Zambezia (2000), XXVII (i).SOME ASPECTS OF THE ARCHITECTURE OF THEPOSSESSIVE NOUN PHRASE IN BANTUFRANCIS MATAMBIROFADepartment of African Languages and Literature, University of ZimbabweAbstractPossessive noun phrases are minimally made up of at least two Nounphrases (NPs) that are possessively related to each other. In this article weargue that the syntactic order in which these noun phrases appear is not asrandom as it may look on the surface. Semantic constraints, in a largemeasure, determine the syntactic order of NPs constituting possessive phrases.Some such constraints include the Person-Animacy Hierarchy (PAH), genericand part-whole relationships of nouns among others that will be explored inthe article. It shall also be argued that the said semantic constraints are notsyntactic rules as such bat are strong cross-linguistic tendencies to which thelinear ordering of the possessive phrase more often than not conforms andespecially within the performance as opposed to the competence domain oflanguage ability.INTRODUCTIONThis article is an attempt to explain the syntactico-semantics of one ofthe most frequent linguistic structures in Bantu (and dare I say in otherlanguages of the world as well).1 The Bantu family of languages is fairlywide, so much so that we do not claim to represent all of it in this article.Such an attempt would be as undesirable as it is impracticable. We willdraw illustrations from such Bantu languages as Shona, Ndebele, Zulu,Chichewa, Haya, Kinyarwanda, Kikuyu and Sotho, among others and the1 I wish to register my sincere appreciation of the useful comments that I got when Ipresented a version of this article in the Department of Linguistics' Staff Seminar Series inOctober 1998.1 also got equally important feed back from Mr Mkanganwi and Dr CarolynHarford. Apart from that, I would like to register my many thanks as well as acknowledgethe valuable support that I got from Professor Sam Mchombo of University of California,Berkley (USA) from whom 1 got materials that gave me the breakthrough in the Theory ofPossessive Phrases in Bantu that I was developing. Many thanks also go the University ofZimbabwe for allowing me the Contact Visit that made it possible for me to meet and gaininspiration from some world luminaries of linguistics such as Sam Mchombo, LarryHyman, Joan Bresnan, Charlse Fillmore, George Lakoff and John Mugane. Last but notleast, I wish to make it clear that although I have incorporated Mr Mkanganwi and DrHarford's suggestions in this article, the conceptual and organisational mistakes remainmine alone.7172 ARCHITECTURE OF THE POSSESSIVE NOUN PHRASE IN BANTUresults yielding therefrom, shall, by extrapolation, be presumed to standin the stead of others.The syntax of the Possessive Noun Phrase (henceforth PNP) hasbeen made reference to in Shona by Fortune (1957; 1985) and Harford(1985} among others. It should be noted that the latter looks at thisaspect in relation to two other Bantu languages, which are Gikuyu andKiruundi. Much more work has been done in other Bantu languages inrespect of the PNP; by Mugane (1997) in Gikuyu; Hyman (1977) in Haya;Guma (1971) for Southern Sotho; and Taljaard and Bosch (1988) for Zulu,among others. In much of this work, most of the researchers have notconcerned themselves with the semantic aspect underlying the syntacticstructures of this construction. Taljaard and Bosch (1988) and Fortune(1985) follow the Dokean descriptivist method of analysing the possessiveconstruction in Zulu and Shona respectively. Harford (1985, 32) pointsout in passing that the possessive phrase has a 'variety of semanticrelationships which it may denote, including possession, characteristic,part of, kinship, etc'. It is not however, within the scope of her thesis toexplore the syntactico-semantic nature of this relationship, an aspectwhich we propose to undertake in this article. Nonetheless, the passingobservations she makes with regards to the shades of semanticrelationships which possessive constructions refer to are illuminating forour purposes here.This article makes the following claim as its point of departure,namely that the linear structure of the PNP is largely sensitive to a clusterof semantic constraints which, in turn, govern its syntax. Some suchsemantic constraints, following Hawkinson and Hyman (1974), Hayman(1977), Lyons (1967) and Anschutz (1997) include: the person-animacystatus of possessively related nouns, the type ot NPs involved, i.e. whetheror not they are common, among others. We need to. point out from thebeginning that it is virtually futile to try and give all the semanticunderpinnings that determine the nature of the PNP, let alone the countlessvarieties of such constructions. At least the tenets of generative grammarrender an exercise such as that pointless. Moreover, Sager (1990, 29)warns against such an exercise in futility when he makes the followingobservation:The size of conceptual fields and the complexity of the relationships tobe declared within it, is matter for practical assessment of the purposefor which the conceptual analysis is undertaken ... it is not concernedwith absolute conceptual systems ... (the emphasis is mine).The said semantic constraints influence the syntax of these languagesthrough what we may refer to as a hierarchy of significance. Given thatminimally the PNP is configurationally constituted by at least two NPsF. MATAMBIROFA 73that are in a logical semantic relationship, the claim here is that one ofthe NPs of the PNP is hierarchically more significant relative to the other.This semantic hierarchical order is perceptually built on the understandingthat the NP that is perceived of as the possessor (or psr) is