CC BY
55
S. Luraghi
Transitivity, intransitivity and diathesis in Hittite
Резюме. Статья посвящена категории переходности/непереходности в хеттском языке и некоторым вопросам, касающимся ее взаимоотношений с категориями каузативности, неэргативнос-ти/неаккузативности и залога в исторической перспективе. Среди прочего, рассматриваются такие вопросы, как признаки переходности/непереходности у глаголов движения и связь между контролируемостью действия и залогом глагольных форм.
Ключевые слова: хеттский язык, глагол, переходность, непереходность, залог, медиальный залог, неаккузативность, неэргатив-ность, каузатив, антикаузатив.
1. Introduction
The aim of this paper is to explore some transitivity related phenomena in Hittite. As well known, Hittite has a variety of features that affect or are affected by transitivity, such as the distribution of nominative and accusative third person clitics, affixes used for verb derivation, the so-called reflexive particle -z(a), and verbal diathesis. Arguably some of these features are connected with degrees of transitivity or intransitivity. After surveying some relevant concepts related to (in)transitivity (sec. 2), I will especially concentrate on the rise of split intransitivity (sec. 3), and on (pairs of) verbs that display a causative vs. anticausative semantic opposition (sec. 4). I will then discuss the relevance of the findings in sec. 4 for the reconstruction of the PIE voice system (sec. 5). Section 6 contains the conclusions.
2. Transitivity, intransitivity and types of event
Transitivity is a semantic property of predicates and clauses: as such, it reflects properties of events.1 It is a scalar notion, which is defined in terms of a number of features
1 I use the word 'event' as synonym of 'state of affairs', rather as the label for one specific type of state of affairs, as some others do.
which presence/absence determine varying degrees of transitivity. Highly transitive predicates, which may be regarded as prototypical (N^ss 2007), refer to controlled states of affairs in which a volitional, and typically human, agent causes a change of state on the side of a patient. Other features of high transitivity involve telicity, realis modality, punctuality, and affirmative polarity (Hopper, Thompson 1980).
Semantic transitivity has reflexes on argument structure and on syntactic transitivity, even though such reflexes are indirect. Indeed, many verbs that indicate states, that is, predicates which are positioned at the lower edge of the transitivity scale, are typically transitive in the Indo-European languages. This is also true for Hittite, in which verbs such as 'see' and 'know' are syntactically transitive.2
Degrees of intransitivity may also have reflexes on the syntactic behavior of certain predicates. In many languages there is evidence for split intransitivity (Van Valin 1990), whereby subjects of certain intransitive verbs, called unergative, share properties of subjects of transitive verbs, while subjects of other intransitive verbs, called unaccusative, share properties of direct objects. In Hittite split intransitivity affects the distribution of clitic subjects. However, even though many middle verbs are unaccusative, middle voice in itself does not seem to be a manifestation of unaccusativity in Hittite. Split intransitivity reflects properties of participants coded as subjects only partially. It might be thought that subjects of intransitive verbs that share object properties are semantically patient-like, i.e. they are not the controller. Note however that this is not always the case: in particular, directional motion verbs such as 'come' and 'go' are typically
2 Note that these particular verbs belong to the -hi conjugation in Hittite, which is possibly connected with the PIE middle (see Rose 2006). Since in Hittite such verbs do not build a semantically motivated group and do not display semantic features of middle voice synchronically, I am not going to consider their possible origin here. However, when discussing the Indo-European origin of the Hittite middle, one should keep in mind that PIE middle verbs might be reflected either by Hittite media tantum, or by (often morphologically active) -hi verbs.
unaccusative in languages for which thorough descriptions of unaccusativity are available, such as Italian, German, and Dutch. As we will see, Hittite motion verbs provide interesting evidence in this respect.
Transitivity cross-cut parameters are relevant for lexical aspect. In particular, atelic and non-dynamic predicates (states) are on the lower edge of the transitivity scale, while activities and transitions may display different degrees of transitivity. Activities are atelic, thus not highly transitive, but, contrary to states, they can be controlled, as in (1), (2) and (3), or uncontrolled as in (4) and (5); syntactically, they are often intransitive, but they may also be transitive when the direct object is non-referential as in (1):
(1) The child eats pasta.
(2) Paul is drinking.
(3) Mary is walking.
(4) The door squeaks.
(5) The river flows.
Controlled activities are often denoted by unergative verbs in languages which display split intransitivity, while uncontrolled ones are unaccusative.
Transitions3 are telic. They may be controlled or uncontrolled and syntactically transitive or intransitive. They can further be divided into achievements and accomplishments:
(6) The snow melted.
(7) The door opens.
(8) Mary walked home.
(9) John is eating a dish of pasta.
(10) The wind broke the window.
