TITLE "The rise and fall of a rogue relative construction"
SUBJECT "SL, Volume 25:3"
KEYWORDS ""
SIZE HEIGHT "220"
WIDTH "150"
VOFFSET "4">
The rise and fall of a rogue relative construction1 Guy Deutscher St John’s College, Cambridge
In the earliest attested stage of the Akkadian language, relative clauses were introduced by a pronoun which agreed in case with the head noun in the main clause, rather than with the relativized NP in the relative clause. Such a system is extremely rare across languages, is demonstrably dysfunctional, and has been termed ‘inherently disfavoured’. This article attempts to explain how Akkadian acquired this rogue relative construction, and how the language then managed to get rid of it. I argue that this construction was only an unstable way-station in the emergence of a new relative clause in the language. The final section of the article examines the few parallels from other languages to the Old Akkadian system.
1.
Relative clauses in Old Akkadian
Akkadian is a Semitic language which was spoken in ancient Mesopotamia, an area that roughly corresponds to today’s Iraq. The history of Akkadian spans two millennia: it is attested in writing from about 2500 BC, and it died as a spoken language about 500BC. The earliest historical stage of the language, before 2100 BC, is conventionally called ‘Old Akkadian’, and is the main focus of this article. Since the first texts which can be used for syntactic investigation are from around 2300BC, we are in fact dealing with a period of only two centuries. In Old Akkadian, relative clauses were introduced by a pronoun šu, which declined for case, number and gender, as shown in (1) below.
However, although šu was a pronoun which introduced relative clauses, it was not a ‘relative pronoun’ in the European sense. The pronoun šu agreed in case with its antecedent, that is, with the head-noun in the main clause. It did not have the case of the relativized NP in the relative clause (henceforth NPrel). In other words, the case of the pronoun šu was not determined in the relative clause, but in the main clause. Consequently, the pronoun šu played no role in indicating the role of NPrel in the RC. This NPrel was either simply deleted, or indicated by anaphoric (‘resumptive’) pronouns. The Old Akkadian relative construction is illustrated below. The examples show both the case agreement between šu and the head-noun, and the pattern of anaphoric pronoun retention within the relative clause. The NPrel is deleted when it is the subject (2) and (optionally) direct object (3), whereas in more oblique positions, the NPrel is generally indicated by an anaphoric pronoun (4, 5).2 (2) head-noun: S, NPrel: S (deleted) hç ubull-um šu al X ibašše’-u debt-nom rel.nom(m.sg) on X exist.3m.sg-sub3 ‘(this is) the debt that is upon X’ (HSS 10:109:22) (3) head-noun: Topic (nominative), NPrel: DO (deleted) še’-um šu ana ¶e.ba ašı¯t-u barley-nom rel.nom(m.sg) to rations I.left-sub ‘(concerning) the barley that I had left over for rations’ (SAB: Ga 3:4) (4) head-noun: S, NPrel: IO (represented by anaphoric pronoun -šum) Šarru-kı¯n šar ma¯t-im šu Enlil ma¯hç ir-a Sargon king.of land-gen rel.nom(m.sg) Enlil rival-acc la¯ iddin-u-šum not he.gave-sub-to him ‘Sargon, king of the land, that Enlil has not given him a rival’ (~Sargon … to whom Enlil has given no rival) (RIME 2.1.1.6. Caption 1) (5) head-noun: S, NPrel: gen (represented by anaphoric pronoun -šunu) šı¯b-u¯ šu¯t maçhar-šunu enma X ana Y witness-nom.pl rel.nom(m.pl) in front.of-them quot X to Y ‘witnesses that in front of them X said to Y…’ (Gelb OAIC 12:16)
The rise and fall of a rogue relative construction 407
(6) head-noun: DO, NPrel: DO (deleted) eql-am ša k·… nı¯tiq-u lišqi’u¯ field-acc rel.acc(m.sg) gate we.passed-sub they.should.water ‘they should water the field that we passed (by the) gate’ (SAB: Pu 3:5) (7) head-noun: gen, NPrel: GEN (time adverbial, deleted) in šant-im šališt-im ša¯ti Enlil šarru¯t-am in year-gen third-gen rel.gen(f.sg) Enlil kingship-acc iddin-u-šum he.gave-sub-to him ‘In the third year that Enlil gave him the kingship’ (RIME 2.1.2.6:68)
2. Cross linguistic rarity The internal structure of the Old Akkadian relative clauses is not unusual. The pattern of anaphoric pronoun retention follows Keenan and Comrie’s ‘accessibility hierarchy’, and has exact parallels in many other languages (Keenan and Comrie 1977). Gender and number agreement between the head-noun and the relative particle is also common enough. (Since gender and number are trivially the same for the head-noun and for NPrel, there is no question of whether the relative particle agrees with the head, or with the NPrel.) However, the picture with case agreement is quite different. The Old Akkadian construction, in which the relative particle always agrees in case with the head-noun in the main clause, seems to be extremely rare across languages. Givón’s typology of relatives (1990, Chapter 15) does not mention any language with such a system. Lehmann’s comprehensive typology only mentions one language with traces of such a system, standard literary Arabic (1984: 97). Nichols, in her typology of relatives, says that relativization ‘inherently disfavours’ a system like that of Old Akkadian (1984: 534).4 She also mentions only one counter-example, namely Arabic. Indeed, it is very hard to find genuine parallels to the Old Akkadian system. The Old Akkadian RC has its roots in structures that go back to proto-Semitic, as will be discussed below. But similar constructions have entirely disappeared from the Semitic languages, except for marginal traces in literary Arabic. Arabic and other partial parallels are discussed in the appendix below, but I have not found any language with an exact parallel to the Old Akkadian construction: a fully functional system of finite relative clauses, in which the relative particle regularly agrees in case with the head-noun. Of course, such a language may be found on further investigation, but the Old Akkadian construction can safely be
408 Guy Deutscher
assumed to be rare. The rarity of this system is not surprising, and it is not difficult to see why the Old Akkadian construction should be ‘disfavoured’. This system is simply not a very sensible way of forming relative clauses: it is inefficient and dysfunctional. First, the case marking on the pronoun šu does not serve any useful purpose. Clearly, this case marking does not help to recover the role of NPrel, because the pronoun shows the wrong case for that purpose. In RCs with a nominal head, the only potentially useful information that this case agreement could provide would be to clarify which noun in the main clause is the head of the RC.5 This information could be useful if RCs were often extraposed, and separated from their head noun. But since RCs in Akkadian almost invariably stand directly after their head noun (with any adjective or genitive modifiers), this case agreement is entirely redundant. But the problem with the case marking on the pronoun šu is more serious than just redundancy. In fact, this case marking is counter-productive, because it clashes with natural clause boundaries and obstructs the recovery of the role of NPrel. Except for the unusual case marking on the relative particle, the RCs in Old Akkadian behave precisely like ‘normal’ RCs of the embedded post-posed type with invariable relativizer.6 In this type of RCs, the relative particle naturally belongs to the relative clause, not to the main clause.7 From the synchronic perspective of Old Akkadian, therefore, the case of the pronoun šu is determined in the ‘wrong’ clause: the pronoun should be a part of the RC, but its case is assigned by the main clause. Within the RC, the case of šu is entirely arbitrary, and can easily be in conflict with the role of NPrel. The RC can thus contain two pronominal elements: the ‘relative particle’ šu with an arbitrary case marking, and in addition, an anaphoric pronoun with the potentially different case of NPrel. Such a conflict of case marking must be detrimental to the quick and efficient processing of the RC, and to the recovery of the role of the NPrel. Indeed, the (limited) data from Old Akkadian indicate such processing difficulties, especially with oblique cases.8
3. The fall It should not come as a great surprise, then, that the Old Akkadian system suffered an early death. The case-system in general remained stable in Akkadian until 1000BC.9 But the case distinctions on the pronoun šu had entirely disintegrated by 2000BC, quickly taking gender and number distinctions with
The rise and fall of a rogue relative construction 409
them to the grave. From around 2100BC, we generally find only one fossilized form of the pronoun šu, the accusative masculine singular ša. Some case, number, and gender distinctions lingered in poetic texts (as well as in personal names, and in other frozen expressions), but outside this archaizing context, the form ša became the universal, invariable, relative particle.10 The internal structure of the RC in later stages did not change, and the role of NPrel was indicated precisely as in Old Akkadian. The system in later stages of Akkadian thus conforms to the very common cross-linguistic pattern of relativization, in which the RC is introduced by an invariable particle, and the NPrel is either deleted, or indicated by an anaphoric pronoun. This younger system remained entirely stable until the death of the language around 500BC. The younger construction is demonstrated with a few examples below, all taken from the law code of Hammurabi (ca. 1750 BC): (8) awı¯l-um ša ana bull-îm illik-u man-nom rel to extinguish.inf-gen went.3sg-sub ‘the man that went to extinguish (the fire)’ (CH §25) (9) ina tupp-im ša iš»tur-u-šim in tablet-gen rel he.wrote-sub-to her ‘in the tablet that he wrote to her’ (CH §179) (10) še’-am ša imçhur-u iriab barley-acc rel he.received-sub he.(will)replace ‘he will replace the barley that he received’ (CH §254) (11) almatt-um ša ma¯r-u¯-ša seç » hhç er-u¯ widow-nom rel son-nom.pl-her very young-nom.pl ‘a widow that her sons are very young’ (CH §177)
The story so far is thus very satisfactory: as proper fiction would have it, the good ended happily and the bad unhappily. The Old Akkadian construction died a deserved early death, whereas the later system with the invariable particle ša remained stable for fifteen hundred years, and was halted only by the death of the language in the first millennium BC.
