Although possessors internal to an argument DP do not qualify as canonical controllers of verbal agreement, in some languages an internal possessor may be cross-referenced on an applicative verb. The aim of the paper is to propose a historical scenario for the emergence of this pattern, following the basic insights of the constructional approach to language change. The paper argues that this pattern is a historical innovation. It emerged when the external benefactive argument was reanalysed as internal possessor, a process that has parallels in some languages with dative possessors. The change was motivated by cross-constructional analogy, namely, formal and semantic assimilation to the class of internal possessive constructions. When constituency was reanalysed, the location of agreement remained intact creating a non-local configuration.
Morphological applicatives create an applied object argument with the regular properties of primary objects. This can include the ability to be indexed on the verb, subject to various language-internal conditions. Several types of applicatives are relevant for the present paper. Some are multifunctional and I will call them ‘general applicatives’, following Kiyosawa & Gerdts (
Evidence for the internal status of such possessors comes from three domains: (i) case or other grammatical marking of regular possessive constructions in the language, and/or (ii) word order facts, i.e. linear adjacency, and/or (iii) behavioural properties, e.g. the impossibility of the possessor to undergo movement-like processes separately from the possessed noun. To cite one example, in Mi’gmaq (Eastern Algonquian) the derived applicatives in -
(1)
a.
Ges-atm-u-i-t-l
love-
’nt-gij-l.
1
‘S/he loves my mother.’
b.
Gnasg-m-u-a-t-l
wear-
[Piel-oq
Peter-
ugt-atlai’-m-l].
3
‘S/he is wearing (deceased) Peter’s shirt.’
c.
Mali
Mali
nem-it-u-a-pn
see-
[ji’nm-uei
man-
atlai].
shirt
‘Mary saw the man’s shirt.’
The possessor cannot assume the subject status in passive applicatives. The passive counterpart of (1a) is illustrated in (2), suggesting that the possessor cannot be passivized separately from the possessed noun.
(2)
*Ges-atm-u-ugsi-eg-p
love-
’nt-gij-inen.
1
‘Our mother was loved.’
Hamilton (
The pattern illustrated in (1) appears infrequent and is in fact predicted to be unusual or even impossible in theoretical terms because the internal possessor does not qualify as an obvious controller of verbal agreement. Many grammatical frameworks, e.g. Minimalism, HPSG and LFG, take agreement to be a local relation that obtains in specific syntactic domains, in particular, between the verb and the head of an argument phrase, so the pattern in question is non-canonical (in the sense of
I will argue that the possessor in such possessive applicatives is diachronically related to the benefactive argument. The connection between internal possessors and benefactives is well-known cross-linguistically (
My analysis follows the basic insights of the constructional approach to language change without adopting a particular implementational version of Construction Grammar. This approach, referred to by a cover term ‘Diachronic Construction Grammar’, has gained some popularity in recent historical research (
To the best of my knowledge, applicative verbs indexing internal possessors have only been attested in some languages of the Americas (cf.
The discussion of Nez Perce (a member of the Sahaptian family, spoken in the American Northwest) is mainly based on Deal (
(3)
Haama-pim
man-
hi-nees-wewkuny-e’ny-Ø-e
3
ha-haacwal-na
lawtiwaa.
friend.
‘The man met the boys’ friend.’
In (3) the plural on the verb indexes the semantic possessor of the theme, while the possessee does not agree. The possessor has two other properties of primary objects. First, it is marked by the objective case, while the possessee stands in an unmarked case termed “nominative” in Deal (
(4)
’Isii-ne
who-
’e-sewleke’yk-ey’-se-Ø
3
’aatoc?
car.
‘Whose car are you driving?’
This contrasts with the behaviour of regular genitive possessors, for which linear contiguity is required, suggesting that the possessor and the possessee in (3) are independent constituents.
According to Deal (
Alternatively, the possessor can be located within the possessive phrase and be marked by the genitive, with no obvious semantic difference.
(5)
’A-ax-nay’-sa-qa
3
[’ip-nim
3
huukux].
hair.
‘I saw her hair.’
The genitive version of the applicative construction “presents a mix of properties we expect from an external possession structure and those we expect from its internal possession counterpart” (
(6)
a.
Angel-nim
Angel-
paa-’ya
3/3-find-
Tatlo-na/Tatlo-nm
Tatlo-
taaqmaaɬ.
hat.
b.
Angel-nim
Angel-
Tatlo-na/*Tatlo-nm
Tatlo-
paa-’ya
3/3-find-
taaqmaaɬ.
hat.
‘Angel found Tatlo’s hat.’
Crucially for the purpose of this paper, the genitive construction appears to be recent (
The Mayan family consists of about 30 languages, with some controversy regarding further subgroupings (
Central Mayan languages.
Mora-Marín (
a time during which Greater Tzeltalan (Tzotzil-Tzeltal) and Western Ch’olan (Ch’ol-Chontal) were in very close interaction and innovated the obligatory use of
None of the available sources explicitly discusses surface constituency in Western Ch’olan, but there are indications that in Greater Tzeltalan the agreeing possessor is internal to the theme object.