Lexical aspect identifies event types. In some classifications of event types, the feature(s) of control and/or of causation are considered basic, as for example in Croft (1991), where events are divided into causative, as in (9) and (10), inchoative as in (6) and (8), and stative. Inchoative events typically take place spontaneously. Following this approach, one might further distinguish between causative and non-
3 See Pustejovsky (1995: 68) for this terminology.
causative (including stative and spontaneous) events. Note however that activities are left out of this classification. Indeed, Croft discusses this issue, and concludes that activities should be grouped with inchoative events. According to Croft (1991: 265), evidence for this conclusion is provided by the fact that activities may lead to a resulting state, in much the same way as inchoative events:
(11) The ice is melting / The ice is melted.
(12) Torey is dancing / Torey is all danced out.
In addition, Croft (1991: 266) also suggests that activity verbs all have an understood direct object: while this view is shared by other authors regarding those activity verbs which have a transitive counterpart, such as eat or paint, Croft extends the analysis to other verbs. However, this analysis obscures the similarity of activities and states, which lies in their being both atelic; it also shows that combining telicity and causation is not an easy task. Note further that causation does not mean control in Croft's approach:4 activity verbs may indicate controlled events, but they are treated as inchoative (i.e. non-causal) in this classification.
The parameter of causation is relevant not only for semantic, but obviously also for syntactic transitivity, as shown by cross-linguistic evidence based on the causative/ anticausative alternation (or causative/inchoative, following the terminology in Haspelmath 1993). Such alternation may be expressed through voice. Indeed, one of the functions typical of middle voice is to indicate inchoative events, as in Italian:
(13) Paolo ha chiuso la finestra.
"Paul closed the window."
(14) La finestra si e chiusa.
"The window closed."
The Italian particle si functions as a detransitivizing device in (14), which refers to a spontaneous, uncontrolled event. As we will see in section 4, middle voice is connected with this
4 A classification of states of affairs in which control is considered basic is the one originally devised in Dik's Functional Grammar, on which see Luraghi, Parodi (2008: 174-175).
type of alternation in Hittite as well, even though the contrary strategy, namely transitivization, is also widely employed.
3. Intransitive verbs in Hittite
In Hittite the distribution of subject and object third person clitics is sensitive to transitivity: while third person forms of intransitive verbs cannot occur without an overt subject, third person forms of transitive verbs can. On the other hand, omission of a referential direct object is less frequent than in the other ancient Indo-European languages. As shown in Luraghi (1990), third person subject and object clitics have a complementary distribution: subject clitics can only occur with intransitive verbs, and are virtually obligatory for third person when the subject is not a full NP or an accented pronoun; they never occur with transitive verbs, which allow zero third person subjects in case no subject NP occurs.5 In addition, transitive verbs used intransitively, i.e. unergative verbs, share the distribution of subject clitics typical of transitive verbs (see example (22) below). As argued in Garrett (1990) this can be taken as an indication of split intransitivity. Examples are the following6:
(15a) nu= kan INA KUR URUarzauwa parranda paun
CONN PTC in country A. upwards go:PRT.1SG
b) nu INA URUapasa ANA URULIM SA muhha-LU andan paun CONN in A. to city of U. into go:PRT.1SG
c) nu= mu muhha-LU-is UL mazzasta
CONN 1SG.OBL U.-NOM not resist:PRT.3SG.M/P
d) n= as= mu= kan huwais CONN 3SG.NOM 1SG.OBL PTC escape:PRT.3SG
5 The distribution of obligatory subject clitics follows the animacy hierarchy (see e.g. Comrie 1989: 189-195): 1st and 2nd person pronouns, which rank highest on the hierarchy allow null subjects with all types of verbs; third person subjects, which rank lower, can have null subjects with transitive (and unergative, see below) verbs but must have overt subjects with intransitive (unaccusative) verbs.
6 Examples are glosse following the Leipzig glossing rules; added abbreviations include: n/a = nominative/accusative neuter; PREV: preverb; PRT: preterite.
e) n= as= kan aruni parranda gursawanza pait
CONN 3SG.NOM PTC sea:D/L toward island-DlR go:PRT.3SG
f) n= as= kan apiya anda esta
CONN 3SG.NOM PTC there in be:PRT.3SG
"I went up to the country of Arzawa. In the city of Apasa I went into Uhhaziti's quarters and Uhhaziti did not make any resistance. He escaped me and went to the island and stayed there." AM 50.28-32;
(16 a) sallanun= war=an kuit ammuk
promote:PRT.1SG PTC 3SG.ACC because 1SG.NOM
b) nu= war=an huwappi Dl-esni huwappi DINGIRLIM-ni UL para
CONN PTC 3SG bad:D/L tribunal:D/L bad:D/L god:D/L NEG PREV
UL kuwapikki tarnahhun
never hadle:PRT.1SG
c) kinuna=ya= war=an karapmi now and ptc 3SG.ACC take:PRS.1SG
d) nu= war=an ANA DUTU URUTUL-na ASSUM ™SANGA-
CONN ptc 3SG.ACC to sungod Arinna for priesthood
UTTIM tittanumi
install:PRS.1SG
"Because I promoted him, I never handled him over to a bad tribunal or to a bad god; and now I will take him and make him priest for the sungoddess of Arinna." StBoT 24 iv 11-15.
