Stability in Subject-Verb Word Order: From

0 downloads 0 Views 2MB Size Report
clauses. That this tendency reflects literary artifice alone (as an attempt to .... Subject occurs in dependent adverbial clause: ... The independent variables were chosen to reflect potential intralinguistic .... latter isolated from other types of nouns in view of their unique favoring of VS ...... underlines the dominant significance.

Trustees of Indiana University Anthropological Linguistics

Stability in Subject-Verb Word Order: From Contemporary Arabian Peninsular Arabic to Biblical Aramaic Author(s): Jonathan Owens and Robin Dodsworth Source: Anthropological Linguistics, Vol. 51, No. 2 (SUMMER 2009), pp. 151-175 Published by: The Trustees of Indiana University on behalf of Anthropological Linguistics Stable URL: http://www.jstor.org/stable/40730842 . Accessed: 08/12/2013 15:38 Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp

. JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected]

.

Trustees of Indiana University and Anthropological Linguistics are collaborating with JSTOR to digitize, preserve and extend access to Anthropological Linguistics.

http://www.jstor.org

This content downloaded from 132.180.10.132 on Sun, 8 Dec 2013 15:38:29 PM All use subject to JSTOR Terms and Conditions

Stabilityin Subject-VerbWordOrder: FromContemporary ArabianPeninsularArabicto Biblical Aramaic Jonathan Owens ofBayreuth University

RobinDodsworth NorthCarolinaState University fromtraditional treatments ofsubject-verb Abstract. This articlediffers word orderin Semiticin two respects.First,we take as our pointof departurea ArabianPeninsularArabic,which detailedstudyofwordorderin contemporary showsthattherespectiveorderofthesubjectand verbin thatvarietyis deterand bydiscourse-immanent factors.Fromthisstarting minedbymorpholexical the same to subjectwork we backwards,applying analyticalframework point, verbwordorderin BiblicalAramaic.Secondly,we use corpus-basedquantitativemethodsand regressionanalysisto determinethe degreeofsimilarity betweenArabianPeninsularArabicand BiblicalAramaic.It emergesthat,for verbwordorderin ArabianPeninsularArabic all intentsand purposes,subjectand and BiblicalAramaicare governedby an identicalset ofmorpholexical Historicalexplanationsfortheseresultsare discussed;it discourseconstraints. is emphasizedthat,whetherthepatternsare due to commoninheritanceorto is sustainedoverat a complexpatternofwordorderdetermination diffusion, time. least 2,500yearsofchronological

1. Introduction. Semiticlanguagehistoryis markedby a numberofbroad, is reflexofa formor construction whereina Proto-Semitic motifs, normalizing form the and shift from withitsfurther away original established, development The following characterized summaryfrom bylaws ofuniversalchangeor drift. to Protothe applicationofthisviewpoint a Semiticlanguagehandbooktypifies Semiticsentencewordorder. In Old Aramaic(withthe exceptionofstrongly AkkadianizingMesopotamian texts),wordorderis generallyofthestandardSemiticVSO [verb-subject-object] ofanyelement.ImperialAramaic type,allowingforemphasisbythe fronting evinces a tendencytowardverb-finalorder,especiallyin infinitivalverbal clauses. That this tendencyreflectsliteraryartificealone (as an attemptto mimictheAramaicofnativePersianspeakers)is evidencedbythefactthatit does notcontinueintolaterperiods,whichreflect, rather,thenormalSemitic driftfromVSO to SVO [subjectverb-object]type. In the classical Eastern dialects,in particular,wordorderin theverbalphrasemaybe said to be free. [Kaufman1997:127]

151

This content downloaded from 132.180.10.132 on Sun, 8 Dec 2013 15:38:29 PM All use subject to JSTOR Terms and Conditions

152

Anthropological Linguistics

51 no. 2

Similar assessments are found throughout Semitic language history studies (Bergsträßer 1977:16; Rosenthal 1961:56; Muraoka and Porten 2003:29g).1 The statement deserves greater scrutiny,as it encapsulates a number of stereotypicizedtruthsabout word orderin Semitic. The very firstis that there obviously exists a Proto-Semiticverb-subjectobject (VSO) word order. However, this is not self-evident,as a briefsurvey of Semitic languages makes clear. Akkadian has subject-object-verb(SOV) order, as do all contemporaryEthiopie Semitic languages. Many Arabic dialects have subject-verb-object(SVO) order. It is argued in this article that Aramaic has a variable, morpholexicallyand discourse-determined,subject-verb/verb-subject (SV/VS) word order, referredto here as "lexico-discourse-basedword order." This word orderis similar in essential respects to that ofvarious contemporary Arabic dialects, which will serve as the analytical basis ofthis study. A partial overviewofSemitic languages gives the orders seen in table 1. Table 1. Word Order of Selected Semitic Languages VSO Hebrew ClassicalArabic Gafaz

SOV Akkadian Amharic, etc. Tigrinya,

SVO EgyptianArabic (Edwards2009)

Lexico-discourse SV/VS Aramaic(as arguedbelow) ArabianPeninsular dialects

A basic survey of Semitic languages alone does not obviously point to an originalVSO word order.While contactplausibly accounts forthe SOV orderof Akkadian (fromSumerian) and modernEthiopie Semitic (fromCushitic), there is no internalevidence as to what orderthe SOV shiftedfrom. A second imprimaturofvaliditycontained in the summaryis the claim that there is an ineluctable drifttowards SVO among Semitic languages. This is obviously contradictedby the hypothesis that SOV order diffusedto modern Ethiopie,and it is contradictedby the lexico-discourse-determined SV/VS order in contemporaryArabic varieties. If driftis a factor,it presumably needs to apply across the board, in a single direction. In this article,we do not attemptto solve the problemofword orderin Semitic. However,by comparingsubject-verbword orderin two Semiticlanguages we are able to establish that lexico-discourse-controlled SV/VS is a word ordertype that goes back at least to the Official Aramaic period and, by inference,is present in a "classical" Semitic language fromits earliest attestation. The methodologyfollowed in this study is somewhat unusual in Semitic linguisticsin that, ratherthan beginningwith an older varietyoflanguage and workingtowards the present, the basis of the argument rest on observations fromthe contemporaryspoken Arabic ofthe Arabian Peninsula, which is then extendedto OfficialAramaic, a varietyattested some 2,500 years ago. This comparison produces a strikingexample of a complex mechanism for determiningSV/VS word orderthat holds fortwo languages overmore than two

This content downloaded from 132.180.10.132 on Sun, 8 Dec 2013 15:38:29 PM All use subject to JSTOR Terms and Conditions

2009

Jonathan Owens and RobinDodsworth

153

millennia. It underscoresthe pointthat Semitic languages are as interestingfor the long-termstability of their grammatical systems as forthe changes that occur in them.

2. Subject-verb word order in Arabian Peninsular Arabic. The basis of

this comparative study is extensive research on word order in spoken Arabic describedin greater detail elsewhere (Owens, Dodsworth,and Rockwood2009), whose main pointswe summarize here. It is shown that subject-verbword order in the examined varieties ofArabian Peninsular Arabic (APA)- Kuwaiti, Emirati, and Jedda- is based on two broad pragmatic parameters- availability and discourse status ofthe subject noun.2Availabilityis manifestedin morpholexical class and can be representedon a cline, as in figure1. Available Pronouns Quantifiers(termed"pronominals") Personal and definitenouns Indefinite nouns Less Available/Presentational

classes. ofmorpholexical Figure 1. Availability The quantifiers referredto in figure 1 include numerals and quantifiers such as kull 'air, baW 'some of, waahid 'anyone', (ma) had 'no one' and had 'someone'. The idea ofavailability,followingPrince (1981) and Naro and Votre (1999), is that subjects whose identity is either known (pronouns, by definition),or irrelevantas individuated entities (quantifiers such as 'all of,some of) constitute a class whose relative discourse unimportanceis encoded in a characteristic word order. In APA, this is SV order,as in (la), whereas, by way oftypological contrast,in Brazilian Portuguese (Naro and Votre 1999), this is VS, as in (lb). Subjects of greater discourse relevance work in the opposite way in the two languages: they tend to exhibit the VS position in APA, but the SV position in Brazilian Portuguese. (Here the subject is given in boldface.) ti-xtilif min baffl (la) wd baîD il-alîaab and some DEF-PL.game3F-differfrom some fromothers' 'Andsomeofthegamesdiffer

(EmiratiArabic)