Like in other Mayan languages, core arguments in both Tzotzil and Tzeltal are indexed on the verb by two sets of person markers, usually glossed as Set A (ergative) and Set B (absolutive). Internal possessors are expressed by possessive clitics on the possessed noun. Possessive applicatives have no obvious semantic effects and differ from canonical external possession in that the possessor does not need to be animate or affected. It bears no thematic relation in the clause and is not sensitive to the semantic class of the verb. The theme object must be syntactically possessed; the applicative construction is impossible if it is not. As Aissen (
(7)
Lah
a-we’-bon
*(k)-waj.
1
‘You ate my tortilla.’
A lexical possessor co-occurs with the 3
(8)
a.
7a
[li
the
s-tot
3
li
Xun-e],
Xun-
7i-j-k’opon-be.
‘I spoke to Xun’s father.’
b.
[K’usi
what
s-tzekil-al]
3
ch-a-jal-be?
‘What are you making the skirt for?’
Shklovsky (
(9)
a.
[Mach’a
who
s-mut]
3
lah
aw-il-be?
‘Whose chicken did you see?’
b.
*Te
Pedro-he’
Pedro-
x-Marta
lah
s-kuch’-be
y-ala’al.
3
‘As for Pedro, it was Marta’s that he hugged the baby.’
So at least in Greater Tzeltalan the possessor is internal to the theme object phrase in terms of surface constituency, although it triggers regular object agreement on the applicative verb.
The Salish family consists of 23 languages, some with several dialectal varieties. There are several applicative suffixes, but I will only address two types of applicatives in Interior Salish, relying mainly on Kiyosawa (
Northern Interior Salish languages (Lillooet, Thompson and Shuswap) display the general applicative in -*
(10)
M-sté(t)Ɂǝ-x-t-sm-s
tǝ
beer
‘She drank the beer on/for me.’
Four Southern Interior Salish languages (Okanagan, Coeur d’Alene, Columbian a.k.a. Nxaʔamxcin, and Spokane a.k.a. Kalispel), also have the general applicative expressed by the reflexes of -*
(11)
a.
Mary
Mary
ʕac-x-ít-s
tie-
[iɁ
t
snikłcʾaɁsqá
horse
[iɁ
ttwʾit].
boy
‘Mary tied the horse for the boy.’
b.
Kʾwúlʾ-š-t-ǝn
make-
[łuɁ
Agnes]
Agnes
[łuɁ
t
yám
basket
‘I made a basket for Agnes.’
The distribution of grammatical functions in these constructions appears to be identical to Northern Interior Salish: the applied primary object is not marked for case (or sometimes marked by an absolutive), whereas the theme takes oblique marking. Theme objects in Okanagan undergo extraction just like regular direct objects, albeit other obliques cannot extract. This led N. Mattina (
In addition to general applicatives, all Southern Interior Salish languages exhibit more specialized possessive applicatives which go back to
(12)
a.
Way’
yes
kaʔkíc-ł-t-s-әn
find-
an-q ̓aʔxán.
2
‘Yes, I found your shoes.’
b.
Mús-ł-t-ǝn
feel-
łuɁ
Albert
Albert
sǝnɁuršícti-s.
stove-3
‘I felt Albert’s stove.’
Possessive applicatives are usually referred to as “external possession” in the literature on Salish, and there is indeed evidence that the theme and the applied object are independent constituents in Okanagan. The applied object can be expressed by a clause-level DP coreferential with the internal possessor (13a), or it can even be referentially distinct from the possessor of the theme (13b).
(13)
a.
Lut
kwu
1
a-ks-naq ̓w-m-ł-t-әm
2
in-kәwáp.
1
‘Don’t steal my horse from me.’
b.
NɁiy-ł-t-s-n
buy-
Fred
Fred
iɁ
pʾúyxǝn-s.
car-3
‘I bought you Fred’s car.’
The possessor can be pronominalized on the verb separately from the possessed, much like the applied object of the general applicative, which clearly does not form a constituent with the theme, cf. the possessive applicative in (14a) and the general applicative in (14b). In both examples the theme is a referential null and the applied object is expressed by a pronominal marker incorporated into the verb.
(14)
a.
Kwu-p’íc’-ł-t-xw.
1
‘Lit. You pinched it of mine.’
b.
Kwu-p’íc’-x-t-xw.
1
‘You pinched it for me.’
This indicates that the Okanagan possessive applicative is a true external possessor construction.
(15)
a.
Kwu-ʕ’áw ̓m-ł-t-xw
1
yaʔ
ʕacsqá
reins
‘Let loose my reins.’
b.
Gwnit-ł-t-sel-s
call-
k̓wít̓n.
mouse
‘He asked for my mouse.’
However, the situation seems different in Spokane and Columbian. In Spokane possessive applicatives, a single determiner precedes the possessor and possessed, suggesting that they form one DP. The possessed theme must bear possessive agreement targeting the possessor. Carlson (
(16)
a.
Ɂíłi-ł-t-ǝn
ate-
[łuɁ
Albert
Albert
sqéltč-3].
meat-3
‘I ate Albert’s meat.’
b.
Ɂíłi-š-t-ǝn
ate-
[łuɁ
Albert]
Albert
[łuɁ
t
sqéltč].
meat
‘I ate some meat for Albert.’
In Columbian the possessor and possessed behave as one syntactic unit. In (17) the whole possessive phrase ‘John’s brother’ is promoted to subject when the applicative is passivized.
(17)
[John
John
l
qáck-s]
brother-3
x̣ə́lq’-ɫt-m
kill-
t
sʕʷáʔʕʷaʔ.
cougar
‘John’s brother was killed by a cougar.’
The internal status of the possessor is supported by the presence of the genitive, the usual (albeit optional) marker of internal possessors (
Thus, there is initial evidence that in Spokane and Columbian the possessor belongs to the same phrase as the theme, although it is impossible to reach a definitive conclusion based on a limited number of available examples.
Chimane, spoken in Northern Bolivia, is closely related to Mosetén (
Chimane has several applicatives. The applicative in
(18)
a.
Juan
Juan(
taj-je-’
touch-
[un
hand(
[mu’
Sergio-s]].
Sergio(
‘Juan touched Sergio’s hand.’
b.
Juan
Juan(
taj-je-bi-te
touch-
[un
hand(
[mu’
Sergio-s/*Sergio]].
Sergio(
‘Juan touched Sergio’s hand.’
Ritchie argues that the notion of DP is primarily defined by gender agreement. Adnominal dependents (determiners and attributive modifiers) may be DP-internal on the surface; their position in the phrase is fairly free, the only real restriction being that determiners must be DP-initial. Alternatively, they may be located discontinuously to the head noun due to the non-configurational nature of Chimane syntax. In either case, however, adnominal dependents agree with the semantic head in gender. Possessors do not differ from attributive modifiers in this respect: they exhibit the same type of gender agreement and follow relevant positional restrictions. Crucially, in both examples (18) the possessor
There are also bound possessive pronominals that cliticize to any sub-constituent of the DP apart from the determiner. Bound possessors have the same form as free pronominal possessors but do not show attributive agreement; both are glossed here using English possessive pronominals. Free-standing possessors can co-occur with bound possessors. Examples (19) show the three options available for pronominal possessors: a free agreeing possessor, a bound possessive clitic and the double-marking option.
(19)
a.
Cạv-e-bi-te
see-
[ococo
frog(
mu’-si’].
his-
‘I saw his frog.’
b.
Yụ
I
nại-j-bi-te
see-
[ococo=mu’].
frog(
‘I saw his frog.’
c.
Chị-ya-cse-bi
know-
[mọ’
dyijyedye’=mu’in
thought(
mu’in-si’=in].
their-
‘He knew their thoughts.’
Irrespective of how the internal possessor is expressed, the applicative verb indexes its gender and number. Ritchie concludes that possessive applicatives contrast with external possessive constructions in that the possessor forms a syntactic constituent with the possessed noun. The external possessive construction is actually available in Chimane too, but it exhibits different properties: the verb bears no applicative, and the possessed noun must be a body part and stand in an oblique form.
Mosetén, closely related to Chimane, exhibits the cognate applicative in
Based on the evidence from Salish and Mayan, for which comparative data are available, we can conclude that, first, the distribution of possessive applicatives is more restricted than that of general applicatives. Possessive applicatives are only observed in the youngest linguistic subgroupings, Southern Interior Salish, Greater Tzeltalan and Western Ch’olan. Second, possessive applicatives which target internal possessors have even more restricted distribution. They are only available in a subset of these languages, Greater Tzeltalan and possibly Spokane and Columbian, so they are more recent. The same appears to be true of Chimane-Mosetén. The most direct evidence for the claim that internal agreeing possessors are historically novel comes from Nez Perce, where the possessor can be either external or internal and the latter construction is a recent innovation. The next two sections will explain the diachronic path I propose for these structures.
This section studies applicative constructions in more detail and proposes that possessive applicatives go back to general applicatives.
All constructions addressed in the previous section show the typical properties of “low applicatives” in the sense of Pylkkänen (
While the core meaning of canonical possessive constructions such as
However, even canonical low applicatives can imply a non-directional possessive relation, and particular verbs and contexts strongly favour a non-directional interpretation. It may be less frequent or in fact impossible in other situations, while there are also situations in which the construction appear truly ambiguous: (i) the applied object is understood as the prospective or former possessor of the theme object, or (ii) the applied object is understood as the actual possessor, e.g.
The literature contains various accounts of this ambiguity. According to one line of thinking (e.g.
The same kind of ambiguity is amply documented in the Northern Interior Salish general applicatives. When interpreted as possessed, the oblique theme is marked for possession in most available examples (
(20)
a.
Młmałqw-x-t-s
paint-
tǝ
citxw-s.
house-3
‘Hei paints the/hisk house for himk/Hei paints hisk house.’
b.
ɁúqweɁ-xi-t-sem-es
drink-
tǝ
tíy.
tea
‘She drank my tea up on me.’
c.