The above examples contain coordinated clauses with a high degree of continuity, both for the third person subject in (15d-f) and for the third person direct object in (16a-d). In such conditions, other ancient Indo-European languages usually display coordination reduction, i.e. omission of both the subject and the direct object (see Luraghi 2004). The fact that such context does not trigger omission in Hittite demonstrates that Hittite third person clitics do not only have anaphoric function, but also have the function of indicating verbal transitivity.
Note that I am not saying that null direct objects never occur in Hittite. Indeed they do occur; what is remarkable, however, is that they are much less frequent than in the other ancient Indo-European languages. In particular, coordinated clauses which present the coordinating conjunctions et or -que never allow for repetition of direct objects in Latin, as argued
in Luraghi (1997), (1998), unless specific discourse factors intervene (i.e. if the object is emphatic, a condition which obviously cannot hold for a clitic), and apparently the same conditions hold for Greek (see Luraghi 2003). Possibly, obligatory coordination reduction of direct objects was a feature of PIE (see Luraghi 2004). Thus, the syntax of object clitics in Hittite hints at an ongoing process: null objects can still occur as an inheritance from PIE, but they are no longer obligatory in coordinated clauses, as shown in example (16). This is especially clear in (16c), where the coordinating conjunction -(y)a- occurs, which corresponds to Latin -que but, as remarked above, does not share its behavior (contrariwise, shared by other ancient Indo-European languages).
As for clitic subjects, renownedly they do not occur in other ancient Indo-European languages, and are a Hittite innovation. Garrett (1990) mentions a few occurrences of intransitive verbs with what he regards as discourse conditioned null subjects; he also notes that in most occurrences motion verbs are involved (Garrett 1996), beside a small number of stative verbs. Goedegebuure (1999) has gone a step further: by carefully analyzing all passages where intransitive verbs occur without a third person subject in Old Hittite, she has shown that they all contain motion verbs, while no change-of-state or stative verb can occur without subject clitics. Apparent exceptions are, as reported by Goedegebuure, the verb es- 'sit', which, however, occurs without a third person subject in passages in which it means 'to sit down' (and thus, must rather be taken as a motion verb), and the verb es- 'be', which she leaves out of consideration on account of the fact that "the expression of subject clitics in nominal clauses is based on different principles" (p.c.).7 In
7 Although I have not studied the distribution of subject clitics in nominal clauses, I suspect that it might be affected by negation. Passages where no subject clitics occur in Old Hittite are from negated sentences, such as natta GUD.MAH-as '(that) is not an ox' HG § 57, in which the negated predicate is non-referential. This is remarkable in the light of the fact that non-referential third person subjects of unaccusative verbs are not cliticized (that is, such occurrences have zero subjects), as shown in Garrett (1990: 106).
Middle and Late Hittite, motion verbs take subject clitics as consistently as other unaccusative verbs.
Goedegebuure (1999) further observes that the distribution of subject clitics in Old Hittite "indicates that OH represents a period in which the language is changing. Originally the language grouped transitive and motion verbs together (subject = agentive, expressed as 0), and change-of-state and stative verbs (subject = non-agentive, use of the enclitic Subject)." This is an interesting conclusion in the light of what has been remarked above in sec. 1 regarding motion verbs. Motion verbs are somewhat different from other unaccusative verbs because their subject is volitional. However, they denote a change of position undergone by their subject: displacement in space can be conceived as equivalent to a change of state. In this case, only the feature of change of position/state is focused, and not the feature of volitionality. Hence, subjects of verbs of directional motion can be treated as patients, rather than as subjects of transitive verbs.
Indeed, in languages with split intransitivity, directional motion verbs are usually unaccusative; however, the double nature of the subject (undergoes a change, but acts intentionally) makes them non-prototypical. Note further that in some languages in which motion verbs are unaccusative, there appears to be a difference between verbs that indicate directional motion vs. verbs that indicate manner-of-motion, whereby the former are unaccusative and the latter unergative. This contrast may involve different constructions of the same verb, as shown in German:
(17) Peter hat (zwei Stunden lang) geschwommen.
"Peter swam for two hours."
(18) Peter ist bis ans andere Ufer geschwommen.
"Peter swam to the other bank."
These two types of verbs are also distinct by lexical aspect: while manner-of-motion verbs, as in (18), are atelic and indicate activities, directional motion verbs, as in (17), are telic and indicate transitions.
In Hittite, all types of motion verbs offer evidence both for use or for possible omission of subjects clitics, with subject clitics being more frequent than omission with virtually all
verbs. However, the feature of telicity might have played a role in Old Hittite as well. Indeed, the only relevant exception, if one considers the data collected in Goedegebuure (1999), is constituted by the verb huwai- 'run', which occurs nine times with a null subject and never with a third person clitic.8 This contrasts with other frequently attested verbs, such as for example pai- 'go', which occurs 21 times with a subject clitic and 7 with a null subject.