hierarchicallyhigher, more significant or more important in comparison to the NP thatis perceived of as the possession (or psn). It is however, instructive atthis juncture to illustrate the syntactic organisation of the constructionunder consideration. Consider the following examples from Ndebele/Zulu, Haya and Swahili respectively:1. a. izimpondo ze-mbuzihorns of-goat'the goat's horns'2b. omukono gw' omwaanaarm of child'The child's arm' [Hyman, 1995, 868]c. risasi ya Jumabullet of Juma'Juma's bullet' [Keach and Rochemont, 1992-1994, 91]In all the above illustrations, we have in each case two NPs that areseparated by a possessive marker or affix, as it is traditionally referred toin the literature (and especially by Fortune). We shall refer to theuninflected NPs in (1) a., b., and c, izimpondo, omukono and risasi asthe bare nominals or the heads. These same NPs shall also be referred toas the possessions (psn) or the possessees, following Guma (1971, 119)even in cases in which this is not semantically so. The said NPs are herealso viewed as being hierarchically less significant in comparison to theirpossessively inflected cousins in (1), respectively imbuzi, omwaana andJuma. Note that these NPs have different names. Mugane (1997, 92)refers to them as 'Non-argument associative phrases.' The possessivelyinflected NPs in (1) a. and b. which are -imbuzi and omwaanarespectively, are the possessors (psr). Mugane (1997) refers to these as,'Argument associative phrase's.' They are, on account of that, viewed asbeing hierarchically more significant relative to the bare nominals or theheads.The syntactic structure of the PNP fits into the following input templatewhich is a rather slight modification of Hyman's (1977) which he formulatesas X of Y.Unless otherwise stated, all examples which do not acknowledge source are the author's.74 ARCHITECTURE OF THE POSSESSIVE NOUN PHRASE IN BANTU2. XofYofQ-n [where of can be substituted for: from, made of, ~'s,etc, while Q~n stands for an infinite number ofPossessive NPs that can hypothetically follow oneafter another]The above input template also fits in with Anschutz's (1997, 2)description of the PNP's linear order when he notes that:The term "Possessive NP" refers to any NP, NP construction included inthe study [i.e. his study] (NP's NP; NP from NP, etc.In distributional terms, we should situate the PNP in the same locationsas those of the bare NP. Consider (3) belowe3. X of Y of Q~n ŠVPŠ R of S of Q~n =>In this formulation, we posit that syntactically, the PNP can be madeup of a concatenation of possessively related NPs which, theoretically,range from at least two ad infimtum, both, before and after the verbphrase (VP). The number of the hypothesised concatenation of NPs isrepresented by the notation ~n, while =»Ł stands for other phrases thatmay be part of the sentence, but which, for our purposes here are outsidethe scope of this discussion.PART-WHOLE PNPSPart-whole PNPs as defined by Lyons (1977, 312) show a relationship:which holds between the separate or separable components of a thingand the whole thing of which they are components.Here both Sager (1990, 32) and Lyons (1977) give the following as theconceptual formulation of the part-whole PNP:34. X is a constituent of Aor X, Y, and Z are constituent parts of Aor A consists of Xor A consists of X, Y and ZThe above semantic conceptualisation of the part-whole relationshipmaps onto the syntax of the PNP in Bantu by according hierarchicallyhigher status to the 'argument associative phrase' which we have alsoreferred to as the psr. The illustrations given below support this view.Note that the values, X, A, Z etc. are substituted in (6) further down.F. MATAMBIROFA 755. a. michira ya mbewatails of mice'The mice's tails' Chichewa [Bresnan and Mchombo,1987, 774]b. amaisho g omwaanaeyes of child'The eyes of the child' Haya [Hyman, 1995, 873]c. izindlebe ze-njaears of dog'The dog's ears' Zulu [Taljaard and Bosch, 1988, 87]d. denga re-mbaroof of-house'The roof of the house' Shona [ Author's example]e. mororo oa-tauroar of-lion'The roar of the lion' S. Sotho [Guma, 1971, 120]Before we go into the explanation aspect of the discussion, we needto first look once more at the formulation in (4) with a view to substitutingthe values provided with concrete examples. Below is the substitution ofvalues X, Y, Z and A by NPs from one of the PNPs in (5). We propose to dothis exercise using PNP (5) d. as shown in (6) below.6. denga [roof] is a constituent of imba [house]or denga, chidziro [wall] and musiwo [door] are constituent partsof imbaor imba consists of dengaor imba consists of denga, chidziro and musiwoNote that although constituent NPs, chidziro and musiwo, are notincluded in the original PNP, they are nevertheless in a part-wholerelationship with the main NP, imba. All the other Bantu PNPs in (5) canbe cast in a substitutive format of values along the same lines as has beendone in (6) above.The above examples in (5) which are taken from five different Bantulanguages in which the constituent parts of the whole are invariablysyntacticised as the psn while the wholes are syntacticised as the psr,cannot, in all fairness, be attributed to some accidental organisationaloccurrence. It is true to say that the whole is perceived of as beingsemantically superior relative to its constituent parts. This explains theprevalent grammaticalisation of the associated NPs where the bare psnNP is located on the right hand side of the psr NP. Cast in a hierarchical76 ARCHITECTURE OF THE POSSESSIVE NOUN PHRASE IN BANTUordering that we argue in support of in this article, NPs, mbewa and tauin (5) a. and e. respectively, are therefore hierarchically higher as