(BrazilianPortuguese) (lb) Ele estavaperdido,assim,apareceuuns homens 'He was lost,likethat.Somemenappeared'(lit.'appearedsomemen') (Naroand Votre1999:80)

This content downloaded from 132.180.10.132 on Sun, 8 Dec 2013 15:38:29 PM All use subject to JSTOR Terms and Conditions

154

AnthropologicalLinguistics

51 no. 2

A characteristic functionofVS orderin APA is to introducesubjectsthat formsthe climaxof are centralto an unfoldingevent,or whose introduction orderis thus function. an episode.We termthisa presentational Verb-subject discourse in certain a status of the to the discourse sensitive text, subject highly in a set In events force VS order. momentstendingto sequenced particular, framewilltendtobe orderedas VS chains,as in (2). abuu-ha il-farah w ii-ji (2) baldeen i-siir father-her and then 3M-come 3M-happenDEF-party

(JeddaArabic)

axwaana-ha w yii-j-u and 3-come-PLbrothers-her Then comesthe (wedding)partyand her (bride's)fathercomesin and thenher brothers come'

examplesfromEmiratiArabic,and (See thetextsin Al-Rawi[1990]forfurther and Rockwood[2009]fora further Owens,Dodsworth, "highpoint"function.) These conclusionsaboutAPA werederivedfroma multivariateregression analysisof SV/VS orderusing a corpusofconversationalspeech.The initial variables: analysisincludedeightindependent class ofthesubject: Morpholexical pronoun pronominal noun Transitivity: 'be' intransitive withobject(pronominal ornominal) transitive transitive withoutobject Numberofwordsin clause: up tofour greaterthanfour ofgaal 'say': Subjectin complement occursas subjectofreportedspeechclause doesnotoccuras subjectofreportedspeechclause Dialect: Emirates Kuwait Jedda Same/different subject: same as previoussubject subjectreferent fromprevioussubject referent different subject Subjectoccursin dependentadverbialclause: yes no Followsspeakerswitch: yes no

This content downloaded from 132.180.10.132 on Sun, 8 Dec 2013 15:38:29 PM All use subject to JSTOR Terms and Conditions

Jonathan Owens and RobinDods worth

2009

155

The independent variables were chosen to reflectpotential intralinguistic and pragmatic influences. Relative clauses were excluded fromthe analysis, given that dependencies exist between antecedent and extraction site in the relative clause, and the study was restrictedto predicates with finiteverbs as predicates; hence, no active participles were included. Four ofthe independent variables emerged as significantcorrelates ofword order,as shown in table 2. Factor weights above 0.5 favorSV, while factorweightsbelow 0.5 favorVS. Table 2. Four Variable Correlates ofWord Order in APA Factor

Weight

SV

N(SV/VS)

Transitivity copula intransitive transitive transitivewithoutobject

0.279 0.447 O.65O 0.653

54-1% 73.7% 88.4% 89-9%

105/89 269/96 312/41 62/7

0.910 O.47O

97-9% 75.1%

47/1 7OI/232

O.426 O.615

71.7% 834%

431/170 317/63

O.3OO 0.688 0.669

59.8% 89-7% 90.3%

269/181 70/8 409/44

gaal COMPLEMENT

yes no Numberof words Up to 4 Greater than 4 Morpholexical class Noun Pronominal Pronoun Total

981

NOTE:Applicationvalue = SV; input = 0.827; log likelihood= -433-309; chi-square/cell = I.O6OO.

clause,shouldit variable,the complement Regardingthegaal complement have an overtsubject,is almostinvariablyin SV order.Here, apparently,the ofdirectorreportedspeech,marking ofsubjectsignalsthebeginning expression level a contrasttothenarrator bythematrixverbgaal. represented (3) hatta law gaal

ana a-bya

l-bint

haaÕi

3SG.say 1SG lSG-want DEF-girl this 'Even ifhe says I want this girl'

even if

(JeddaArabic)

One may develop a single account ofthe otherthree variables: all in all, VS clauses are geared towards introducing an unavailable subject into the discourse. Clauses with VS order are less encumberedby words- the clauses they occur in are short;related to this,the predicate itselftends to be intransitiveor a copular verb (kaan, saar), and the referentofthe subject noun itselfis drawn froma potentiallyinfinite,though individuallydelimitable,universe.

This content downloaded from 132.180.10.132 on Sun, 8 Dec 2013 15:38:29 PM All use subject to JSTOR Terms and Conditions

156

51 no. 2

Anthropological Linguistics

For morpholexicalclass, the percentages and N values in table 2 indicate that nouns admit a greater degree ofvariabilityin SV/VS orderthan pronominais and pronouns. Looking more closely at the nominal subjects reveals an internal differentiation.Table 3 shows their division into the four categories of proper noun, definite noun (marked by the definite article il-), possessed noun, and indefinite(nonpossessed) nouns. Indefinitenouns are the only categoryfavoringVS; these are also the "unavailable" nouns in discourse-pragmatic terms. Table 3. Differentiation of Nominal Subjects in APA Subject Type

SV

VS

N (SV/VS)

Proper Definite Possessed Indefinite

81% 65% 68% 36%

19% 35% 32% 64%

22/5 157/84 55/26 38/69

Total

59%

41%

271/I86

ofArabicor Semiticlanguagesas VSO fitsthe Clearlythecharacterization APA findings onlymarginallyat best.VS(O) certainlyoccurs,but it is an order thatcan onlybe understoodwithinthe contextofdetailedmorpholexical and discourse-based factors. We takeup thelargerArabicand Semiticcontextofthe findingsin section6. First,we extendthe analysisbackwardsin chronological timeto anotherSemiticlanguage,BiblicalAramaic. 3. Biblical Aramaic. Thereare twofairlylongportionsoftheHebrewBible writtenin Aramaic(Kittelet al. 1971). Ezra 4.8-6.I8 and 7.12-7.26 contain abouttwelvehundredwords,whileDaniel 2.4-7.28 containabout fourthousand words.The date ofcomposition forthesechaptershas been estimatedto be between500-200 BCE,withEzra usually consideredan oldercomposition (Collins 1993:14~19).3The Aramaicof this era is generallytermed"Official Aramaic"(GermanReichsaramäisch), theAramaicoftheBiblehavingmuchin common withthelanguageemployed forofficial purposesin thePersianEmpire, mostnotablyas attestedin the Aramaicofthe Egyptianpapyribetween500400 BCE(Muraokaand Porten2003; see section6 below).Forpresentpurposes, exactchronology is notimportant; in therangeof2,000-2,500years something lies betweenthevarietiesbeingcompared. 3.1 Comparability of categories. A problemforwhichno singlesolutionis ofgrammaticalfeatures.Arabicand Aramaic possibleis thatofcomparability are structurally close to one anotherand forthegreatmajorityoffeatures very formaland semanticcomponentscoincideto a large degree.The Arabicverbal prefixconjugationcorrespondsto the Aramaic,forinstance,withrespect to all inflectionalclasses. Both varietiesalso mark subject-verbagreement

This content downloaded from 132.180.10.132 on Sun, 8 Dec 2013 15:38:29 PM All use subject to JSTOR Terms and Conditions

2009

JonathanOwensand RobinDodsworth

157

"naturally";that is, regardlessofthe SV or VS order,thereis fullagreement betweensubjectand verbin termsofperson,number,and gender. For instance, At the same time,thereare significant pointsofdifference. finitepredicate becausetheparticiplein Arabicis clearlynota morphologically In BiblicalAramaic,however,theparit was excludedfromconsideration. form, as verbsare. was a passiveparticiple,is inflected ticiple,orwhathistorically The passive participlecan be lookedat fromtwoperspectives.Relativeto an adjective,theAramaicpasAPA,wherethepassiveparticipleis functionally shouldbe excludedfrom and therefore sive participlehas a different function, offinitepredicates,howthe comparison.Relativeto a generalcategorization " to a passiveverb(usuallythe ever,theAramaic"passiveparticiplecorresponds formin ourcorpus,rarelytheinternalpassiveshownbyinternalvocalic infamai but from change)in APA. Fromone perspective,there is noncomparability, anotherthereis. As thereis no singlebest solution,we describethe resultsof bothincludingand excludingrelevantsubcategories. multiplecomparisons, 3.2. Variables in Biblical Aramaic analysis. Our approachis to investigate theSV/VSparameterin BiblicalAramaic(BA) as determined bythe same independentvariablesthat were describedabove forthe corpusof spokenAPA. To this end,the Aramaicportionsofthe HebrewBible wereinitiallycodedas follows: class: Morpholexical pronounorquantifier noun definite noun indefinite Wordsin subjectphrase: up tofour greaterthanfour Transitivity: 'be' intransitive ornominal) withobject(pronominal transitive transitivewithoutobject Book: Daniel Ezra Verbform: perfect imperfect

fromthatused forthe APA studyfora numberofreasons. The list differs to The variables"speakerswitch"and "complement oigaal" are notappropriate thepragmaticsorthesyntaxofBA.4The APA variable"dialect"is analogousto the BA variable"book,"thoughonlyin the sense that each variable encodes discretevarieties,dialectalin the one case, temporalortextualin the other.It

