3.3 Initial Reduplication P-2.

We are here concerned with the primary reduplication rule, applying either at the beginning of the noun or verb root or, more regularly, at the beginning of the first morpheme following inflectional prefixes. This rule is most common with verbs (indicating repetition or prolongation) but occurs with some nouns (plural). Some roots, often having both verbal and nominal forms and with generally adjective-like meaning, are often reduplicated regardless of aspect or plurality. Details are covered in §4.14, §8.4, §13.5, §14.14-15. Many roots (flora-fauna nouns, for example) have no reduplications. A distinct reduplication affecting demonstrative pronouns is dealt with in the following section.

The point in the unreduplicated input where the reduplicative segment (-Rdp-) goes is determined by morphology. For verbs it is characteristically immediately after the inflectional (pronominal) prefix, and the segment reduplicated may therefore be the root or a preceding derivational/compounding element (except that epenthetic /-n^gu-/ from P-1 may not be at the beginning of the reduplication). In the case of the bisyllabic -Rdp- type (see below), the input segment may involve an internal morpheme-boundary. For example, in the input form /CV-C₁V₁=C₂V₂C3V3-/ where /-C₁V₁=/ is a derivational or compounding element, the reduplication may be /CV-C₁V₁C₂V₁-C₁V₁- =C₂V₂C3V3-/ (the initial /CV-/ representing an inflectional prefix not considered in the reduplication). Similarly, if the input is /CV=C₁V₁-C₂V₂/ with monosyllabic verb root /=C₁V₁-/ and a following inflectional suffix, the latter may be included in the reduplication if the bisyllabic -Rdp- type is found, hence /CV-C₁V₁C₂V₁=C₁V₁-C₂V₂/ In some cases, reduplication applies at the beginning of the verb root and disregards earlier derivational/compounding elements; there are also a few cases (mostly in narrative/song style involving a specific set of stems) where both the root and an earlier derivational/compounding element are reduplicated. These types are lexically restricted. See "internal" and "double" redup. (§8.4).

The only suffixes which can constitute input are derivational and inflectional verbal suffixes (mainly marking tense/aspect). Other suffixes (including all nominal suffixes) are not involved in the reduplication even when the preceding root is monosyllabic.

Basically, the form taken by the -Rdp- segment is as follows:

Thus /=bura-/ 'to sit' reduplicates as /-bu=bura-/; /=lhan^garma-/ 'to reach, find' as /-lhan^ga=lhan^garma-/ and /=agi-/ 'to fetch' as //-aga=agi-// (which then becomes /-aga:='gi-/ by P-49). Note that in the bisyllabic types the vowel of the second syllable is /V₁/ not /V₂/ as in the root; a further example is /=ma-n^gi/ 'got' (including Past₂ suffix) becoming /-man^ga=ma-n^gi/

Of the three types listed above, clearly the second and third are simply subtypes of a general bisyllabic type, the only difference between them being whether the input (and hence -Rdp-) has or lacks an initial consonant. The basic rule is thus: monosyllabic -Rdp-before input beginning with stop, otherwise bisyllabic type.

While the rule works fairly well as stated, there are some exceptions: a) monosyllabic inputs; b) cases of /-C₁V₁-/ instead of expected bisyllabic /-C₁V₁C₂V₂-/; c) special / C₁V₁C₂-/ used with a few stems. We now discuss these in turn.

With monosyllabic input, not surprisingly we are limited to a monosyllabic -Rdp-. There are not many inputs that are this short. For nouns I can cite only /yilg/ 'silly', rdp. /yi-yilg/ For verbs, we do have some monosyllabic roots which can take inflectional suffixes of the shape /-Ø/ and/or /-C/ and since the input to reduplication ends with the inflectional suffix the inputs here are monosyllabic. The actual inputs are /=C₁V₁-C₂/ with short vowel and /=C₁V₁: -Ø/ with no final consonant but with long vowel. (In some cases the latter type is arguably derived from the earlier type, or from //=C₁V₁-V₂// but this need not worry us here.)