thepossessors vis-a-vis the bare NPs, michira and mororo, that are thepossessions.GENERIC POSSESSIVE RELATIONSHIPSOne equally important constraint that determines the syntacticorganisation of the psn and psr within the PNP, is the nature of therelationship that binds the two or more elements involved. One suchconstraint is the generic relationship. Sager (1990) defines this kind of arelationship in the following manner when he writes:the generic relationship establishes a hierarchical order; it identifiesconcepts as belonging to the same category in which there is a broader(generic) concept which is said to be superordinate to the narrower(specific), subordinated concept/concepts (p. 30).The said generic relationship as given by Sager (1990) can beformalised in more or less the same way as was done for the part-wholerelationships as indicated in the input template below (note that thesubstitution of values R, T, P etc. is done in (9) further down).7. R is a type of Tor R, P and Q are types of Tor T has the specific concept of R, P and Qor T has the sub-type RConsider the examples from Shona that are provided in (8) below.8. a. chibagwe che-mbeumaize of-seed'seed-maize'b. nzungu dze-mbeugroundnuts of-seed'groundnut seed'c. mapfunde e-mbeusorghum of-seed'sorghum-seed'd. benzi ro-munhufool of-person'a foolish person'e. simbe yo-munhulazy of-person'a lazy person'F. MATAMBIROFA 77f. mhutse yo-munhudestroyer of-person'a destroyer of people'As has been the case in (6) above, we will first do a practical insertionof NPs into value slots, R, P, Q etc., that are the input template of thegeneric PNP. Use is made of PNPs a., b., and c. which are taken from (8)above. In (9), we substitute 'abstract' values of (7) with NPs.9. chibagwe [maize] is a type of mbeu [seed (or crop)]or chibagwe, nzimgu [groundnuts] and mapfunde [sorghum] aretypes of mbeuor mbeu has the specific concept of chibagwe, nzungu, mapfundeor mbeu has the sub-type chibagweThe examples that are given above in (8) instantiate cases where thegeneric forms, 'seed' and 'person' are semantically perceived of as beingmore significant than their sub-types. The nature of this semanticrelatedness grammatically transmutes to yield the present syntacticordering of the PNP, where the generic form is grammatically the barenominal (semantically the psn) while on the other hand the sub-type isgrammatically the possessively inflected NP (which semantically is theHere we do not have a typical case of possession in the strict sense ofthe word. Semantically what is represented by the generic relationshipabove leans more towards 'type of than 'possessed' or 'owned by' kindof PNPs. The argument that is being pursued in this article is that it isprecisely the nature of the generic relationship that constraints the linearor syntactic organisation of the PNP. Here the generic forms arerepresented as being semantically superordinate to their sub-types thatare then viewed as hierarchically subordinate to them.Note however, that from (7) d. down to f., the hierarchicallysubordinate NPs refer to people who may be morally undesirable (atleast as judged by the moral standards of the Shona people although wecannot conceive of a society in which people with such character traitsare condoned). One may only venture to suggest that this scenario,where NPs indicative of negative character traits assume the subordinateposition vis-a-vis the superordinate generic human NP, prevails, chieflybecause such people are generally loathed by others. Owing to this, theyare therefore not allowed to 'possess' or to take precedence of significanceover the generic human NP which remains semantically superordinate tothem.78 ARCHITECTURE OF THE POSSESSIVE NOUN PHRASE IN BANTUPERSON-ANIMACY HIERARCHYHyman and Morolong (1977), Hyman (1977) and Anschutz (1997) indicatethat the Person-Animacy (PAH) dichotomy has pervasive influence insyntactic organisation cross-linguistically. For instance, given any twoarguments in Bantu that are associated with the same predicate andwhere one of them is human and the other is animate, in most such cases,the syntax almost invariably accords the human argument more accessto direct object status vis-a-vis the other. The basic assumption is thatdirect object status is more significant than say the indirect or secondaryobject status.Consider the following example from Swati:10. Tsine sakh-ela indvodza sibayawe build-Ben man cattle byre'We are building a cattle byre for the man' Swati [Kunene, 1987, 6]In (10), the human or person element is largely responsible for theadjacency of the NP indvodza to the predicate and the subsequentperipheralisation of the inanimate argument sibaya. It must also be bornein mind that at the lexical semantic level, that is, prior to the morpholexicalinsertion of the applicative morpheme -el-, sibaya was one of the originaltwo arguments present. This would somewhat intuitively lead us to assumethat sibaya must retain its position adjacent to the predicate and thenthe newcomer, beneficiary argument indvodza be logically pushed to theperiphery as the grammatical indirect object. If we were to follow 'naturaljustice' as we see it fit, and reverse the above order, an ungrammatical orat the very least, an unintended sentence would be the one given below:11. *? Tsine sakh-ela sibaya indvodzawe build-Ben byre manThere are different versions of the PAH found in the literature, suchas the Quirk's Gender Scale (in Anschutz, 1997), Hawkinson's AnimacyHierarchy, Hyman and Morolong and Hyman. We shall for the purposesof this presentation, use Hyman's (1995) version which is as follows:12. 