This content downloaded from 132.180.10.132 on Sun, 8 Dec 2013 15:38:29 PM All use subject to JSTOR Terms and Conditions

158

Anthropological Linguistics

51 no. 2

was decided not to encode the "same/differentsubject" variable because the frequentverse switches introducea fixedthematichiatus at frequentpoints in the text; no comparable segmentation is found in the APA texts. Clauses in portionsofthe two books can be famouslylong and involved,so it was decidedto countthe numberofwords in the subject,ratherthan those in the entireclause. The morpholexicalclass variable was also redefined:pronouns and quantifierswere combined in one class because of a very low subject pronoun count (eighttokens),and nouns were brokeninto definiteand indefinitecategories,the latter isolated fromothertypes ofnouns in view oftheir unique favoringofVS order in the APA results (table 3). Definite nouns are those marked by the definitearticle (determined state), possessed nouns, and personal names (see section 3.3 forfurtherdiscussion). Definite nouns, as in (4), were classified according to form,so that even where standard translationshave an indefinitereading,so long as the noun has the definitearticle it is considered definite. (4) minn-ii ta?m-aa yi-tt§am from-medecree-DEF3M-issued 'a decreeis issuedbyme' (Ezra 4-21)5

(BA)

Any noun phrases headed by a quantifier(koll 'air, numeral), such as in (5), were considered quantifiers. (5) wi tlet min qarn-ayy-aa qadmaay-aat-aa et?qar-ew and three fromhorn-PL-DEFfirst-PL-DEF uprooted-PL 'and threeofthefront hornswereuprooted'(Daniel 7.08)

(BA)

Possessed nouns are those in a direct genitive relation, with either a pronoun or noun possessor, as in (6) and (7). (6) w hubad gism-ah and destroyedbody-its. M 'and itsbodywas destroyed'(Daniel 7.11)

(BA)

allaah sm-ayyaa (7) y-qiim 3-establishGod heaven-PL-DEF 'Godoftheheavenswillestablish'(Daniel 2.44)

(BA)

Indefinitenouns with a qualifyingphrase (includingthe genitivediy), such as that in (8), were consideredindefinites. (8) melek l-i§raa?el rab bnaa-hiy king to-Israel great built-it.M 'A greatkingof[lit.,'to'] Israelbuiltit' (Ezra 5.11)

This content downloaded from 132.180.10.132 on Sun, 8 Dec 2013 15:38:29 PM All use subject to JSTOR Terms and Conditions

(BA)

2009

JonathanOwensand RobinDodsworth

159

Whenthesubjectwas an appositivephraselike "Nebuchadnezzarthe king/'it was categorizedaccordingtothefirstnoun,so in thiscase as a propernoun. betweenthe APA and BA morpholexical The correspondences class catein are summarized table 4gories Table 4. Morpholexical Class Categories forAPA and BA APA CODING

BA CODING

noun,possessednoun propername,definite

definite noun

phrase pronoun,quantifier indefinite noun

quantifier indefinite

Finally, the "verb form"variable, which encodes perfectversus imperfect tense-aspect,was added as a new variable. A numberoffundamentalstudies in Semitic word order,such as Givón (1977) on Hebrew and Cook (1986) on BA, have found a dependency between word order and perfector imperfectverb form. In this section, we brieflystate our 3.3. Coding the dependent variable. groundrules forcountingand classifyingSV/VS in BA. Only verbs with overt subjects are counted. When there are sequences of verbs, as in (9), onlythe firstverb is counted as having an overtsubject,so that in (9) onlyone token,the subject ofthe firstverb,is counted. (9) gubr-ayy-aaillek hargis-uw w haskah-uw man-PL-DEFthese came-PL and found-PL 'Thesemencamein and (they)found'(Daniel 6.12)

(BA)

The categorical status of the participial formpaaïil is problematic with regard to establishing equivalences between Arabic and Aramaic. In both languages, the (etymological)active participle functionsparadigmatically within the verbal system(Owens and Yavrumyan [2007] forArabic; Rosenthal [I96I] forAramaic), though to differentdegrees and in differentways. In spoken Arabic, corpus counts show participlesto be less frequentthan in BA. Activeparticiples account for8.8 percent (105/1092) of all predicates with overt subjects in the APA data. In a spoken Nigerian Arabic corpus of about thirty-threethousand words,active participles account for11.5 percent(210/1762) ofpredicates with overt subjects.6Moreover, the passive participle in Arabic has never assumed a paradigmatically verbal function,so there is no formalcomparability between the Arabic passive participle and the (historically) Aramaic passive participle.By contrast,in the Aramaic data the incidenceofthe (historical)participles, active and passive, is much higher- 30.6 percent foractive participles and 2.9 percentforpassive participles. In lightofthese interpretivedifficulties, we have not included participles ofthe formpaa?il in our counts.7

This content downloaded from 132.180.10.132 on Sun, 8 Dec 2013 15:38:29 PM All use subject to JSTOR Terms and Conditions

AnthropologicalLinguistics

i6o

51 no. 2

ForAramaic,Rosenthal(1961:62)considersthepdúil form(theetymological passive participle)to be the regularpassive ofthe basic perfectstem,and, in otherwisefoundontheperfect fact,it doestake suffixes verb,as in (10). (10) w sipr-iinptiih-uw and book-PLopened-PL 'Thebookswereopened'(Daniel7.10)

(BA)

and formally on the basis ofthe suffixforms,thepdîiil passive Functionally, could therefore be considereda verbal predicate.Our analysis includesthe passive participle,but also examinesthe statisticalresultswhen etymological thisformis excluded. On theotherhand,theAramaicformpaafel thatcorresponds to theArabic activeparticiplefaaïil was notincludedin thisanalysisbecausetheAPA study didnotincludethisform.This is a pointwherea closercomparativefunctional studyofpredicatesis necessary.8 Lookingat specificlexicalitems,we considertheverb§iim'be issued' (as in minnii§iimt?am'a decreewas issuedbyme') tobe a regularverb,nota passive participle(see Rosenthal1961:49).The verb'answer',on the otherhand,we have classifiedas an activeparticiplewhenit has theformray(?aaneh),which in factis byfaritsmostfrequentguise(see n. 7).9 The verb?mr'say,speak'was considered comparabletogaal 'say' in Arabic, and therefore was classifiedin termsoftransitivity in thesame way,namely,as a transitiveverb,whoseobjectis theclause encodingtheproposition ofwhatis said orordered. As always,thereare individualinterpretive issues- decisionsthataffect the overallstatistics.Whilethosementionedhere are toofewto affectthe overall statisticaloutcome,it is worthmentioning themin ordernotto lose sightofthe needfordetailedlinguisticanalysisand philologicalinterpretation. In (11),theorthography fortheverbnyi (b?h)'seek',iyi, givesbïi-w,which is hereinterepreted as a full-fledged finiteverb. Daniel w (11) wi Mi-w habruu-hiy and sought-PL Daniel and friends-his 'Daniel and his friendswere sought (forexecution)' (Daniel 2.13)

(BA)

It could conceivablybe read as a defectivespellingforvyi (b?iyw);such a spellingoccurs,forinstance,in (12) (fora different verb),wherethe writing makesclearthata passiveparticiplestemis used. (12) kaarsaaw-aan throne-PL

rmiy-w thrown-PL

'throneswere thrown'(Daniel 7.09)

This content downloaded from 132.180.10.132 on Sun, 8 Dec 2013 15:38:29 PM All use subject to JSTOR Terms and Conditions

(BA)

2009

Jonathan Owens and RobinDodsworth

161

Similarly,we read the verb in (13) as a finiteverb. (13) kpit-uw bound-PL They werebound'(Daniel 3-21)

(BA)

