The evolution of the Mawes Aas’e (Omotic-Mao) pronouns: evidence for Omotic Lineage

2.2.1 Mawes Aas’e

MA has three sets of bound pronominals^[9] which mark the person and number of subjects: a set of prefixes which occur on the realis verb, a set of suffixes which occur on the irrealis non-future negative verb, and a second set of suffixes which occur on the irrealis future verb (on both the affirmative and negative verbs). These bound pronominals and their corresponding free pronoun forms are provided in Table 2.

The prefixes, which are found on realis verbs (column 2), bear the clearest resemblance to the free pronoun forms (compare column 1 in Table 2 with examples 1–9, below). The primary distinctions include the presence of the terminal nominal vowel [-e] on all free pronouns, the [-el] plural suffix on the 1st and 2nd plural free pronouns, and the fact that the 3rd person series is zero-marked as a prefix (ex. 3, below) while the (3rd person) non-singular forms are marked with a suffix /-and/ following the verb stem (4).

There is a non-obligatory /ha-/ AFF (affirmative) prefix that very frequently appears on affirmative realis verbs with 1SG and 3rd person subjects (1–4).^[10] The /ha-/ is also included in the base (3SG) realis form which is used as the citation form.

(Table 2 is from Ahland 2014:63).

On realis verbs with 1st and 2nd person non-singular subjects (i.e. 1DU, 1PL, 2DU and 2PL), the subject prefix begins with a [ha] sequence and no additional /ha-/ affirmative prefix can be affixed to the verb (5–8, below). This sequence [ha] is always present on the 1st and 2nd non-singular person realis prefixes and is also found on the corresponding free pronouns (Table 2).

It will be demonstrated in Sections 5.1.2 and 5.2.3 that the [ha] sequence which is today part of these subject markers and their corresponding free pronouns is etymologically connected to the /ha-/ affirmative prefix through an earlier stage of development.

The 2SG subject marker on realis verbs is /hì-/, corresponding to the pronoun /hìj-è/. The /ha-/ prefix is never attested preceding the /hì-/ subject marker.^[11]

MA irrealis verbs, on the other hand, carry bound pronominal suffixes. The first paradigmatic set is similar to the realis prefix form and is used for the irrealis non-future negative verb (column 3 of Table 2). The second set exhibits a more divergent paradigm (column 4 of Table 2); this paradigm is found on the irrealis future verb (regardless of polarity). Neither of the subject-marking suffix paradigms includes any [ha] sequence associated with any person-subject. Each set is further discussed and illustrated below.

The irrealis non-future negative verb, like all negative verbs, never carries the /ha-/ affirmative prefix. Any attempt to use the /ha-/ prefix on negative verbs results in ungrammaticality (Ahland 2012:393–394). Other features of the MA negative verb include the use of the infinitive verbal stem (marked as such by tonal melody), a negative suffix, the subject-marking suffix and then a bound auxiliary (cf. Ahland 2012:355–374). Examples (10–13) illustrate the 1st person (SG, DU, PL) and 2SG suffixes.

The 2DU and 2PL suffixes are illustrated in (14–15), respectively. The dual/plural distinction in 2nd person is expressed by a tone difference realized on the bound auxiliary (high tone for DU and low tone for PL).

When subjects are 3rd person, a different negative suffix and bound auxiliary (in this case a copula) are used. As with the realis subject prefixes, the 3rd person is zero marked for the 3SG (16) and the DU and PL are expressed as non-singular through the /-and/ suffix (17).

It’s important to highlight that while the subject-marking paradigm associated with the irrealis non-future negative exhibits consistent correspondence to the forms found in the realis subject prefixes, there is no hint of any [ha] sequence on the 1st and 2nd person non-singular forms, as is found on the realis prefixes and also on the free pronouns (Table 2, above).