1st pers. > 2nd pers. > 3rd humans > 3rd pers. Animal > 3rd pers.Inanimate.It must however, be noted that the hierarchy that is provided heremust not be taken as the laying down of rules as such. This hierarchy,probably like any other, is supposed to indicate fairly strong tendenciesby which Bantu languages syntacticise semantic phenomena associatedwith the PNP.F. MATAMBIROFA79Consider the examples taken from various Bantu languages that areprovided below:13. a. masimo a- murenafields of-chief'The chief's fields'b. khomo tsa-ba-lona?cattle of-yours'Your cattle'c. izimvu zikamalumesheep of/uncle'Uncle's sheep'd. nyumba ya mfumuhouse of chief'The chief's house'e. pa-mudzi w-athuat-house of-our'At our house'f. kitabu cha Zeinabubook of Zeinabu'Zeinabu's book'S. Sotho [Guma, 1971]Zulu [Taljaard and Bosch, 1988, 87]Chichewa [Bresnan and Mchombo,1987, 773]Kiswahili [Carstens, 1991, 84]g. urugo rwo [kwaa] daatacompound of at father'Compound of father' Kiraundi [Harford, 1985, 35]i. shuro ya- babahare of-father'Father's hare'j. inkunzi ye-nkosibull of-chief'The chief's bull'k. enkoni y' omwaanastick of child'The child's stick'ShonaNdebeleHaya [Hyman, 1995, 868]I. nyungu ya cucupot of grandmother'Grandmother's pot' Gikuyu [Mugane, 1997]80 ARCHITECTURE OF THE POSSESSIVE NOUN PHRASE IN BANTUm. libizo la-mushumani yoname of-boy this'the name of this boy' Lozi [Fortune, nd, 31]The examples given above are drawn from ten different Bantulanguages that cover a wide geographical expanse of Sub-Saharan Africa.What is striking regarding the data provided above is the systematicstructural patterning of PNPs in conformity with predictions of the PAHas given in (12) above. We argue as we have done earlier, that thissystemic patterning cannot be left to some fortuitous explanation. Fromour view, the only plausible explanation for this common architecturetraceable in different languages is that the miscellaneous PNPs are allconstrained by the provisions of the PAH as outlined in (12) above.In the data given above, we notice that the subordinate NPs (also thepsn) are lower in the PAH scale as compared to the hierarchicallysuperordinate or significant NPs (also the psr) which are higher on thesame scale. Note that the following bare NPs drawn from (13) are allinanimate and lower than their human counterparts on the PAH scale: a.inasimo, d. nyumba, e. pamudzi, f. kitabu, g. urogo, 1. nyungu and m.libizo. While the latter examples cited show the position of inanimateNPs in comparison to human NPs, the following also bear testimony tothe contention that even animate NPs which are lower than human on thePAH scale are still syntacticised as hierarchically subordinate to thelatter. In the following examples from (13) b. khomo, c. izimvu, i. shuroand j. inkunzi, we note that although the latter are animate, they arenonetheless semantically lower than human, hence their less significantstatus relative to human NPs.Two examples which illustrate the supremacy of the persons overother variables as indicated on the PAH scale are from Southern Sothoand KiSwahili. The PNPs, khomo tsaba-tona and pamudzi w-athu fromSouthern Sotho and KiSwahili respectively show the persons as -lonaand -athu. In the case of Southern Sotho, the NP khomo which is animateis the psn of the second person pronoun -lona. In KiSwahili, the locativepamudzi is inanimate and it is subordinated to the first person plural -athu. The persons are the more significant elements of the PNPs becausethey assume psr roles. The syntactic organisation of these PNPs assumesthis order according to the predictions posited by the provisions of thePAH.This state of affairs as described in the preceding paragraph above,as indicated earlier, is a cross-linguistic phenomena. Hawkins (1981) ascited by Anschutz (1997, 3) makes the following incisive observation inrespect of the nature of the syntax obtaining from this semanticorganisation of NPs when he comments:F. MATAMBIROFA 81The division between human and non-human nouns is realised as arather interesting semantic category whereby human nouns have linearprecedence over non-human nouns [...] If one of the nouns is humanand the other is not, the surface form corresponding to the structure inwhich the human noun comes first will be more acceptable than thesurface form corresponding to the structure in which the human nouncomes second (the emphasis is Anschutz's).Note however, that when applying this to Bantu, we must reverse theabove statement and say, 'non-human nouns have linear precedence overhuman nouns'. This is necessitated by the reverse order in which Bantuand English consistently syntacticise the same possessive phrases.The prs are invariably either humans or the persons. Hyman (1995,875) whose version of PAH has been used here explains this in terms ofwhat may be referred to as the human-centric or anthropocentric natureof language when he notes:Call it a natural person hierarchy or a hierarchy of egocentricity,efficiency, or empathy, but the fact remains, namely that persons higher[in the hierarchy] will be more susceptible to possessor promotion.This explains why human language in general accords superordinate(or psr) status to humans and the pronouns/persons while relegating onthe other hand, non-humans to the subordinate (or psn) status. It wouldhowever, be interesting to explore how Bantu languages syntacticisesemantic cases in which both the psr and the psn are human, an exercisethat we explore in the next section of this article.Human Possessive Noun PhrasesPNPs involving humans are interesting when looked at from the viewpointof the PAH that we have so far been dealing with. It has already been seenfrom the viewpoint of the said hierarchy that humans take first positionon that scale. This, it was observed, tends