Txwus-min̓-xí-c-kaxw
look-
ni
n-c̓qá
1
‘Look out for my horse for me!’
More examples of ambiguity are cited in Kiyosawa & Gerdts (
This is also true of Western Ch’olan. Coon & Henderson (
(21)
a.
Mi
k-mäñ-b-eñ
waj
tortilla
aläl.
child
‘I buy the child the tortillas.’
b.
Tyi
a-ts’äk-ä-be
k-alo’bil.
1
‘You cured my son for him/her.’
c.
Tyi
k-il-ä-be
[y-alo’bil
3
aj-Betui]
‘I saw Beto’s son.’
The external status of
(22)
a.
Tyi
il-ä-b-eñ-tyi-y-oñ
see-
k-alo’bil.
1
‘My son was seen.’
b.
Tyi
k’ajty-i-b-eñ-tyi-y-oñ
ask-
ts’ak
medicine
(joñoñ).
1
‘I was asked for the medicine.’
In this analysis the same applicative structure is responsible for different readings of the applied object, while coreferentiality of the external
(23)
a.
Mi
k-mäñ
waj
tortilla
cha’añ
aläl.
child
‘I buy tortillas for the child.’
b.
Tyi
k-il-ä
[y-alo’bil
3
aj-Betu].
‘I saw Beto’s son.’
Coon & Henderson (
Similar facts are observed in Chontal, where the applied object takes a variety of roles including benefactives and possessors,
(24)
7u-jäk-sä-b-0-ón
7u-cho7án.
1
‘They lowered the price for me.’
Unlike non-applied objects, the applied object must be affected (
Deal (
(25)
a.
’E-hiteeme-ney’-se
3
Cecil-ne
Cecil-
tiim’es.
book.
‘I’m reading Cecil the book.’
b.
Kaa
and
pé-e’wi-èn-ye
3
tilóqa-na
people-
wewúkiye.
elk.
‘And he shot the elk for the people/And he shot the people’s elk.’
In (25a) the applied object is translated as a non-subcategorized benefactive, but Deal (
Thus, possessive applicatives are not a distinct construction in Northern Interior Salish and Western Ch’olan. All existing work to date suggests that they are just one of several readings conveyed by general low applicatives. The possessive relation between the applied object and the theme is normally directional and involves the prospective or former possessor, but the non-directional possessive meaning is obtainable under certain conditions. The Nez Perce raising construction is also multifunctional, although the possessive reading appears more frequent.
This section deals with the languages that distinguish between benefactive and possessive applicatives as separate constructions. In benefactive applicatives the relation between the theme and the applied object is directional; the applied object is interpreted as the affected prospective or former possessor of the theme. In contrast, in possessive applicatives the relation between the theme and the applied object is non-directional possession.
The difference between the two constructions is most clearly seen when they employ distinct verbal morphology. As discussed in Section 2.3, all Interior Salish languages have the applicative in -*
Unlike these languages but somewhat similarly to the languages addressed in Section 3.1, Mi’gmaq and Greater Tzeltalan employ the same applicative marker in both possessive and benefactive applicatives; still these are not constructionally equivalent. The difference goes beyond the obvious fact that in possessive applicatives the applied object must be referentially identical to the possessor of the theme.
Consider Mi’gmaq first. As seen in Section 1, -
(26)
Elugw-atm-u-i-t-l
fix-
a’pi-l.
net-
‘S/he fixes a/the net for me.’
Unlike the possessor, the benefactive applied object is an independent surface constituent and can be passivized, cf. the ungrammatical (2) above and the grammatical (27).
(27)
Elugw-atm-u-ugsi-eg-p
fix-
a’pi-l.
net-
‘A/the net was fixed for us.’
Another difference is that applicativization is required for all ditransitive verbs. However, for monotransitives, possessive applicatives are optional in the sense that they alternate with the regular transitive construction. In (28a) the possessor must be topical, unlike in (28b).
(28)
a.
Ges-atm-u-i-t-l
love-
’nt-gij-l.
1
‘S/he loves my mother.’
b.
Ges-al-a-t-l
love-
’nt-gij-l.
1
‘S/he loves my mother.’
Hamilton’s (
Similar to Western Ch’olan, the Greater Tzeltalan -
(29)
Lah
k-il(-be)
s-nah
3
Pedro.
Pedro
‘I saw Pedro’s house.’
Such alternation are due to the factors that Aissen (
The Nez Perce genitive possessor is located in a non-thematic position, and existing literature does not cite any examples where it may be interpretable as benefactive. So unlike the ambiguous possessor raising construction, the genitive construction is a specialized possessive applicative. It encodes a non-cancellable non-directional possessive relation between two non-subject arguments as part of its core meaning, or at least there is no evidence to the contrary.
This section argues that possessive applicatives developed out of general low applicatives due to the conventionalization of the implied meaning.