This reading implies an internal passive marked by internal vowel alternation this could also be read as a defective in the basic stem ofBA verbs.Alternatively, spelling ofa passive participle. More generally,overtsubjects withinrelative clauses were not consideredin the original study,given that dependencies exist between the antecedent and the nominal site in the relative clause, and overtsubjects withinrelative clauses similarlyhave not been included in the Biblical Aramaic. Subjects of adverbial and conditionalclauses are included. The followingwere not counted as subjects: vocatives associated with imperativeverbs CO king,live forever'[Daniel 5.10]), and noun clauses introduced by diy, as in ydiy?le-hwa l-malkaa diy (known that-hebe to-king-DEFthat) 'let it be known to the king that' (Ezra 4.12). Imperative verbs associated with vocativeshave no explicitpersonmarker,while the putative noun clause subject is also interpretableas a noun clause associated with an impersonal verb ('it is knownthat'). Finally, in both the APA and BA analyses, onlythe grammatical categories exhibitingSV/VS variation are included. Significantly,in both the APA and BA corporathe categoriesofinterrogativesubject (Arabic minu,BA man 'who') and headless or open relatives (Arabic Uli, BA man diy) are invariably SV (seven tokens in total in BA), and hence both are excluded. However, where BA displays significantrestrictionson variation that do not obtain in APA, the BA invariance has, in this study,been ignored.The major categoryin this regard is the modifyingdemonstrative,as in (14). (14) baadayin gubr-ayy-aaillek kpit-uw man-PL-DEFthese bound-PL then Then thesemenwerebound'(Daniel 3-21)

(BA)

Any subject noun phrase with a modifyingdemonstrativehas SV order in the BA corpus (twenty-twotokens in all, five of which with active participle predicates).By contrast,modifyingdemonstrativesin APA occurin both SV and VS subject phrases.

4. Statistical results and comparison. Severalstatisticalanalyseswererun using GoldvarbX (Sankoff,Tagliamonte,and Smith2005), a tool developed forregressionanalysisofnominaldata. Because theindeamongsociolinguists to correlatewithwordorderin APA and BA are pendentvariableshypothesized notidentical,as notedabove,we presenttheresultsfrommultipleperspectives, of definitions mixturesofindependentvariables,and different withdifferent

This content downloaded from 132.180.10.132 on Sun, 8 Dec 2013 15:38:29 PM All use subject to JSTOR Terms and Conditions

51no. 2

AnthropologicalLinguistics

162

wordcategories(e.g.,withand withoutetymological constituent passive para finding ticiplesin the BA data). The resultsconvergein the same direction, theircredibility. whichin ourviewreinforces The initialmultipleregressionanalysis of subjectand verb orderin BA variableslistedabove:morpholexical includedthefiveindependent class,tranwordsin the subjectphrase,verbform,and book.The bookvariable sitivity, no (Daniel vs. Ezra) was removedwhen it was foundthat Ezra contributed for Ezra out of The low token count tokens. or (47 quantifier pronounsubject 230),ratherthan anymorphosyntactic generalizationaboutits varietyofAramaic,probablyaccountsforthegap. in the class and verbformwerefoundto be significant Onlymorpholexical with these two The results of a one-level only regression stepwiseregression. variablesare shownin table5independent Table 5. Goldvarb Analysis ofBA Corpus Factor Morpholexical class definite indefinite quantifier/pronoun Verbform perfect imperfect Total

Factor weight

SV

N(SV/VS)

O.516 0.206 0.752

71% 38% 88%

110/47 13/24 32/4

0.436 O.648

64% 81%

99/63 56/12 230

NOTE:Input = O.698, chi-square/cell= 0.174

In ordertobettercompare theseresultswiththeAPAregression results (table2), a newregression analysisoftheAPAdatawas run.TheAPAtokens ofgaal werefirstremovedbecausetheyare nearly complements involving In addition, theAPA SV, and thegaal variablewas removed. categorically classvariablewas recodedintotheBA categories ofindefinite, morpholexical thesameindependent andquantifier. variablesthatwere definite, Otherwise, testedin table2 wereincludedin theresultsshownin table6. Theseresults resemble thoseintable2, allowing fornogaal variableandtherecoded closely classvariable: andclauselength class,transitivity, morpholexical morpholexical arethesignificant factors. Fromourcomparative theresultsformorpholexical classare perspective, themoststriking, as theresultsforthisvariablearenearlyidentical acrossthe favorSV, definites are either twodata sets.Forbothlanguages,quantifiers neutral(inthecase ofBA),ormoderately favorVS (in thecase ofAPA),and disfavor SV. Thisis thecase despitethefactthatdefinites indefinites clearly intheAPAcorpus. dominate theBAcorpus, whilequantifiers aremorefrequent in BA,morpholexical classis thestrongest as determined Further, factor, by

This content downloaded from 132.180.10.132 on Sun, 8 Dec 2013 15:38:29 PM All use subject to JSTOR Terms and Conditions

2009

JonathanOwensand RobinDodsworth

163

factorweightfor the "inhibitor" factorweightfromthe "promoter" subtracting measures(see Horvathand thetwosignificant variables,yieldingtwo"contrast" Horvath2003). For morpholexical class, the contrastis 0.752 - .206 = 0.546, a scorethatHorvathand Horvathconsider"verystrong"(2003:149),thehighest level ofcorrelationin theirscale. This is much strongerthan the verbform fromtheBA corpus contrast: O.648- 0.436= 0.212.Removing passiveparticiples same section and the (see 3.1) analysisyieldsthesame signiregression running the class contrast:0.795 ficantfactorsand further magnifies morpholexical = class (0.677 O.I5O= 0.527) showsa For APA,morpholexical O.I5I O.644.10 and clause lengthand is also considered strongercontrastthan transitivity set Horvath and Horvath. the criteria by "verystrong"by Table 6. Goldvarb Analysis of APA with Same Independent Variables as forTable 2 butwithoutGoal Tokensand withRecoded MorpholexicalClass Factor Transitivity copula intransitive transitive withoutobject transitive ofwords Number up tofour greaterthanfour class Morpholexical definite indefinite quantifier Total

Factor weight

SV

N(SV/VS)

0.290 0.454 0.649 0.625

53% 73% 88% 89%

101/89 256/95 287/41 57/7

O.423 0.620

70% 83%

399/170 302/62

O.36I O.I5O 0.677

65% 36% 90%

219/116 36/65 446/51 933

= O.9247 NOTE:Input= O.8O6,chi-square/cell Finally, and crucially,in order to test forsignificantdifferencesin SV/VS orderingbetween the two languages, the BA and APA data were combinedfora single regression in which the independent variables were the same as above, with the addition of language as an independent variable (BA vs. APA). Only morpholexicalclass was selected as significant(input = O.76I), with the same pattern as was found in the previous analyses: quantifiers favor SV (factor weight = O.6OO),while definitenouns very slightlyfavorVS (0.429), and indefinitenouns stronglyfavorVS (O.I46). The model foundno significantdifference between APA and BA. Given the analytical categories as input as described in this and the previous two sections, and fromthe statistical perspective developed here, the constraintsgoverningVS/SV word order in BA and APA are basically identical. These results indicate that morpholexical class is strongly correlatedwiththe orderofverb and subject in both contemporaryAPA and BA, and that the relationshipbetween word orderand morpholexicalclass shows the

This content downloaded from 132.180.10.132 on Sun, 8 Dec 2013 15:38:29 PM All use subject to JSTOR Terms and Conditions