Further characteristics of huwai- deserve to be mentioned in this connection. In the first place, one may observe that this verb never takes a complement in the directive case (or in the dative-locative) in Old Hittite: rather, it either occurs without a complement, or with a complement in the accusative. Remarkably, the accusative does not occur with other motion verbs in Old Hittite original texts (direction complements in the accusative are attested only starting from texts in Middle Hittite script). The meaning of the participle of huwai-provides evidence for lack of telicity: it indicates a state, rather than a result, in much the same way as the participle of other verbs that denote atelic events. In (19) the participle huyanza means 'running', in much the same way as the participle of huis- 'live' in (20), which means 'living', 'alive':
(19) LULUM-ma kuis piran huyanza esta ...
man CONN REL.NOM.SG before run:PTCP.NOM.SG be:PRT.3SG n= an= kan GIM-an kuenun
CONN 3SG.ACC PTC when kill:PRT.1SG "When I killed the man who was running before (them) (i.e. who was leading them)" StBoT 24 ii 39-40;
(20) man MugENharanan husuwandan appanzi n= an udanzi when eagle:ACC.SG live:PTCP.ACC.SG take:PRS. 3PL CONN 3SG.ACC
carry:PRS.3PL
"When a living eagle is captured, it is carried (inside)" StBoT 8 ii 19.
8 The verb salik- 'approach' also only occurs with a null subject, but only one occurrence is available, and cannot provide enough evidence for its habitual syntax.
On the contrary, the participle of directional motion verbs, which denote telic events, indicates a result:
(21) [ ] INA URUpuranda sara pan esta
in P. upwards go:PTCP be: 3SG.PRT "[The population] had gone to Puranda." AM 60.55.
Another manner-of-motion verb provides evidence for atelic aspect: the verb iya- 'walk', 'march', as observed by Goetze (1925: 73) never occurs with a direction complement, and must be taken as referring to the activity of walking (see further Neu 1968b: 87). The participle iyant- likewise indicates a state and means 'walking'. Unfortunately, Old Hittite texts offer no evidence for the use of subject clitics with this verb.
Thus, in Old Hittite most verbs of motion were still partly treated as unergative: at the time when special syntax for unaccusative verbs was being established, they were the last to acquire it. In particular, while directional motion verbs had already started acquiring unaccusativity features, the verb huwai- was consistently treated as unaccusative, and its subject shared coding properties with direct objects rather than with subjects of transitive verbs. I have remarked above that the reason why verbs of directional motion are often unaccusative across languages is that displacement (change of location) is understood as change of state, but this did not work for huwai-, which, at least at the Old Hittite stage, was a manner-of-motion verb. Unaccusative syntax extended to manner-of-motion verbs apparently only after the Old Hittite stage. The relative chronology of the extension of unaccusativity features is given in (21):
(21) states, spontaneous events > directional motion verbs > manner-of motion verbs
To sum up, one can see that unaccusativity started from uncontrolled events; it then extended to motion verbs, which indicate controlled events, starting from directional motion verbs, and only later extending to manner-of-motion verbs. This chronology partly reflects the findings from other languages, in which directional motion verbs are unaccusative while manner-of-motion verbs are unergative.
4. The causative/anticausative alternation
Transitive verbs can be detransitivized: in the Indo-European languages, this is typically a function of middle voice. Detransitivized verbs have a single argument, the subject, which corresponds to the patient of their active counterpart. Thus, they are typically unaccusative. However, middle voice does not seem to be especially connected with unaccusativity in Hittite: among Old Hittite unaccusative verbs listed in Goedegebuure (1999) just a few are middle, with the exception of stative verbs, which are by th most part media tantum. The connection between middle verbs and stativity is well known, and it does not only concern Hittite, but it was most likely a feature of PIE as well.
Beside detransitivization, transitivization is also possible, through the formation of causatives (see Comrie 1995). Causatives of stative verbs are well known from Hittite, they are formed though the suffix -nu-. Examples are: a-/e- 'be warm'/ inu- 'warm up (tr.)'; tarra- 'be powerful' / tarranu-'make powerful'. In addition, various intransitive verbs that indicate change of state have causative counterparts; many of them are also media tantum, such as war- 'burn (intr.)' / warnu 'burn (tr.)'. As remarked by Neu (1968b: 53), -nu- causatives build transitive counterparts of ancient media tantum, whose diathesis was connected with either states or change-of state verbs, that is, typically uncontrolled events. Thus, with these verbs the basic form is intransitive and indicates an uncontrolled event, while the transitive form is derived.
Haspelmath (1993) describes causative and anticausative derivation in a typological perspective. He singles out four possible situations (some languages, as English and Hittite, have more than one):9
i. causative alternation: the anticausative verb is
unmarked, the causative verb is marked: this is the case of the Hittite verbs described above;
ii. anticausative alternation: the causative verb is
9 Note that Haspelmath uses the term 'inchoative' for anticausative. The examples below are partly form Haspelmath (1993) partly mine.
unmarked, the anticausative verb is marked, as in Italian: rompersi/rompere (break), Russian: rasplavit'/rasplavit'sja (melt), German: verandern/sich verandern (change) (These three
examples involve the usage of the reflexive particle for anticausative derivation.);
iii. equipollent alternation: both the causative and the anticausative verb have the same number of morphemes, but they are different, as in English fall/fell, Japanese atumaru/atumeru 'gather';
iv. suppletive alternation: the alternation involves lexically different verbs: English: die/kill; Russian: goret'/zec (burn);
v. labile: there is no overt marker of the alternation:10 English: break, melt, open,...