This can be best illustrated for Salish. The general low applicative *-
This process arguably followed a frequent diachronic path that constructions go through over time proposed in Diachronic Construction Grammar. In a nutshell, the notion of construction refers to a symbolically linked pair of form (SYN) and meaning (SEM), which creates an integrated whole. Diachronic development consists of a succession of (possibly abrupt) micro-changes that can involve any aspect of the internal organization of the construction, i.e. SYN, SEM or the linking between the two (
This paper does not aim to examine all details of the constructions in question, so I will abstract away from many aspects of their form and meaning and only represent those that appear relevant for the diachronic story. I take general applicatives to be the pairing of SYN1 with SEM1. SEM1 indicates the (abstract) transfer of possession between two non-subject arguments. SYN1 refers to the syntactic structure in which two syntactically independent objects stand in a directional possessive relation, i.e. [
We have seen in Section 3.1 that in Proto Interior and Northern Interior Salish SEM2 is cancellable, so it is not the encoded meaning of the general applicative SYN1/SEM1. In contrast, I propose that in Southern Interior Salish SEM2 generalized and was eventually paired with the new formal structure SYN2, giving rise to possessive applicatives. This process followed three steps claimed to be common in the creation of a new construction (
In the first step, the construction undergoes semantic expansion. Frequent implicatures often get conventionalized with time, since frequency gives constructions a stronger mental representation (
The rise of the innovative SEM2 motivates the second diachronic step, the reanalysis of form SYN1 → SYN2. That is, once the non-directional possessive meaning generalizes and a possessive relation is no longer cancellable, the applied object gets reanalysed as the semantic possessor of the theme. No visible structural changes occur at this intermediate stage; “the change in SYN is posited on the basis of the prior change in SEM plus the knowledge that observable changes in SYN are forthcoming” (
At the third step SEM1/SYN1 continue to combine to constitute the original construction, while SEM2/SYN2 combine to make a new construction. As a result, the general low applicative splits into two novel constructions in Southern Interior Salish. The benefactive applicative SEM1/SYN1 expresses directional possession, while the new possessive applicative SEM2/SYN2 expresses non-directional possession. The two constructions are now representationally distinct and can be simultaneously present within one language.
This processes removed the inherent ambiguity of general applicatives, but the possessive applicative can still carry the benefactive implicature reflecting its historical origin. In Okanagan, for instance, speakers often translate the possessive applicatives in
The proposed scenario is summarized in Table
Applicative constructions in Interior Salish.
Languages | Construction | Form | Encoded meaning | Implicature |
---|---|---|---|---|
Proto Interior Salish, Northern Interior Salish | general applicative | SYN1 |
SEM1 |
SEM2 |
Southern Interior Salish | benefactive applicative | SYN1 |
SEM1 |
SEM2 |
Southern Interior Salish | possessive applicative | SYN2 |
SEM2 |
SEM1 |
Step (iii) was accompanied by actualization, i.e. structural changes that align form with the new semantic analysis bringing “the surface into line with the innovative underlying structure’’ (
Given the gradual nature of syntactic change, we can expect to find various intermediate stages. In Spokane and Columbian the possessive applicative in -
(30)
Kʾwúlʾ-ł-t-ǝn
make-
[łuɁ
t
yám
basket
[łuɁ
Agnes].
Agnes
‘I made a basket for Agnes.’
This suggests that actualization went through two distinct steps, the emergence of the new applicative suffix and the change in case marking. (30) reflects the stage which preserves case marking (and semantics) of general applicatives but the suffix -
Thus, the possessive applicative construction in Southern Interior Salish emerged from the general applicative. The main driving force behind the change is the conventionalization of a non-encoded but frequently implied non-directional possessive relation between
As seen above, in some possessive applicatives the possessor is an independent syntactic argument (external possessor), whereas in other possessive applicatives it belongs to the phrase headed by the theme (internal possessor). The goal of this section is to argue that the latter construction goes back to the former. In the absence of direct historical evidence, the diachronic scenario I will propose is based, first, on apparent parallelism with the languages for which historical records are actually available, and second, on interpreting cross-linguistic synchronic variation as stages in language change. This kind of methodology, recommended e.g. in Greenberg (
Below I summarize the facts from a number of languages where the dative possessor goes back to the dative argument due to the reanalysis of constituency relations.
Old Church Slavonic, known from the manuscripts from the late 9th until the 12th c., is the first literary account of the history of Bulgarian and for practical purposes can be considered its immediate ancestor. According to Krapova & Dimitrova (
(31)
da
to
pokryjǫtъ
cover
sę
emou
3
[děla
deeds.
ego]
3
‘to cover for him his deeds’
There were also external dative possessors inherited from early Slavic and usually called “doubly bound datives”. Typically they are pronominal (94% of occurrences in the corpus studied by
(32)
a.
jegda
when
že
imŭ
3
prěbivaachǫ
break.in.two
golěni
knees
‘when they broke their knees in two’
b.
Ōtpuštajǫtŭ
forgive
ti
2
sę
[grěsi
sins
tvoi].
your
‘Your sins will be forgiven/Sins will be forgiven to you.’
The claim that Krapova & Dimitrova (
The tendency for replacing the genitive with the dative is a statistical preference already in
Krapova & Dimitrova (
(33)
a.
Ne
not
by
be.
mi
1
umrĭlŭ
died
bratŭ.
brother.