The more divergent subject-marking paradigm associated with the irrealis future forms is illustrated below (18–25). The details of this divergence are not central to the argument in this paper (but they do play a role), so they are summarized only briefly here. In short, there is an intrusion of an [m] in 2SG and 3rd person suffixes in the irrealis future paradigm. This has been shown to be the result of perturbations from a preceding future suffix (/*gàm/) which had a final [m] and which immediately preceded the bound pronominals (see Ahland 2014 for the details). Essentially, while the final [m] was lost before stops such as /-t́/ 1SG, /-ń/ 1DU, /-m̀/ 1PL, and before /w/ in 2DU (/-ẃ/) and 2PL (-/ẁ/), due to phonotactic constraints, the [m] remained in other environments (see examples 21, for 2SG, and 24–25, for 3rd person, below) and entered the paradigm, being reanalyzed as part of the bound pronominal itself.

In a later development, the 2DU and 2PL eventually weakened, resulting in compensatory lengthening of the vowel on the future suffix. The tone of the future suffix serves as the DU versus PL number distinction (22–23).

2.2.2 Ganza

Ganza’s pronouns exhibit relevant SG and PL categories across 1st, 2nd, and 3rd persons as well as binary gender distinctions in 3SG (Table 3).^[12] Ganza’s bound pronominals are all clitic forms which attach to a variety of hosts (preverbally) as well as the verb itself in future and negative verbal constructions.^[13] Ganza’s free pronouns and bound pronominals are provided in Table 3.

Throughout this section, data that are not cited have been taken from the author’s own fieldnotes. The sources of all other data are cited accordingly.

Ganza’s bound prononominal distribution is relative to non-future versus future tense domains on verbs as well as polarity within non-future tense verbs. Table 4, below, summarizes the full range of distributional possibilities identified thus far.

In non-future affirmative verbs, the bound pronominal person clitics can attach as enclitics to several hosts (all of which precede the verb); these include the affirmative marker /hǎ/, overt nouns (NPs), and overt (full) co-referential pronouns. Ganza’s person clitics can also procliticize to the verb in affirmative non-future constructions. Interestingly, negative non-future tense verbs do not take pronominal clitics at all. Rather, they require the full overt pronoun. The future tense verb constructions position the bound pronominal clitics after the verbal root; in these constructions, the clitcs are found encliticized to the future tense suffix on the verb itself. Examples of all these distributional distinctions are illustrated and discussed below.

Let’s begin with the affirmative non-future tense verbs and their four possibilities. The most frequently attested preverbal host for Ganza’s clitic pronominals is Ganza’s affirmative form /hǎ/; this is, of course, very likely cognate with MA’s /ha-/ AFF prefix and the form [ha] that has entered into MA’s pronominal inventory. While Ganza’s affirmative form is not bound as a prefix, it does precede bound pronominal markers of subject person in a preverbal position.^[14] Examples (26–32) illustrate the behavior of Ganza’s pronominal enclitic subject markers in non-future, affirmative utterances where they are hosted by the /hǎ/ form.

While the AFF + enclitic complex is always preverbal, it is not always immediately so. There are examples in my data of instances where intervening material is clearly present between the host + clitic and the verb (as in 33). This example, with an intervening overt subject NP, shows that the host + clitic can be positioned outside the verb phrase.

In instances where there is no overt NP (as in 33) and when the host + enclitic complex is not immediately preverbal, the full/free pronoun appears to be required (34–37).

Ganza’s pronominal clitics can also be hosted by overt nouns/NPs. This is particularly common in interrogative utterances, where the affirmative prefix is not found (38). In (38), the host is a full case-marked object pronoun.

Alternatively, in some polar interrogatives, /ha/ appears in the same position as the host /wǎ/, which has been found as a subject marking host only in interrogatives thus far.

Polar interrogatives in Ganza can take the /ha/ form (as they can in similar constructions in MA, where the expected answer is affirmative).

Example (43) shows a preverbal NP object serving as the host for the pronominal clitic in a declarative utterance--illustrating the possibility for hosts (other than /hǎ/) in declarative sentences as well.