Below I give a non-exhaustive list of Hittite verbs that display either causative derivation or causative/anticausative meaning connected with diathesis, whereby the middle voice is typical for states and spontaneous events; in addition, a few verbs display suppletion. Note that causative verbs in -nu-may have middle voice, see sec. 5.11
SUPPLETION:
1. die/kill = ak- (active -hi verbs, infreq. middle=active);/ kuen-
2. fall/fell = maus- (active=middle )/ pessya-
3. appear/show = dug- (middle) /tekkussai- (active)
CAUSATIVE DERIVATION:
4. be warm / warm up = e- (middle) / inu-
5. collapse/destroy = hark-( active=middle) / harganu-
6. dry = hates- (mostly active ) / hatnu- (both verbs from OS)
7. delay = istnatai- (active) / istandanu-
8. go out/put out = kist- (middle) / kistanu-
9. grow = makkes- middle (from MS also active=middle)/
10 Some verbs can be labile or have anticausative derivation, as Italian fondere 'melt': Il metallo fonde/ il calore fonde il metallo 'the metal melts/heath melts the metal' (labile); Il metallo si fonde (anticausative) 'the metal melts'. The same is also true for some Hittite verbs, as I will show below.
11 The data are partly from Neu (1968a), partly from CHD, and, for causative verbs, from Luraghi (1992). Listed verbs are arranged in Hittite alphabetical order.
maknu- both verbs from OH
10. get lost/loose = mer- active=middle (middle only after OH/OS)12/marnu-
11. fear/frighten = nahsariya- (active=middle more frequent) /nahsarnu- ('be afraid' is nahh-, a -hi verb, mostly act.)
12. become/make drunk = nink- (active= middle) /ninganu-
13. become/make tall = parganes-( active=middle) /parganu-
14. become/make tall = parkiya- (active=middle) /parkiyanu- (in OH apparently LABILE, causative derivation from MS)
15. grow/raise = salles- (active) /sallanu-
16. disappear/eliminate = samen- (active) /samenu-
17. sleep/cause to sleep = ses- / sassanu-
18. burn = war- (middle)/warnu-
19. turn = weh-( active=middle)/wahnu-
20. fear/scare = werite- / weritenu-
21. boil = zea- (middle) /zanu-
VOICE ALTERNATION:
22. split = harp- (active=transitive; middle= anticausative)
23. finish = irha-(active=transitive; middle= anticausative)
24. prosper, flourish/set straight = lazziya- (active=transitive; middle= anticausative; middle also 'recover', from OS; more frequent than act.; as 'heal' also huisnu- caus. from huis-, 'live')
25. survive/sustain = luluwai- (active=transitive; middle= anticausative; both from OH with same frequency)
26. melt down/melt = mariya- (active=transitive; middle= anticausative; middle from OS, active from OH/NS less frequent)
27. to be invisible/hide = munnai- (active=transitive; middle= anticausative; active from OH/NS, middle from MH/NS, less frequent)
28. turn = nai-(active=transitive; middle= anticausative; both from OS)
29. break = pars- (active=transitive; middle= anticausative or transitive; active only from OH/NS; thus in OH possibly LABILE)
30. fill = suwai- (active=transitive; middle= anticausative)
12 See below sec. 5.
31. break = duwarnai- (active=transitive; middle= anticausative)
32. finish = zinna- (active=transitive; middle= anticausative; partly LABILE: active can also be anticausative)
Verbs such as 6, 9, 13, and 15 build apparently equipollent pairs with verbs in -es, which are taken as derived from adjectives; however, such -nu verbs are usually taken not as derived from adjectives, but from other de-adjectival verbs. I do not wish to enter a discussion of word formation here, and consider them part of the group of verbs with causative derivation.
Haspelmath (1993) provides cross-linguistic evidence for certain verbs to display more frequently either causative or anticausative derivation, depending on their meaning, reproduced below in Figure 1.13 Even though the list does not cover all Hittite verbs surveyed above, it can be seen that, with few exceptions, verbs that have causative derivation in Hittite are the same or are semantically similar to verbs that tend to have causative derivation across languages, while verbs that have a causative/anticausative alternation based on voice are close to those that tend to have anticausative derivation. Thus, voice seems to be connected with the causative/anticausative alternation in Hittite.