As I hear them, the rdp. forms from these inputs are as follows:

Examples of the first type are a handful of Nonpast₃ verb forms from /=CV-/ verbs, with suffix /-i/ or /-u/ being absorbed by the root vowel by P-49. Thus /=ni:-'/ 'will not see' becomes /-ni:=ni:-'/ 'will never see', and /=yu:-'/ 'will not give' forms /-yu:=yu:-'/

Examples of the second type involve Nonpast₁ forms in /-n^g/ or /-n^y/ and Past₂ forms with /-y/ added to monosyllabic verbs in certain inflectional classes. Examples are /=na-n^g/ 'will see' or 'does not see', rdp. /-na=na-n^g/ and /=lha-y/ 'was standing', rdp. /-lha=lha-y/

I should add that since actual vowel-length contrasts do not occur in these contexts, and since phonetic vowel length is highly subject to intonational skewing, these transcriptions of length are impressionistic. Since Nonpast₁ and Nonpast₃ are infrequently reduplicated, the only common forms are the Past₂ ones, and here the suffix /-y/ occurs only with two monosyllabic roots, /=lha-/ 'to stand' and /=yi-/ 'to sleep'.

We now consider cases of /-C₁V₁-/ reduplication for stems which should give bisyllabic, -Rdp- by the rules given above. Essentially what is going on here is that bisyllabic reduplications which would involve several identical or very similar syllables are simplified. For example, there are some roots which already happen to have /C₁/ and /C₂/ identical (whether or not the vowels are also identical) , as in the cases of the adjectival noun /mamar/ 'empty' and the verb /=lala-/ 'to tear'. Expected reduplications /mama-mamar/ and /-lala=lala-/ with four consecutive identical /CV/ sequences are in fact simplified to /ma-mamar/ and /-la=lala-/ However, the form /=na-ni/ Past₂ of /=na-/ 'to see', always reduplicates as /-nana=na-ni/ with bisyllabic -Rdp-. It appears, then, that this particular alternation in reduplicative shapes may be lexically specific even though a phonological basis is detectable. (We could attempt to avoid this, suggesting that /=na-ni/ behaves differently because its two vowels are different, or because of the internal morpheme boundary, but we do not have a large range of forms which might permit us to identify clearcut patterns of this type. A form /-lhalha=lhalhara-/ rdp. of /=lhalhara-/ 'to fall out', was obtained in an elicitation session, and if valid this clearly shows lexically based variation.)

There are a few other cases of /-C₁V₁-/ for expected bisyllabic -Rdp-. These include /=lu:lha-/ 'to wade', rdp. /-lu=lu:lha-/ where /C₁/ and /C₂/ are distinct but both laterals, and two important verbs beginning in /ma/ /=ma:ndha-/ 'to make' and rdp. /-ma=ma:ndha-/ /=ma:jga-/ 'to make' and rdp. /-ma=ma:jga-/ (Vowel length in root may be a factor here, but many other verbs with long vowel in first syllable behave normally in reduplication.)

A special /-C₁V₁C₂-/ reduplication is used for a small number of roots whose simple form already appears in what looks like a reduplication of this same syllabic shape. Exx. are /=gujguji-/ 'to hold up', rdp. /-guj=gujguji-/ /=bilwila-/ 'to be weak', rdp. /-bil=wilwila-/ Others can be gleaned from §14.5.

This pattern is restricted lexically as well as phonologically, and some former reduplications of this type are being reinterpreted as separate roots. Thus /=w₁uldha-/ 'to cut, sever' (*-gulk-dhu-) and /=w₁ulguldha-/ 'to cut up (wood, etc.)' (*-gulk-gulk-dhu-) are now essentially independent of each other but were once related as simplex to reduplication (the reconstructed forms still occur as such in Ngandi). They now each have their own synchronic reduplication, /-w₁uldhu=wuldha-/ and /-w₁ulgu=wulguldha-/ respectively, and note that the latter shows full bisyllabic reduplication.