164

Anthropological Linguistics

51 no. 2

same pattern in both languages. Extrapolating fromour broader discoursebased analysis (Owens, Dodsworth,and Rockwood2009; see section5 below),we proposethat underlyingthe morpholexicalpatternis a shared set ofbasic pragmatic constraints: available subjects, most prominentlypronouns and quantifiers,migrate towards SV order, while less available, new subjects, prototypicallyindefinitesubjects, tend towards VS. In both data sets, definiteand possessed nouns populate a middle ground. Before moving on to a discussion of the pragmatic constraints, some explanation concerningthe other independent variables is in order. Regarding numberofwords,the results are not completelycomparable, as the APA analysis counts words per clause, whereas the BA analysis counts words per subject noun phrase. Nonetheless, in the BA corpus, 70 percent of VS clauses have a single element in the subject noun phrase, as opposed to only 52 percentin the SV clauses. Furthermore,SV subjectshave a largerrange, containingup to fourteen words. So even here a tendencycan be discerned wherebySV is associated with heavier constituents,VS with lighter,as in the APA corpus. However, BA and APA differin terms of the relationship between word order and transitivity.Althoughthe transitivityvariable is not significantfor BA, the verb 'be' (DID (hwh)) occurs in SV order in eleven out oftwelve tokens, whereas in APA the verb 'be' (kaan/saar) favorsVS. Regarding the verb form variable that is mildlysignificantin the BA data, imperfectfavoringSV, perfect VS, in a number of studies on Semitic languages, a linkage has been drawn between perfectand VS order,on the one hand, and imperfectand SV, on the other. This has been argued forin Hebrew (Givón 1977; Fox 1983), BA (Buth 1987), and various Arabic dialects (Dahlgren 1998) on the basis ofan analysis of published texts. In addition, Cook (I986) draws a correlationbetween perfect and VO order and between imperfectand OV order. Holes (2009), on the other hand, notes that VS with imperfectverbs is common in Bahraini Arabic. In termsofpercentages,both perfectand imperfectin BA have dominantS V order. In Semitic languages, this variation has been explained by the discourse propertyof foregrounding,that is, highlightingwhat a story is about. Foregroundedeventsfollowa sequential timelinein a narrative,and, it is argued,VS and VO word order is chosen to representthis sequentiality. Given that narratives are oftenabout events that happened in the past, VS/VO associates with the perfect. 5. Event-based determinants. Again, takingthe lead fromour originalAPA study, we propose that SV/VS order in BA is sensitive to larger discourse segments,a crucial one being event-based frames(Myhill 1992; Buth 1987:140; Holes 2009 for Bahraini Arabic). Descriptions and narratives that sequence events and procedureswill favorVS. The VS ordersignals the introductionofan essential, event-relevantreferent.

This content downloaded from 132.180.10.132 on Sun, 8 Dec 2013 15:38:29 PM All use subject to JSTOR Terms and Conditions

2009

JonathanOwensand RobinDodsworth

165

Abouthalfofall theBA VS tokens(37/75)occurin narrativecontexts,and seventokensoccurin the theseare typicallymarkedbyVS chains.The thirtypassages. following Daniel2.34,2.35,2.35(interpretation byDanielofa dreamsequence); recounted Daniel4.O8,11 4.09,4.09,4.O9,4.O9(dreamsequence byNebuchadnezzar); Daniel5-24,12 5-30 5.26,5.28,5.29, (interpretation by Daniel ofdreamto King Belshazzar); Daniel6.I8,6.I8,6.I8,6.I&136.21,6.21,then6.24,14 6.25,6.25:Daniel'simprisonthe nextday; mentin the lion'sden;his reunionwithKingNebuchadnezzar of the againstDaniel; punishment conspirators Daniel7.11,7.II,157.12,7.I4,16 7.15(dreamsequencerecounted byDaniel). Daniel7.20,7.22,17 7.22(dreamsequencerecounted byDaniel). In addition,thereis a set ofthreeVS verbsin Daniel 2.44>2.45,2.45 (interpretationbyDaniel ofa dreamsequence). In Ezra, thereis one prominentpassage whereVS dominates:5-12,5.14> thereturnoftherichestoJerusalemfromBabylon. 5.I6, 5.I7, recounting In mostinstances,the gaps in VS tokenswithintheseevent-basedsequences are due to therebeingno predicateat all, a nonverbalor activeparticiple (as in Daniel 5.25,7.13),or a verbalsubjectwith'all' or a predicateintervening SV in thenounphrasesubject.In a few,thereis an intervening demonstrative on thatappearswherea VS mightbe expected.These cases are commented in thefootnotes. Withoutgoingintofurther APA,VS stylisticdetail,just as in contemporary set of events related a of discourse closely clusteringsignalsa particulartype in BA.18 function orprocedures so,too,doesVS have thisas itsprominent to the There are considerableramifications 6. Comparative discussion. comparativeconclusionsderivedfromthis study,and theyneed to be contextualizedin a numberof scholarlytraditions.However,in this articlewe on threepoints:a briefsurveyofotherearlyAramaiccorpora,a concentrate discussionofotherstudieson wordordervariationin BA, and a generalconsiderationofthe implicationsof our studyforArabicand Semitichistorical linguistics. 6.1. Aramaic. Lookingat otherexamplesfromthe OfficialAramaicera or older,the Ahiqarnarrativefromthe fifthcenturyBCE,givenin fullby Buth (1987:412-26),has bothVS and SV structures.The initialnarrativesequence (lines 1-12) is VS, all instancesofS beingfullnounsor personalnames.Beginningat line 13,SV beginsto set in,withS beingeitherthefirstpersonsingular ofovert pronounor an openrelative,'whoever'.A nearlyidenticaldistribution from 8OO Zakkur in older the is attested BCE, (about Zenjirli epigraph subjects the in VS name with in south-central order, pronoun personal subjects Turkey), T and openrelativesin SV (Buth 1987:350-62),and the same applies to the

This content downloaded from 132.180.10.132 on Sun, 8 Dec 2013 15:38:29 PM All use subject to JSTOR Terms and Conditions

166

Anthropological Linguistics

51 no. 2

Bar Raakib inscription(ca. 730 BCE, fromZenjirli). In these three texts, open relatives (three tokens) and pronouns (either T or 'he') are always SV (eight tokens), while definite,indefinite,possessed, and proper nouns all are nearly always VS. Buth (1987) attributesthe consistentVS orderto a narrative structure with,in his terms, S ofVS signaling a foregroundedsubject. The morpholexical distributionof these word orders mirrorsour own, with pronouns and open relatives (see section 3-5) in SV order, and nominais in VS order in a narrative context. Analyzing the corpus of Egyptian Aramaic fromthe fifthcenturyBCE, a varietyprobablycontemporarywith the Aramaic ofEzra, Folmer (1995:524-34) gives examples with headless relatives always in SV orderand pronounsalmost always in SV order. There is thereforeevidence that all varieties up to and contemporarywith BA have a fundamental SV/VS dichotomybased on the category of morpholexical class. Pre-OfficialAramaic (Old Aramaic) is too sparsely populated, both in terms of absolute corpus size and in terms ofgenre variability,to draw more direct inferences.What it does reveal, however, is compatible with the interpretationsofferedhere forBA. 6.2. Other interpretations. Besides the traditional view ofAramaic, along with otherClassical Semitic languages, as being a VSO language, quoted at the beginning of this article, three other interpretationsof word order in BA or varieties older or contemporarywith BA may be brieflymentioned. In their extensive study ofthe Egyptian papyri,Muraoka and Porten point out, "However, in view of a substantial number of cases in which the verb occupies a non-initialposition,it is more accurate to speak offreeword-orderin our idiom" (2003:299). That is, they findso many instances of non-VSO structures, ofvarious types,that theyresortto freeword orderas an explanation. Rosenthal similarlyobserves that forBA "the position ofwords in a verbal sentence is free and does not follow any hard and fast rules" (1961:56). He speaks of a "more ancient" VSO sequence that occurs after particles such as 'then' and 'now' (see Buth [1987:190-94] for a more adequate discussion of these), but nowhere adduces evidence of where the ideal ancient prototype exists, outside ofreconstructedProto-Semitic. Our own data and interpretationobviouslycontradictthe "freeword order" as much as it does the VSO view, but it is interestingthat leading Aramaicists themselvesfindit impossibleto apply a classical VSO reading to a large body of early Aramaic texts. Taking a verydifferentapproach, Buth (1987) devotes a large and detailed studyto interpretingword orderfroma functionalperspective,using the grammatical approach ofthe late Simon Dik. He is much aware that both SV and VS ordersoccur,and that SV is a statisticallydominantpattern.Nonetheless,Buth concludes that BA is fundamentallyVSO (1987:230). In his analysis, instances

This content downloaded from 132.180.10.132 on Sun, 8 Dec 2013 15:38:29 PM All use subject to JSTOR Terms and Conditions