At the core of the causative alternation, we find stative verbs: indeed, such verbs do not even indicate a dynamic event, thus, causativization does not only indicate that they are brought about by an agent, but also adds a dynamic dimension. Other verbs (often, but not necessarily middle) which exhibit the causative derivation indicate events that are most likely to occur spontaneously, that is, that are most likely to be uncontrolled. Lack of control is indicated by middle voice in cases of middle/active alternation; however, with some of these verbs there is evidence for late development of the alternation. In the first place, some of them seem to be labile in Old Hittite, such as 29 and 32 (there is similar evidence also for 14, with causative derivation); in the second
13 These are only a part of the verbs surveyed in Haspelmath (1993), for which I have found a correspondence in Hittite.
place, with some of these verbs the active seems to be as basic as the middle, or in some cases even older, as I will show in sec. 5.
Figure 1.
(causative)
boil
go out/put out learn/teach turn melt dissolve burn destroy get lost/lose fill finish improve break split (anticausative)
To sum up, causative derivation is older that voice alternation.14 Note that according to Haspelmath (1993) the tendency shown in Figure 1 for verbs to display either type of derivation depends on likelihood for events to occur spontaneously (causative derivation) or be brought about by an agent (anticausative derivation).
5. Diathesis, control and unaccusativity
Below are some observations that can be made regarding to the distribution of Hittite diatheses:
• Inchoative verbs that have morphological causative counterparts are either media tantum or they have active=medium, i.e. no matter the morphology, they denote uncontrolled events.
14 See also ffla^oB (2008).
• Morphological causatives with the suffix -nu- can often be inflected in the middle. In the middle, they have reflexive (mostly with the reflexive clitic =z(a)) or passive meaning; i.e. they denote controlled events. The main function of -nu- then can be said to change the meaning of the predicate from [control] to [+control].
• The Hittite middle can have passive meaning, but this semantic extension was only at the onset in the Old Hittite period (cf. Neu 1968a, b). In some cases, suppletive verbs that appear in the causative/anticausative alternation can be used as lexical pairs of active/passive (see below, the remarks from Justus 2000).
• This alternation is parallel to the alternation transitive/causative found with action verbs as in kank- 'hang (tr.)' / ganganu- 'cause to hang'.15
• Intransitive verbs denoting controlled events are very often active (pai- 'go', uwa- 'come').
• Stative or change-of-state verbs with typically human subjects (as 'live' or 'die') are often active.
Benedetti (2002) has argued that the Indo-European middle was connected with unaccusativity. As we have seen in sec. 2, unaccusativity is connected with the syntax of subject clitics in Hittite; the Old Hittite data points in the direction of an ongoing extension of unaccusativity, which started with stative and change-of-state verbs, that is, with verbs that denoted lack of control. Indeed, these two groups of verbs cover the original functions of middle voice in Hittite, according to Neu (1968a).
According to Kemmer (1993), middle voice is a semantic
15 Intransitive verbs with causative derivation outnumber transitive ones. Note that there is very little evidence for causatives of transitive verbs to be ditransitive: usually, the causee remains unexpressed even with causatives. The whole Hittite corpus only contains one occurrence (in HG § 43) of a ditransitive causative, as I have shown in Luraghi (1992), in which I checked all available occurrences. Thus, the statement that "Double accusatives ... are particularly common with verbs which are causatives of transitive verbs" (Hoffner, Melchert 2008: 247) is false, since it is based on a single occurrence.
category intermediate in transitivity between one-participant and two-participants events. It is typically used with verbs that denote spontaneous events ([-control]), or to indicate subject-affectedness ([±control]). Kemmer's definition has recently been shown to fit the PIE middle (see Meiser 2010), as well as morphological middles of many Indo-European languages, as e.g. Ancient Greek (see Allan 2003). In Hittite the semantic properties of middle voice are distributed between morphological middle and reflexive constructions, as shown in Figure 2:
Figure 2.
spontaneous events [-control] = middle, s' active non-causative
Hittite
" subject involvement [±control] = reflexive particle =za (or pron. clitic) with middle or active morphology
The use of the particle =z(a) to indicate subject involvement developed especially after Old Hittite, as shown in Boley (1993). Josephson (2003) argues, among other things, that the subject of an intransitive verb with =z(a) is more actively involved in bringing about a certain event than the subject of the corresponding verb without =z(a). Both Josephson and Boley compare the use of =z(a) with some usages of reflexives in the Romance languages: Josephson refers to Maldonado (2000) on Spanish se, while Boley has some non-systematic remarks on Italian si. Non-reflexive usages of si have been connected with unaccusativity. Note however that the notions of 'typical' middle voice meaning and unaccusativity do not coincide: as remarked by various authors, middle voice is not necessarily connected with intransitivity . Some typical usages of middle voice are found with transitive verbs. See for example the following passage from Ancient Greek:
(23) oute bomous poieuntai oute pur anakaiousi mellontes thuein
"When about to sacrifice, they do not build altars or kindle fire"
(Hdt. 1.132.1).
In the above example, the form poieuntai is middle. As well known, active voice could also occur in this same context,
without any changes in transitivity: by the choice of middle voice the author indicates that the agent performs an action in his/her own interest. A further example from Italian can illustrate this point, and also offer a parallel with Hittite:
(24) Mi sono mangiato tutta la torta che era rimasta
"I ate up (lit.: I ate + reflexive) all the cake which had been left over."