We have discussed vowel length in the monosyllabic-root reduplications (preceding page, top), but not for the nonmonosyllables. When the reduplicative segment itself is monosyllabic, its own vowel is clearly short in all cases, while the root retains the same vowel length it has in the simplex: /-lu=lu:lha-/ from /=lu:lha-/ to wade'; /-bu=bu:la-/ from /=bu:la-/ 'to be smoking'; /-ju=jura-/ from /=jura-/ 'to push'. When the -Rdp- segment is bisyllabic, both of its vowels are always short, and if the root has /V₁/ long this is shortened (subsequent vowels are not affected). Thus from /=ya:-ri:/ 'goes' we have rdp. /-yara=ya-ri:/ and from /=yuri:-’/ transports' we have rdp. /-yuru=yuri:-'/ When the root starts with a vowel the same pattern can be presumed, but later rules (P-49) eventually combine adjacent vowels into new long vowels. Thus both /=a:gi-/ to return' and /=agi-/ 'to fetch' have reduplications which can be represented as underlying //-aga=agi-// surface /-aga:='gi-/ (and with the first vowel in the -Rdp- segment usually combining with the final vowel of the prefix to form another long surface vowel).

A few remarks concerning nominal reduplication are appropriate here. In general, the formal pattern is the same as already seen with (mostly) verbal examples. As far as the point where -Rdp- begins, we should note the following. To a greater extent than verbs, nouns tend to disregard compounding initials and derivational prefixes (aside from noun-class prefixes) and begin the -Rdp- with the root. This is particularly true of high-frequency, sometimes irregular, plural reduplications of certain adjectival nouns such as /dhama-lhamun^gur/ from /dhamun^gur/ 'short', /ala:-'la:di/ from /ala:di/ 'bad', and irregular /w₂ura:yun^g/ from /w₂irig/ 'small'. Thus pattern carries over to predicative uses (with verbal prefixes), hence /wu-ran^ga=dhamun^gur/ 'the tree /ran^gag/ is short', plural /wu-ran^ga-dhama=lhamun^gur/ 'the trees are short', not /wu-ran^ga-ran^ga=dhamun^gur/ (However, double rdp. /wu-ran^ga-ran^ga- -dhama=lhamun^gur/ is possible.)

The major exception is that the derivational noun-class prefixes are included in the reduplication. Since reduplication in this context is possible only for semantically human nouns, only human plural derivational prefix /-uG-/ is affected, and only a few stems are involved (since most roots which take derivational noun-class prefixes are strictly nonhuman). Thus from /uG-muwaj/ 'name(d)' (syntactically a noun) we get human plural /uGmu-uG-muwaj/ surface /-umu:-'-muwaj/ Similarly, Gentilic prefix /nuN-/ is included in reduplications, as in /nun^yju-nun^y-jul/ from /nun^y-jul/ bush (inland) person'.

In reading texts, the major difficulty in recognising and analysing reduplications is that the underlying form may be obscured by later rules, most notably VV-Contraction P-49. Suppose we find a form /n^gawa:ra:ru:/ 'I kept leaving it (behind)'. To clarify the structure we first retranscribe it with punctuation as /n^gawa:-'ra:='ru:/ following the system used in my earlier publications. Here / ' / indicates elision of morpheme-initial vowel.

It is now easier to see the underlying structure //n^gawu-ara=aru-u// (the root being actually /=a:ru-/ . As another example, consider two forms /n^gani:na:nani/ 'he saw (it) for me' and /n^ganinananani/ 'he kept seeing me'. The latter is straightforwardly /n^gani-nana=na-ni/ but the former is less simple. By first rewriting it as /n^gani:na:-'=na-ni/ we can see that the input form differs from the input to the other form in having Benefactive prefix /-aG-/ just before the verb root, hence //n^gani-Rdp-aG=na-ni// The underlying form of the reduplication is thus //n^gani-aGna-aG=na-ni// which ends up as /n^gani:-'na:-'=na-ni/ Readers working through my texts will find that the punctuation helps clarify word structure, but those transcribing new material or working through texts lacking such punctuation may sometimes find it difficult to identify and correctly analyse some reduplications.

We summarise the formulation of the rule as follows: