Reduplicated Nouns in Hausa

Hausa is a language that makes extensive use of reduplication in word formation. Reduplication operates productively in constructions such äs nominal plurals and adjectival past participles, e.g. gunoonii 'melons' (< günäa\ däfaffee 'cooked' (< däfäa), and is found frozen in synchronically simplex noun and verb roots, e.g. marmaräa 'latente', bäbbäkaa 'to grill'. It sometimes is total, e.g. köoke-köoke 'complaints', and sometimes partial, e.g. büuzuuzüu 'düng beetle'. Some reduplicated words preserve the underlying tone of the component parts, e.g. düddufäa 'white ibis' (< *dufä HL + dufa HL), while some have an independent superseding tonal pattern, e.g. jinä-jinä LLHL 'a red juiced weed' (<jinii HH 'blood'). In most cases in Hausa the reduplication is evident just looking at the surface form; in others, however, it is more obscure, e.g. jinjirii 'infant' (<*//>/ + jiri). Hausa reduplication äs a general phenomenon has been the subject of an important study by Gouffe (1975a), a work characterized by its extensiveness in scope and its carefulness and accuracy in detail. In the present paper, I propose to take a closer look at synchronically frozen reduplicated nouns. More specifically, I shall be limiting myself to an analysis of the some two hundred reduplicated nouns exhibiting the canonical shapes and surface tone patterns illustrated in (l).

The paper is thus, in effect, an exercise in internal reconstruction.A number of issues dealt with in the course of the analysis should, nevertheless, have a wider application in Hausa synchronic äs well äs historical phonology and morphology, e.g. the direction of reduplication, the length of final vowels, and the validity of certain tone rules.As a comprehensive case study, the paper should also contribute to the recently awakened theoretical interest in reduplication äs a morphophonological process (Marantz 1982, Odden andOdden 1985).

SEGMENTAL MODIFICATIONS
The canonical shape of three syllable reduplicated nouns (henceforth RNs) in Hausa is CVCCVCV. 3Because of syllable structure constraints, the vowel in the closed syllable in an RN undergoes certain automatic changes.Similarly, because of restrictions on syllable final consonants, the first C of an abutting pair also undergoes certain changes.These changes are accounted for by the following rules.

W -> V /
C$.A long vowel becomes short in a closed syllable.This is an automatic, synchronically active rule, e.g.
(2) bambäamü < *baambäamii 'upper part of deleb palm' kükkuukii < *küukkuukü 'a gum tree' 1.2.e, o -+ a / C$.The mid vowels /e/ and /o/, which would necessarily be short because of the previous rule, shift to /a/ in a closed syllable. 4Any palatalization or labialization of the consonant conditioned by the underlying front or back vowel remains after the vowel quality shift, e.g.
1.4.2.(Optional) w -> C^ / C^.When /w/ is the first of an abutting pair, it either "hardens" and forms a geminate with the following consonant or, like /y/, weakens and forms a diphthong with the preceding vowel (see Newman and Salim 1981).As in the previous rule, the gemination is a characteristic of RNs and is not required by general P rules.Unlike /w/, the corresponding semivowel /y/ normally does not exhibit gemination, e.g.1.4.4.Non-gemination of sonorants.In noting the prevalence of gemination in RNs, it is worth pointing out that apart from the case of /r/ just described, which is strictly conditioned, and the case of /w/ 9 which is optional, sonorants and semivowels normally do not undergo gemination.These fixed reduplicated nominals thus do not behave like synchronically productive reduplicated pluractional verbs (the so-called "intensives"), in which underlying sonorants and semivowels often form a geminate with the following abutting consonant, e.g.
In some frozen RNs, underlying coronal liquids (and occasionally /s/ and /ts/ äs well) show up äs a nasal.(The position of articulation is determined by the just described assimilation rule.)This change is manifested not only in RNs, but also in fixed reduplicated verbs, where the coronal consonant is more often an obstruent.Unlike the other changes that have been discussed, all of which occur to some extent or other in present-day Hausa, this change is manifested only in frozen, lexically fixed forms.The historical nature of this nasalization rule is illustrated more clearly by looking at verbs.Whereas lexically fixed reduplicated verbs, which originally were pluractional forms of simple roots, exhibit nasal replacement for non-nasal coronals, see (l Ib), synchronically generated pluractionals never apply the nasalization rule.The indication "sporadic" on the rule simply means that the nasalization takes place in some words and not in others and that so far the conditioning factor hasn't been identified. 8 2.1.Forms such äs fämfamii 'wood trumpet', jinjirii 'infant', and färfeetsii 'a shrub' are usually described äs involving reduplication of the initial CVC-of the root (Gouffe 1975a:297).The actual surface forms are then derived simply by applying rules such äs those given in sec.l above.(The tones of RNs, which will be taken up in sec.4, are in previous accounts mostly left unexplained.)With forms such äs bälbeeläa 'cattle egret' and kwärkwataa 'lice', on the other hand, the reduplication at the deepest level is assumed to involve copying of the entire root including the second vowel, i.e. kwärkwataa < *kwatä(a)kwatäa.This analysis is prompted by the presence of the falling tone on the first syllable of the RN.In Hausa, falling tone is nothing but a surface manifestation of high plus low on a single (heavy) syllable.This being the case, RNs with the tone pattern Falling-High-Low (= HL-H-L) can be seen äs exhibiting füll tone copying along with the segmental reduplication.The vowel that is later dropped is postulated in order to carry the low tone.Given modern nonsegmental approaches to tone (e.g.Goldsmith 1979), it is clear that, theoretically speaking, the postulated vowel is not necessary.One could easily generate kwärkwataa and similar forms by reduplicating only the initial CVC-segments but the füll tone pattern.From a specifically Hausa perspective, however, the analysis involving the postulated vowel is probably correct.First, in less problematic cases, we can see falling tones being created by the deletion of vowels carrying low tones, e.g.min < mini 'to me'; zai < zaayä 'he will'; gyäffaa < *geefä(a)fäa 'edges' cf.turäakaa 'tethering posts' (plural of same class).
Second, äs we shall see in the following section, assuming an initial CVCV-rather than CVC-enables one to account for the direction of reduplication and for the shape of longer RNs.If HLHL forms such äs bälbeeläa manifest CVCV-reduplication with subsequent vowel dropping, there seems to be no reason to adopt a different analysis for the RNs with other surface tone patterns.Better is to take a unified approach and postulate the same basic word formation rule for all the RNs.Thus, famfamn and färfeetsii, for example, should be analyzed äs being derived from *fami(i)famii and *feetsi(i)feetsn, respectively, in accordance with the general rule for all RNs even though the subsequently deleted vowel is not recoverable from the surface form of these particular words.2.2.Having settled on the general analysis of the CVC-syllable in RNs äs being derived from CVCV-, where the underlined V henceforth indicates the "doomed" (obligatorily deleted) vowel, one can ask what was the length of this postulated vowel.) specifies a long vowel, which has the advantage of allowing a straightforward process of complete reduplication, e.g.kwärkwataa < *kwatqakwataa.In another paper (Gouffe 1975a:297n), however, he leaves the length of the doomed vowel unspecified, perhaps attracted by a short vowel analysis because of the ease with which short vowels in Hausa are deleted äs compared with long ones.In fact, from a historical perspective one can have a short, easily deletable doomed vowel and füll copying!A major advance in our understanding of Hausa historical phonology and mo hology during the past decade has been the discovery that common nouns (and most other words for that matter) originally ended in a short final vowel (Greenberg 1978, Schuh, 1984[presented at a 1978 congress]).As explained by Greenberg and Schuh, the long vowels that one now finds on most common nouns are the result of diachronic morphophonological processes related to the suffixation of article-like elements (see also Newman 1979).If the lexically frozen RNs were formed at a time when the final vowels were still short, then one would naturally get the hypothetical forms that make the most sense, i.e. forms that have a short doomed vowel and füll reduplication of all the elements (consonants, vowels including vowel length, and tone).Our reconstructed underlying form for famfamn would thus be *famifami (not T.famüfamii nor V.fämifamii) and, similarly, the reconstructed form for kwarkwataa would be *kwatakwata.Subsequently the doomed vowel was dropped and the word final vowel was lengthened (with tonal consequences to be discussed in sec.4).2.3.So far we have concluded that the basic form of RNs (limiting ourselves to trisyllabic surface forms) was CVCV-CVCV, where the final vowel of the root äs well äs that of the copy was short.No mention was made of the quality of the final vowel since this poses no fundamental word formation problem.Nevertheless, it is probably worth pointing out -even though I do not understand the significance of the observation - that RNs in Hausa exhibit a much more skewed distribution of final vowels than do simple non-reduplicated nouns.Of the roughly 200 RNs that I have culled from the dictionaries, only two end in /e/ 9 four in /u/ 9 and fourteen in /o/, with the rest evenly distributed between /// and /a/.By contrast, a rough count of an arbitrary list of 200 non-reduplicated nouns in Hausa turned up some twenty words ending in /u/ and twentyfive in /e/ and /o/. 9While /// and /a/ are the most common final vowels in non-reduplicated äs well äs reduplicated nouns, the other vowels appear grossly underrepresented in RNs.It is particularly curious why /e/ should be so rare, especially since it occurs commonly in other reduplicated structures.

DIRECTION OF REDUPLICATION
Up to this point we have been proceeding äs if the reduplication in RNs were to the left, i.e.Copy + Root rather than Root + Copy.At first sight this order seems correct.In düddufäa 'white ibis', for example, the syllables /dufäal exhibit the root shape while the CVC syllable /düd/ has a missing vowel and a changed final consonant.The fact that the first syllable is phonologically deformed is not, however, conclusive evidence regarding its structural Status.Starting with an underlying form *dufqdufa, the deletion rule for the doomed vowel need only specify its position in the RN: it is not important whether the doomed vowel is part of the root or of the copy.Similarly, the change from /f/ to /df (äs with all the other changes described in sec. 1) is conditioned by the consonant's surface position äs a syllable final abutting consonant without regard to its morphological origin.Since in *dufädufä (-» düddufäa), *famifami (-* famfamii), *kwatakwata ( -> kwarkwataa), etc., both appearances of the underlying CVCV are absolutely identical, one could just äs easily reject the usual viewpoint and argue that the order in RNs is Root + Copy, i.e. that the reduplication is to the right.Because the forms themselves throw no light on the issue one way or the other, one has to look elsewhere for evidence in support of the Root + Copy order.First, Hausa is primarily a suffixing rather than a prefixing language and thus reduplication is more likely also to occur to the right (Wilbur 1973, Marantz 1982).Second, most examples of reduplication in Hausa where the direction is evident because the reduplication is partial show reduplication to the right, e.g.With these longer forms built on three syllable roots, it is clear that the reduplication is to the right rather than to the left, i.e. cililligaa < a root *ciliga + a partial copy *liga.In the absence of any evidence that would indicate that RNs built on disyllabic roots behave differently from longer forms, we must assume that the direction of reduplication is the same, i.e. famfamii < a root *fami + a copy *fami.Having established a direction of reduplication that is the same in all cases, we can now formulate a single word formation rule for all RNs regardless of their syllable structure.
(14) RN = Root-V + Copy2S [where V means that the final vowel of the root is "doomed" and Copy 2S means copy the two rightmost syllables of the root including their tones] In the case of three syllable roots, copying the two rightmost syllables results in an RN exhibiting partial reduplication.With two syllable roots the resulting output exhibits füll reduplication (prior to the deletion of the doomed vowel); but exactly the same rule is involved in both cases, In the rule, the overt specification of the doomed vowel -V by underlining is a mnemonic that strictly speaking is probably unnecessary.One of the general characteristics of word formation to the right in Hausa is the dropping of the root final vowel, a process that is found in inflection äs well äs derivation and in verbs äs well äs nouns, e.g.

TONI;
Hau sä treats the tone of reduplicated forms in two different ways.One mechanism, generally applied in inflectional forms, is to assign a discrete tone pattern to the resultant reduplicated form without regard to the intrinsic tone of the root.The word duddugee 'heel', for example, has HHL tone since this is the set pattern for reduplicated words ending in /ee/ designating things that occur in pairs.The other means is to copy the tones of the root along with the segmental units comprising the syllables that are reduplicated, e.g.birbiröo HLHL 'adornment' < *biro HL + birö HL.This is the method employed in the RNs being treated in this paper.Although this type of tonal formation has been recognized with words of certain tonal shapes (particularly HLHL words), it has not been recognized in others.This is because the tone copying has been masked on the surface by the Operation of certain tone changing rules.Nevertheless, one can show that all of the RNs included in this study are formed by tonal reduplication äs well äs segmental reduplication.Each of the four tone classes will be described in turn.

4.1.
(L/H)HHH.RNs with HHH surface tone are derived from a root HH plus a copy HH. (When the doomed vowel is dropped, HH on the resulting closed syllable simplifies to H.) HHHH nouns come from a root HHH + a copy HH, while LHHH nouns come from a root LHH + a copy The tonal analysis presented here may also account for the tone of the reduplicated words termed by Parsons (1963:192)  In this case, however, an equally plausible explanation would be that this class of phonaesthetic words has a set LHLH tone pattern associated with it rather than exhibiting a surface tone pattern derived from a combination of the constituent parts.

(L)LLH.
RNs with (L)LLH surface tone are derived from roots with all low tone plus an LL copy.The final high tone that occurs is related to the length of the final vowel.In sec.2.2, it was explained that at the historically earlier period when the RNs were formed, Hausa nouns ended in short final vowels.Reduplicating an LL root would naturally produce a form with all low tones and a short final vowel, e.g.*fala + fälä -> *fälfälä 'a basket'.Subsequently the final vowel of RNs was length-ened, äs happened to all common nouns, whereupon a tone raising rule came into play.This rule, first described by Leben (1971)  I would like to offer a couple of suggestions so äs not to leave us with the idea that these forms are hopelessly anomalous.
One possibility for the LHL forms is that they represent HLHL RNs that have undergone tonal simplification of the surface falling tone.This is consistent with the existence of tonal doublets cited in the dictionaries such äs bälbeeläa = bälbeelaa 'cattle egret', and karRaaräa = karkaaräa 'a thorny acacia'.The other possibility, which applies to both LHL and LHH forms, is that these tonal classes of RNs manifest a set tone pattern belonging to the resultant RN which is not dependent on the tone of the underlying root.Viewed this way, LHL RNs would reflect a reduplicative pattern *LL.HL (cf.occurring forms such äs bidä-bidä 'a dark viper'), while LHH would reflect a *LL.HH pattern.