(25) NINDA-ann=a= z TI-annas azikkizzi bread:ACC and PTC live:lNF eat:PRS.3SG
"she eats the bread of life" (from Boley 1993: 168).
Thus, the Hittite data seems to point toward a division in the domain of middle voice: in particular, subject involvement is connected with the particle -z(a), while spontaneous events are connected with grammatical middle.
The development of Hittite middle voice also involves morphology. Morphological marking of middle voice became heavier after Old Hittite: Neu (1968a) has shown that middle endings containing -r are later than middle endings that did not contain it. Justus (2000) argues that the extension of -r endings owed to a re-structuring of the voice system, in the transition from phase I to phase II illustrated in Table 1.
Justus provides examples with the verb mer-/mar- which is active in Old Hittite and later acquires middle morphology (but active morphology still continues, with no apparent change of meaning):16
(26) man=kan LuMESHEDI=ma arha mirzi
when PTC M. CONN PREV go.missing:PRS.3SG
'When the MESHEDI men go entirely missing ...' (IBoT I 36 I 53)
(27) martari= war=at= kan nu= war=at= kan aszi
disappear:PRS.3SG.M/P PTC 3PL.N/A PTC CONN PTC 3PL.N/A PTC
remain: PRS.3SG
"(Some) things get lost, (others) are left over"
(KUB XIII 35 IV 45-46).
As Justus points out, in Hittite "counterpart intransitives or
16 Table 1 and examples (26) and (27) are from Justus (2000).
transitive actives tended to be expressed suppletively (ak-hi 'die': kuen- 'strike'; kisa(ri) 'come about, happen', iya- 'make'; kitta(ri) 'lie, be placed': dai- 'put, place')". One can add that derivation, that is word formation, rather than inflection was used to create causative counterparts to basic anticausative verbs, as we have seen above.
Table 1.
Stage I Verb classes: person endings have a classifying function. Verbs are semantically either (intransitive) active or stative, transitive actions are syntactically or morphologically derived.
Stage II Transition: verb classes are collapsing together. Various transitivity-changing strategies appear.
Stage III One overt verb class with voice transformation. Verbs can be transformed from (transitive) active to passive.
Such evidence points toward a lexical distribution of PIE diatheses, as represented in Table 1, which summarizes a view shared by numerous scholars who subscribe to the reconstruction of PIE as an active language (see Maiser 2010 with references). In the Indo-European languages, and possibly in late PIE, the transition from active to nominative/accusative alignment had already been completed, but there were remnants of the old alignment type, among which the media tantum. In the meantime, diathesis became inflectional, and middle forms started to appear with verbs which were formerly only active. This was the onset of the process that led middle voice to acquire passive meaning. Presumably, in PIE passive was expressed by activizing intransitive (stative) verbs (Meiser 2010: 331). With its array of lexical means to implement the causative/anticausative alternation, which supplement inflectional diathesis, Hittite offers evidence for this transitional stage.
6. Conclusion
Middle voice in Hittite is typical of stative and change-of-state verbs, as shown in Neu (1968a, b).17 The same verbs constitute the core of Hittite unaccusatives, as shown on the evidence offered by the distribution of subject clitics. However, unaccusativity as we can observe it in the Hittite texts was an expanding phenomenon, as it extended to include not only verbs whose subject is more readily comparable with a direct object because it is nonvolitional, but also verbs that indicate directional motion, whose subject, being volitional, is semantically closer to an agent. When this extension took place, unaccusativity was no longer connected with middle voice, either semantically, because it was no longer limited to verbs indicating uncontrolled events, or morphologically, because it extended to active verbs. The starting point however was provided by the close connection of middle voice with unaccusativity, favored by the presence of =z(a), which covered non-unaccusative usages typical of middle across languages. In its further development, middle inflection extended to once only active verbs, and the middle voice acquired passive meaning. Causativization strategies remained, as did media tantum, as vestiges of older stages.
References
Allan, Rutger 2003. The Middle Voice in Ancient Greek. Amsterdam: Gieben.
AM = Goetze, Albrecht 1933. Die Annalen des Mursilis. Leipzig: Hinrichs.
Benedetti, Marina 2002. Radici, morfemi nominali e verbali: alla ricerca dell'inaccusativita indoeuropea. Archivio Glottologico 87, 20-45.
Boley, Jacqueline 1993. The Hittite Particle -z/-za. Innsbruck: Institut fur Sprachwissenschaft.
17 Note however that a number of such verbs, as es- 'be' and huis-'live' are active -mi verbs, some others, such as ak- 'die', sagahh-'know' are (active) -hi verbs. As I remarked above, fn. 2, the fact that verbs of the -hi conjugation are clearly connected with the Indo-European middle does not simplify things when one tries to reconstruct the history of the Hittite middle voice.
CHD = The Chicago Hittite Dictionary, Chicago, 1980-.
Comrie, Bernard 1985. Causative verb formation and other verb-deriving morphology. In T. Shopen (ed.) Language Typology and Syntactic Description vol. 3 Grammatical Categories and the Laxicon. Cambridge: CUP, pp. 309-348.