4. The rise There is only one snag in this story: why should a rogue construction like the Old Akkadian relative clause arise in the first place? This section argues that the Old Akkadian system was a historical ‘accident’: an unstable way-station in the
410 Guy Deutscher
diachronic emergence of a new type of relative clause in the language. The Old Akkadian system was a relic of an earlier stage, in which the relative clauses with šu were not yet fully embedded, but rather appositional. To understand how the new construction emerged, we need to take two short steps backwards. The relative clauses with šu were modelled on an older relative construction in Akkadian, which they later superseded. Akkadian, in common with the other Semitic languages, has a special form of the noun called the ‘construct state’.11 The main function of the construct state is to mark the head noun in the genitival complex, and I therefore gloss it ‘.of’. As demonstrated in (12) below, the head noun is in the construct state, whereas the genitive is in the genitive case: (12) dı¯n šarr-im judgment.of king-gen ‘the judgment of the king’
At some prehistoric stage, this genitival complex was extended to a relative clause, following a well trodden path of syntactic extension.12 This extension must have occurred already in Proto-Semitic, since the relative clause with the construct state is attested in the earliest strata of most Semitic languages (Lipinski 1997: 324). An example of this relative clause from Old Akkadian is given in (13) below. Here, the onset of the relative clause is not marked by any independent particle, but by the construct state on the head noun. (13) tuppi addin-u-šum tablet.of I.gave-sub-to him ‘the tablet that I gave to him’ (SAB: Ki 2:13)
This older construction was being supplanted by the new RCs with the pronoun šu already in Old Akkadian.13 The pronoun šu was originally a type of determinative/demonstrative pronoun, meaning ‘this/that/the one’. As the head of a genitival or a relative construct, the pronoun itself must have been perceived to be in the ‘construct state’, and it meant ‘the one of ’ (14) or ‘the one that’ (15). (14) Šu-Enlil the one(nom.m.sg).of-Enlil ‘the one of (the God) Enlil’ (an Old Akkadian personal name). (15) šu¯t in tu.ra u-ù-çhi-ru-un lı¯hç uz the ones(acc.m.pl).of in illness were.delayed.sub he.should.take ‘he should take the ones who were delayed because of illness’ (SAB Gir 3:7)
The rise and fall of a rogue relative construction
If one ignores the historical origin of the pronoun šu, one could be tempted to call the examples above ‘headless relatives’. But at least originally, examples such as (14) and (15) above were not ‘headless’: the pronoun šu simply acted as the head-noun. The onset of the RC was marked by the construct state on this pronoun, precisely as it would be marked on the nominal head ‘tablet’ in (13) above. The development of šu from the head of a RC to a ‘relative particle’ must have its origin in appositional patterns. The whole genitive/relative complex headed by šu could be added as an apposition after a noun, probably for the purpose of emphasis (perhaps because the construct state was not salient enough on its own as a clause boundary). A sequence such as (16) below initially must have meant ‘the judgment, the one of the king’, and the situation would have been similar with relatives such as (17).14 (16) dı¯n-um šu šarr-im judgment(m.sg)-nom the one(nom.m.sg).of king-gen ‘The judgment, the one of the king’ (17) dı¯n-um šu idı¯n-u judgment(m.sg)-nom the one(nom.m.sg).of he.rendered-sub ‘The judgment, the one that he rendered’
This use of the pronoun šu to introduce appositional RCs must go back to Proto-Semitic, because most Semitic languages use cognates of šu to introduce RCs. The appositional construction with the ‘determinative’ pronoun could have lived happily for many years in the prehistory of Akkadian. As an apposition, the ‘determinative’ pronoun šu naturally agreed with its antecedent in case, number and gender. As long as the pronoun šu was the head of an independent clause, its case marking was still motivated: the pronoun was a part of the main clause, and its case-marking was thus correctly assigned by the clause to which it belonged. Nevertheless, presumably because of frequent use, the appositional clauses underwent a process of ‘condensation’: they lost their independence and became embedded. This diachronic process can be represented as a reanalysis of clause boundaries, as in (18) below. The pronoun šu, which had been part of the main clause and the head of the RC, was reanalysed as merely the marker of relativization, while its antecedent became the head of the RC.
411
412 Guy Deutscher
(18) Reanalysis dı¯num šuhead [idı¯n-u]RC judgment the one.relative marker he.rendered-sub ‘the judgment, the one that he rendered’ Ø dı¯numhead [šu judgment relative particle ‘The judgment that he rendered’
idı¯n-u]RC he.rendered-sub
This is a fairly modest and not uncommon type of clause boundary reanalysis (cf. Harris and Campbell 1995: 62 ff., 287ff. for similar examples). Nevertheless, the consequences of this reanalysis in Akkadian were spectacular. Once the pronoun šu was reanalysed as the relative particle and became a part of the RC, its case marking became maladaptive, because it was assigned by the ‘wrong’ clause. In consequence, the whole system of case marking on this pronoun disintegrated, and only one fossilized form of the pronoun remained as an invariable relative particle, as we saw above. Since the pronoun šu was used in genitives as well as with relatives, one could sketch precisely the same scenario of reanalysis with a genitival complex, as demonstrated in (19) below. Indeed, the reanalysis of genitives and relatives must have gone hand in hand. After the Old Akkadian period, ša became not only the invariable relative particle, but also the invariable genitive particle. Nevertheless, the more pressing motivation for the disintegration of the case distinctions on the pronoun šu probably came from the RCs, not from the genitives. In the genitive complex, the case marking on the pronoun šu may be redundant, but at least there is no conflict of case-roles, since there is no NPrel.15 (19) dı¯n-um šuhead šarr-im judgment-nom the one.of king-gen ‘the judgment, the one of the king’ Ø dı¯n-umhead šu šarr-im judgment-nom of king-gen ‘the judgment of the king’
"deu-r8">
The rise and fall of a rogue relative construction 413
5. Conclusion We have followed the Old Akkadian construction from the cradle to the grave. I argued that the Old Akkadian system was an unstable way-station in the historical emergence of a new type of relative clauses in the language. It was a transitional stage between the older relative construction without a relative particle, and the new type of relative clause, with an invariable relative particle. The relative construction with the pronoun šu originated as an apposition (in all probability already in Proto-Semitic). The pronoun belonged to the main clause, and was the head of the relative clause. In this earlier incarnation, the construction could have been stable for a long period: as long as the construction remained appositional, the case marking on the relative particle was still motivated. But at some stage the appositional RCs underwent reanalysis. The clause boundary was shifted, and the pronoun šu was ‘demoted’: from being the head of the clause, to the status of a mere marker of subordination, and consequently, from the main clause to the relative clause. I have tried to show that there is nothing untoward in this diachronic path. There is also no reason why similar processes should not happen frequently elsewhere. Nevertheless, after reanalysis, the resulting system is not only unmotivated, but maladaptive, and therefore unstable. This instability is surely the reason why it is so difficult to find parallels to the Old Akkadian system in other languages. In fact, we are lucky to have caught the Old Akkadian construction red-handed in the beginning of the historical period. Had the Akkadians started writing just two centuries later, the only traces of this construction would have been isolated occurrences in archaizing poetry and fossilized personal names. It would then have been very difficult to believe that this system had been fully functional just a short time before.
6. Appendix: Parallels to the Old Akkadian construction As I argued above, there is nothing unusual about the appositional source construction from which the Old Akkadian RCs emerged. We find similar nonembedded RCs in various other languages (Lehmann 1984: 122, Givón 1990: 651). One example can be given here to demonstrate such constructions, from Bambara (Givón 1990: 653, cf. Lehmann 1984: 135):
"deu-r13">
414 Guy Deutscher
(20) c¸e ye muru san n ye min ye man past knife buy I past rel see ‘the man bought the knife, the one that I saw’
The only relevant difference between examples such as (20) above and the source of the Old Akkadian RCs is that the ‘relative particle’ min in Bambara is case-invariable, and so the issue of case agreement does not arise. It is precisely this problem of case agreement, however, which makes finding ‘real’ parallels to the Old Akkadian RCs such a difficult task. My attempts to find at least partial parallels to Old Akkadian are summarised below.
Semitic The origins of the Old Akkadian construction go back to common Semitic structures. The use of the construct state as a relative marker is a Semitic inheritance, and so is the use of the inflected ‘determinative’ pronoun šu as the head of appositional RCs. We can thus safely assume that the original appositional construction, with a case-marked ‘determinative’ pronoun, existed in the prehistory of most Semitic languages.16 However, this case marking on the determinative-relative pronoun disintegrated in almost all the Semitic languages, even before their earliest attestations (Lipinski 1997: 324, Pennacchietti 1968). Standard Arabic is the only current Semitic language in which traces of the original system have been maintained. An extended form of the determinative pronoun is used in Standard Arabic as the relative particle ’allaðı¯, but only traces of the case declension have survived.17 The particle ’allaðı¯ is case invariable in the singular and the plural. Only in the (rare) dual, a diptotic distinction is maintained between nominative and other cases. The particle agrees with its head-noun in case (’allaða¯ni — nominative.m.dual, ’allaðayni accusative/ genitive.m.dual). Nevertheless, the status of standard Arabic as a frozen written language should be remembered. No such distinctions have survived in the spoken dialects of Arabic. The reason for the disintegration of the case distinctions on the determinative-relative pronoun all over Semitic must have been the same as the one outlined for Akkadian.
Greek and Latin Outside Semitic, there are some similarities to the Old Akkadian construction in older Indo-European languages. In ancient Greek and (less commonly) in
The rise and fall of a rogue relative construction
Latin, there is a construction, traditionally called ‘case attraction’, in which the relative pronoun takes the case of the head-noun, rather than that of NPrel: (21) sùn toîs the¯sauroîs hoîs ho pate¯r katélipen with def.article.dat treasures.dat rel.dat the father left literally: ‘with the treasures with which my father has left’ for: ‘with the treasures which my father has left’ (Touratier 1980: 213) (22) atque eadem ratione qua ante dixi in Caesaris and same.abl way.abl rel.abl before I.said in Caesar’s legionarios impetum faciunt legionaries attack they.make literally: ‘and they charged Caesar’s legionaries in the manner in which I already mentioned’ for: ‘and they charged Caesar’s legionaries in the manner which I already mentioned’. (Touratier 1980: 218)
But the similarity between these constructions and the Akkadian system is only partial: as opposed to Old Akkadian, ‘case attraction’ in Greek and Latin is not a fully functional and productive system. ‘Case attraction’ is the exception, and can only occur in limited syntactic contexts. In particular, it is found only when the head noun in the main clause is in a more oblique case than NPrel. Moreover, case attraction also occurs in the opposite direction, and again seems to be motivated by the power of more oblique case-roles to attract less oblique ones. In Greek and Latin, therefore, it seems that we are dealing with a phenomenon of a different kind, that of ‘oblique’ attraction.
Germanic A closer parallel is found in the older Germanic languages: Old Norse, Old English, and Old High German, as well as traces in Gothic. (See Stong-Jensen 1977, Mitchell 1985, Hock 1991, Harbert 1992, Pittner 1995.) The situation in Germanic can be illustrated with Old Icelandic examples. In Icelandic, there was an invariable relative particle es, which could introduce relative clauses on its own. As can be seen in (23) below, the noun ‘men’ is the head of the relative clause, and it is followed directly by the invariable relative particle es:
415
416 Guy Deutscher
(23) …vóro þar þeir menn es Norðmenn kalla Papa there were there those menhead rel Northmen call Papa ‘there were there those men that Northmen call Papas’ (Stong-Jensen 1977: 14)
However, the particle es could also be preceded by a demonstrative pronoun, as in (24) and (25). The demonstrative pronoun was the head of the relative clause, and it agreed with its antecedent in the main clause in case: (24) ok blótaðe hrafna þriá þá es hánom and he worshipped ravens three.acc.m.pl those.acc.m.pl rel him skylldo leið vísa should way show ‘and he worshipped three ravens, those that should show him the way’ (Stong-Jensen 1977: 13) (25) ok fiórer tiger nauta með henne þeirra es aoll vóro and forty cattle.gen.n.pl with her those.gen.n.pl rel all were frá henne komen from her come ‘and forty cattle with her, those that were all come from her’ (StongJensen 1977: 13)
The particle es in Icelandic can be compared to the construct state in Akkadian, while the demonstrative pronouns in Icelandic can be compared to the pronoun šu. Although es is an independent particle, whereas the construct state is marked phonetically only by the omission of suffixes on the noun, the function of the two is the same: they both mark the onset of the relative clause. Examples like (23) are parallel to the older construction in Akkadian without the pronoun šu (13 above), whereas (24) and (25) are parallel to the Akkadian RCs with the pronoun šu, as in (2)–(7) above. Note, however, that the Icelandic constructions with the demonstrative pronoun are still appositional. Constructions like Icelandic (24) and (25) existed also in Old English, with the invariable relative particle þe/ðe corresponding to Icelandic es (26). Similar structures were also used (in more restricted contexts) in Old High German (27), but with no overt relative particle corresponding to es or þe. Traces of such a system may also be reconstructed for Gothic (with the particle ei). (26) Ða wæs æt ðam geongan grim andswaru eðbegete then was for the.dat.m.sg youth.dat.m.sg grim answer easy.to.get
The rise and fall of a rogue relative construction 417
þam ðe ær his elne forleas him.dat.m.sg rel earlier his courage lost ‘then for the youth a grim answer was easy to get, (for) the one that earlier lost his courage’ (Hock 1991: 56) (27) bistu nu zi uuare furira Abrahame ouh then ∆ are.you now really older(than) Abraham.dat and those.pl.dative man hiar nu zalta? one here now named ‘are you now really older than Abraham and named here?’ (Hock 1991: 62)
those (that) one has