2009

JonathanOwensand RobinDodsworth

167

ofSV are derivedvia a rulethatshiftsa VS subjectto a prefieldposition(Px), includyieldingthe structurePXV(Px= S), undercertainpragmaticconditions, ingtopicalizingand questioning(1987:90,94, 173)It is hardto compareourownapproachwithButh'sfora numberofreasons. First,Buth categorizesthe entireBA clause in functionalterms,whereaswe more modestlylimitour inquiryto the relationbetween subject and verb. on whathe terms"narrativepassages" that Second,Buth'sstudyconcentrates containa subsetofall SV/VS clauses. Finally,thereis, as always,a general different analyticaland theoreticalframeworks. problemofcomparing A basic criticismthat we have ofButh's approachis that what we have accountedforby a numberof definedas SV vs. VS orderis, in his framework, disjointprinciples.These include: • A clause centeris theverb. • Languages tend to be either "prefield"languages, resultingin SV, or postfield, resultingin VS (BA is postfield). • Thereexistsan initialposition,the P1?intowhichitemscan be movedfor varioussemanticand pragmaticreasons[1987:90]. observedVS and SV structures theserulesare adequatein generating Certainly ifit is that,ofdoingso in termsof in BA, and theydo have theaddedattraction, generalprincipleswhichdo morethanexplaintheissue at hand.However,this comesin ourtermsat theexpenseofhavingdifferent explanationsforpreverbal is thatthereis no SV or to thecontrary, and postverbal subjects.Our argument, and discourse-based VS orderas such,but rathera complexofmorpholexical NeitherAPA nor tendto favorone orderortheother.19 thattogether constraints wordorderfallsoutfromthe BA is basicallySV orVS. Rather,theirsubject-verb wider discoursecontextofthe and the noun of the nature subject morpholexical excellence. word It is a lexico-discourse-based order,par text.20 6.3. The historical interpretation. Withthese pointsbehindus, we now that set the stage forthe article. cometo the historicallinguisticframework the lexico-discourse-based analysisofthis studyis its situationin Underlying Semiticlanguage history.Given our claim that SV/VS word orderin both an obvious APA and in BA is governedbythe same constraints, contemporary historical what is their further origin. questionis One answer,a typeoftengivenin Semiticlinguistics,as Retsö(2000:112) The has pointedout,is thattheyarose via parallelindependentdevelopment. basis ofsuchan assumptionis moreoftenthannot,nota well-arguedoutlineof butrathersimplythe theconditions development, parallelindependent favoring assumptionthat once the individual Semitic languages arose, they each in this perspective, by definition developedin theirownway. Commonalities, mustbe due to independent development. However,the likelihoodof such a preciselyparallel set of constraintsin languageswhoseattestationis some2,500yearsapartis minuscule.Certainly

This content downloaded from 132.180.10.132 on Sun, 8 Dec 2013 15:38:29 PM All use subject to JSTOR Terms and Conditions

Anthropological Linguistics

168

51 no. 2

there is some evidence that morpholexical class and word order do have commonalities across completelyunrelated languages. As pointed out by Owens, Dodsworth,and Rockwood(2009), forexample, whereas languages like Klamath and Nez Perce have pragmaticallycontrolledword order,certain morpholexical subclasses in them,pronouns in Nez Perce, referentialindefinitesin Klamath, invariablyor dominantlyhave S V order. However, in the present case we are not looking at the chance identityof certain classes ofitems in two data sets, but rather at an identical morpholexical mold characterizing each. Added to this is a comparable discourse-based conditioning.It is very difficultto imagine two languages as closely related as Aramaic and Arabic21developingan identical set ofconstraintsindependentlyof each other. Note in passing that this identityof constraints is all the more striking consideringthe well-knownextremitiesofword ordervariationfoundin BA. For Daniel alone, Cook (1986:4) documents that of the six logically possible word orders of SVO (subject and object both nominal), all six, in fact,do occur, SVO, SOV, OVS, etc. As already noted, this variability is due to a combination of stylisticchoice and influencefromthe adstratal SOV languages Akkadian and Persian. Whatever the explanation forthis overall variation, however, in our view, the choice of SV and VS reduces to the basic factorsthat we have defined above. Moreover,fromour perspective the continuationof this morpholexicaldiscourse based model into contemporaryAPA provides evidence that it is deeply embedded in the language communities of the region, and that it has been forat least 2,500 years. Eliminating parallel development leaves two furtherpossibilities. Both of them have ultimatelyto be integratedinto the originofword order in Classical Arabic, so we will firstbrieflysketchthis varietyofArabic. Classical Arabic is, in the tradition of Arabic grammarians (Owens 2000), said to have two word orders,eitherVS (traditionallytermed"verbal sentence") as in (15), or Topic-Comment,as in (16a) and (16b). (15) jaa?a ir-rijaal-u came DEF-men-NOM

(VS)

'The mencame.' (16a) ir-rijaal-u jaa?-uw DEF-men-NOM came-M.PL 'The mencame.'

(Topic-Comment)

ra?ay-tu-hum (16b) ir-rijaal-u DEF-men-NOM saw-I-them 'As forthemen,I saw them.'

(Topic-Comment)

In the VS sentence (15), the verb is invariably sentence-initialand always singular, irrespective of the number of the subject. In the Topic-Comment

This content downloaded from 132.180.10.132 on Sun, 8 Dec 2013 15:38:29 PM All use subject to JSTOR Terms and Conditions

2009

JonathanOwensand RobinDodsworth

169

sentences(16a) and (16b),the Topicelementoccursfirst,the Commentafterelementin the Comment.In (16a), wards.The Topicmustfinda coreferential the coreferential elementis said to be the masculineplural suffix-uw on the verb;in (16b),it is theresumptive objectpronoun-hum. ofthebasic characterization Thus,lookingbeyondtheArabicgrammarians' whichis usuallythe one assumedin moderngrammarsof sentencestructure, StandardArabic,it can be said thatClassicalArabichas bothSV and VS order. The latteris consideredthe unmarkedstructure,while SV occurswhen the Topicand Subjectare the "same" noun,as in (16a). However,the grammarof fromthatobservedin our data, Classical ArabicSV and VS ordersis different in singularform(comparesingular sincewhentheverbis initial,it is invariably formin (I5) withpluralin (16a)). can in mind,thetwofurther historicalinterpretations Withthisbackground wordorderof be described.On theonehand,it couldbe thatthelexico-discourse APA and BA are reflexesofthe same wordorderin the protolanguage.On the otherhand, it could also be that the ancestralvariantofAPA witnessedthe wordorderfromAramaic.This is plausiofthelexico-discourse-based diffusion orderand also giventhe hisble giventhe linguisticidentityin subject-verb toricalrealitythatAramaeansand Arabsare knownto have been in close and contactsinceat least thetenthcenturyBCE(Lipiñski2000:101, uninterrupted Retsö 2003: 352). 495; If it is assumed that the Classical Arabicpatternis the originalone for explanationis the mostlikelyone for Arabic,thenobviouslythe diffusionist APA.However,ClassicalArabicis aloneamongthe Semiticlanguagesin having sentencestructureslike (15). All otherSemiticlanguages,including,as seen regardlessoftheword agreement here,Aramaicand APA,havefullverb-subject orderpatternassumed.It mightthusbe thatClassicalArabicis itselfinnovative amongthe Semiticlanguagesin thisrespect.Ifone tookthisposition,the case forattributingthe APA SV/VS word orderto the protolanguagewould be strengthened. Thereis no needto resolvethisissue here,norcan it obviouslybe resolved withinthecontextofthe data presentedhere.A keypointfromeithersolution thatdoes emerge,however,is theremarkablestabilityofa wordorderpattern acrossat least 2,500years.Whereashistoricallinguistsin general,and Semion language change,22 ticistsand Arabicistsin particular,have concentrated instances of almost remarkable thereare equally perfectstabilityacrosslong eras. chronological of the It appears that the clearlypartitionedfunctionaldifferentiation to its has contributed class to current stability. patternaccording morpholexical At the statisticalextremes,quantifiersand pronounsfallintoSV order,indeassociated finitenounsintoVS. The ordersSV and VS thus are prototypically withclosedlexicalclasses,on the one hand,and nounswitha particularmorphologicalmarking(or lack thereof),on the other.In the case of APA, the

This content downloaded from 132.180.10.132 on Sun, 8 Dec 2013 15:38:29 PM All use subject to JSTOR Terms and Conditions