Comrie, Bernard 1989. Language universals and linguistic typology. 2nd edition. Chicago: University of Chicago Press.
Croft, William 1991. Syntactic Categories and Grammatical Relations. Chicago: University of Chicago Press.
Garrett, Andrew 1990. The Syntax of Anatolian Pronominl Clitics. Ph.D. diss., Harvard University.
Garrett, Andrew 1996. Wackernagel's Law and inaccusativity in Hittite. In A. Halpern (ed.), Aproaching Second, Chicago: CSLI Publications, pp. 85-133.
Goedegebuure, Petra 1999. The use and non-use of enclitic subject pronouns in Old Hittite", paper read at the IV. Internationaler Kongress fur Hethitologie, Wurzburg, October 1999.
Goetze, Albrecht 1925. Hattusilis. Leipzig: Hinrichs.
Haspelmath, Martin 1993. Inchoative/causative verb alternations. In B. Comrie, M. Polinsky (eds.), Causatives and Transitivity, Amsterdam/Philadelphia: Benjamins, pp. 87-120.
HG = Friedrich, Johannes 1959. Die hethitischen Gesetze. Leiden: Brill.
Hoffner, Harry A. and H. Craig Melchert 2008. A Grammar of the Hittite Language. Winona Lake: Eisenbrauns.
Hopper, Paul and Sandra Thompson 1980. Transitivity in Grammar and discourse. Language 56/4, 251-299.
Josephson, Folke 2003. The Hittite reflexive construction in a typological perspective. In B. L. M. Bauer, G. Pinault (eds.) Language in Time and Space. A Festschrift for Werner Winter on the Occasion of his 80th Birthday. Berlin/New York: Mouton de Gruyter, pp. 211-232.
Justus, Carol 2000. The Age of the Indo-European Present -R Person Endings. Proceedings of the 11th UCLA Indo-European Conference. JIES Monograph 35, 267-292.
Kemmer, Suzanne. 1993. The Middle Voice. Amsterdam; Philadelphia: Benjamins.
Luraghi, Silvia 1990. Old Hittite Sentence Structure. London/New York: Routledge.
Luraghi, Silvia 1992. I verbi in -nu- e il loro valore causativo. In O. Carruba (ed.) Per una grammatica ittita, Pavia: Iuculano, pp. 153180.
Luraghi, Silvia 1997. Omission of the direct object in Classical Latin. Indogermanische Forschungen 102, 239-257.
Luraghi, Silvia 1998. Participant tracking in Tacitus. In B. Garcia-Hernandez, ed., Estudios de Linguistica Latina, Madrid, Ediciones Clasicas, pp. 467-485.
Luraghi, Silvia 2003. Definite referential null objects in Ancient
Greek. Indogermanische Forschungen 108, 2003, 169-196. Luraghi, Silvia 2004. Null Objects in Latin and Greek and the Relevance of Linguistic Typology for Language Reconstruction. In Proceedings of the 15th Annual UCLA Indo-European Conference, JIES Monograph 49, 234-256.
Luraghi, Silvia and Claudia Parodi, 2008. Key Terms in Syntax and Syntactic Theory. London/New York: Continuum.
Maldonado, Ricardo 2000. Conceptual distance and transitivity increase in Spanish reflexives. In Z. Frajzyngier, T. Curl (eds.) Reflexives: Forms and Functions, Amsterdam/Philadelphia: Benjamins, pp. 153-185.
Meiser, Gerhard 2010. Zur Typologie des urindogemanischen Mediums. In R. Luehr, S. Ziegler (eds.) Protolanguage and Prehistory: Akten der XII. Fachtagung der Indogermanischen Gesellschaft, Wiesbaden: Reichert, pp. 318-334.
Naess Ashild 2007. Prototypical transitivity. Amsterdam;
Philadelphia: Benjamins.
Neu, Erich 1968a. Interpretationen der hethitischen mediopassiven Verbalformen. Wiesbaden: Harrassowitz.
Neu, Erich 1968b. Das hethitische Mediopassiv und seine indogermanischen Grundlagen, Wiesbaden: Harrassowitz. Pustejovsky, James 1995. The Generative Lexicon. Cambridge, MA: MIT Press.
Rose, Sarah R. 2006. The Hittite -hi/-mi Conjugations: Evidence for an Early Indo-European Voice Opposition. Innsbruck: Innsbrucker Beitrage zur Sprachwissenschaft.
StBoT 24 = Otten, Heinrich 1981. Die Apologie Hattusilis III.
Wiesbaden: Harrassowitz StBoT 25 = Neu, Erich 1980. Althethitische Ritualtexte in Umschrift.
Wiesbaden: Harrassowitz.
Шацков, А. В. 2008. Хеттские глаголы с суффиксом -nu-. In Индоевропейское языкознание и классическая филология -XII Санкт-Петербург: Нестор-История, pp. 471-474.
Van Valin, Robert 1990. Semantic parameters of split intransitivity. Language 66, 221-260.
