Subject: sum : object affixes

dear linguists the time has come for me to post a summary based on replies to a query i posted a few weeks ago . this summary is rather long , and has the following structure : 1 ) the original query 2 ) acknowledgement of respondents 3 ) discussion of a few languages 4 ) general discussion 5 ) refs 6 ) outline of relevant details of roviana 7 ) summary 1 ) the query ) i am studying roviana , a western oceanic lg . there are pronominal object ) suffixes which occur on the verb indexing person , number etc . however , ) there are no affixes on the verb which would correspond to 's ubject ' ) ( however we might construe the concept ) . dixon ( 1994 ) claims that there ) are no lgs with object affixes but which lack subject affixes . does ) anybody else know of any other counter-examples ? 2 ) acknowledgements . my thanks to the following people ( listed in no particular order ) for their responses ( and my apologies if i have left anyone out ) : matthew dryer , karen wallace , mark campana , eloise jelinek , leo connolly , geoffrey nathan , johannes heinecke , malcolm ross , mark mandel , jaejung song . first , a point of clarification . by western oceanic i mean a branch of eastern austronesian . some of the replies seemed to indicate that people interpreted it to mean western austronesian . ( not that it is really important ) . some people replied with examples of languages with object affixes , whether or not the languages in question also had affixes indicating other grammatical relations . matthew dryer replied with the following list of languages which have object affixes but no subject affixes ( taken from an a database mentioned by him in language 1992 ) : africa : central khoisan ( nama ) , ijoid ( kolokuma ijo ) , gur ( bimoba ) , east chadic ( kera ) , biu-mandara ( margi , mbara ) , west chadic ( ngizim ) . eurasia : nax ( ingush ) , avaro-andi - dido ( avar ) , lezgic ( archi ) , munda ( mundari ) . se asia oceania : nicobarese ( car nicobarese ) , central-eastern malayo-polynesian ( kiribatese , ponapean , woleaian ) . north america : pimic ( papago ) . south america : warao , maipurean ( palikur ) . i have not been able to follow up on all of these refs . i have however , looked at the micronesian languages . 3 ) discussion of a few languages sohn ( 1975 ) re woleaian : object suffixes , but no affixes corresponding to any other grammatical relation . 3sg has three forms : zero and - y / - w . it seems that verbs fall in to two classes , with zero being added to verbs in one class ( many of whose members appear to have a tr suffix ) and - y / - w to members of the other class ( the choice in forms being phonologically conditioned ) . interestingly , 3pl distinguishes animate vs inanimate , whereas all other persons and numbers do not . these affixes can occur with or without an overt independent object np . these affixes occur on certain post-verbal advbs , suggesting that they might in fact be clitics . zewen ( 1977 ) re marshallese . pronominal objects immediately follow the verb . 3sg and 3pl are enclitics and undergo vowel harmony . these enclitics can co-occur with lexical np object . ' subject ' prons are all independent . groves , groves and jacobs ( 1985 ) re kiribatese . it seems from their description that the language has object suffixes but no other affixes indicating grammatical relations . however , there are things which could perhaps be analyzed as subject prefixes ( p1 08 ) . exactly which object suffixes get attached ( out of 11 classes ) seems to be lexically determined . 4 ) general discussion it would appear that there are no cross-linguistic studies of object affixes per se . some respondents questioned whether the object suffixes in roviana are really suffixes and not just clitics . they are suffixes , but even if they were n't the focus of my interest would be the same . i am focussing on two facts about roviana : ( a ) the case marking on np 's is on an ergative-absolutive basis , but the only grammatical relation indicated by verbal affixes is object . ( b ) of the object affixes , only 3pl has zero form . being functionally inclined , i am interested in finding a motivation for ( b ) . it is not simply a case of referentiality ( see the brief sketch below ) . 5 ) refs ( with diacritics mangled by the internet ) corston , simon h . 