Participial Relatives and Ergativity Er-Xin Lee
IntroductionIn order to investigate the structure of participles, we use their appearance in
reduced relatives as a probe. The very fact that in various reduced relatives can be
constructed out of certain participles and not others leads us to investigate the
organization of participles with respect to reduced relatives in linguistically distant
languages, specifically English and Hindi-Urdu (HU).
In Section 1, we provide a brief overview of a descriptive generalization – the
have-be generalization – of participles’ ability to function as reduced relatives according
to their ability to occur with the auxiliary be. The strengths of the generalization are
discussed, specifically using English data to provide evidence. The weaknesses of the
generalization are also laid out, specifically with respect to data from HU, leading to an
exploration of a new generalization in Section 2.
The new generalization – the Ergative-Nominative generalization – seems at first
to be on the right track in positing what the have-be generalization fails to account for.
However, as further exploration into HU constructions reveal, the distinction that is
drawn between subjects that are Ergative and those that are Nominative is dependent on
whether subjects’ are External Arguments of the verbal predicates.
Thus, in Section 3, we investigate the role of External Arguments of participial
predicates with respect to reduced relative formation. Based on the facts from English
and HU reduced relatives, we argue that participles bear different structural features. In
particular, External Arguments in Perfective participles need to be externally licensed,
whereas External Arguments in Imperfective participles do not need such a requirement.
As for participles of reduced relatives, we show that for most cases, only those with no
External Arguments are able to form acceptable reduced relatives.
1
Participial Relatives and Ergativity Er-Xin Lee
1. The have-be generalizationIn many Indo-European languages, the formation of reduced relatives has
restrictions based on the particular kinds of participles occurring in them. The languages
in which such restrictions have been attested are predominantly European languages such
as English and Italian. A summary of Burzio (1986), Iatridou, & Izvorski (1999),
Iatridou, Anagnostopoulou & Izvorski ((=IA&I) 2000), all of which address the
relationship of the auxiliaries occurring in these languages to the participles, is provided
in Sections 1.1.1-1.1.3. In those languages, the Perfect participle stands out as having
three options for the auxiliary – (i) have as the auxiliary, of which English is an example,
(ii) be as the auxiliary (cf. Bulgarian, as described in IA&I 2000 and Marvin 2000), and
(iii) auxiliary-selection for languages such as French and Italian, wherein have and be are
both candidates for the auxiliary although the exact use have- or be-Perfects is dependent
on the whether the verb is Transitive, Unaccusative, or Unergative.
The distribution of reduced relatives for many languages is thus captured by the
following generalization:
(1) The have-be generalization: an XP can function as a reduced relative only if it can appear as the complement of be. (cf. Embick 1997, 2001, Anagnostopoulou, Iatridou, & Izvorski 1999, IA&I 2000, a.o.)
This generalization is exemplified by the paradigm below:
(2) a. Active Verb, Present Partciplei. The student is eating an apple.ii. The student [eating an apple] is tall.
b. Active Verb, Past Participlei. The student has eaten an apple.ii. *The student [eaten an apple] is tall.
c. Passive Participlei. The apple was eaten by John.ii. The apple [eaten by John] was sweet.
d. Unaccusative Verb, Past Participlei. The parcels have come by mail.
2
Participial Relatives and Ergativity Er-Xin Lee
ii. *The parcels [come by mail] are heavy.
That the English Imperfective and Passive participles are able to appear as
reduced relatives seems to be directly correlated with the fact that in the Imperfective and
Passive cases, that is, (2a) and (2c), the auxiliary in the unreduced clause is be. In
contrast, the inability of Perfective participles of both Transitive and Unaccusative verbs
to form reduced relatives in (2b) and (2d) can be linked to the XP appearing as a
complement of have.
1.1 Background on the have-be generalizationThe proponents of the have-be generalization include IA&I (2000), Embick
(1997, 2000), Mahajan (1994, 1995). The have-be generalization has been supported by
their data from a diversity of languages as well as been criticized for not being able to
adequately account for reduced relative formation.
The generalization that an XP can function as a reduced relative only if it can
appear as the complement of be which we have termed the have-be generalization, has
for IA&I been a powerful generalization in revealing the distribution of Reduced relatives
in the many European languages that they study.
Reduced relatives prove to be useful in studying whether the meaning of the
Perfect is spread over the auxiliary and its participial complement because the formation
of reduced relatives involves the separation of the participial clause from the auxiliary,
such that only the participial clause surfaces in the reduced relative. That the auxiliary
does not appear in a reduced relative is critical to the investigation of how localized the
meaning of a Perfect is.
3
Participial Relatives and Ergativity Er-Xin Lee
1.1.1. BulgarianAs an example, the data from Bulgarian provide evidence that the auxiliary be
does not contribute to the meaning of the Perfect and hence can be isolated from the
participle. As a be-language, Bulgarian participles naturally occur with be across the
board, seen below in (3ia and c) and (3iia). In (3ia and b), the semantics of the Perfect is
Existential and the participles involved are Perfective, whereas in (3ic) the Perfect
reading in the reduced relative is not subsumed, presumably due to the difference of the
participle. The Imperfective participle in (3ii), too, can produce a Universal Perfect
reading in the reduced relative (IA&I 2000). Hence, it is strictly the type of each
participle that determines the meaning of the Perfect in the participial reading of the
reduced relatives.
(3i) Transitive:a. Ženata e pročela knigata
woman-the is read-Pfv book-the‘The woman has read the book.’
b. Ženata pročela knigata...woman-the read-Pfv. book-the‘The woman who has read the book…’
c. Ženata čela knigata...woman-the read-Neut. book-the‘The woman who has read the book…’
Unaccusative:c. Ženata e dosla navreme...
woman-the is arrive-Pfv. on-time‘The woman has arrived on time…’
d. Ženata dosla navreme...woman-the arrive-Pfv. on-time‘The woman who has arrived on time…’
(3ii) a. Ženata e celuvala Ivan ot sutrinta nasamwoman-the is kiss-Pst./Impfv. Ivan from morning-the till-now‘The woman has been kissing Ivan since this morning’
b. Ženata celuvala Ivan ot sutrinta nasam...woman-the kiss-Pst./Impfv. Ivan from morning-the till-now‘The woman who has been kissing Ivan since this morning…’
(IA&I 2000)
4
Participial Relatives and Ergativity Er-Xin Lee
1.1.2. EnglishOn the other hand, Transitive and Accusative data from English, a have-language,
presented below in (4), show that reduced relatives cannot be formed unless the
participles are complements of be. In comparison to Bulgarian, the semantics of the
Perfect in the reduced relatives is therefore less clearly distinguished as to whether it is
attributed to the participle or auxiliary, or a combination of both. However, it is clear that
have as an auxiliary plays no role at all in reduced relatives as participles that regularly
take have cannot form reduced relatives.
(4) a. I saw the boy *(who has) [eaten the fish]b. I saw the boy *(who has) [left on time]c. I saw the boy *(who has) [walked through the park] vs.
I saw the boy *who is [walked through the park]d. The boy (who is) [singing the Marseillaise] is my brothere. I saw a house (which was) [built in 1925] (IA&I 2000)
1.1.3.ItalianBurzio (1986) presents data for Italian that are also in line with the have-be
generalization. In Italian, an auxiliary-selection language in which be and have are both
able to function as auxiliaries, participles of Unaccusative verbs that take be as the
auxiliary are able to form reduced relatives (cf. 5a and b). Unergative participles that take
have as the auxiliary conform to the have-be generalization – they are unable to form
reduced relatives (cf. 5c and d).
(5) Unaccusative verbs take be:a. Giovanni è arrivato
Giovanni be arrived‘Giovanni has arrived.’
b. il treno [arrivato entro le 3] é ripartito subitothe train arrived by 3 left again immediately‘The train which had arrived by 3 left again immediately’
5
Participial Relatives and Ergativity Er-Xin Lee
Unergative verbs take have:c. Giovanni ha telefonato
Giovanni has telephoned‘Giovanni has telephoned.’
d. * un ragazzo [telefonato a Maria] non puòa guy telephoned to Maria cannot come
‘A guy who telephoned Maria cannot come.’
Passive verbs take be:e. Uno studente che era ammirato...
a student who was admired‘A student who was admired...’
f. Uno studente ammirato... a student admired‘A student admired...’ (Burzio 1981, 1986)
Burzio notes that the ability for reduced relatives to be constructed out of Passive
participles requires be to be the relevant auxiliary (cf. 5e and f). Crucially, Burzio also
points out a common syntactic characteristic of the Passive participles and Past
participles of Unaccusative verbs. The commonality is rooted in the “non-vacuous loss of
Subject θ-role or perhaps the possibility of a Thematic-Subject role” for Passives, and the
“vacuous loss of Subject θ-role” in Unaccusatives (not forgetting that Unaccusatives are
not assigned Thematic-Subject roles). In other words, participles of Passive reduced
relatives and of Past Unaccusatives involve arguments which do not have Subject θ-roles.
We shall return to this subject in Section 3.3 where the licensing of External Arguments
will be discussed further.
Also prevalent at the time was an analysis of reduced relatives known as Whiz-
deletion which was based on the deletion of a wh-phrase and be. Burzio’s data seem to
provide strong evidence that participial complements of be behave differently from
complements of have, hence agreeing with the basic tenets of Whiz-deletion. Burzio
(1986) draws from a Past participial clause such as (6a) with two potential readings for si
and the subsequent absence of reading (i) as in (6b) to show that Wh-be deletion should
6
Participial Relatives and Ergativity Er-Xin Lee
not be able to cause the reading (i) to be precluded in the reduced clause. (See Appendix I
for a more detailed discussion of Whiz-deletion.)