‘My brother would not have died on me.’
b.
Ne
not
by
be.
umrŭlŭ
died
mi
1
bratŭ.
brother.
‘My brother would not have died.’
The dative clitic typically preceded the theme when in the postverbal position. Following earlier authors, Krapova & Dimitrova (
Modern Bulgarian shows essentially the same patterns as Middle Bulgarian. As described above, the external possessor construction developed a novel variant, in which the dative clitic appears inside a DP without apparent interpretational difference. So dative pronominals either function as postnominal possessive clitics (34a) or are placed next to the verb (34b).
(34)
a.
Znam
know.1
[adresa
address.
mu].
3
b.
Znam
know.1
[mu]
3
[adresa].
address.
‘I know his address.’
Cinque & Krapova (
(35)
Tja
3
[mu]
3
ščupi
broke.3
[malkija
little.
prăst].
finger
‘She broke his little finger (on him).’
The true external possessor and the benefactive compete for the same position and are not compatible.
Similarly, it is widely accepted that internal dative possessors are a relatively recent development in Germanic. They are found in several languages which maintain a case system, i.e. some varieties of German, Swedish dialects, possibly Gothic, Old Norse, Middle Dutch and, marginally, Old English (
The change was due to the reanalysis of the clause-level datives as DP-internal, but its exact nature is a matter of discussion. The scenario standardly assumed for Germanic goes back to as early as Havers (
This is also observed in dative-like prepositions. In Old Norse inalienable external possessors were marked by the dative case or the preposition
(36)
a.
[Leveren]
liver.
matte
must
de
they
fjerne
remove
[på
on
ham].
him
‘They had to remove his liver.
b.
Det
there
floy
flew
en
a
fugl
bird
[over
over
hodet
head.
på
on
ham].
him
‘A bird flew over his head.’
Lødrup (
A similar development took place from Old Norse to Modern Icelandic. Pronominal datives are usually preposed in the language of Poetic Edda but occur in a post-nominal position in the later Old Norse prose (
Going back to possessive applicatives, this section studies variation in the expression of possessor.
No overt expression of the internal possessor is necessary in external possessor constructions. This is equally true of Bulgarian, and of Okanagan and Coeur d’Alene, as seen above. What these languages have in common is that the possessor does not belong to the theme object phrase, but corresponds to a clause-level DP. Depending on analysis, the unpronounced internal possessor can be understood either as implied or as a null pronominal which stands in an anaphoric relation with the external possessor.
In contrast, we saw that in Mi’gmaq, Greater Tzeltalan, Nez Perce, Chimane, Spokane and Columbian the possessor forms a syntactic constituent with the possessed theme. There is no evidence that such internal possessor has any form of external representation in Mi’gmaq, as witnessed, among other things, by the fact that it cannot be passivized. However, the internal possessor has been claimed to have a silent external counterpart in the rest of these languages. That is, various analyses postulate a null pronominal element located externally to the theme phrase but coreferential with the internal possessor. This dual representation of possessor resembles backward raising, i.e. covert subject-to-subject movement which leaves the overt subject in the complement clause but produces a silent copy in the matrix clause (
According to Aissen (
(37)
a.
Ch-i-toyilan-b-at
j-jol.
1
‘My head was lifted over and over.’
b.
*L-i-cham-be
j-tot.
1
‘My father died.’
Another argument is that possessor raising has syntactic restrictions. In particular, it is impossible in clauses that already contain an indirect object. This is because, in Aissen’s analysis, the possessor anticopy is DP-external and, similarly to Bulgarian, it competes with the benefactive for the applied object status—even though possessive and benefactive applicatives are representationally distinct as far as their synchronic constructional properties are concerned (Section 3.2).
The empirical facts from closely related Tzeltal are largely the same. (38) demonstrates passivization.
(38)
Il-bot
see-
s-nah
3
Mariya.
Maria
‘Maria’s house was seen.’
Shklovky’s (
(39)
Lah
k-mil-bat.
‘I killed it (lit. I killed your it).’
Shklovsky (
There is (admittedly, inconclusive) evidence that an analysis along similar lines may be required for at least a subset of agreeing possessors in Columbian. All available examples of Columbian possessive applicatives contain an internal possessor. As seen in Section 2.3, the agreeing possessor can be genitive and the possessive phrase as a whole is promoted to subject. However, the situation is different for 1st and 2nd person possessors. Regular 1st person (and possibly 2nd person) objects do not undergo subject promotion via passivization. Likewise, one cannot passivize a possessive applicative if the possessor of the theme is lst or 2nd person, cf. (40a) and (40b) constructed based on the discussion in Willett (
(40)
a.
*Kn
1
c’aw’-nt-m
wash-
t
John.
John
‘I was washed by John.’
b.
Ɂin-qack
1
x̨ǝlq’-(*ɫt)-nt-m
kill-
t
sʕʷáʔʕʷaʔ.
cougar
‘My older brother was killed by a cougar.’
In (40b) the possessive applicative is impossible. This seems to suggest that the 1st person possessor of the theme behaves exactly like a regular 1st person grammatical object for the purpose of passivization, as would be expected if it had some kind of silent clause-level representation similar to Greater Tzeltalan.