1993 . ergativity in roviana . ma thesis : auckland university , new zealand . ( to appear in the pacific linguistics series ) . groves , terab ' ata r , groves , gordon w . and jacobs , roderick . 1985 . kiribatese : an outline description . pacific linguistics , series d , no . 64 . anu printing service : anu . sohn , ho - min with the assistance of tawerilmang , anthony f . 1975 . woleian reference grammar . pacific and asian languages institute : micronesia . honoloulu : university press of hawaii . zewen , francois - xavier n . 1977 . the marshallese language : a study of its phonology , morphology and syntax . veroffentlichungen des seminars fur indonesische und sudseesprachen der universitat hamburg , band 10 . berlin : verlag . 6 ) outline of relevant details of roviana the following is a brief description of roviana based on my own field work ( corston 1993 ) . case marking on np 's is split ergative , in a typologically unusual way : absolutive is marked ( si / se ) , ergative is unmarked . pronouns , proper np 's and enumerated np 's in main clauses distinguish erg vs abs by means of special particles . all other np 's are neutral ( ie do not formally distinguish a / s / o ) . as well as using special particles , prons distinguish erg vs abs in the form of the pron . a , s and o are always specific . person object suffix 1sg - u , - au 2sg - xo 3sg - a 1pl . incl - xita 1pl . excl - xami 2pl - xamu 3pl - 0 ( zero ) these affixes are similar to or the same as various indep pronouns , e . g . compare rau 1sg xoi 2sg xita 1pl . incl xami 1pl . excl xamu 2pl but , asa 3sg abs or sa 3sg erg sarini 3pl abs or ri 3pl erg ( there is quite a story to sarini historically , see the ma ) if there is a do np , you get the affixes e . g . meke doxor-i - a ri si keke ixana and see - tr-3sg . do they abs one fish ( erg ) ' and they saw a fish . ' ( animals , 034 ) ( the paper i am writing now clarifies why i only put erg in parentheses - it is really unmarked , and you have to infer in this context that the pron = erg because there is no particle ) kote arina tie mae magu-i - 0 sarini fut pl man come carve - tr-3pl . do 3pl . abs ' the men will come and carve them up . ' ( feast , 016 ) ( them = pigs in a feast ) nb , here the do is the pron sarini . however , you can get the affixes even if there is no overt do np , suggesting that they have independent reference , e . g . ai lul-i - u mo intj follow - tr-1sg . do dt ' hey ! follow me . ' ( animals , 011 ) and the following longer example . e saimone sa tie he-hegere hoirana . pers simon def man dup - laugh there ' that 's simon , the man who 's laughing over there . ' avos-i - a xoi ? hear - tr-3sg . do you ( erg ) ' do you hear him ? ' ( day , 029-030 ) note that the suffixes do not ' index ' absolutive etc , only o . ( see ma ) roviana has case marked and quantified np 's . case ( erg , abs , obl , etc ) is indicated by articles before the np . quantifiers ( numbers , 's ome ' , ' all ' etc ) precede the head n . non - specific undergoers obligatorily occur in a backgrounded object construction . compare the following two examples . raro talo si gami cook taro abs 1pl . excl ' we cooked taro ' / ' we did some taro cooking ' it is difficult to say if ' talo ' is ' incorporated ' since there are no other indep elements assoc with the verb which ever come after . note that the backgrounded constr is intr , with no transitive marking , and with the actor as s . compare to the ' usual ' order vao raro-a gami sa talo cook - 3sg . do 1pl . excl def taro ' we cooked the taro . ' undergoers which are pragmatically backgrounded can also occur in the backgrounded object construction , irrespective of whether they are referential or not . 7 ) summary * languages with object affixes but which lack affixes indicating other grammatical relations do occur , but would appear to be uncommon . * i feel that having 3pl as the only zero form is unusual , but am unable to find a cross-linguistic study to bear this out . even so , the zero form requires a functional explanation ( cf du bois , john w . 1987 ' absolutive zero : paradigm adaptivity in sacapultec maya ' , lingua 71 : 203-222 ) . again , my thanks to those who replied . ps i should have a draft of my paper ' featuredness and core arguments in roviana ' available shortly . simon corston corston @ humanitas . ucsb . edu