170

Anthropological Linguistics

51 no. 2

semanticnature ofdefinitenouns furthercomes into play, with nouns ofgeneral All in all, meaning,such as 'people', patterningwith quantifiersand pronouns.23 the S of SV has a higher degree of availability, and its precise identityis less important,than the S of VS. The properties of each are leveraged into a discourse complementarity:narratives achieve compactness in part by movingthe storyalong with the introductionofspecificreferentsin VS order.24 To conclude, we emphasize that the pattern of lexico-discourseword order described here was identifiedfroma corpus of contemporaryspoken Arabic, fromArabic dialects. Whereas these are traditionallyviewed as younger offspring of Classical Arabic or of a formof Arabic contemporarywith Classical Arabic,the data and analysis presented here argues furtherforthe position of Owens (2009), namely,that contemporaryspoken Arabic is a reservoirofstructures and linguisticpractice that affordinsights deep into the historicalpast of Arabic,into an era older than Classical Arabic,and, as seen here in conjunction with furtherdata, intothe past ofSemitic languages in general. Notes APA = ArabianPeninsularArabic;BA = BiblicalAramaic;DEF = Abbreviations. definite;F = feminine;M = masculine;NOM= nominative;O = object;PL = plural; S = subject;SG = singular;SS/DS= same subject/different subject;V = verb.Wordorder = are abbreviated in the usual fashion: VS patterns order,SVO = subjectverb-subject = VSO and so on. order, verb-object order, verb-subject-object In transliterations of Biblical Aramaic,§ is used for'y. For the Transcriptions. sometimesambiguousAramaicvocalizationwe followRosenthal(1961:11), though transcribe both..and ..as /e/.Rosenthaldistinguishes theseas closed(longorshort)and and short We do notmarkpostvocalicspirantization in theAramaic open (respectively). ofArabic,D is used fora voiced,interdentalemphatic examples.In transliterations fricative. 1. Interestingly, Brockelmann thatit is (1982:170)is morecircumspect, suggesting "doubtful"(fraglich)whetherProtoSemitichad a fixedsubjectwordorderin verbal sentences.Garr(1985:191),whofollowsthe standingviewthatProto-Semitic was VS, observations on nominal,not supportsthisposition(n. 172) in partwithBrockelmann's verbalsentences.It can be noted,however,thatin a slightlylaterwork,Brockelmann withoutassociating speaksofa "strictly regulatedwordorder"(strengundfestgeregelt), thiswithanyspecificgrammatical categories(1916:20). AboutGafaz,Gragg(1997:255)saysthatthebasicwordorderis VSO, butthatother ordersare attestedas well. No characterization ofbasic wordorderis offered. Givón (1977) arguesthat BiblicalHebrewmovedfromVSO to SVO fromits earlierto later period. A further encountered motifin thedescription ofsubjectverbwordorder frequently in Semiticis thatifSV shouldbe encountered, it is because the subjectis givengreater emphasis(e.g., Garr 1985:189;Gragg 1997:255;the quotationfromKaufmann1997 above).This explanationis rarelyfollowedup withcorpusstudy,however,and in most cases itfollowsa different tackto ourfindings in regardsto BiblicalAramaic explanatory and ArabianPeninsularArabic,whereSV ordermarksgreaterreferential of availability thesubject. 2. We use thetermArabianPeninsularArabicas a heuristiconly,leavingopenthe questionwhetherthepeninsulashouldbe considereda singledialectarea.

This content downloaded from 132.180.10.132 on Sun, 8 Dec 2013 15:38:29 PM All use subject to JSTOR Terms and Conditions

2009

Jonathan Owens and RobinDodsworth

171

3. On linguisticgrounds,forinstance,Ezra has theoriginalthirdpersonmasculine -hum,whichin Daniel and in mostsubsequentvarietiesofAramaicchanges objectsuffix to-hoon(Grabbe1998:131). as complement of'say' one couldencodea clausal complement ofthe 4. Conceivably, verbs?nhand ?mr. ofthispassage: "eineAnweisung 5. We followBückersinterpretation gegebenwurde was issued]"(1953:45),withan indefinite [a directive subjectreading. different numbersofovert 6. The NigerianArabicand APA corporahave strikingly the of active relative to finiteverbsis On the other hand, participles proportion subjects. comparable. we termtheseparticiples,bymorphological 7. Conventionally analogyto the comon functional maybe appropriate grounds,but parableArabicforms.Otherterminology thisis notan essentialpointin thisarticle. 8. A brieflookat therawstatisticspresentedin tablesA and B suggeststhatrelative are much theAramaicactiveand passiveparticiples totheSV/VSwordorderparameter, morelikeverbsthanis theactiveparticiplein spokenArabic. Table A. Aramaic (Daniel) Activeparticiple Passive participle Verb

sv

vs

49 10 II6

37 8 48

Table B. Arabian Peninsular Arabic Activeparticiple

SV 95

VS 10

ofthe active However,thefiguresforAramaicare qualifiedbythefactthattwenty-nine tokensaccrueto oneform,ïneh 'answering',all in VS order.The potentialfor participle of 'answeredX' = 'X answered',to skewstatisticaltreatments thisonefixedexpression, BA has beennotedbyresearchers(Cook 1986:3;ButhI987). VS. He 9. Buth(1987:128)notesthatïaaneh 'say' withan overtsubjectis invariably also considersthisformto be a finiteverb,not a participle,despiteits spellingin the wouldappear HebrewBible(1987:141;see also,e.g.,Steuer2003:435).HereBergsträßer as Buth(see, e.g.,1977:74)tohavethesame interpretation As Buthnotes,in the voweledHebrewBible version,a literalreadingofthethird personmasculinesingularformgivesan activeparticipleform.He arguesfora basic verbalreading,notingthatthepluraland feminine singularforms(onetokeneach) have a verbalvoweling(in Daniel 3.16 and 5.10).He explainstheattestedvowelingas due to a collocates verb ?mr,whichfn/ifrequently later analogicalinfluencefroma following with,and whichis in activeparticipleform. thetwo"verbal"tokensof'answer' a literalistvowelinginterpretation, In following thatoccurin the corpusare classifiedamongthe verbsand includedin the statistics, whereasthemasculinesingularreflexesare classifiedas activeparticiplesand are not includedin the analysis. While Buth's interpretation may well be correct,thereis as an activeparticiple thata priorirulesoutthelexemefor'answer'as occurring nothing forms.In anycase,itwouldbe an in a masculinesingularformand as a verbin inflected of'say' as a finiteverb. easymattertorerunthestatisticsunderan interpretation

This content downloaded from 132.180.10.132 on Sun, 8 Dec 2013 15:38:29 PM All use subject to JSTOR Terms and Conditions

172

Anthropological Linguistics

51 no. 2

doeseliminatethepotentially thecurrenttreatment skewingeffect By coincidence, ofincludinga highlyfrequent lexicalitemthatinvariablyoccursin VS format. 10. A regressionwiththe same independentvariables,but runoveronlythe data as significant, with fromtheBookofDaniel,selectedmorpholexical class and transitivity class class havingthehighercontrastscore.The factthatmorpholexical morpholexical in theDaniel subset,whenverbformwas tradedfortransitivity, remainedsignificant underlinesthedominantsignificance ofmorpholexical class. withno obviousreasonforSV rather 11. An SV clause 'its heightwent'intervenes, thanVS. 12.A nominalsentence'thiswas thewriting'intervenesin 5.25. 13. In 6.19, 'his sleep fled fromhim' has SV order,and in 6.20, there is an SV clause 'the kingarose'. Giventhatthesephrasesfitintothenarrative intervening VS sequence, mightbe expectedin bothcases. PerhapstheSV setsofftheendofthefirst in the lion's den fromthe secondpart,wherethe king partofDaniel's imprisonment hurriesoffthenextdayto see whetherhe survived. 'all' as subject(kellhbellaa 'all 14. An SV clause intervenes here,witha quantifier not was found'). injury 15. Thereis an SV clause in 7.11 containingthe quantifier'all'- 'all the peoples, nationsand languageswillserve'. 16.An SV clause 'spiritsofmyhead disturbedme' comesin 7.15. 'that horn'.Demonstratives, as 17. In 7.21 the subjectcontainsa demonstrative notedin section3-3,invariablyfallintoSV orderin BA. 18. Additionally, examplescouldbe citedwhereVS appearsto usherin a dramatic as in Daniel 5.24,wherea handsuddenlyemergesbeforeBelshazzarbearing highpoint, in writing thenewsthathe has losthis kingdom. effect, Definingthedramatichighpoint however,requiresmoreanalysisinbothcorpora. 19. This formulation is, in fact,compatiblewithButh's study.Whilearguingthat Aramaicis basicallyVS, he concentrates entirelyon narrativepassages.Thesetend,in ourinterpretation as well,towardsVS order.Fromourperspective, "narration" however, involvesinteralia choosingthosemorpholexical classes thatlend themselvestowards SV/VSmanipulation, i.e.,nouns(definite, indefinite, possessed,proper).In neglecting the basic parameterof morpholexicalclass, Buth (perhaps along withmany other Semiticists) undulyprivilegesonegenrein thesearchfora basic wordorder. 20. Our approachdovetailswiththe typological observationsofDryer(1991) and Tallerman(1998),whoarguethatthereis no prominent set ofproperties thataccruesto a VSO languagetype.TallermanthusarguesagainstVSO as a discretetypological class ofwordorder. 21. Aramaicand Arabicbycommonconsensusbelongto theWestSemiticbranchof Semitic.OpiniondivideswhetherArabicshouldbe further subclassified as SouthSemitic,alongwithSouthArabianand EthiopianSemitic,or as CentralSemitic,alongwith Hebrew,Aramaic,and Ugaritic(see Faber [1997]forgeneraldiscussion). 22. Note,forinstance,the near obsessionofArabicistsand Semiticistswiththe "old"vs. "new"(neu vs. alt in the Germantradition):Old Arabicand Neodichotomy Arabic,Old Aramaicand Neo-Aramaic, EpigraphicSouthArabianand ModernSouth Arabian.Historicallinguisticcategorieshere appear to be based on chronology, not comparative linguistics. In manyparticularcases, therecertainlymaybe historicallinguisticreasonsfor supposingthatone linguisticvarietyis innovative("neo-")relativeto a anothervariety, but historicallinguisticsas usuallyunderstoodcertainlydoes not make it a general principlethat a chronologically youngervarietymust be innovativerelative to a oldervariety. chronologically