Unfortunately, the situation in Germanic is quite complex, not least because the different languages present significant variations in their use of such constructions. Some scholars have suggested a path of emergence which is identical to the one I sketched for Akkadian. In this scenario, the demonstrative pronoun originated as the head of a relative clause, which stood as apposition after the noun (see Hock 1991: 64). But other paths have been advanced as well.18 Whatever the precise path which created the Germanic RCs above, it is striking that they disintegrated so early in Germanic as well. The RCs in which the demonstrative pronoun agreed in case with the head were ‘on the way out’ already by the earliest records (Hock 1991). Given the discussion above, this early death should not come as a great surprise. It seems that the main clause case marking on the demonstrative pronoun could not be maintained for any length of time, once the appositional RCs became embedded. Nevertheless, the actual routes by which the construction disintegrated in Akkadian and in Germanic are quite different. In Akkadian, the case declension on the pronoun šu became defunct, and only one invariable form survived. In Germanic, the case system on the pronouns did not disintegrate, but the pronoun switched allegiance: from agreeing with the head in case, it came to represent the case of NPrel. The Germanic demonstratives thus became ‘real’ relative pronouns in the European sense. This diachronic development appears somewhat less natural than the Akkadian one. The only explanation that has been suggested for the Germanic route is the reanalysis of ‘equi-case’ constructions. It is claimed that in instances where the head-noun and NPrel had the same case-role in their respective clauses, a reanalysis of clause boundary occurred, by which the demonstrative pronoun became a part of the relative clause, rather than the main clause. At least at first sight, this route seems quite
"deu-r2"> "deu-r12">
418 Guy Deutscher
problematic. Unless one assumes that instances of ‘equi-case’ were overwhelmingly more frequent than those of different cases, one can wonder how such a reanalysis could have proceeded in the face of constant contradictory evidence. Perhaps one has to remember that this process occurred in a European context, where relative pronouns were the dominant strategy of relativization. We know that ‘relative pronouns’ in the European sense are typologically quite rare, and generally confined to Europe (Comrie 1998, Lehmann 1984: 109). Nevertheless, relative pronouns are also very ‘contagious’, and seem to spread easily on contact. It may therefore be that the switch of alliance of the Germanic pronouns should be viewed against the areal background of the dominant relativization strategy in other European languages.
Tümpisa Shoshone Beyond Indo-European, I am only aware of one other language with a system which appears similar to that of Old Akkadian. In Tümpisa Shoshone (UtoAztecan), relative clauses are introduced by a particle which has two forms, subjective (atü) and objective (akka), depending on the role of the head noun in the main clause. Two examples are given below (Dayley 1989: 358):19 (28) Wa’ippü nia pusikwa atü woman.subject me know relative(subjective) hupiatüki-tü sing-participle.subjective ‘the woman who is singing knows me’ (29) Wa’ippü-a nüü pusikwa akka woman-object I know relative(objective) hupiatüki-tünna sing-participle.objective ‘I know the woman who is singing’
On closer look, however, the similarity with Old Akkadian appears superficial. In Tümpisa Shoshone, RCs are participial rather than finite. These participial clauses, as nominalized noun-modifiers, agree with their head-noun in case. Tümpisa Shoshone (rather unusually) uses a relative particle to introduce the participial clause, and since the whole clause agrees in case with the head-noun, the relative particle does so as well. Nevertheless, there is a vital difference between the finite RCs in Akkadian and the participial clauses in Tümpisa Shoshone. The participial clauses only allow the subject position to be relativized.
"deu-r12">
The rise and fall of a rogue relative construction 419
In Tümpisa Shoshone, when the object needs to be relativized, the clause has to be transformed into a genitival structure: a sentence such as ‘the woman whom I know is singing’ would be expressed as ‘the woman of my knowing is singing’. The fact that only subjects can be relativized with participial clauses is crucial. The dysfunctionality of the Old Akkadian system does not apply in this case, precisely because recovering the role of NPrel is not a problem. The Tümpisa Shoshone construction is thus of a different nature, and offers only a spurious parallel to Old Akkadian.