SUMMARY
The aim of this paper was to provide a historical account of word formation in Hausa involving synchronically fixed reduplicated nouns.From the analysis of these particular forms, a number of findings emerged that contribute to our understanding of Hausa reduplication in general äs well äs other areas of Hausa tonology and morphophonology.
(a) Reduplication was to the right rather than to the left for all forms regardless of the number of syllables in the root, i.e. the basic Hausa pattern is Root + Copy, not Copy + Root.
(b) The two rightmost syllables of the root were reduplicated in füll, including their two tones.(b f ) The reduplication of the tone along with the segmental reduplication is one means of word formation in Hausa.The other means is for the resultant reduplicated form to have a discrete tonal pattern that supersedes the tone of the underlying root and copy.
(c) The closed syllable that appears in the ...CVCCVCV cannonical shape of RNs resulted from the dropping of the root final vowel.When the vowel was lost, its tone nevertheless remained.
(d) Syllable final consonants in reduplicated forms evidenced differences in behavior from the same consonants in non-reduplicated forms.Specifically, RNs manifested a common, but irregulär, change of coronal consonants (usually sonorants) to /n/ a change not attested in non-redplicated forms, while they failed to undergo the change of velar obstruents to /u/, which is the regulär, historically well-attested change for non-reduplicated forms.
(e) At the time these now-frozen reduplicated nouns were formed, common nouns all ended in short final vowels.In the original RNs, the final vowel of the root and of the copy would have been short.(g) The simplification of LH on a single syllable to H also operated äs a regulär tone rule.NOTES 1.The following transcription conventions are employed.Long vowels are indicated by double letters.Acute, grave, and circumflex accent marks indicate high, low, and falling tone, respectively.The accent marks are placed only on the first of a double vowel.The letters /c/ and /// represent alveopalatal affricates; /ts/ indicates a coronal ejective, which in the Standard Kano dialect is usually a fricative [s '].When phonemes represented by digraphs occur doubled, only the first letter of the diagraph is repeated, e.g.-t t s-= {-t st s-], -ssh-= [-shsh-], -ggw-= \-g w g w -], etc.The rolled /r/is written with a subscript dot to distinguish it from the flap /r/, written without a diacritic.The asterisk * indicates presumed historical forms.Unacceptable forms are marked ??. 2. The examples cited in this paper were taken from the two major Hausa dictionaries: Abraham (1962) and Bargery (1934).Many are obscure words that are unknown to present-day Hausa Speakers.Since the paper is concerned with morphophonology rather than lexical semantics, I provide only abbreviated glosses for the Hausa examples.Glosses such äs *a fig tree' or 'a grasshopper' mean that the words in question denote a particular kind of tree or grasshopper, for which one should turn to the major dictionaries for a detailed scientific description.3. To simplify the presentation, I shall generally describe things in terms of RNs of three syllables, the most common length.The same description is also applicable to longer forms.My use of the cover symbols C (= consonant) and V (= vowel) is not parallel.Whereas C represents a single consonant only, V Stands for any vowel whether long or short.4. The centralling and lowering of /e/ and /o/ in closed syllables is often described äs a regulär, feature-changing P-rule.Whether the contrast between /e/, /o/ and /cr/is really neutralized in closed syllables, äs has been claimed, or whether the consonants rcmain distinct, at least statistically if not in every particular case, is an unsettled question.5. Most descriptions of Hausa include d -*• / in this automatic palatalization rule.This is not justified, especially when viewed historically.While /d/ has undergone palatalization in a number of common words, the application of the rule is much less general for /d/ than for the other coronal obstruents.The phoneme /d/ thus Stands half-way between /cf/, which does not palatalize, and /t/, which always does (a few modern loanwords excepted).6. Marantz (1982) has argued that reduplication is nothing other than affixation in which the affixed material just happens to be phonological identical to the stem.To sustain his extreme position that "there is nothing special about reduplication"' (p.436), he goes to some length to try to explain away the many instances cited in the literature of reduplicated forms behaving differently phonologically from nonreduplicated forms.In my opinion, his rationalization of the contrary evidence is unconvincing.While Marantz may be right that reduplication should be subsumed under the general rubric of affixation, I would argue that reduplication, nevertheless Brought to you by | Indiana University Bloomington Authenticated Download Date | 4/26/16 4:33 PM qualifies äs a distinct process/mechanism with indentifiable phonosemantic and phonological properties (cf.Carrier 1979).7. The failure of reduplicated forms to obey Klingenheben's law for labials and velars was first brought to my attention a number of years ago in an unpublished paper by Schuh (1968).The special Status of reduplicated words in Hausa with regard to phonological rules is further illustrated in Al-Hassan (1983), an interesting thesis concerned primarily with intensive (i.e."pluractional") verbs.8.If it is true that the consonant affected by the rule must be a coronal, then the word yanyaawaa Yennec', presumed to be derived from *yaawa-yaawa, must have some explanation other than the direct replacement of/w/ by /n/. 9.The count of non-reduplicated words was based on the first 200 items (excluding English loanwords and active verbal nouns) in a mimeographed alphabetical list of disyllabic nouns having distinct plural forms prepared by F.W. Parsons.10.Synchronically, pluractional verbs in Hausa are generally formed by reduplication to the left, e.g.gaggaskataa 'reverify' < gaskataa Verify\ Historically, however, the reduplication was almost certainly to the right since this is the formation evidenced in all the lexically frozen pluractional verbs, e.g.
Most important, however, is the form of RNs that are more than three syllables in length, e.g.
10 (12) däfaffee 'cooked', adjectival past participle of dafäa gunoonii 'melons', plural ofgunaa dumäamaa 4 to heat' < dümii 'warmth' büuzuuzüu 'düng beeile' < *buuzu Brought to you by | Indiana University Bloomington Authenticated Download Date | 4/26/16 4:33 PM Leben (1971)h a falling tone on the first syllable are derived from a root HL + a copy HL.When the doomed vowel is dropped, the low tone of the root is preserved and combines with the high to produce a falling tone, e.g.(L/H)HLH.RNs with HLH surface tone (in which the tonal reduplication is far from obvious) come from a root LH + a copy LH.With the apocopation of the doomed vowel, the H attaches to the L in the preceding syllable in the same jnanner that the L attached to the H to produce a falling tone in the HLHL pattern.Hausa, however, does not have a surface rising tone.As first mentioned byParsons (1955:385n)and described more fully byLeben (1971), LH on a single syllable is realized äs H, and thus LHLH appears äs HLH.This HLH class, which really should be conceptualized äs the LHLH classes, is thus to LH roots what the HLHL class is to HL roots.As is generally the case, longer forms simply display the initial tone (L or H) of the root in the first syllable of the RN, e.g.