This content downloaded from 132.180.10.132 on Sun, 8 Dec 2013 15:38:29 PM All use subject to JSTOR Terms and Conditions

2009

Jonathan Owens and RobinDodsworth

173

and Rockwood(2009). The corpusofBA is too 23. DiscussedbyOwens,Dodsworth, thisparameter. smallto investigate 24. It followsfromthe analysispresentedhere that the assumptionthat ProtoSemiticfollowedVSO order,as describedin section1, needs closerassessment.It does of notapplyto BA, norprobablyto anystageofAramaic,and underthe interpretation notto all varietiesof APA subjectand verborderas goingback to the protolanguage, Arabiceither. References Al-Rawi,Rosina-Fawzia und -lexik, StudienzumarabischenDialektvonAbuDaby:Nominalbildung 1990 Texte.Heidelberg:J. Groos. Gotthelf Bergsträßer, und grammatische in die semitischen 1977 Sprachen:Sprachproben Einführung Skizzen[1928].Darmstadt:Wissenschaftliche Buchgesellschaft. Carl Brockelmann, Berlin:G. J. Gösching. SemitischeSprachwissenschaft. 1916 Grammatikdersemitischen Grundrissdervergleichenden Sprachen.Vol. 1982 2: Syntax[1913].Hildesheim:Georg01msVerlag. Bückers,Hermann Vol. 4, Part 2: Die BücherEsdras, Nehemias, HerdersBibelkommentar. 1953 Tobias,JudithundEsther.Freiburg:Herder. Buth,John ofFunctionalGrammarand WordOrderin AramaicfromthePerspectives 1987 ofCalifornia,Los Angeles. DiscourseAnalysis.Ph.D. diss.,University Collins,JohnJoseph on theBookofDaniel.Minneapolis:FortressPress. Daniel: A Commentary 1993 Edward Cook, WordOrderin theAramaicofDaniel. Afroasiatic Linguistics9:111-26. 1986 Dahlgren,Sven-Olof ActaUniversitatis WordOrderin Arabic.Göteborg: Gothoburgensis. 1998 Matthew Dryer, JournalofLinguistics27:443SVO Languagesand the OV:VO Typology. 1991 82. Edwards,Malcolm StrucWordOrderin EgyptianArabic:Formand Function.In Information 2009 turein SpokenArabic,editedbyJonathanOwens and Alaa Elgibali,93106.London:Routledge. Faber,Alice ofthe SemiticLanguages.In The SemiticLanguages, GeneticSubgrouping 1997 editedbyRobertHetzron,3-15. London,Routledge. Folmer,Margaretha The AramaicLanguagein theAchaemenidPeriod.OrientaliaLovaniensia 1995 Analecta68. Leuven:Peeters. Andrew Fox, 1983 Topic Continuityin Biblical HebrewNarrative.In Topic Continuityin Discourse:A QuantitativeCross-LanguageStudy,editedbyTalmyGivón, JohnBenjamins. 215-54.Amsterdam: Randall W. Garr, Dialect GeographyofSyria-Palestine, IOOO-586B.C.E.Philadelphia:Uni1985 versityofPennsylvaniaPress.

This content downloaded from 132.180.10.132 on Sun, 8 Dec 2013 15:38:29 PM All use subject to JSTOR Terms and Conditions

174

Anthropological Linguistics

51 no. 2

Givón,Talmy The DriftfromVSO to SVO in BiblicalHebrew:The PragmaticsofTense1977 Aspect.In MechanismsofSyntacticChange,editedbyCharlesN. Li, 181ofTexas Press. 254. Austin:University Grabbe,LesterL. Ezra-Nehemiah.London:Routledge. 1998 Gragg,Gene G8Î8Z. In The SemiticLanguages, edited by RobertHetzron,242-62. 1997 London,Routledge. Horvath,BarbaraM., and RonaldJ.Horvath A Closer Look at the ConstraintHierarchy:Order,Contrast,and Geo2003 graphicalScale. LanguageVariationand Change15(2):143-70. Clive Holes, WordOrderand TextualFunctionin GulfArabic.In Information Structure 2009 in SpokenArabic,editedby JonathanOwens and Alaa Elgibali,6I-74. London:Routledge. Kaufman,Stephen Aramaic.In The SemiticLanguages,editedby RobertHetzron,II4-30. 1997 London,Routledge. eds. Kittel,Rudolf,Paul Kahle,Albrecht Alt,OttoEissfeldt,and WalterBaumgartner, BiblicaHebraica.I6th ed. Stuttgart: Bibelanstalt. 1971 Württembergische Lipiñski,Edward 2000 The Aramaeans: Their AncientHistory,Culture,Religion.Orientalia Lovaniensiaanalecta100.Leuven:Peeters. Muraoka,Takamitsu,and Bezalel Porten A GrammarofEgyptianAramaic.2d rev.ed. Leiden:Brill. 2003 Myhill,John WordOrderand TemporalSequencing.In PragmaticsofWordOrderFlexi1992 bility,editedbyDorisPayne,265-78. Amsterdam: Benjamins. and SebastiãoVotre Naro,Anthony, DiscourseMotivationsforLinguisticRegularities:Verb/Subject Orderin 1999 SpokenBrazilianPortuguese.Probus11:75-100. Owens,Jonathan 2000 The StructureofArabicGrammaticalTheory.In HistoryoftheLanguage Sciences: An InternationalHandbookon the Evolutionof the Studyof Languagefromthe Beginningsto the Present,editedby SylvainAuroux, 286-300.Berlin:Walterde Gruyter. A LinguisticHistoryofArabic.Oxford:OxfordUniversity Press. 2009 Owens,Jonathan,and MaratYavrumyan The Participle.In Encyclopedia ofArabicLanguageand Linguistics, edited 2007 byC. H. M. Versteegh, 541-46. Leiden:Brill. and TrentRockwood Owens,Jonathan,RobinDodsworth, Verb Order in Spoken Arabic:Morpholexicaland Event-Based 2009 SubjectFactors.LanguageVariationand Change21:39-67. Prince,Ellen F. Towarda Taxonomyof Given-NewInformation. In Radical Pragmatics, 198I editedbyPeterCole,223-55. NewYork:AcademicPress. Retsö,Jan 2000 ofAncientArabia.Oriente Kaskasa, T-Passivesand theDialectGeography Moderno,n.s.,19:111-18. The Arabsin Antiquity: TheirHistoryfromtheAssyrians totheUmayyads. 2003 London:RoutledgeCurzon.

This content downloaded from 132.180.10.132 on Sun, 8 Dec 2013 15:38:29 PM All use subject to JSTOR Terms and Conditions

2009

Jonathan Owens and RobinDodsworth

175

Rosenthal,Franz A GrammarofBiblicalAramaic.Wiesbaden:O. Harrassowitz. I96I Sankoff, D., S. A. Tagliamonte,and E. Smith GoldvarbX: A MultivariateAnalysis.Retrievedfromhttp://individual. 2005 utoronto.ca/tagliamonte/Goldvarb/GVJndex.htm Steuer,RitaMaria,ed. Das alte Testament.Vol. 5: Sprichwörter, Ruth,Das Hohelied,Prediger, 2003 Klagelieder,Esther, Daniel, Esra, Nehemia, 1. Chronik,2. Chronik. Hänssler. Neuhausen-Stuttgart: Tallerman,Maggie CelticWordOrder:Some TheoreticalIssues. In ConstituentOrderin the 1998 LanguagesofEurope,editedbyA. Siewierska,599-647.Berlin:Moutonde Gruyter.

This content downloaded from 132.180.10.132 on Sun, 8 Dec 2013 15:38:29 PM All use subject to JSTOR Terms and Conditions