In affirmative non-future verbal construction, there are also instances where the pronominal clitic in Ganza is hosted by a co-referential, free pronoun (encoding the same person, number, gender and case information). In example (44), the 3SG.F pronoun hosts the 3SG.F enclitic, while the /hǎ/ AFF marker is phonologically bound to the beginning of the verb.^[17]

Finally, and perhaps most perplexing of all, there are at least a few instances in my own notes (and also in Smolders’ fieldnotes), where the person clitics have procliticized to the verb (45–46).

Evidence that these forms are procliticized is based on speakers’ word-boundary pauses in hyperarticulated speech as well as asking speakers to repeat words (through questioning) and the procliticized form is provided with the verb.

My fieldnotes also contain a few instances where the affirmative marker is included with the person clitic and both are together bound to the beginning of the verb (47) and in some of these cases, phonological reductions in the person clitic are clear (48).

It’s worth noting that the phonologically reduced subject-marking complex (hǎ=gu AFF=3PL > hà=w) which precedes the verb in example (48) is strikingly similar to the 2DU and 2PL subject marking forms found in Mawes Aas’e (/háw-/ and /hàw-/, respectively)--see Table 2, above. Presumably, these four different positional/hosting distinctions for the person clitics in non-future verb constructions involve some pragmatic conditioning. But at this point, no fully interlinearized texts are available in Ganza, so analysis of pragmatics is not possible at this time.

Bound pronominal clitics are prohibited in negative non-future verb constructions. They have not been found in any negative non-future constructions and were not acceptable to speakers when attempted. Rather, speakers prefer the overt, free pronoun and position it preverbally (49).^[18]

While Ganza’s affirmative non-future constructions (above) generally encliticize person clitics to hosts positioned preverbally, future tense verb constructions position these clitics after the verbal root, following the future tense suffix. In (50–52), the affirmative future form of the verb is formed with subject clitics positioned after the future tense suffix, before the declarative suffix.^[19]

The negative future tense verb also positions the bound person clitics after the future tense suffix (53).^[20]

After examining subject marking in Seezo, the discussion will again return to Ganza, as the affirmative marker in Mawes Aas’e and Ganza are carefully examined.

2.2.3 Seezo

Seezo’s pronouns exhibit SG versus PL opposition across 1st, 2nd, and 3rd person categories; the 1PL category is subdivided by clusivity, showing the only inclusive versus exclusive distinction found in the Mao languages (Girma Mengistu 2015:216). Seezo’s bound pronominals have been described as proclitics (Girma Mengistu 2015:216). Table 5, below, provides the full set of subject-marking clitics as well as the full pronouns (bound roots in Seezo) for comparison.

Apart from phonological boundedness, the only differences between the full pronouns and the clitics involve the lack of vowel length in the 1SG and 1PL.EXCL forms as well as unique 3SG forms (a zero on main or final verbs and the form /hí= / on dependent verbs, cf. Girma Mengistu 2015:216). Girma Mengistu describes the distribution of the clitics relative to realis/irrealis category: “When they occur with the realis verb, they are procliticized to the verb root. But when they occur with the irrealis verb form, they are procliticized to tense marking auxiliaries” (2015:216).^[21] It is important to note then, that on the irrealis verbs, the subject clitics follow the verb root (as is the case with Mawes Aas’e and Ganza). This distribution is summarized in Table 6.

Examples (54–55) illustrate the proclitic subject markers on realis verbs, while examples (56–58) illustrate the subject markers procliticized to the auxiliary (following the lexical verb root) in the irrealis verbal construction.