to influence the syntacticorganisation of the PNP by showing a greater or less tendency of grantingpsr status to human NPs thereby relegating other NPs to the subordinatepsn status. This, as has already been demonstrated, enters into thesyntax of the PNP by granting 'non-human nouns linear precedence overhuman nouns'.It will be argued here that human PNPs, that is, where the head andthe possessive modifier are both human, are constrained by an intrinsichierarchical organisation. This claim already excludes the persons (i.e.,the 1st person singular/plural, 2nd person singular/plural, etc.). It hasalready been shown in (13) above, using examples from both SouthernSotho and Swahili, that the persons take precedence over human nouns82 ARCHITECTURE OF THE POSSESSIVE NOUN PHRASE IN BANTUin general. The claim that is being advanced here pertains to humans ofequal nominal status such as the following from Southern Sotho (Guma,1971,120-122), batho (people), mosali (wife), mora (son), ngaka (healer),monna (man) etc. Consider the following PNPs involving human nounsonly.14. a. abaana b'oomushaijachildren of man'the man's children' Haya [Byarushengo, et. al., 1977, 38]b. abantwana bo-malumechildren of-uncle'uncle's children' Zulu [Taljaard and Bosch, 1988, 87]c. muranda wa-mamboservant of-king'the king's servant'd. vadzidzi va-Jesudisciples of-Jesus'Jesus' disciples' Shonae. adui za askarienemies of soldiers'the soldiers' enemies' Swahili [Harford, 1979, 1]f. cucu wa Murang'agrandmother of Murang'a'Murang'a's grandmother' Gikuyu [Mugane, 1997, 96]g. ntat'a-Masilofather of Masilo'Masilo's father'h. ngaka ea-Mosothodoctor of-Mosotho'a Mosotho doctor/healer' S. Sotho [Guma, 1971, 121-122]i. umsekeli ka-mphathintambovice of-leader'vice chairperson/president etc' Ndebele [MatshakayileNdlovu (personalcommunication)]j. uumw aana w'umugorechild of-woman'the woman's child' Kinyarwanda [Kimenyi, 1978,99]F. MATAMBIROFA 83k. njau y - a ng'ombecalf of cow'calf of cow' Kikuyu [Perez, 1986, 37]The examples in (14) are drawn from a number of different Bantulanguages. All of these examples with only the exception of (14) k., dealwith human possessive relationships. The last possessive phrase involvesanimals and it will be used subsequently in support of some of the claimsthat we will advance in the discussion.The first observation that we are drawing attention to in this sectionrelates to a hierarchy of human nouns that run along lines of social statusand/or institutional seniority either within the family institution, churchor even the political organisation of a people. The point being made hereis that people who enjoy high social standing such as those we have in(14) c. mambo (chief/king), d. Jesu, e. askari (soldiers) and i.umphathintambo (leader), are syntactically arranged within the PNP asthe psr. The reverse is generally true of people with relative lower socialstatus. The bare NPs occurring with those given in (14) c, d., e., and i.suffice as illustrations of this phenomenon. In grammatical terms, thislatter group is syntacticised as psn within the mechanics of the PNP. Itgoes without saying though that this type of a PNP does not indicate truepossession as such; rather it indicates what Harford (1986, 32)characterises as 'kinship'. Anschutz (1997, 9) describes this PNP asindicative of an instance whereby, 'the possessor and the possession arerelated through some social or genetic bond'. As can be argued, not all ofthe examples given above fall into this pattern.The claim that has been made in the preceding paragraph in respectof seniority can be demonstrated with reference to 'kinship' relationships.It is evident from the data that children are, in terms of 'possessivelanguage', perceived of as being possessions. This kind of perceptionseems to be cutting across all the Bantu languages, judging with theexamples that are at hand. The following nouns in (14) a. abaana(children), b. abantwana (children) and j. uumw aana (child) are juniorsvis-a-vis the other human nouns that they are pitted against. The seniorpersons semantically enjoy superordinate status hence they are thegrammatically possessively inflected NPs. The reverse holds true forjunior members of society. This extra-linguistic conceptual organisationof Bantu world view seems to touch on NPs other than humans only.Illustration (14) k. is a case in point. Here we witness a case whereby thecalf (njau) is the semantic possession while on the hand the cow (hg'ombe),is the possessor. In human relations, this is akin to mwana (sibling) andmai (mother) respectively.This kind of a possessive relationship can be explained either interms of what has been said above or alternatively in terms of what84 ARCHITECTURE OF THE POSSESSIVE NOUN PHRASE IN BANTUAnschutz (1997, 9) refers to as Origin. The said writer explains Origin inthe following way when he writes:The possessor created the possession; or the possession originatedout of the possessors, usually a place or institution.As is clear, not all aspects of Anschutz's argument can be used toexplain the semantics of the senior/junior possessive phrase dichotomy.The clause,'... originated out of the possessors ...' is most applicablein the above scenario (i.e. 14 k.) Children, it can be argued, and hopefullywithout much controversy, biologically originate out of their parentsthrough the twin processes of conception and birth. Fitting this intoAnschutz's logic, parents are therefore the possessors while the childrenare the possessions. When this kind of a relationship is cast in terms ofthe PNP, it has already been seen that parents are grammaticalised as thesuperordinate possessively inflected NPs while the children are thesubordinate