Notes 1. I am grateful to Tom Givón, Peter Matthews, Theo Vennemann, and referees for Studies in Language for many helpful comments and corrections on an earlier draft of this paper. 2. References to text editions follow the standard abbreviations of The Assyrian Dictionary of the Oriental Institute of the University of Chicago (CAD) vol. R, except SAB which appears there as ‘Kienast-Volk SAB’. 3. -sub stands for the ‘subordinative’ form of the verb, which appears in all subordinate clauses, and which is normally marked by a suffix -u. 4. The system in Old Akkadian is an example of what Nichols calls a ‘clause affecting, headmarking’ strategy. She argues that ‘clause affecting relativization inherently favors dependent marking, and disfavors head-marking’. 5. When there is no nominal head to the RC (i.e. with ‘headless relatives’), the case marking on the relative particle indicates the syntactic role of the RC in the main clause, and so could be said to perform a useful function. But since such headless relatives tend to be topicalized, the importance of this function even with headless relatives is debatable. 6. This is the type that Lehmann (1984:85) calls ‘postnominaler Relativsatz mit einleitendem Subordinator’. 7. This is the case not only in our syntactic representations, but in intonation patterns. It is interesting that in Arabic, intonation patterns suggest that the relative particle ‘allaðı¯ is a part of the RC, not the main clause, even when it agrees in case with the head-noun (Nichols 1984:534). 8. There are only about 25 different examples of relative clauses in the extant Old Akkadian texts, if we ignore repeated examples of identical, or near identical constructions. Nevertheless, it is interesting to note that combinations of different oblique roles (e.g. ‘I saw the house of the mangen that I gave silver to himdat’) are not attested at all. Oblique cases are attested either with ‘equi-case’ (the same case for the head noun and NPrel), or in examples where at least one of the NPs is the subject. This does not mean that such combinations were never used, but they were clearly rarer, and were probably avoided precisely because of the processing difficulties explained above. 9. The only other area of weakening in case distinctions was with the construct state of the noun (see n. 11 below).
10. It seems that in the spoken language, the number distinction lingered a little longer than case and even gender distinctions, at least in the northern dialect of Ešnunna. There, we find the number distinction (singular ša, plural šu¯t) in the earliest texts from the twentieth century BC. For examples see Deutscher (2002: n. 14). 11. The morphology of the construct state changed during the early history of Akkadian. Initially, the construct state differed from the normal (‘free’) state of the noun only by the lack of the final consonant -m. For example, in the nominative case, the noun ‘judgment’ in the ‘free’ form was dı¯n-um, and in the construct state dı¯n-u. But already by the earliest period, some case vowels on the construct state have disappeared, and the construct state in many (but not all) instances became case-invariable: dı¯n (cf. von Soden 1995: 101). Since the pronoun šu must originally have been in the construct state (see further below), one could claim that the loss of case distinctions on the pronoun is in some sense related to the loss of cases on the construct state of nouns, at least on a superficial level. However, the motivation for the two processes should be sought independently, because the two processes of reduction proceeded along independent paths, both in time and in nature. (Cf. Deutscher (forthcoming: n. 15), and also n. 15 below.) 12. See Aristar 1991 for discussion. The most likely route for such an extension was through verbal nouns such as the infinitive. 13. The older construction never disappeared entirely in later periods, but was used only in limited configurations (for example, only with simple verbal relative clauses), and probably as a hallmark of a more elevated style. 14. Constructed examples are used here (rather than attested ones) simply to demonstrate the parallel between genitives and relatives more clearly. There are dozens of attested examples of genitival complexes with the pronoun šu in the Old Akkadian period, and examples of relatives were given in (2)–(7) above. 15. Indeed, the (limited) Old Akkadian data support this conclusion. For most of the Old Akkadian period, the case on the pronoun šu was used ‘correctly’ in the genitival complex. But with relatives, we find the form ša making inroads into the territory of the other forms already in the early Old Akkadian period. 16. In fact, the texts from Ebla (which are contemporaneous with early Old Akkadian, all from the 24th century BC) seem to show the same system of relativization as in Old Akkadian (Fronzaroli 1987). However, the language of the Ebla texts is very close to Akkadian, and its status as an independent language is controversial. 17. The early history of the relative particle in Arabic is somewhat complex. Classical Arabic poetry appears to have still maintained the case distinctions on the ‘determinative-relative’ pronoun (cognate with Akkadian šu). But this is probably a conscious archaism (see Lipinski 1997: 326), because the case declension of the determinative-relative pronoun had already disintegrated in the older Tayyitic dialect. 18. Hock (1991) argues that the demonstrative pronouns in front of the invariable relative particle are a direct continuation of the old Indo-European correlative construction. An entirely different path has been suggested for the Old High German constructions such as (27) above. Lehmann (1984: 378 ff.) argued that the relative clauses headed by a demonstrative pronoun originated as simple adjectival attributes (such as kunningin thia
The rise and fall of a rogue relative construction 421
richun ‘queen dem mighty’, that is, ‘the mighty queen’), which were extended to finite relative clauses via participial or other non-finite forms. However, Lehmann’s scenario cannot account for Old English and Old Icelandic, because it cannot explain the use of the invariable relative particle es or þe after the demonstrative pronoun. 19. I am grateful to Holger Diessel for pointing out this example.