It is not known whether the genitive possessor is passivizable in Nez Perce, but Deal (
(41)
Haacwal-nimagent
boy-
pee-kiwyek-ey’-se-Ø
3>3-feed-
[’ip-nim
3
ciq’aamqal]
dog.
[Meli-nm
Mary-
ke’niks].
leftovers.
‘The boy is feeding his/herj/*i dog Maryi’s leftovers.’
In this sense the genitive in (41) behaves like an external possessor. Deal (
Chimane offers the clearest example of internal possessors doubled by external pronominals because these pronominals may actually be pronounced. In Ritchie’s (
(42) illustrates that the proxy clitic is external to the possessive phrase. I gloss it with grammatical abbreviations such as e.g. 3
(42)
Mi
2
naͅij-bi-te
see-
[ococo
frog(
Juan-si’]
Juan(
munja=mu’.
yesterday=3
‘You saw Juan’s frog yesterday.’
Importantly, the clitic cannot occur in non-applicative constructions or at least it is strongly dispreferred by the speakers, cf. (42) and (43).
(43)
Mi
2
naͅijtye-‘
see-3
[ococo
frog(
Juan-si’](?*=mu’).
Juan(
‘You saw Juan’s frog.’
Additional pieces of evidence used to demonstrate that the possessor has external representation are passives and pronominalization. Example (44a) illustrates that the passive applicative verb cross-references the internal possessor or, more precisely, the proxy clitic coreferential with the internal possessor, as per Ritchie’s analysis. (44b) shows that the possessor and possessed noun are pronominalized separately, just like in Tzeltal (39). In this example the possessee refers to the part of a tree and is left unpronounced. The possessor refers to the tree itself and is indexed by means the masculine object marker on the verb.
(44)
a.
[Maria-ty
Maria(
vojity]=mọ’
brother(
ja’-ĉat-bu-ti-’
(Juan).
Juan(
‘Maria’s brother was hit (by Juan).’
b.
Judyeya’
and
aca-dyes
house(
jäc-bi-te
remove-
tsun.
1
‘And we take it (lit. its it) to make houses.’
Thus, the doubling clitic is only possible in possessive applicatives and is therefore analysed as the optional clause-level representation of the internal possessor of the theme. However, unlike in Greater Tzeltalan but similar to the languages with external possessors, the internal expression of the possessor is not required.
(45)
Chị͔-ya-cse-bi
know-
mọ’
dyijyedye’.
thought(
‘He knew their thoughts.’
Table
The expression of possessor in possessive applicatives.
Languages | Agreement controller | Possessor on the theme |
---|---|---|
Okanagan, Coeur d’Alene | external possessor | optional |
Chimane | optional | |
Nez Perce genitive, Greater Tzeltalan, Columbian | obligatory | |
Mi’gmaq, (Spokane) |
internal possessor | obligatory |
As shown, possessive applicatives where the applied object is overtly expressed externally to the theme DP and stands in a non-directional possessive relation to it are only attested in Okanagan and Coeur d’Alene. This kind of construction seems cross-linguistically infrequent (cf.
The hypothesis I would like to put forward is that non-directional possessive applicatives are historically unstable if the possessor and possessed are independent constituents. They tend to be reanalysed: the possessor loses its argument status and becomes internal to the theme object DP assimilating to the class of internal possessors. This process was observed in the languages with dative possessors (Section 4.1), and I propose that Chimane, Greater Tzeltalan, Columbian, Spokane, Mi’gmaq and the Nez Perce genitive construction followed a similar path. It went through several stages reflected in cross-linguistic variation (Section 4.2).
This historical change has no obvious semantic effect and only targets the formal aspect of the construction, namely, constituency and grammatical relations. This can be schematically represented as follows: [
Since the history of the languages discussed in this paper is poorly documented, it is impossible to judge whether the change SYN2 → SYN3 occurred through rebracketing. This possibility cannot be excluded for Mayan and Spokane because of constituent order: the lexical benefactive invariably follows the theme object, whereas the lexical possessor follows the possessed noun, which could have facilitated rebracketing. However, rebracketing is more problematic for Nez Perce, Chimane and Mi’gmaq, where the order of the benefactive and the theme is not fixed, while in Columbian the benefactive precedes the theme but the genitive is either prenominal or postnominal (
This suggests that rebracketing did not play a major role in these languages. I propose instead that, similarly to what was suggested for Icelandic, a new possessor slot was created in the theme object DP. This was motivated by analogy with regular possessives.
Constructions are known to be organized into networks; they are related to other constructions of varying degrees of complexity through shared formal and/or functional properties. Although many changes result from the reanalysis of meaning which can further cause a change in form, some changes are driven by external dimensions, i.e. the place of the construction in the constructional network and its analogical links to other constructions. Such changes occur through the generalization of a pre-existing pattern to new instances and the redeployment of old constructional pieces for new purposes, and can involve form as well as meaning. The relevant mechanism of change has been referred to as “analogical extension” (
As discussed in Section 3, possessive applicatives convey the meaning of regular possessive constructions as far as the relation between non-subject arguments is concerned. To put it differently, they are related to possessive constructions, independently present in grammar, through the shared semantic component (SEM2). Arguably, users of language can perceive this association and generalize over it (cf.