bare nominals.This may not be that much surprising when one considers the factthat in most Bantu cultures the young are expected to show respect tothe elders. In the said societies, age, seniority, wisdom and authority areall rolled up in one package which the young are expected to takecognisance of.The last semantic constraint governing the syntax of the PNP, as it isbeing articulated here, concerns itself with weighting what we may referto as the 'semantic precedence status' of common nouns vis-a-vis propernouns. Where Hawkins says, 'non-human nouns have linear precedenceover human nouns', we would say, proper nouns have semantic precedenceover common nouns. It is instructive to give brief working definitions ofthe terms of contention, that is common and proper nouns. Crystal (1991,64) defines a common noun in the following manner:In traditional grammar, 'common nouns' were semantically definedsub-class nouns (referring to 'general concepts' . . .[the emphasis onsemantics is Crystal's].The same author defines a proper noun as follows:The alternative term, proper name, reflects its traditional semanticdefinition: the name of an individual person, place, etc (p. 282) [theemphasis is Crystal's].Given these definitions, it becomes much easier to argue in supportof the semantic constraints that yield linear grammatical organisation ofNPs in the PNP such as we have in (14) d., f., g. and h. In all theseinstances of human PNPs, it is true that proper nouns are taking semanticprecedence over their common noun cousins. Thus in (14) f. and g. forinstance, proper nouns, respectively, Murang'a and Masilo are takingsemantic precedence over common nouns, cucu and ntate. Note that thisF. MATAMBIROFA 85phenomenon is holding true for languages as geographically far and wideapart as Southern Sotho in South Africa and Gikuyu in Kenya. Hereproper names are accorded the possessor status while their commonnoun counterparts are perceived of as the possessions.In view of this observation and its pervasive occurrence in Bantulanguages, we propose to slightly re-modify the PAH as it is given in (12)in order for us to accommodate this phenomenon. Consider the revisedformulation of the PAH in (15) below.15. lsl prs. >2nd prs. >3rd prop. Humans >3rd comm. Humans >3rd prs.Animals >3rd prs. InanimateThis kind of order would help solve a number of cases where onefinds the junior/senior order being upset by cases in which we have asenior but 'common' person subordinated to a junior person with aproper name. Phrases (10) f., g. and h. clearly exemplify what we arearguing in support of here. Many more examples could be cited to bearwitness to this claim. However, these should suffice for the moment.We also wish to point out that the hierarchy that we suggest above isconceptually and formally the same as the animacy hierarchy given byCroft (1990,112) which is as follows:first, second-person pronoun<third-person pronoun<propernames<human common noun<non-human animate commonnoun<inanimate common nounIt is also worth noting that Croft reaches this conclusion using, interalia, Takelma and Quiche, languages that are vastly different from Bantuor at least not known to be excessively similar.On the Reversability of the PNP's Linear OrderIn this section, we want to dispel possible counter-claims to the theory ofthe PNP that has been developed in this article. The claims that havebeen made above regarding the efficacy of semantic constraints shouldin the most be viewed in the following light which is used by Anschutz(1997, 3) to qualify predictions of the animacy hierarchies when heobserves:It is no difficult task to come up with counter-examples to the animacyhierarchies, and by using phrases such as 'chiefly used with' or 'moreacceptable' their creators make it clear that the hierarchies are to beunderstood as tendencies rather than rules.It is quite evident in this article that we have used more than only theanimacy hierarchies in making claims about the architecture of the PNPin Bantu. We have used part-whole as well as generic relationships.86 ARCHITECTURE OF THE POSSESSIVE NOUN PHRASE IN BANTUamong others. And although Anschutz specifically absolves the predictionsof the animacy hierarchies in particular, what he says is equally applicableto the other constraints that have been advanced in the article.We would like to briefly discuss some of the salient issues that arecontained in Anschutz's observation above because they raisefundamental questions regarding the notion of language andgrammaticality or correctness. If we are to repeat the phrases used, suchas, 'chiefly used with' and 'more acceptable', it becomes clear thatgrammaticality does not always neatly bifurcate into a parochial, 'this isgrammatically correct' and 'this is grammatically incorrect' sort ofdichotomy. The question of 'more acceptable' brings in the notion of ascale or continuum of acceptability with hypothetical lower and upperreaches which, not surprisingly, competent native speakers of a languagecould argue over without coming to a conclusive resolution. This ushersinto the argument the incongruity between the competence and theperformance modules of language ability. While stoutly resisting thetemptation to delve into debates' surrounding the efficacy of engaging inlinguistic studies from the view of either module, we should nonethelesspoint out that this article has largely looked at the PNP from theperformance module vis-a-vis the competence one.The point we are making here is that it is hypothetically possible toreverse almost all the PNPs that have been provided as illustrations inthis article without necessarily creating ungrammatical possessives. Whatshould be