(6) a. Gli individui che [si erano presentati] al direttore…The individuals that self be.Pfv. introduced to the director‘The individuals that (i) self had introduced/ (ii) had introduced themselves to the director...’
b. Gli individui [presentatisi] al direttore…The individuals self-introduced to the director‘The individuals (ii) self-introduced to the director…’ (Burzio 1986)
c. La donna che è stata presentata al direttore…The girl who is been introduced to the director‘The girl who has been introduced to the director…’
d. La donna presentata al direttore…The girl introduced to the director‘The girl introduced to the director…’ (Figari1, p.c)
1.1.4. Accounts of the have-be generalizationGiven the apparent dichotomy between have and be in their ability to form
reduced relatives with their corresponding participles, IA&I (2000) raise the question of
whether participles involving the auxiliary have share features of the Perfect with the
regular have, unlike participles which do not share Perfective features with be, thereby
suggesting that the nature of participles is not uniformed across different types of
participles. Therefore, in proposing a structural analysis for reduced relatives, two
separate structures for Past participles are construed by IA&I.
The structural account, which stems from Kayne’s (1993) proposal, assumes that
be is the canonical auxiliary. Other accounts by Burzio (1981, 1986), Hoekstra (1984),
Roberts (1986) and Cowper (1989a,b) take have as the auxiliary instead, although IA&I
argue that such an approach would only account for languages whose Passive and Perfect
participles do not differ in form and would fail to work for languages such as Bulgarian
1 Thanks to Gianni Figari for providing additional Italian data.Figari also remarked that (39b; below) is antiquated and is not in currency in Italian presently:
Gli individui [presentatisi] al direttore…‘The individuals (ii) self-introduced to the director…’
7
Participial Relatives and Ergativity Er-Xin Lee
and Greek whose participles do differ. Hence, IA&I analyze have as be + an incorporated
nominal head, X. They propose that X, being right below be and right above the
participle, either incorporates to be giving rise to have (7a) or has the participle raise to it,
leaving be as it is (7b).
(7a) …
be XP (+N)
X PerfPIncorporation lose nominal trait Perf AspP
-en Asp VP
(7b) …
be XP (+N)No Incorporation Nominal featureretained X PerfP
Perf AspP
-en Asp VP
An advantage of such an analysis is that the nominal trait of reduced relatives can
be linked to the occurrence of be. Under the assumption that reduced relatives are
nominal in nature, when the nominal X incorporates to be to result in have, the nominal
feature is lost in the process and hence have does not host participles that form reduced
relatives. On the other hand, if the participle merely raises to X, the resulting complex
undergoes declension for nominals, which thus leaves it possible to function as a reduced
relative.
8
Participial Relatives and Ergativity Er-Xin Lee
With respect to research on Hindi-Urdu (HU) reduced relatives, Mahajan (1994,
1995) also follows Kayne’s (1993) proposal that have is derived from an empty
preposition and is incorporated into be. It is further stated that in HU the empty position
is actually a postposition, the Ergative marker, which does not incorporate into be. The
main argument against the incorporation is based on HU having its auxiliary position
head-finally, thus prohibiting the auxiliary from being adjacent to the subject and
inhibiting incorporation of the postposition to be. Therefore, be remains as the auxiliary
in HU and the overt Ergative marker becomes the subject postposition.
1.2. Reduced relatives in Hindi-UrduParticiples can also be used in HU to form reduced relatives, which are always
prenominal. The ability to form reduced relatives for Imperfective as well as Perfective
participles, as shown in (8), appears to be related to the auxiliary used in HU. The clausal
counterparts of the Reduced relatives in (8) appear as complements of be, shown in (9).
(8) a. Perfective participialmẼ-ne [vo [pi:la: paR gaya:] phu:l] utha: liya:I-Erg DEM yellow ‘become’ GO-Pfv flower lift TAKE-Pfv‘I picked up the flower that had become yellow.’ (from Kachru 1973)
b. Imperfective participial[[chal-ti:] ga:ri:]-se mat utromove-Impfv.f vehicle.f-from Neg descend-Imp‘Do not descend from the moving vehicle.’ (Hook 1979)
(9) a.phu:l pi:la: paR gaya: tha:flower yellow fall GO-Pfv be.Pst.Sg‘The flower had become yellow.’
b. ga:ri: chal-ti: thi:vehicle.f move-Hab/Impfv.f be.Pst.f‘The vehicle used to move.’
Given these data, the have-be generalization seems to make correct predictions for
HU. HU is a be-language – there is no auxiliary that corresponds to have. According to
9
Participial Relatives and Ergativity Er-Xin Lee
the have-be generalization, it is to be expected that HU is able to form reduced relatives
on Accusative verbs, thereby giving rise to reduced relatives that are impossible in
English. Indeed, the prediction is borne out as the auxiliary that appears with
Unaccusatives in the Perfect in HU is be (in contrast to have in English).
(10) a. parcel mel-se aa-ye theparcel2 mail-by come-Pfv.Pl be.Pst.Pl‘The parcels had come by mail.’
b. [[mel-se aa-ye] parcel] bhaarii hẼmail-by come-Pfv.Pl parcel heavy.f be.Prs.Pl‘The parcels that had come by mail are heavy.’
In the case of the Perfect, the auxiliary that appears in HU is be, hence the ability
to form reduced relatives from Perfect participles is again expected, as shown in (11).
(11) a. The auxiliary for the Perfect is be:kal tiin peR kaat-e ga-ye theyesterday three trees.M cut-Pfv.MPl Pass-Pfv.MPl be.Pst.Mpl‘Three trees had been cut yesterday.’
b. [[Avi-dwaaraa kal kaat-e ga-ye] peR] neem-ke the Avi-by yesterday cut-Pfv.Pl Pass-Pfv.Pl trees Neem-Gen.Pl be.Pst.Pl‘The trees which had been cut by Avi yesterday were Neem trees.’
English, on the other hand, involves have as the Perfect auxiliary, thus [the trees have
been cut by Avi] is acceptable whereas [the trees are been cut by Avi] is not. Thus, * [the
trees [been cut by Avi]] follows from *[the trees are been cut by Avi].
2 Unless indicated overtly by the Ergative case suffix –ne for subjects of Transitive Perfective sentences, the case on all other Subject NPs in HU sentences is Nominative. The Nominative case is not realized by overt case marking.
10
Participial Relatives and Ergativity Er-Xin Lee
1.3. Inadequacies of the have-be generalizationThe have-be generalization seems to be useful in accounting for HU reduced
relatives formed from Unaccusative and Perfect participles. However, the generalization
breaks down in the case of Transitive constructions in which be appears as the auxiliary.
(12) a. *The student [read a book] is new to the school.b. *The student is read a book.
In English, the Past participle of read categorically occurs with have, not be, hence (12b)
is unacceptable. Since have is not expected by the have-be generalization to allow the
corresponding Past participle to form a reduced relative, the participle of a Transitive
verb is therefore unacceptable. The unacceptability of (12a) therefore appears to follow
from that of (12b).
(13) a. Natasha-ne kitaab paRh-ii thii Natasha-Erg book.f read-Pfv.f be.Pst.f ‘Natasha had read the book.’b. *[[kitaab paRh-ii] laRkii aa-yii hai
book read-Pfv.f girl come-Pfv.f be.Prs‘*The girl [read the book] has come.’
In HU, the Past participle of read takes be as an auxiliary (13a), which should be possible
to form a reduced relative according to the have-be generalization. However, a reduced
relative involving the Past participle is not possible. The unacceptability of (13b) is
unexpected, and reveals a critical flaw with the have-be generalization: in spite of be
being the auxiliary in the HU Transitive construction, a reduced relative formed from the
participle is still impossible. Barring the fact that a construction such as (10) involves an
Intransitive Unaccusative verb whereas (13) involves a Transitive verb, the have-be
generalization should predict identical acceptabilities for (10b) and (13b), since its basic
premise makes no mention of differences in the transitivity of the verbs. The fact that
there is a syncretism in the generalization’s prediction demonstrates that it is not a
11
Participial Relatives and Ergativity Er-Xin Lee
sufficient generalization for the construction of reduced relatives. Hence, is there more to
be said about the use of auxiliary be as the decisive factor as to whether reduced relatives
can be formed? Or should an appeal to the differences between Transitive and
Unaccusative constructions be made? In regards to the transitivity of the verbs, which is
very much connected to the type of theta-roles that the verbs can assign to arguments,
should the ability of participles to form reduced relatives therefore be dependent on the
types of arguments that are subcategorized or licensed? We will return to these questions
in Section 3.
The above questions are vital because HU Transitive constructions are not the
only exceptions to the generalization. Marvin (2000) also shows that the participle of a
Transitive verb does not result in an acceptable reduced relative in Slovenian, as in (14a),
even though Slovenian is strictly a be-language. On the other hand, (14b) shows that a
reduced relative is possible if the participle comes from an Unaccusative verb.
(14) a. *Videl sem žensko, napisalo knjigo seen am woman.Acc.FSg written.Acc.FSg book ‘I saw the woman who wrote the book.’
b. Videl sem žensko, prispelo danes zjutraj seen am woman.Acc.FSg arrived.Acc.FSg today morning ‘I saw the woman who arrived this morning.’
Hence, we find evidence not just from HU but also from a non-Indo-Aryan
language that the presence of be as the auxiliary does not guarantee the acceptability of a
reduced relative formed from the participle that is a complement to the auxiliary.
The have-be generalization is also insufficient in that it does not apply to Passive
constructions in HU. HU being a be-language, is expected to consist of be as the Passive
auxiliary, just as in Bulgarian. The ability of the Passive participles in HU to form
reduced relatives would therefore be predicted by the have-be generalization to be in the
12
Participial Relatives and Ergativity Er-Xin Lee
similar to English. In English, the acceptability of “the apple [eaten by the student]”
seems to be related to the Passive auxiliary in English being be. As it turns out, the
auxiliary used for HU Passives appears to be more complicated. In HU and many other
Indo-Aryan languages, there is an auxiliary that is a root related to the lexical verb go
which contributes to the Passive semantics of the Passive participle (15a). There is also a
regular be auxiliary that indicates the Perfect of the Passive. The have-be generalization
does not make any stipulation for other auxiliary verbs apart from be. In the case of
reduced relatives formed from HU Passives, the have-be generalization should not be
capable of applying to the participle as it is a complement of go and not of be. There is
the possibility that the reduced relative construction in (15b) is licensed by the Perfect
auxiliary be (which has been demonstrated in (11)), but it would have to be assumed that
the Perfect auxiliary and the Passive auxiliary are both direct complements of the
participle. If this were indeed the case, then it is a curious phenomenon that (i) a unique
auxiliary for the Passive occurs in HU and yet does not govern the Passive participle
syntactically, and (ii) the Passive auxiliary has to co-occur with the Perfect auxiliary.
Given the limitations of the current study, and for lack of more in-depth descriptions of
the Passive auxiliary in HU grammars, we can only speculate the status of go at best.
(15)a. Avi-dwaaraa kal do peR kaat-e ga-ye the Avi-by yesterday two trees cut-Pfv.Pl GO-Pass.Pl be.Pst.Pl ‘Two trees had been cut by Avi yesterday.’
b. [[Avi-dwaaraa kal kaat-e] peR] neem-ke the Avi-by yesterday cut-Pfv.Pl tree Neem-Gen.Pl be.Pst.Pl ‘The trees cut by Avi yesterday were Neem trees.’
The discussion in this section has raised a slight anomaly with the have-be
generalization with respect to HU reduced relatives formed from Passive participles
whose auxiliary is not be. As we later go on to discuss the relationship between HU
13
Participial Relatives and Ergativity Er-Xin Lee
participles and their resultant reduced relatives in terms of the verbs’ abilities to license
External Arguments, the use of go as the auxiliary associated with the Passive does not
seem to be a significant factor in determining whether a reduced relative can be formed
from the Passive participle in HU. What is thought provoking about the discussion in this
section has been that active Past participles of Transitives in HU cannot function as
reduced relatives even though they can be complements of be. Given the differences in
English, Italian and Bulgarian reduced relatives on the one hand, and HU and Slovenian
reduced relatives on the other, it is apparent that the have-be generalization does not
provide an adequate account for languages such as those of the latter group.
However, we do still acknowledge that the have-be generalization is able to
account for English data quite sufficiently, particularly where the External Argument
account that we propose in this paper does not prove to be feasible in explaining the
unacceptability of reduced relatives formed from participles of Unaccusative Intransitive
verbs.
(16) a. *The people very recently come to the party…b. *The people been arrested…
(17) a. *The people are very recently come to the party.b. *The people are been arrested.
It can be argued that for some English dialects noun phrases such as (16a) which include
an adverb modifier as part of the reduced relative may not be absolutely unacceptable
(that is, have a ? rather than a * reading) (cf. Pesetsky 1995). Even so, it is unlikely that
the acceptability of the Passive Perfect participle in (16b) would be improved with an
adverb modifier. The ungrammaticality of (16a,b) follows directly from (17a,b). As we
present and develop on the External Argument account in Section 3 of this paper, it
14
Participial Relatives and Ergativity Er-Xin Lee
should become clear that an analysis based on External Arguments would tend to be able
to provide a more complete account of reduced relatives not just of English or HU, but of
reduced relatives in general. We use the term External Argument here in accordance to its
customary sense in the literature wherein Transitive and Unergative Intransitives like
laugh have External Arguments, while Unaccusative Intransitives do not. However, we
concede that Unaccusatives in English pose a problem for the External Argument
analysis. Hence, for English, the have-be generalization cannot be done away with.
2. The Ergative-Nominative Generalization2.1. A new generalization
So far, we have identified that the have-be generalization characterizes the
suitability of a predicate to appear as a reduced relative. A predicate is hence suitable if it
can appear as the complement of be. The main reason that the have-be distinction has
been highlighted as a possible link to reduced relative formation is because it was first
found to be pervasive in some Indo-European languages. HU, an Indo-European
language from the Indo Aryan sub-family, is a be-language but HU reduced relatives are
not fully accounted for by the have-be generalization in the way that English reduced
relatives are.
HU is a split Ergative language, in which Ergative case-marking appears on
subjects of transitive sentences, provided that the subjects bear person features and the
sentences are in the (Past) Perfective aspect (Mahajan 1994, 1995, Masica 1991). Hence,
the Ergative case occurs only where there is a Perfective participle, as exemplified by
(18a, from Mahajan 1995). When the participle is not in the Perfective aspect, subjects in
HU either take a Dative suffix in the context of a psych verb or a Genitive or Locative
15
Participial Relatives and Ergativity Er-Xin Lee
suffix in possessive constructions, as reflected in (18b and c). Alternatively, the subject
could simply be Nominative, as shown in (18d).
(18) a. Raam-ne vah kitaabe parii thii Ram-Erg those book-Pl. read-Pfv.f. be.Pst.f.Pl. ‘Ram had read those books.’
b. Raam-ko sar dard hẼ Ram-Dat head ache.f. be.Prs.f. ‘Ram has a headache.’
c. Raam-kii do bEhne hẼ Ram-Gen two sister.Pl. be.Prs.f.Pl. ‘Ram has two sisters.’
d. laRkii kitaab paRh-tii haigirl book read-Hab.f be.Prs‘The girl reads the book.’
In summary, the distribution of Ergative case-marking in HU is as follows:
(19) The distribution of Ergativity in HU: the subjects of Transitive verbs in Perfective aspect receive Ergative case.
The difference between HU and the other Indo-European languages that follow
the have-be generalization, as pointed out by Mahajan (1995), stems from the fact that
HU is an SOV language, which is a type of verb peripheral language (VSO languages are
also verb peripheral but not SVO languages), whereas languages like English are verb-
medial languages. A characteristic of verb peripheral languages such as HU is that they
generally do not consist of the lexical verb have, much less use have as an auxiliary verb.
Another unique characteristic is that they tend to display Ergative case-marking, whereas
verb medial languages like English display only Nominative case-marking on subjects of
Transitive predicates. Mahajan relates the two characteristics by positing a version of the
have-be generalization, outlined above in Section 1.1.4. Specifically, Mahajan states that
the source from which have is derived is inherently the incorporation of be and an empty
preposition (or in the case of HU, a postposition). Be-languages like HU lack have
because the Ergative case marker –ne, which is regarded as a postposition by Mahajan
16
Participial Relatives and Ergativity Er-Xin Lee
and McGregor (1995) and is a sister to the subject NP, does not become incorporated
with be, whereas languages that use have as an auxiliary verb have an empty preposition
that combines with be to derive have. Hence, the presence of have in a language
necessarily rules out the presence of an overt pre- or post-position morpheme (the
Ergative case in HU); conversely the presence of the Ergative case marker rules out the
use of have as an auxiliary.
Languages like English whose reduced relatives can be predicted by the have-be
generalization rule out participles that appear with have as the auxiliary. On the other
hand, be-languages such as HU, which do not make use of have as an auxiliary and yet
only allow for reduced relatives in certain constructions but not others, should be ruled
out by Ergative case-marking on subjects, that is, assuming that Mahajan’s (1994, 1995)
account of the interplay between the auxiliary have and the Ergative marker in HU holds.
There seems to be a parallel between the have-be alternation of the have-languages and
the choice between Ergative and Nominative in an Ergative language such as HU. That
is, participles that occur with have would be predicted to not form reduced relatives in the
have-languages, and participles that occur in the same clause as subjects bearing Ergative
case would be predicted to not form reduced relatives in HU.
Hence, instead of attributing the formation of reduced relatives from participles to
the presence or absence of the auxiliary be, we could determine from the distribution of
Ergativity in HU, that is, where subjects are Ergative in the Perfective aspect, we posit
that case on the predicate’s subjects plays a role in the actual distribution of reduced
relatives. The predictability of reduced relative formation in HU follows the
generalization in (20).
17
Participial Relatives and Ergativity Er-Xin Lee
(20)The Ergative-Nominative Generalization (for HU): Only participles whose subjects receive Nominative case can function as reduced relatives.
Let us now revisit the HU constructions from Section 1.3 that are problematic for
the have-be generalization. In the case of the Transitive past participle, in (13) and
repeated below in (21), the have-be generalization could not predict why (21a) would be
unacceptable since the auxiliary is be and not have. The Ergative-Nominative
generalization, on the other hand, is able to because the subject of a Transitive past
participle receives Ergative (and not Nominative) case.
(21) a. *[[kitaab paRh-ii]] laRkii aa-yii hai book read-Pfv.f girl come-Pfv.f be.Prs ‘*The girl [read the book] has come.’b. Natasha-ne kitaab paRh-ii thii
Natasha-Erg book.f read-Pfv.f be.Pst.f ‘Natasha had read the book.’
The second problem with the have-be generalization was that it incorrectly
predicted that the Passive participle could not be used to form a reduced relative because
the Passive auxiliary in HU is go rather than be. The acceptability of the reduced relative
in (22) follows from the fact that the subject of a Passive participle receives Nominative
case, which falls within the prediction of the Ergative-Nominative generalization. The
use of go as the Passive auxiliary is simply not a problem for the new generalization.
(22) Passivesa. Avi-dwaaraa kal do peR kaat-e ga-ye the
Avi-by yesterday two trees cut-Pfv.Pl GO-Pfv.Pl be.Pst.Pl‘Two trees had been cut by Avi yesterday.’b. [[Avi-dwaaraa kal kaat-e ga-ye] peR] neem-ke the Avi-by yesterday cut-Pfv.Pl GO-pfv tree Neem-Gen.Pl be.Pst.Pl‘The trees cut by Avi yesterday were Neem trees.’
The Ergative-Nominative generalization demonstrates that it too can account for
HU reduced relative constructions that do support the have-be generalization. In (23a),
the subject of the Unaccusative predicate, ‘parcels,’ bears covert Nominative case. A
18
Participial Relatives and Ergativity Er-Xin Lee
reduced relative, in (23b), can also be formed. The Passive and Imperfective or Habitual
participles in (24) and (25) respectively can also form reduced relatives. The cases on the
subjects, ‘Avi’ and ‘laRkii,’ are both Nominative, hence the fact that the participles can
function as reduced relatives is predicted.
(23) Unaccusativesa. parcel mel-se aa-ye the
parcel mail-by come-Pfv.Pl be.Pst.Pl‘The parcels had come by mail.’
b. [[mel-se aa-ye] parcel] bhaarii hẼ mail-by come-Pfv.Pl parcel heavy.f be.Prs.Pl‘The parcels that had come by mail are heavy.’
(24) Passivesa. Avi-dwaaraa kal do peR kaat-e ga-ye the
Avi-by yesterday two trees cut-Pfv.Pl GO-Pfv.Pl be.Pst.Pl‘Two trees had been cut by Avi yesterday.’
b. [[Avi-dwaaraa kal kaat-e ga-ye] peR] neem-ke the Avi-by yesterday cut-Pfv.Pl GO-pfv tree Neem-Gen.Pl be.Pst.Pl‘The trees which had been cut by Avi yesterday were Neem trees.’
(25) Imperfectives/Habitual Participlesa. laRkii kitaab paRh-tii hai
girl book read-Hab.f be.Prs‘The girl reads the book.’
b. [[kitaab paRh-tii] laRkii] book read-Impfv.f girl‘The girl [reading the book]…’
So far, we have shown that the Ergative-Nominative generalization, is also
capable of making the right predictions about the ability of the participles to form
reduced relatives. However, in the next section, a weakness of the Ergative-Nominative
generalization is revealed as we show that the Ergative-Nominative generalization is too
strong: it overgeneralizes that all Nominative subjects in HU can be relativized as
subjects of reduced relatives. The Ergative-Nominative generalization (as we have it up
to this point) does not state anything about Ergative case; it suggests that since subjects
never bear Nominative case in HU Perfective verbs, and given that the distribution of
19
Participial Relatives and Ergativity Er-Xin Lee
Ergativity occurs in conjunction with Transitive Perfective verbs, that therefore reduced
relatives cannot be formed when the subject is Ergative.
2.2. Problem with the Ergative-Nominative GeneralizationThe Ergative-Nominative generalization seems to be able to account for reduced
relatives formed from Imperfective (or Habitual), Unaccusative Intransitive and Passive
participles. In this section, we turn to another class of Intransitive verbs in HU, the
Unergatives and show that the Ergative-Nominative generalization is unable to make the
right predictions. In HU Unaccusative and Unergative constructions, shown in (26)
below, the past participles license Nominative subjects. While the participles of
Unaccusative verbs form reduce relatives, as in (27a) below, those of Unergative verbs
do not, as in (27b). As stated in (20), only participles whose subjects receive Nominative
case can function as reduced relatives. According to the conditions of the Ergative-
Nominative generalization, since the subjects in (26) are Nominative, both of the
participles should be expected to undergo reduced relative formation. Thus, just as the
subject, kameez ‘shirt’, in the Unaccusative construction in (26a) is relativized as phat-ii
kameezein ‘torn shirts’in (27a), the Ergative-Nominative generalization should also allow
for the subject in (26b), kuttaa ‘dog’, to be relativized. However, the fact that bhOnk-aa
kuttaa ‘barked dog’ in (27b) does not fit into a similar type of construction as in (27a)
reveals a flaw in the Ergative-Nominative generalization: it appears that not all subjects
with Nominative case can appear with the participles in reduced relatives. Hence, in this
section, we will show that Unergative verbs differ from Unaccusatives in that subjects of
Unergatives can in fact be interpreted as having similar properties as Ergative subjects.
20
Participial Relatives and Ergativity Er-Xin Lee
(26) a. Unaccusative Intransitivekameez phat ga-yii shirt.f tearintr GO-Pfv.f‘The shirt tore.’
b. Unergative Intransitive Kuttaa bhOnk-aa dog.M bark-Pfv.MSg‘The dog barked.’
(27) a. Unaccusative Intransitive mujhe [phat-ii kameezein] pasand hẼ
I-Dat tearintr shirt.FPl like be.Prs‘I like torn shirts.’
b. Unergative Intransitive *mujhe [bhOnk-aa kuttaa] pasand hai I-Dat bark-Pfv.MSg dog.MSg like be.Prs ‘*I liked the barked dog.’
The Ergative-Nominative generalization, when it works, works because it
captures an important aspect of the argument structure of the predicates involved: only
predicates which have External Arguments allow for Ergative subjects. Unaccusative
Intransitives do not license External Arguments and therefore the subjects can never
receive Ergative case. On the other hand, Unergative Intransitives do have External
Arguments and are thus capable of receiving Ergative case. In (28) the presence of an
Ergative subject in HU is not entirely unacceptable, whereas (29) reflects the
improbability of an Ergative subject showing up in an Unaccusative construction.
(28) ?kuttõ-ne bhOnk-aa dogs-Erg bark-Pfv‘The dogs barked.’
(29) *kameez-ne phat-aa shirt-Erg tear-Pfv‘The shirt tore.’
The flip side to the Ergative-Nominative generalization lies in the fact that when
it fails, it fails in those environments where the predicates sanction External Arguments.
In HU, we have noted that subjects of Unergative constructions are candidates for
21
Participial Relatives and Ergativity Er-Xin Lee
Ergative marking. What seems to be postulated therefore, is the possibility for subjects of
Unergative verbs to have either Nominative or Ergative case marking. In other words, it
appears that the type of verbs plays a part in determining if the subjects can take both
Nominative and Ergative markings. In (26b), we have noted that ‘dog’ is morphologically
Nominative, whereas in (28), ‘dog’ can be Ergative. One way of looking at this
alternation in case marking is perhaps to state that the Nominative subject in (26b)
involves the simultaneous non-overt realization of Ergative case-marking on the subject.
However, such an account proves to be problematic as it would have to also account for
why there are two different cases that are assigned to a single subject, as well as the
motivations and/or ramifications for constructions involving different morphological case
markings on the subjects (that is, when the subject appears with Nominative marking, as
in Kuttaa, as opposed to Ergative marking, as in kuttõ-ne).
Another way of analyzing the two possibilities of case marking is to assume that
certain classes of verbs have the option of assigning either Nominative or Ergative case
on subjects. Kachru and Pandharipande (1979) argue against the possibility of any
systematic semantic property that might govern when Ergative case would be used
instead of Nominative case and simply leave the optionality of Ergative marking on verbs
such as ‘cough’ and ‘sneeze’ in HU as an artifact of historical development. Mohanan
(1994), on the other hand, provides an account of the predictability of Ergative versus
Nominative marking on subjects based on a semantic property that she calls ‘Conscious
Choice:’
22
Participial Relatives and Ergativity Er-Xin Lee
(30) a. ravii davaaii pii gayaa. (*ravii-ne)Ravi medicine drink go-Pfv.‘Ravi (impulsively) drank up the medicine.’
b. ravii-ne davaaii pii daalii (*ravii)Ravi-Erg medicine drink pour-Pfv.‘Ravi (deliberately) drank up the medicine.’ (Mohanan 1994, p.74)
Other than some verbs in HU that do not seem to constitute the property of
“deliberateness,” which she labels “lexical exceptions,” Mohanan contends that subjects
of verbs can be predicted to have Ergative case when “an argument is the grammatical
subject, is associated with the meaning of ‘conscious choice,’ and the predicate is
associated with PERF aspect” (p.77). This notion of ‘conscious choice’ pertains in part to
purposeful action and in part to “the intuition underlying the term ‘agentive marker’ used
to refer to the ergative (Kachru (1980), Kachru et al (1976))” (p.73). Taking Kachru et. al
and Mohanan’s observations of the fact that depending on the type of verb, Ergative case
marking may be optionally used in subjects which would otherwise take Nominative
case, we propose that the fact that the Unergative construction in (26b) cannot form a
reduced relative is due to the subject’s ability to bear Ergative case marking.
Hence, in addressing the speculation from the previous section as to whether
Ergative subjects can be relativized as a reduced relative, and in order to provide a
consistent analysis of the data presented, we need to state more accurately that reduced
relatives can never be formed where the Transitive subject in the Perfective aspect is able
to receive Ergative case. As such, we provide a modification of the Ergative-Nominative
generalization.