Analogical extension represents the first step in the emergence of a new construction and can be accompanied by further morphosyntactic changes (
As shown in Section 4.2, languages differ in this respect because the internal possessor can still have a partial external representation as an overt or covert pronominal. In Chimane the external possessor was pronominalized as an optionally pronounced clitic coindexed with the internal possessor. The expression of the possessor on the theme is optional, which arguably suggests that Chimane is at a less advanced stage of the grammaticalization of the internal possessor than other languages: the internal expression of the possessor has not yet become obligatory. In Greater Tzeltalan, Nez Perce and possibly Columbian, the internal expression of the possessor is required, whereas the external possessor is represented by some kind of referential null with syntactic effects but no visible phonological reflex. Finally, in Mig’maq external possessor was altogether eliminated from the clausal structure.
Another consequence of the analogical change SYN2 → SYN3 is morphological actualization. This involves, first, the newly created possessor assuming the form of the canonical internal possessor in the language, as seen e.g. in Nez Perce. According to Deal (
The second morphology-related change has to do with the form of the possessed noun being remodelled based on the form of the regular primary object. As discussed in Section 3.3, the theme lost its oblique marking when the subset of general applicatives was reanalysed as possessive in Southern Interior Salish. Kiyosawa (
To summarise, I proposed that possessive applicatives in which the verb agrees with the internal possessor of the theme were modelled after possessive constructions in terms of the syntactic relation between the possessor and possessed. The analogical transfer of structure removed the non-iconic mismatch between SYN2 (two syntactically independent verbal arguments) and SEM2 (the non-cancellable non-directional possessive relation between them), by aligning the form of possessive applicatives with that of canonical internal possessives.
As has been claimed in the literature, the constructional approach to language change offers a number of useful advantages. Among other things, it proves fruitful in
accounting for seemingly unmotivated syntactic patterns that do not easily fit in a syncronically attested grammatical network for a given language, or that present a typologically odd and inexplicable pattern. (
The present paper contributed to this general enterprise by offering a tentative historical scenario for the emergence of an unusual construction that violates standard assumptions about agreement domains: an applicative verb agreeing with the internal possessor of its theme object.
The paper proposed that this construction goes back to a more canonical structure. It developed relatively recently from the construction in which a non-directional possessive relation holds between two independent clause-level arguments, the theme object and the applied primary object, and the latter triggers object agreement on the applicative verb. Applicatives that link non-directional possessive semantics with this kind of syntactic structure are typologically rare and are likely to undergo structural reanalysis whereby the applied object gets reanalysed as the internal possessor of the theme. For the languages addressed in this paper the reanalysis consists in introducing a new possessor slot on the theme object and the subsequent elimination of the external possessor, fully or partially.
The change is motivated by analogical assimilation to another pattern which independently exists in grammar, the canonical internal possessive construction. Alignment in terms of the syntactic relation that holds between the possessor and possessed is ensured by semantic affinity between the two constructions. When constituency is reanalysed, the location of agreement remains intact on the verb, creating a non-local configuration. Agreement is then non-canonical in the resulting applicative construction essentially because this construction inherits properties of two parent constructions which served as its historical source, surface constituency of internal possessives and agreement pattern associated with clausal arguments. This scenario supports the view that grammatical constructions with mixed properties arise as the consequence of partial borrowing from other constructions at different levels of representation (cf.
My proposal implicates analogy as a factor constraining the organization of language systems and triggering diachronic change. Analogical reasoning as a cognitive skill is based on pattern recognition ability, so the explanatory role of analogy has repeatedly been highlighted in construction-oriented grammatical research, cognitive and acquisition studies, which maintain that grammar is pattern-based (e.g.
1 = 1st person, 2 = 2nd person, 3 = 3rd person,
All phrases with the nominal lexical content are referred to as DP for the sake of consistency and following some of the literature. I preserve the transcription of the sources, but glosses have been regularized and modified in places.
Kiyosawa (
Applicatives derived from a small closed class of verbs have distinct properties and tend to lexicalize (
Interestingly, Linzen’s (
This is also observed in Tzotzil (
Little Spokane data are available. In the absence of the evidence to the contrary, I assume that the agreeing possessor is not silently represented at the clause level.
Ackerman & Nikolaeva (
This research was carried out as part of the AHRC-funded project “Prominent Possessors” (AH/M010708/1). The AHRC’s support is gratefully acknowledged. I am grateful to Ilja Serzants, András Bárány and three anonymous reviewers for their comments on the earlier versions of this paper, as well as audiences at the 9th International Conference on Construction Grammar (Juiz de la Fora, 2016), the workshop “Diachrony of Differential Object Marking” (Paris, 2017), and the 34th annual conference of the Israel Association for Theoretical Linguistics (Beer Sheva, 2018). I also thank Brad Montgomery-Anderson for sending me his paper.
The author has no competing interests to declare.