitut (Fortescue and Olsen 1992; Allen 1996), and some Mayan languages (Pye 1990, 1991; Mateo Pedro 2010, 2011). For DP- ergative languages, there is work on the acquisition of Basque, Warlpiri, and some Papua New Guinean languages, which I will discuss briefly below. Within this literature, no evidence is provided to suggest that the acquisition of A- bar move- ment in PP- ergative languages (e.g., Samoan, West Greenlandic, or Q’anjob’al) is dif- ficult. For West Greenlandic, Fortescue and Olsen (1992: 216) note that both active (intransitive) and passive participial morphemes involved in the formation of rela- tive clauses are acquired very early. In some Mayan languages, one finds a special agent focus (AF) construction used in the extraction of ergative DPs (recall discus- sion in chapter 4). Pye (1990, 1991) reports that K’iche’- speaking children have dif- ficulty acquiring antipassive constructions,5 but he does not discuss AF. Mateo Pedro (2010, 2011) shows that Q’anjob’al- speaking children successfully acquire the AF suf- fix - on in transitive embedding contexts.6 The following wh- question with the AF morpheme on the verb was produced by a child at 2;2:
5. Antipassivization also involves detransitivization, but crucially differs from the AF construction in that it lacks extraction and demotes the patient argument to oblique sta- tus. See Stiebels 2006 and Aissen 2011 for a detailed discussion of the difference between AF and antipassives in Mayan.
6. Mateo Pedro (2010, 2011) also shows that children distinguish the morphology of intransitive and transitive verbs in embedded and split- ergative contexts at an early age
(34) Child form: Axha ma? Q’anjob’al Underlying form: Maktxel x- ứ- maq’- on- i?
who asp- abs.3sg- hit- af- intr
‘Who hit him?’ (Pedro Mateo Pedro, pers.comm.)
The syntactic structure of wh- questions in Q’anjob’al is open to interpretation; it is possible that they involve genuine wh- movement, as in (35a) below (irrelevant struc- ture not shown), or that they are clefts or pseudo- clefts, as schematized in (35b). On the latter analysis, wh- questions actually include a genuine relative clause, which may be located in the subject of the pseudo- cleft.
(35) a. [CP maktxeli [TP x- ứ- maq’- on- i ti] ] ? Q’anjob’al b. [PredP maktxel] [DP [CP x- ứ- maq’- on- i e]]?
However, in addition to wh- questions, child Q’anjob’al also provides evidence of bona fide relative clauses with AF. The next example illustrates the errorless use of AF in relativization by another child, aged 2;7:
(36) Child form: Manon tx’at jun dominga. Q’anjob’al
Underlying form: A jun dominga [CP x- ứ- man- on tx’at e]
top Dominga asp- abs.3sg- buy- af bed
‘It was Dominga that bought the bed.’ (Pedro Mateo Pedro, pers. comm.)
Of course, these are production data, and there is no evidence so far as to how chil- dren perform on comprehension of A- bar dependencies with ergatives and absolu- tives in such languages. Yet it seems clear that young learners of these languages do not have any problems producing wh- questions and relative clauses.
The data available for Q’anjob’al contrast with those reported for Basque, where several researchers comment on children’s protracted omission of ergative morphol- ogy in production (Ezeizabarrena 2012; Austin 2012) and, more importantly, on their mixed results with respect to relative clauses with an ergative gap (as compared to relative clauses with an absolutive object gap). The reason for these mixed results has to do with the difference between production and comprehension. In produc- tion, children give a slight edge to subject relatives— those with an ergative gap. In comprehension, however, when presented with two sentences like the pair below in a picture- matching task, Basque children under age 6 make a higher number of errors with subject relatives involving the ergative gap (37a), as compared to object relatives (37b) (Gutierrez- Mangado 2011; Gutierrez- Mangado and Ezeizabarrena 2012). While the comprehension results appear to contradict the production results,
in Q’anjob’al. They start by optionally producing the AF suffix with embedded transitive verbs (where it is required in the adult grammar) and they never produce it with embedded intransitive verbs (where it is impossible in the adult grammar). This distinction is pres- ent despite the fact that, in some cases, the children do not produce the special person- marking pattern associated with split ergativity.
it is possible that the errors in both domains may index the children’s relative diffi- culty in controlling the relevant morphology, especially since the overt ergative mark- ing is not always consistent, even in the input (Karlos Arregi, pers. comm.).
(37) a. Hau da [ ei amona muxukatzen duen] neskai. Basque this is grandmother.abs kiss.ipfv aux.adn girl.abs
‘This is the girl who is kissing the grandmother.’
b. Hau da [amona- k ei muxukatzen duen] neskai. this is grandmother- erg kiss.ipfv aux.adn girl.abs
‘This is the girl that the grandmother is kissing.’
(Gutierrez- Mangado 2011: 180)
The work on Basque is probably the most extensive acquisition research that explic- itly addresses extraction of the ergative.
Data from Ku Waru also suggest, indirectly, that children may have difficulty with the DP- ergative. Ku Waru is an ergative language of Papua New Guinea that lacks syntactic ergativity. In this language, children produce noun phrases without the ergative marker in response to questions about the referent of the ergative expres- sion. This pattern of non- ergative- use in response to ergative wh- questions persists even when the adult corrects the child, as in the following example:
(38) a. Mother: Nu- nga walpakuyl nai- n baim te- nsi- ri- m? Ku Waru you- gen shirt who- erg buy do- ben- pst- 3sg
‘Who bought you that shirt?’
b. Child (2;9): Papa.
Daddy c. Mother: Papa- ni.
Daddy- erg
Glossed by language assistant as ‘Your daddy bought it for you, ha!’ (Rumsey et al. 2013: 154–155)
Thus, there is evidence that, when learning an ergative language, children start with the PP- ergative as the default. Only once there is positive evidence to drive the child away from this hypothesis does s/ he switch to the DP- ergative type. Let me close this section by emphasizing that the analysis explored here remains a test- able hypothesis, not a foregone conclusion. We have little data on the acquisition of the relevant constructions, and Basque seems to be the only ergative language where children have actually been tested on both comprehension and production of A- bar structures.7 Hopefully, the direct comparison between PP- and DP- ergative
7. The work on the acquisition of Warlpiri (Bavin and Shopen 1986) and Kaluli (Schieffelin 1985, 1990), two more languages without syntactic ergativity, does not include infor- mation on the production or comprehension of relative clauses or wh- questions with ergative gaps.
languages presented in this book will motivate more work on the acquisition of such languages.