(31) The Ergative-Nominative Generalization (modified):Only participles whose subjects receive Nominative case and whose subjects cannot receive Ergative case can function as reduced relatives.
23
Participial Relatives and Ergativity Er-Xin Lee
Having pointed out that Ergative case marking on subjects prohibits the
participles to form reduced relatives and that Ergative case can only be assigned to
External Arguments, we now investigate if the Ergative-Nominative generalization can
be further broken down in terms of the licensing of External Arguments. Hence, in the
next section, we turn to the role played by External Arguments in reduced relative
formation.
3. Analysis based on the role of External ArgumentsBefore going into the specifics of the External Argument analysis, the following
is a brief characterization of the term ‘External Argument.’ As mentioned earlier,
Transitive and Unergative Intransitive verbs incorporate an argument that is merged at a
higher node than the predicate. Generally, in sentences such as Kim ate the pie
(Transitive) or Ashley laughed (Unergative Intransitive), the External Arguments (Kim
and Ashley respectively) tend to also be regarded as Thematic Agents. However, it should
be noted that External Argument and Agentivity are two separate notions—the former
pertains to the position of an argument that is syntactically external to verbal predicate
while the latter constitutes the thematic or semantic feature that is borne by an argument.
Agentivity is necessary for the presence of an External Argument, whereas the presence
of an External Argument is not necessary for Agentivity. Hence, the lack of External
Argument does not immediately suggest the lack of Agentivity.
To illustrate, Passives and Unaccusatives are distinguished from each other by the
presence of Agentivity in what occurs with the former but not the latter class of verbs.
The Passive (32a) has agentive semantics in that the ship was sunk implies someone did
the actual sinking. In contrast, the Unaccusative sentence in (32b) does not have the
24
Participial Relatives and Ergativity Er-Xin Lee
implication of someone sinking the ship. Thus, whereas the Passive form of the verb ‘to
sink’ includes Agentivity, the Unaccusative form does not call for it.
(32) a. The ship was sunk.b. The ship sank.
The so-called ‘suppression’ of ‘subject NP’ in Passive constructions and the
‘promotion’ of ‘object NP’ in earlier accounts of D-S Structure Passivization can be
translated to accounts using Merge as the fact that an External Argument is not projected.
In a sentence such as (33) ‘The apple was eaten,’ the logical subject, the apple, is really
an Internal Argument that has been raised. Hence, the lack of External Argument permits
the formation of a reduced relative- the apple [eaten by John]. Unaccusatives, not
projecting External Arguments either, should therefore also be able to form reduced
relatives from their participles, but as pointed out in (16), not all English Unaccusatives
form reduced relatives. To this end, we postulate that the have-be generalization takes
over as the critical determinant as to whether English Unaccusative participles can
function as reduced relatives. On the other end of the spectrum, we have Past participles
of Transitive verbs, such as eat in ‘The boy ate the apple,’ which involves the projection
of an External Argument, the boy. Consequently, reduced relatives cannot be formed, e.g.
*the boy eaten the apple.
(33) The apple was eaten.Stage in the derivation: [vAG
3 [√eat [the apple]]]
No External Argument is merged.
Given the fact that only Nominative subjects may be relativized into reduced
relatives in HU and that the ability of a subject of a Transitive Perfective verb to receive
Nominative case is correlated with the absence of an External Argument, we can thus
3 Embick (2001) stipulates Agentivity as having an Agentive feature within v. Thus, we designate vAG as a category of verb that constitutes semantic agents.
25
Participial Relatives and Ergativity Er-Xin Lee
focus on the distribution of External Argument with respect to the Perfective aspect. In
HU, Ergative subjects never occur when the verbs are in Imperfective aspect. As a
testament that the revised Ergative-Nominative generalization is accurate, the formation
of reduced relatives from Imperfective participles is possible (cf. 25). For Imperfective
participles however, we find that the reverse from that of Perfective constructions is true
about the occurrence of External Arguments, that is, the presence of External Arguments
is actually licensed by the Imperfective participles. As such, we arrive at the following
conditions on the presence of External Arguments in HU reduced relatives.
(34) The External Argument Generalization:In a reduced relative based on aa. Perfective(/Passive) Participle: External Arguments may not be present.b. Imperfective Participle: External Arguments may be present.
The above generalization corroborates Embick’s (2001) findings that languages
whose Passives and Unaccusatives are able to form reduced relatives have a common
underlying factor— both types of participial clauses lack External Arguments. As for
regular Transitive Perfective participles, the formation of reduced relatives is also
prohibited where External Arguments are present.
To sum up, if the Passive or Perfective participle of a verb involves projection of
an External Argument, it is not possible to form a reduced relative on it. In what follows,
we will provide a structural characterization of the External Argument Generalization
proposed in (34).
26
Participial Relatives and Ergativity Er-Xin Lee
3.1. Subcategorization Requirements3.1.1. Some Assumptions about Argument Structure
Following work by Hale & Keyser (1993, 1998), we will make the following
assumptions about argument structure (35). The verbal root is assumed to come under v,
which is further stipulated as either vUNACC and vAG. With vUNACC and vAG, we essentially
distinguish the absence of Agentivity in the former and the presence of Agentivity in the
latter. The External Argument is represented by the presence of an argument to the left
and outside of v. In (35), only the Transitive construction constitutes an External
Argument.
(35) a. Unaccusative: John arrived.Structure before movement for Case takes place:[vUNACC [√arrive John]]
b. Transitive: John hit Bill.Structure before movement for Case takes place:[John [vAG [√hit Bill]]]
c. Passive: Bill was hit.Structure before movement for Case takes place:[vAG [√hit Bill]]
With respect to the licensing of the External Argument, we outline Embick’s
(2000, 2001) discussion on the status of the External Argument Property in the following
section, as well as provide a discussion on the criteria in Section 3.2.
3.1.2. The External Argument Property Embick (1997) examines participles in reduced relatives using IA&I’s have-be
generalization as a starting point, that is, he concurs with IA&I on the observation that
participles in reduced relatives are closely related to their co-occurrence with be,
specifically for English Passive participles. However, he notes that while reduced
relatives involving the Passive occur without be, as is also the case for those involving
the Perfect of a Passive (cf. (36a and b), the Imperfective or Progressive of Passive
27
Participial Relatives and Ergativity Er-Xin Lee
participles however require the presence of the progressive form of be (36c). The
conclusion drawn is that no aspectual be is projected for (36a and b) while for the sake of
retaining the Eventive state of the Progressive, be is projected as an aspectual head for
(36c) to pick up the morphology related to the Progressive state.
(36) a. The coins [discovered in the sunken ship] date back to 1863.b. The man [arrested three times in the last month]…c. The man [being arrested for drunk driving]…
In consequence, Embick proposes that the Passive may involve a Stative reading
for Perfective Passives as well as an Eventive reading for Imperfective Passives.
Following Kratzer (1993), Chomsky (1995) and Hale and Keyser (1993), Embick (2000)
states that the verbal functional head v may contain the property of Agentivity, or AG,
which is related to the ability to license and semantically interpret External Arguments.
The difference in the presence or absence of AG is in turn relevant to the
subcategorization preferences of different aspects. As to how the difference between
Stative and Eventive Passives is played out, let us then consider the subcategorization
differences between Stative and Eventive Passives (Pesetsky 1995, Embick 2000).
(37) a. Stative Passive: Resultative aspect subcategorizes for a v[-AG]b. Eventive Passive: Completive aspect subcategorizes for v[AG] .
Agentivity is incorporated into one but not both of the readings. While the Stative
Passive calls for no AG, which automatically implies no External Argument, the presence
of AG in the Eventive Passive provides a hint for the notion that someone has to still be
performing the incomplete action (of arresting, in the case of (36c)). However, in order to
convey the Passive voice in the Eventive Passive, the subcategorization still needs to
further specify that no External Argument may be licensed. Embick (2000) backs up this
28
Participial Relatives and Ergativity Er-Xin Lee
distinction between the Stative and Eventive Passive by using a Chichewa example from
Dubinsky and Simango (1996), which has separate Passive morphemes for the Stative
and Eventive.
(38) a. Nyemba zi-na-phik-idwabeans AGR-Pst-cook-PASS(Eventive)‘The beans were cooked.’Eventive reading: the beans having been cooked by someone.
b. Nyemba zi-na-phik-ikabeans AGR-Pst-cook-PASS(Stative)‘The beans were cooked.’Stative reading: the beans in the cooked state.
Before concluding that the External Argument is redundant given that AG alone
can differentiate the Passive readings, we have to consider how AG and External
Arguments work in other participles. In an Unaccusative clause such as [[the teardrops
fallen from her eyes]…], v does not comprise AG, unlike the v in Passive constructions,
but at the same time is similar to Passive v in that it is not being associated with an
External Argument. In regards to the case of a v that comprises no AG and yet has the
External Argument Property, we have stated in our general discussion that Unergatives
have External Arguments but no Agentivity.
Embick (2001) however differs in opinion, stating that Unergatives are able to
form Impersonal Passives by virtue of them comprising vAG and not because they have
External Arguments. The assumption that he makes is that Impersonal Passives are
associated with the presence of vAG. The evidence offered is from a particular class of
verbs from Alexiadou and Anagnostopoulou’s (1999) Modern Greek verb class
alternations that lack Agentivity and therefore cannot form Impersonal Passive. There are
essentially two specific classes, the first of which is a class of Intransitive verbs that show
up in the Non-Active and Transitive verbs that are Active. In contrast, there is another
29
Participial Relatives and Ergativity Er-Xin Lee
class of Intransitive, Inchoative verbs such as stegnono ‘dry’ whose active morphology
always reflects the presence of an argument in the specifier of v which does not function
as any kind of Agentive (or Causative) meaning. Nonetheless, the fact that an argument
somehow occupies the structural specifier indicates that this class of verbs somehow
accommodates External Arguments.
Thus, the fact that Agentivity and External Argument are distinct and are
mutually independent brings up the question of whether the External Argument Property
is in fact a [± external argument] feature or whether it is simply addressed somewhere
else in the syntax along with visible structural properties (Embick, 2001). Embick
assumes that if the External Argument can be encoded in a syntactically visible property
such as the specifier in the vP complement of Asp, then the absence of a specifier would
imply that no External Argument is licensed. It has to be assumed also that the vP
complement encodes all the other visible properties. Hence, Embick’s proposal is a
morphological theory in which there is no [-external argument]. The syntax of v itself is
able to determine if the presence or absence of some morphological realization of [-
external argument] is necessary.
3.1.3. Assumptions about Aspectual MorphologyFor concreteness, we will focus on English but the assumptions about the
aspectual morphology hold for HU once the relevant substitutions are made (e.g. –ing
with the Imperfective/ Habitual participial morphology, and –en with the Perfective/
Passive participial morphology in HU). The Imperfective –ing and the Perfect –en are
clearly candidates for being put under an aspectual head Aspo. The question is whether
30
Participial Relatives and Ergativity Er-Xin Lee
the Passive –en should also be put under Aspo or whether it should be put under vAG, in
the spirit of Baker, Johnson & Roberts (=BJ&R 1989).
In BJ&R’s account of Passives, Passive -en essentially starts out at Infl, which is
governed by a VP whose head is the Passive auxiliary, as well as in sisterhood
relationship to the main verb. With –en as a Passive argument originating in a -marked
position, the only likely way for it to receive a -role is through the external assignment
of a -role by the main verb. BJ&R contend that as an argument, -en is also assigned Case
by the verb, which thereby causes the internal argument to move up to subject position to
receive Case. The advantage of such an analysis is that it avoids the complications of
having to account for the downward movement of the Passive –en through multiple VP’s
(especially where there are more than one auxiliary verb involved) to merge with the
main verb (or the upward movement of the verb through multiple maximal projections to
merge with –en). The drawback to having Passive –en start out under vAG is that it
assumes a separate treatment of the Passive participial morphology from the morphology
of Perfective and Imperfective participles.
We will put the Passive –en under Aspo because it allows for a natural treatment
of the Perfect-Passive syncretism in English (and HU), that is, we can assume that the
Passive participial morphology under Aspo is merged with the main verb via secondary
predication, the same way that Perfect participial morphology is combined with the verb
root. An analysis using Aspo also allows us to make the following generalizations:
(i) By having the participial predicates as AspP’s, we are able to keep verbal reduced
relatives categorially uniformed. Thus, in (39), the participial predicates (in
31
Participial Relatives and Ergativity Er-Xin Lee
square brackets) are all AspP’s, rather than having Passive participial predicates
come under a separate category of vP.
(39) a. the boy [AspP eating the apple]…b. the men [AspP recently arrived from London]…c. the apple [AspP eaten by John]…
Other than the fact that there is categorical uniformity, an important motivation
for leaving such participial morphology (be it Passive or Perfect) external to the
vP is that there would be no need for vAG to have to license separate sites where
merge can take place for participial morphology of Transitive verbs on the one
hand and of Passive verbs on the other hand.
(ii) vP’s do not make good secondary predicates, that is, in the examples in (40),
the vP’s themselves cannot independently function as participial predicates. As
primary predicates, they need to undergo at least another level of predication in
order for there to have the right types of participles to serve as reduced relatives.
Should the Passive participial morphology be put under vAG as opposed to Aspo, it
would be difficult to give an account for why the participial morphology in what
is fundamentally a primary predicate would generate an acceptable predicate
without there being any secondary predication.
(40) a. *the boy [vP eat the apple]b. *the boy [vP recently arrive from London]c. *the apple [vP eat by John]
Following the above assumption that all participial morphology falls under Aspo,
we can then state the subcategorization requirements for the various Aspo’s.
32
Participial Relatives and Ergativity Er-Xin Lee
(41)a. Aspo= -ing subcategorizes for Transitive and Unaccusative vP’s(i.e. vP’s headed by vAG with an External Argument or vUNACC 4)
b. Aspo= -enPerfect subcategorizes for Transitive and Unaccusative vP’s(i.e. vP’s headed by vAG with an External Argument or vUNACC)
c. Aspo= -enPassive subcategorizes for Passive vP’s(i.e. vP’s headed by vAG without an External Argument)
Note that we have included the External Argument as part of the
subcategorization requirement. We now need to discuss how the licensing of the External
Argument occurs, although we will remain agnostic about exactly where the External
Argument Property (cf. Embick 2001) occurs.
3.2. The licensing of External ArgumentIn the previous section, we listed the subcategorization requirements for each type
of participial morphology under Aspo. However, subcategorization is only the first step to
determining if a participial predicate can function as a reduced relative. That is, while
subcategorization makes sure that the participial morphology is merged with the
appropriate vP so as to obtain the participle, the ability for the participial predicate that is
obtained to serve as reduced relatives has to do with whether an External Argument is
absent in the full clauses.
As we have pointed out in Section 3 (34), the cases that are problematic for
reduced relative formation in both HU and English are the structures involving Past
participles of verbs that have External Arguments. In (42a), the Transitive vAG is
subcategorized by the Perfect participial morphology. The same vAG occurs with an
External Argument. The reduced relative thus formed (42ai) abstracts over the External
Argument, the boy. However, the External Argument in (42ai), as in (42aii), is not as yet
licensed. Similarly for (42b), abstraction over the External Argument occurs, but in
4 VUNACC necessarily implies that v does not incorporate any External Argument.
33
Participial Relatives and Ergativity Er-Xin Lee
neither of the constructions in (42bi and ii) is the External Argument licensed. Therefore,
we propose that in English, the External Argument has to be licensed by the auxiliary
have. Since have does not occur in any of the constructions in (42), the External
Arguments are all not licensed and hence the constructions are unacceptable.
(42) a. Transitive Perfect Participle:(i) *The boy [eaten the apple]…(ii) * The boy λx [-en [x vAG [√eat [the apple]]]]…
b. Unergative Intransitive Perfect Participle :(i) *The boy [laughed]…(ii) *The boy λx [-en [x vAG [√laugh]]]…
The above cases can be contrasted with (43), where the reduced relative in (i) is
formed from abstracting over the Internal Argument. Because no External Argument is
involved at all, there is no need for an auxiliary to license the External Argument. Hence,
we find that the constructions in (43i and ii) are able to occur without an auxiliary. These
cases reiterate that the formation of reduced relatives from Passive or Perfective
participials hinges on the fact that there should be no External Argument that needs to be
licensed.
(43) Passive Participle(i) The apple [eaten by John]…(ii) The apple λx [-en [x vAG [√eat [the apple]]]]
As (44) and (45) indicate, the clausal counterparts of (42) are acceptable in both English
and HU. With the primary difference between (42) and (44) being the fact that have
occurs in the structures in (44), that the constructions in (44) are acceptable therefore
provides evidence that the External Argument needs to be licensed by have in English.
For HU, on the other hand, External Arguments seem to be licensed by be.
34
Participial Relatives and Ergativity Er-Xin Lee
(44) a. The boy has eaten the apple.…[[the boy]i [have [-en [ti vAG [√eat [the apple]]]]]]
b. The boy has laughed.…[[the boy]i [have [-en [ti vAG [√laugh]]]]]
(45) a. laRke-ne seb khaa-yaa haiboy-Erg apple eat-Pfv be.Prs‘The boy was (has) eaten the/a apple.’
b. laRkaa hãs-aa haiboy laugh-Pfv be.Prs‘The boy was (has) laughed.’
The conclusion that we draw is that it is have and be that license the External
Arguments for English and HU respectively. Since have is unavailable in (42) to license
the External Argument, the subcategorization conditions cannot be fully met, causing the
reduced relative constructions to break down. Along the same lines for HU, the
unacceptability of the reduced relative (in boldface, from 13b, repeated below in (46)),
has to do with the External Argument, laRkii ‘the girl,’ not having the auxiliary be to
license it.
(46) *[[kitaab paRh-ii] laRkii aa-yii hai book read-Pfv.f girl come-Pfv.f be.Prs
‘*The girl [read the book] has come.’
Thus far, the fact that the analysis using External Argument seems to appeal to the
categorial usage of have or be might appear to be reminiscent of the have-be
generalization. The difference between the External Argument analysis and the have-be
generalization is that the External Argument analysis rules in reduced relatives on the
basis that there should be no External Argument that needs to be licensed. The have-be
generalization, takes into account the auxiliary that the participial predicate actually
occurs with, and rules in reduced relatives by virtue of the fact that participles occur with
be. However, the have-be generalization, while being able to descriptively categorize
participles that can serve as reduced relatives according to the auxiliary that they occur
35
Participial Relatives and Ergativity Er-Xin Lee
with, cannot account for why participles that occur with be in HU (such as (46)) are
unable to form reduced relatives.
On the other hand, the External Argument analysis looks at the reduced relatives
themselves rather than the stage of derivation prior to reduced relative formation. To
form acceptable reduced relatives, we are interested in participial predicates that have no
External Arguments to form reduced relatives. Moreover, because the auxiliaries have or
be do not occur as a part of reduced relatives, their unavailability to license External
Arguments as such rules out the formation of reduced relatives from External Arguments.
More importantly, as the analysis is not concerned with which auxiliary the participles
occur with before they serve as reduced relatives, it can account for the acceptable cases
of reduced relatives both in English and in HU.
3.3. The continuing relevance of the have-be generalizationThe External Argument generalization in (34) as laid out through the analysis
based on licensing in Section 3.2 proves to be a powerful generalization in that it seems
to account for the facts in HU. However we have not considered Unaccustive
Intransitives in our discussion in the previous section. Unaccusatives have no External
Arguments and hence the External Argument generalization should predict that
Unaccusatives should be able to form reduced relatives. Yet the unacceptability of the
reduced relatives in (47) points to a shortcoming of the External Argument analysis. To
this end, we offer no compensatory account other than to suggest that an account of
reduced relatives in English would have to fall back on the have-be generalization, seeing
as the have-be generalization is able to rule out the reduced relatives based on the fact
their participles occur with have and not be.
36
Participial Relatives and Ergativity Er-Xin Lee
(47) a. *The people very recently come to the party…b. *The people been arrested…
4. SummaryIn this paper, we started by presenting the have-be generalization and showing its
prolific ability to predict reduced relatives formed from English participles. However, we
pointed out that the have-be generalization proves to be problematic for languages whose
participles occur with be and not have, and yet are not able to have reduced relatives
using those participles. HU being one such counterexample to the have-be generalization,
demonstrates further anomalies by having a morphological case system that is suggestive
of reduced relatives arising from an Ergative-Nominative distinction. We have shown
that this distinction, although accurate to certain extents, first of all only applies to HU
and would thus be problematic as a generalization for languages such as English which
do not possess a split Ergative system. Secondly, and what is more revealing, is the fact
that the Ergative NPs in HU are subjects of verbs in the Perfective aspect, and that are
necessarily also External Arguments.
In HU it appears that only participles whose subjects cannot be marked for
Ergative case can serve as reduced relatives, thus the Ergative-Nominative generalization
translates into, or rather, gives way to a generalization that disallows External Arguments
to be licensed where reduced relatives are to be formed using Perfective or Passive
participles. In regards to what it is in the clausal structure that actually licenses External
Arguments, we propose that it is the auxiliary used Past Perfective constructions, which
is specified by the language.
In general, we found the External Argument analysis to be able to account for
reduced relatives in HU as well as in English, with the exception of English
37
Participial Relatives and Ergativity Er-Xin Lee
Unaccusatives. Therefore, we revert to the have-be generalization as it captures a broader
range of acceptable reduced relatives in English than the External Argument
generalization. Unlike what the External Argument analysis predicts, the have-be
generalization correctly rules out Unaccusative Past participle reduced relatives in
English. The have-be generalization therefore tends to override the application of the
External Argument generalization. It is when the have-be generalization does not apply to
be-languages such as HU that the role of the External Argument generalization might be
more significant.
38
Participial Relatives and Ergativity Er-Xin Lee
Appendix IWhiz-deletion
Much of the discussion in this paper focuses on an XP as a complement of be for
the formation of reduced relatives. This concept stems from a transformational approach
to reduced relatives known as Whiz-deletion, which posits that reduced relatives are
‘reduced’ clauses as a result of the deletion of a sequence involving a wh-phrase and
some (matrix or auxiliary) form of be. Although no longer in currency, the notion of
Whiz-deletion nonetheless lends the ‘reduced’ concept to current discussions of reduced
relatives. Whiz-deletion, in essence, permits participles resulting from the deletion of a
[wh+be] and not a [wh+have] sequence to surface as participles in a reduced relative.
Undoubtedly, we see that the problems raised for a have-be analysis in the present
discussion also pertain to Whiz-deletion, being that the presence of be is strongly argued
in both accounts to be pivotal to the resultant participial relatives.
However, although the have-be and Whiz-deletion analyses may stem from
categorizing participles that occur with be together with the ability to form reduced
relatives, there are fundamental differences between the two approaches- differences that
encroach into the debate of deep- and surface-structures. I present below criticisms
directed at Whiz-deletion. These criticisms do not apply to the have-be analysis by virtue
of the fact that the have-be approach sidesteps the issue of the surface deletion of a wh-
element. While the have-be generalization does not get tangled in the question of whether
a wh-phrase occurs underlyingly, Whiz-deletion tends to be too specific on this count in
that it assumes not only the underlying occurrence of wh-phrases in reduced relatives, but
that all ‘reduced’ clauses have to be formed from the deletion of the wh-element and the
proper accompanying auxiliary.
39
Participial Relatives and Ergativity Er-Xin Lee
Hence, Whiz-deletion is disfavored and has been argued against by Huddleston
(1971), Hudson (1973) and Williams (1975). The main criticism against Whiz-deletion
concerns the question of whether the reduced clauses that this analysis obtains are as
straightforward as the simple deletion of a [wh+be] sequence. For instance, Hudson
(1973) notes that English reduced relative constructions with past participles, such as
(48a), can be identified underlyingly with the corresponding past tense, as in (48b), as
well as with a construction in the Perfect aspect, as in (48c). The reduced form in (48a)
can also correspond underlyingly with (48d), yet the rules for reduced relative formation
at that time fail to account for how (48a) might be derived from (48d) since there is no
wh-element involved in (48d). Therefore, Hudson raised a point that if deletion were to
be considered as the crucial process in forming reduced relatives, then it is necessary to at
least also explore the deletion of sequences of auxiliaries such as having been, and not
just the deletion of [wh+be].
(48) a. Books published before the nineteenth century are very expensive to buy.
b. Books that were published...c. Books that have been published...d. Books having been published...
Burzio’s (1986) Italian data (from Section 1.1.3, reproduced below as (49)) also
provide a counterexample to the deletion of [wh+be]. In (49a), where the relative clause
involving the reflexive nominal element si potentially encodes two distinct readings (i)
and (ii), the straightforward assumption as to the readings that would result from the
Whiz-deletion would be that both (i) and (ii) would be expected in the reduced clause. Yet
the only acceptable reduced relative reading is (ii), as shown in (49b). The criticism is
40
Participial Relatives and Ergativity Er-Xin Lee
therefore that Whiz-deletion is not strong enough to account for the lack of one of the
readings from the full relative clause.
(49) a. Gli individui che [si erano presentati] al direttore…The individuals that self be.Pfv. introduced to the director‘The individuals that (i) self had introduced/ (ii) had introduced themselves to the director...’
b. Gli individui [presentatisi] al direttore…The individuals self-introduced to the director‘The individuals (ii) self-introduced to the director…’ (Burzio 1986)
Another problematic sign is that certain clauses with the wh-phrases do not
correspond in meaning to clauses which are so-called ‘reduced.’ The clauses in (50, from
Williams 1975) illustrate that the Imperfective participles in the first clause of each pair
of clauses appear to stand autonomously in meaning in relation to the second clause,
which resemble full relative clauses. It should be noted that Tense does not play a part in
the ungrammaticality of the second clause in each pair.
(50) a. The first man knowing all the answers will get the prize.b. *The first man who is knowing all the answers…
c. A woman resembling my mother…d. *A woman who was resembling my mother...
e. The company owning the building…f. *The company which is owning the building…
Using other examples involving Imperfective participles, Williams demonstrates
that clauses bearing the wh-phrase and be are differentiated from ‘reduced’ clauses in that
they are able to be extraposed, to host sentential subjects and adverbs and to act as result
clauses, while the ‘reduced’ forms are unable to exhibit the same functions. Thus, the
implication is that the connection between clauses involving [wh+be] and reduced
relative clauses is not as direct as the simple subtraction of the [wh+be] sequence in the
former to produce the latter. The different semantic functions and syntactic properties of
41
Participial Relatives and Ergativity Er-Xin Lee
each type of clauses suggest that reduced relatives do not have to be derived from the full
relative clauses.
42
Participial Relatives and Ergativity Er-Xin Lee
Appendix II Potential Counterexamples
In this paper, while we have presented a generalization, the External Argument
generalization, which seems to apply to reduced relatives of HU and English (with the
exception of English Unaccusatives), and seems to work more generally than the
Ergative-Nominative generalization or the have-be generalization (which work for more
specific languages and not others), we have not been able to show that the External
Argument analysis can function as a stand-alone generalization for reduced relatives, at
least not for English. To account for the English Unaccusative data, we have proposed
that the have-be generalization works in concert with the External Argument
generalization, that is, the have-be generalization will be used where the External
Argument fails to apply completely in a language, such as in English, where the External
Argument generalization works for the most part except for Unaccusatives.
Bulgarian takes be as an auxiliary:
(51) a. Maria e pisala knigataMaria is write-Perf.Prt.FSg the-book‘Maria has written the book.’
b. Zapoznah se sas zena-ta napisala knigatamet.1Sg Refl with woman-the write-Perf.Prt.FSg book-the‘I met the woman who has written the book.’(Lit: I met the woman written the book)
However, evidence from Bulgarian may prove to be problematic as (i) the
External Argument generalization does not seem to apply at all, and (ii) the actual make-
up of the participles may be ambiguous. The reduced relative thus formed diverges from
the prediction of the External Argument generalization since an External Argument is
actually licensed by the Past/Perfective participle of a Transitive verb. Thus, like the
unacceptability of English reduced relatives formed from Unaccusatives that cannot be
accounted for by the External Argument analysis, we will have to assume that the ability
43
Participial Relatives and Ergativity Er-Xin Lee
for Bulgarian to form reduced relatives is predicted by the have-be generalization, since
the acceptability of (51b) can indeed be verified by the fact that the participle occurring
with be in (51a).