pointed out is that, first, such an exercise would be onedevoted to establishing largely one thing, chiefly, to see whether or notungrammaticality results. We therefore argue that althoughungrammaticality may not necessarily result, one would be hard pressedto find a natural or non-contrived context into which the reversed PNPfits. While this is true for almost all of the PNPs that are in this article, wewill nevertheless sample a few for discussion purposes. Consider thereversed PNPs below:16. a. izindlebe ze-njaears of-dog? inja ye-zindlebedog of-ears'the dog of the ears' Zulub. denga re-mbaroof of-house? imba ye-dengahouse of-roof'house of the roof ShonaF. MATAMBIROFA 87c. mororo oa-tauroar of-lion?tau ea-mororolion of-roar'the lion of roar'd. masimo a-murenafields of - chief?murena oa-masimochief of - fields'chief of the fields' S. Sothoe. nyumba ya mfumuhouse of- chief?mfumu wa nyumbachief of - house'chief of house' ChichewaThe above PNPs are striking more by their semantic awkwardnessthan their ungrammatically as such. Note that all these reversed PNPshave already been analysed in respect of the different semantic constraintsthey imposed on the linear order of the possessive phrase. It is evidentthat the reversed possessives convey unintended meanings, at least inrespect of the initial and therefore intended meanings. In other words,even if the reversed PNPs convey meaning, they do not contain the sameinformation as is found in the original possessives. This can bedemonstrated in the formulation of the PNP provided below.17. a. XofYb. YofXIf we agree, as we have already done in (2), that the input template ofthe PNP can be cast in terms of either (14) a. or b., we must also acceptthat the two formulations, if perceived of as being in th^ reverse order ofeach other, are logically and formally different. They contain the sameNPs but they are semantically not identical. In order to illustrate thisdifference, we will take one of the PNPs and reverse the linear order ofthe NPs involved. If we retain the same meaning after doing that, we willbe left with no choice other than to revise our claims regarding meaningand NP status. We propose to use (14) a. in order to achieve that and wewill reconstitute it as (18) below.18. a. izindlebe ze-nja zibuhlungu khakhulu namuhlaears of-dog are painful much today'The ears of the dog are very much painful today'88 ARCHITECTURE OF THE POSSESSIVE NOUN PHRASE IN BANTUb. * inja ye-zindlebe ibuhlungu khakhulu namuhladog of-ears is painful much today'The dog of ears is very much painful today' ZuluThe grammatical disparities between the two examples in (18) clearlysuggest that:1. NPs that constitute a PNP are cast in a strict logical order, theintrinsic dynamics of which we have already demonstrated, and2. That reversal of this logical order, though it may not necessarilylead to ungrammaticality, is not of identical semantic and/or informationstatus as that of the original PNP. This is what precisely renders (18) b.outrightly ungrammatical. The first sentence, i.e. (18) a. is grammaticalbecause it satisfies the semantics of the part-whole possessiverelationship. On the other hand, (18) b. is ungrammatical because itviolates the semantics of the same relationship. Although the notion ofgrammaticality cannot be dispensed with, nevertheless the resultantconstructions are a testimony of the preferences of speakers of Bantu.Here, if two or more nouns are competing for high (or possessive) status,where one element is the whole and the other (or others) is only a part ofthat same whole, the part, more often than not, loses the competition tothe whole. The part becomes the possession while, on the other hand,the whole becomes the possessor, as has been consistently shown in anumber of instances above.Reversed PNPs in (16) d. and e. illustrate the prominence ofhumanness vis-a-vis non-humanness or inanimacy on the PAH as is outlinedin (12) above. In the two PNPs it is highly inconceivable to have inanimateNPs, masimo (fields) and nyumba (house) possessing or owning humans,murena and mfumu (both chiefs). The contexts in which this reverseorder might obtain is quite a task to imagine. The reasons for this areclear, first, the means to convey this extra-linguistic relatedness of NPs isconceptualised through a human semiotic system or signification, i.e.language. Secondly, this is a case where the owner of the means ofsignification has to rate himself in relation to elements other than him.Since the scales of lordship and domination are either way heavily in hisfavour, it would be looking for the unusual to expect his primary meansof communication to run contrary to how he perceives his place inrelation to elements of nature other than himself. This anthropocentricnature of language is what leads to what Hyman earlier described as '...a hierarchy of egocentricity, efficiency or empathy...' This 'unbalanced'relatedness between humans and inanimate nouns finds its way intolanguage in general, and in this particular instance, into the organisationof the PNP by according humans psr status while inanimate NPs areaccorded lower psn status.F. MATAMBIROFA 89In addition to what has been said above, one could also explain thisstate of affairs in terms of what Ferris and Hartmann (1983, 129) refer toas 'semantic expectation'. They describe it in the following manner:Semantic expectation Š speakers have a bias towards constructing forthemselves what seems to be a plausible sentence out of the lexicalitems which they hear.In view of this observation we would therefore say that the scenarioin (16) is improbable or outright ungrammatical because it violates thesemantic expectations of speakers.CONCLUSIONIn