References Aristar, A. R. 1991. ‘On diachronic sources and synchronic pattern: An investigation into the origin of linguistic universals’. Language 67: 1–33. Comrie, B. 1998. ‘Rethinking the typology of relative clauses’. Language Design 1: 59–86. Dayley, J. P. 1989. Tümpisa (Panamint) Shoshone grammar. Berkeley: University of California Press. Deutscher, G. 2002. ‘The Akkadian relative clauses’. Zeitschrift für Assyriologie und Vorderasiatische Archäologie 92:86–105. Diessel, H. 1999. Demonstratives. Form, function and grammaticalization. Amsterdam: John Benjamins. Fox, B. and S. A. Thompson 1990. ‘A discourse explanation of the grammar of relative clauses in English conversation’. Language 66:297–316. Fronzaroli, P. 1987. ‘Le pronom déterminatif-relatif à Ebla’. Mari 5:267–274. Givón, T. 1990. Syntax: a functional-typological introduction (vol. II). Amsterdam: Benjamins. Harbert, W. 1992. ‘Gothic relative clauses and syntactic theory’, in I. Rausch et al. (eds.) On Germanic linguistics. Issues and methods. Berlin: Mouton de Gruyter. Hock, H. H. 1991. ‘On the origin and development of relative clauses in early Germanic’, in E. H. Antonsen and H. H. Hock (eds.) Stæfcræft. Studies in Germanic Linguistics. 55–90. Amsterdam: John Benjamins. Keenan, E. and B. Comrie 1977. ‘Noun phrase accessibility and universal grammar’. Linguistic Inquiry 8:63–99. Lehmann, C. 1984. Der Relativsatz: Typologie seiner Strukturten, Theorie seiner Funktionen, Kompendium seiner Grammatik. Tübingen: Narr. Lipinski, E. 1997. Semitic languages. Outline of a comparative grammar. Leuven: Peeters. Mitchell, B. 1985. Old English syntax. Oxford: Clarendon. Nichols, J. 1984. ‘Another typology of relatives’. Proceedings of the Annual Meeting of the Berkeley Linguistics Society 10:524–541. Pennacchietti, F. A. 1968. Studi sui pronomi determinativi semitici. Naples: Istituto Orientale di Napoli. Pittner, K. 1995. ‘The case of German relatives’. The Linguistic Review 12:197–231. Soden, W. von 1995. Grundriss der akkadischen Grammatik (3rd edition). Rome: Pontificium Institutum Biblicum. Stong-Jensen, M. 1977. ‘The syntax of es-relatives in Old Icelandic’. Cahiers Linguistiques d’Ottawa 5:1–26. Touratier, C. 1980. La relative, essai de théorie syntaxique. Paris: Klincksieck.
422 Guy Deutscher
Author’s address Guy Deutscher St John’s College Cambridge CB2 1TP [email protected]
The rise and fall of a rogue relative construction
it clashes with natural clause boundaries and obstructs the recovery of the role of NPrel. Except ... the NPrel. Indeed, the (limited) data from Old Akkadian indicate such process- ..... of the Oriental Institute of the University of Chicago (CAD) vol.
Oct 4, 2006 - centerpiece of a new theory of sentence comprehension ... foundation for MTS, and so we review the most rele- ...... bridge University Press.
Likewise there have been commonwealths that have perished before they. have achieved any distinction in arms or in laws. Some indeed have been abortive, or perished at. birth, like the city of Münster in the Empire of Germany, dismembered from the E
Mar 17, 2008 - Economic, Military. â Fame/Respect, Influence, Happiness. â Population. â Cultural, Technological advancement. â Who ? â Nations... their leaders, their people. â MNCs...their leaders, their stakeholders. â Groups and All
There was a problem previewing this document. Retrying... Download. Connect more apps... Try one of the apps below to open or edit this item. rise and fall of ...Missing:
There was a problem previewing this document. Retrying... Download. Connect more apps... Try one of the apps below to open or edit this item. (March 2006) ...
The subsequent, foreign ruler invalidated all land titles and thus killed the bank. This unusual case study exemplifies an extension of. Mancur Olson's model of ...
Try one of the apps below to open or edit this item. pdf-1473\babylon-the-rise-and-fall-of-ancient-mesopotamias-greatest-city-by-charles-river-editors.pdf.