Unfortunately, this counterexample occurring in Bulgarian gives further evidence
that the External Argument generalization cannot be a stand-alone generalization for
reduced relatives. At this point, we are unable to account for why the External Argument
generalization fails to account for the Bulgarian data. However, one clue that might point
to the exceptional characteristic of Bulgarian participles used in reduced relatives might
be that there appears to be two different forms of the Past/Perfective participle. We are
not able to tell if the difference in form of the participle in (51a) pisala ‘written’ and the
other participle in (51b) napisala ‘written’ is morphological, with both participles but still
signaling the same participial function, or if they are two separate participles performing
different functions of the Perfect and having different subcategorization requirements.
Consider also the Passive:(52) a. knigata napisana ot ženata
book-the write-Pass.Prt by woman-the‘The book written by the woman.’
b. knigata beše napisana ot ženatabook-the was write-Pass.Prt by woman-the‘The book was written by the woman.’ (Roumyana
Pancheva)
The Passive participle does not take the same form as the Past participle, which
may be suggestive of participles that may function entirely autonomously of the
auxiliary, in which case the have-be generalization may not even come into play.
Alternatively, if we were to pursue the line of argument from the External Argument
generalization, we would have to posit different subcategorization requirements for the
participles, which may be an uneconomical move. Either way, these possibilities are
44
Participial Relatives and Ergativity Er-Xin Lee
merely speculative. The problem with Bulgarian participles definitely needs more
attention than what the discussion afforded in this paper.
The other potential counterexample is Modern Greek. There is a Perfect
participle, which can appear in both Active and Passive forms, that shows up as the
complement of have (c.f. (53)).
(53) a. * H Maria exi xtisi to spiti to 1963 The Mary has built-Act the house the 1963 ‘Mary has built the house in 1963.’
b. To spiti exi xtisi to 1963The house has built in 1963‘The house has been built in 1963.’ (From IA&I 2000)
Embick (1997) deems the auxiliary have atypical, presumably because even
though have occurs with participles, reduced relatives can nonetheless be formed.
However, there is an anomaly in the reduced relatives in that Passive participles used in
reduced relatives in Modern Greek (54c) appear to be different in form than the
participles in regular Passive clauses (54a, b).
(54) a. …pu skediasteke which design-Perf-3Sg.‘which was designed’
b. …pu exi skediastei which has/had design-Perf-3Sg.‘which has/had been designed’
c. to skediasmeno ktiriothe design-Part. building‘The designed building’ (From Mackridge 19855)
Whenever the verbal root is not marked for Tense, it becomes the so-called
Passive participle, or the Passive Past participle, which is synthetic and functions as a
5 Many thanks to Mark Brown for helping with the translation of Modern Greek examples from the Modern Greek grammars.
45
Participial Relatives and Ergativity Er-Xin Lee
verbal adjective. It “indicates action or state that is past or an action completed before the
action indicated by the principal verb.” (Eleftheriades 1985, p.326).
(55) a. exo grapsei to grammahave-1Sg. written-3Sg. the letter-Acc.Sg.‘I have written the letter.’
b. exo [grammeno to gramma]have-1Sg. written-3Sg.Neut.Acc the letter-Acc.Sg.‘I have the letter written.’
Being a synthetic Passive participle, we do not know if participles ending in –
meno(s) do in fact occur in constructions involving any auxiliary at all. We are also
unable to determine if it functions same way, that is, have similar subcategorization
procedures as the Passive participles that were discussed in this paper.
46
Participial Relatives and Ergativity Er-Xin Lee
References
Alexiadou, A., and E. Anagnostopoulou. (1999) “Non-active Morphology and the Direction of Transitivity Alternations,” in Proceedings of NELS 29.
Anagnostopoulou, E., S. Iatridou, and R. Izvorski. (1998) “On the morpho-syntax of the Perfect and how it relates to its meaning,” in P.N. Tamanji and K. Kusumoto, eds., Proceedings of NELS 28, Amherst, Massachusetts, GLSA, 1-17.
Baker, M., K. Johnson, and I. Roberts. (1989) “Passive Arguments Raised,” Linguistic Inquiry 20:2, 219-252.
Burzio, L. (1981) Intransitive Verbs and Italian Auxiliaries, Ph.D. dissertation, MIT.
Burzio, L. (1986) Italian Syntax. A Government-Binding Approach. Dordrecht: Kluwer.
Chomsky, N. (1995) The Minimalist Program, MIT Press, Cambridge, MA.
Cowper, E. (1989a) “Perfective –en IS Passive –en.” In Jane Fee and Katherine Hunt, eds., Proceedings of the Eighth West Coast Conference on Formal Linguistics, 85-93. Stanford Linguistics Association: Stanford.
Cowper, E. (1989b) “Thematic Underspecification: The Case of have[Italics].” Toronto Working Papers in Linguistics 10:85-94.
Dubinsky, S., and S.R. Simango. (1996) “Passive and Stative in Chichewa: Evidence for Modular Distinctions in Grammar,” Language 72:4.
Eleftheriades, O. (1985) Modern Greek: A contemporary grammar. Pacific Books, Palo Alto, CA.
Embick, D. (1997) Voice and the interfaces of syntax, Doctoral dissertation, University of Pennsylvania, Philadelphia, PA.
Embick, D. (2000) Participial Structures and Participial Asymmetries, Talk presented at the University of Texas at Austin Linguistics Colloquium, March 27, 2000.
Embick, D. (2001) “UnAccusative Syntax and Verbal Alternations,” in A. Alexiadou, E. Anagnostopoulou, and M. Everaert, eds., The Unaccusativity Puzzle, Oxford Unversity Press, Oxford.
Hale, K., and S.J. Keyser. (1993) “Argument Structure,” in K. Hale and S. Keyser, eds., The view from Building 20: essays in linguistics in honor of Sylvain Bromberger, MIT Press, Cambridge, MA.
47
Participial Relatives and Ergativity Er-Xin Lee
Hale, K., and S.J. Keyser (1998) “The Basic Elements of Argument Structure,” in H. Harley, ed., The Proceedings of the MIT roundtable on argument structure, MITWPL, Cambridge, MA.
Hoekstra, T. (1984) Transitivity, Grammatical Relations in Government –Binding Theory, Dordrecht: Foris.
Hook, P.E. (1979) Hindi Structures: Intermediate Level, Michigan Papers on South and South-East Asia 16, Center for South and South-East Asian Studies, University of Michigan, Ann Arbor, Michigan.
Huddleston, R. (1971) A problem in Relative Clause Reduction, Linguistic Inquiry 2:1, 115-16.
Hudson, R.A. (1973) Tense and Time Reference in Reduced Relative Clauses, Linguistic Inquiry 4:2, 251-56.
Iatridou, S., E. Anagnostopoulou, and R. Izvorski (2000) “Some observations about the Form and Meaning of the Perfect,” in M. Kenstowicz, ed., Festschrift for Ken Hale (tentative), MIT Press, Cambridge, MA.
Kachru, Y. (1973) “Some aspects of pronominalization and relative clause construction in Hindi-Urdu,” in B.B. Kachru, ed., Papers on South Asian Linguistics, volume 3.2 of Studies in the Linguistic Sciences, Department of Linguistics, University of Illinois, Urbana, Illinois, 87-103.
Kachru, Y., B. Kachru and T. Bhatia. (1976) “The Notion ‘Subject’: A Note on Hindi-Urdu, Kashmiri and Panjabi,” in M.K. Verma, ed. (1976), 79-108.
Kachru, Y. and R. Pandharipande. (1979) “On Ergativity in Selected South Asian Languages,” South Asian Language Analysis 1:193-210.
Kachru, Y. (1980) Aspects of Hindi Grammar, Manohar Publications, New Delhi.
Kayne, R. (1993) “Towards a Modular Theory of Auxiliary Selection,” Studia Linguistica 47, 3-31.
Kayne, R. (1994) The Antisymmetry of Syntax, MIT Press, Cambridge, MA.
Kratzer, A. (1993) “Severing the External Argument from its Verb.”
Mackridge, P. (1985) The Modern Greek Language: A descriptive analysis of Standard Modern Greek. Oxford University Press, New York.
48
Participial Relatives and Ergativity Er-Xin Lee
Mahajan, A. (1994) “The Ergativity Parameter: have-be alternation, Word Order and Split Ergativity,” in Proceedings of the 13th West Coast Conference on Formal Linguistics, Palo Alto, CA, CSLI 286-301.
Mahajan, A. (1995) “Universal Grammar and the Typology of Ergative Languages,” in A. Alexiadou, T.A. Hall (eds.), John Benjamins Publications.
Masica, C.P. (1991) The Indo-Aryan Languages. Cambridge University Press.
McGregor, R.S. (1995) Outline of Hindi Grammar. Oxford University Press.
Marvin, T. (2000) “Past Participles in Reduced relatives,” unpublished manuscript, MIT.
Mohanan, R. (1994) Argument Structure in Hindi. CSLI Publications, Stanford, CA.
Pesetsky, D. (1995) Zero Syntax. MIT Press.
Roberts, I. (1986) The Representation of Implicit and Dethematized Subjects. Dordrecht: Foris.
Williams, E. (1975) “Small Clauses in English,” in J. Kimball (ed.), Syntax and Semantics, vol. 4, Academic Press, New York, 249-273.
49