conclusion, this article has shown that NPs constituting the PNP donot, unlike what meets the eye, align to each other in a random fashion.NPs that are possessively related are intrinsically ordered in a hiei archicalprogression of significance. The basis upon which the hierarchy isconstructed has been demonstrated. A number of Bantu languages suchas, Shona, Ndebele, Kiruundi, Gikuyu, Swahili, Soiunern Sotho and Loziamong others have been cited in support of this argument. It has beendemonstrated that the bare or head nominals are the hierarchically lesssignificant elements of the PNP, contrasting with the possessive qualifierNPs which are the superordinate elements of the two nouns. This, allpervasive, semantic hierarchical relatedness of NPs within the mechanicsof the PNP has been the centrepiece of the foregoing discussion.We have, throughout this article, argued that the syntactic structureof the PNP, are negotiated through the semantic status of the individualNPs that constitute it. To this end, it has been shown that provisions ofthe PAH, the nature of NP relatedness and the sub-categorisation ofnouns into proper and common, among others, play some pivotal role indetermining first, the significant status of nouns and secondly, the resultantsyntactic alignment of the concerned nouns within the internal dynamicsof the PNP. Considerations stemming from the nominal status induced byNP sub-categorisation have persuaded us of the need to recast the PAHas we now have it in (15).Although the above semantic constraints condition to a large extentthe resultant architectural structures of the PNP, as already said earlier,it must again be emphasised that these must not be perceived as syntacticrules that must not be broken. They must be treated rather as strongtendencies largely centred in the domain of linguistic performance asopposed may be to the module of linguistic competence. This explainswhy a 'non-favoured' or 'non-preferred' PNP is not necessarilyungrammatical. Examples (17) and (18) have been employed particularlyto explain this important aspect of the article.90 ARCHITECTURE OF THE POSSESSIVE NOUN PHRASE IN BANTUAlthough this has not been within the scope of the discussion, it isclear to us that the semantic influence of different constraints is not ofequal strength, that is, some are more potent in terms of their influencethan others. For instance, the influence of the part-whole semanticconsiderations are more binding and therefore more difficult to overridethan say those of the senior/junior dichotomy. The latter can, for instance,be easily overridden by the nominal sub-categorisation element of properversus common noun. To illustrate this point, if say a less powerfulmember of a social institution, e.g. a commoner, is pitted against a morepowerful individual, such as a king, the possessive scales tip in favour ofthe latter. However, if the same less powerful individual is identified witha proper name and the other with a common name, the former takespossessive precedence over the latter. This is what has been shown in(14) f., g. and h. which contrast dramatically in this regard with a., b., j.and k. in the same.ReferencesANSCHUTZ, A. (1997) 'How to choose a possessive noun phraseconstruction in four easy steps' Studies in Language, 21 (1), 1-35.BRESNAN, J. and S. MCHOMBO (1987) 'Topic, pronoun and agreement inChichewa' in Language, 63 (4).BYARUSHENGO, E. et al. (eds.) (1977) Haya Grammatical Structure (LosAngeles, Department of Linguistics, University of Southern California).CARSTENS, V. M. (1991) 'The Morphology and Syntax of DeterminerPhrases in Kiswahili' (Los Angeles, PhD. Dissertation, University ofCalifornia).CROFT, W. (1990) Typology and Universals (Cambridge, CambridgeUniversity Press).CRYSTAL, D. (1991) A Dictionary of Linguistics and Phonetics (Oxford,Blackwell Publishers).FERRIS, D. C. and R. R. K. HARTMANN (1983) Understanding Semantics(Exeter, A Wheaton and Co. Ltd.).FORTUNE, G. (n.d.) An Outline ofLozi (Salisbury, Department of AfricanLanguages and Literature, University of Rhodesia).Š (1980) Shona Grammatical Constructions, Parti (Harare, Mercury Press).Š (1967) Elements of Shona, 2nd Edition (Harare, Longman).GUMA, S. M. (1971) An Outline Structure of Southern Sotho(Pietermaritzburg, Shuter and Shooter).HARFORD, C. (1985) 'Aspects of Complementation in Three BantuLanguages' (Bloomington, Ph.D Dissertation, Indiana University).Š (1977) 'A Montague Analysis of the -a Construction in Swahili' (Unpubl.).HYMAN, L. (1977) 'The syntax of body parts' in E. Byarushengo et al.(eds.) Haya Grammatical Structure (Los Angeles, Department ofLinguistics, University of Southern California).F. MATAMBIROFA 91Š (1982) Form and Substance in Language Universals (Los Angeles,University of Southern California).JACKSON, H. (1988) Words and Their Meaning (Essex, Longman).KEACH, C. N. and M. ROCHEMONT (1992-1994) 'On the syntax of possessorraising in Swahili', in R. Botne (ed.) Studies in African Linguistics, 23K1MENY1, A. (1978) A Relational Grammar of Kinyarwanda (Berkeley,University of California).KUNENE, E. C. L. (1987) 'A Re-examination of Some Aspects of theBenefactive (Applied) Extension in Swati' (Harare, paper presentedat the Second Linguistic Association for SADCC Universities, Universityof Zimbabwe).LYONS, J. (1967) Foundations of Language, Vol. 3 (Dordrecht).Š (1977) Semantics II (Cambridge, Cambridge University Press).MUGANE, J. M. (1997) A Paradigmatic Grammar ofGikuyu (Stanford. CSL1Publications).SAGER, J. C. (1990) A Practical Course in Terminology Processing(Amsterdam, John Benjamin Publishing Co.).TALJAARD P. C. and BOSCH S. E. (1988) A Handbook oflsiZulu (Pretoria. J.L. van Schaik (Pvt.) Ltd.).