4.1.1 Proto-Mao pronominal reconstructions

Bender’s pronominal reconstructions for the Mao group (for 1SG, 2SG and 2PL) still work well (Bender 2000:196), despite the new data that are available today from the new Mao reference grammars and other published research. For 1SG, the /*ti/ in Bender’s Mao reconstruction is attested in both MA and Ganza and is close to the /*ta/ in his Proto-Omotic. The 2SG form /*hi / j/ is also Bender’s Mao reconstruction. With respect to 2SG, only Ganza’s 2SG subject enclitic is divergent, with /=na/. But it’s worth noting that Seezo also exhibits an [n] in 2SG: /hín/ for the Seezo pronominals. Perhaps the [t] form for 1SG and these [n] forms are a reflex of an older system which is related to the TN group (so named because of the 1st and 2nd pronominal consonants, see Bender 2000:223). My suggestions for Proto-Mao pronominals are provided in Table 8.

In Table 8, no pronominals are posited for 3rd person. In looking across the Mao languages, it is clear that 3rd person is achieved through grammaticalization of language-internal demonstratives and given the lack of any clear pattern across the group, I assume a variety of demonstratives may have taken on this function over the years.^[36] It is likely, also, that these demonstratives would have expressed a masculine versus feminine gender distinction as is maintained today in both Hoozo and Ganza (see Table 7) and across many non-Mao Omotic languages.

The 1st and 2nd person plural forms are where we find the greatest divergence and challenges for reconstruction. There is evidence for two Proto-Mao forms in the 1PL. First, Bender’s reconstruction for Proto-Omotic /*nu/ does appear to have a reflex in at least two (possibly three) of the Mao languages (Bender 2000:223): Hoozo’s 1PL form /nú/ and also the [n] in MA’s 1DU /hané/. The data in both Hoozo and MA suggest this 1PL form carried a H tone: /*nú/ Proto-Mao 1PL form. Sound change evidence suggests that the Proto-Mao alveolar nasal in 1PL was also possibly inherited to Seezo as well (see Table 9 and the surrounding discussion below). The second 1PL form (/*mù/) is my own reconstruction. I reconstruct this due to the bilabial nasals and low tones found in Ganza’s /mù/ 1PL form, as well as the base [m̀] consonant in MA 1PL: /hambèlè/.^[37] The respective H and L tones of these forms are seen reflected across the Mao group consistently (Figure 3).

Figure 3:

Reflexes of Proto-Mao 1PL forms across Mao group.

As can be seen in Figure 3, I argue the 1PL proto-forms expressed an inclusive versus exclusive distinction; my assumption here is that the inclusive/exclusive distinction reflected in Seezo’s 1PL today is a retention. While MA’s dual distinction is very unusual in Omotic, clusivity distinctions have been noted in multiple languages of different subgroups of the TN branch of Omoticː Koorete, Benchnon, Zayse, Zargulla (Azeb Amha 2017:828). Perhaps the /*nú/ form was actually a minimal inclusive (i.e. an inclusive that does not include a 3rd person: including 1 + 2 but not 1 + 2+3) as a precursor to MA’s dual distinction.^[38] The /*mù/ exclusive is the source for MA’s PL. The full discussion detailing the MA developments from Proto-Mao to today are provided in Section 4.2. Given the fact that each of the Mao languages shows radical divergence in 1PL, this two-form reconstruction provides a tenable solution that brings the group together.

While the focus of this paper is on the developments in MA, discussions of possible changes in each of the languages are mentioned briefly here. First, any clusivity associated with Proto-Mao 1PL was clearly lost in both Ganza and Hoozo, with each exhibiting only single (albeit different) forms for 1PL today (Figure 3).

In Seezo,^[39] we find a more complicated picture. The tonal pattern of H and L on the inclusive and exclusive, respectively holds, but the morphological shapes of the pronominal forms are perplexing. It appears that the Seezo inclusive form underwent a commonly-attested sound change (n > d) and that the exclusive 1PL form in Seezo could have assimilated via analogy. Certainly, there is comparative evidence of an (n > d) sound change attested in Seezo (Table 9).

In Table 9, Seezo cognate words regularly attest the *n > d change. Hoozo forms show inconsistent participation in the sound change with two cognate forms evidencing the change and the other retaining the older *n form. Ganza more consistently retains the *n in cognate words. MA does not show the sound change and maintains *n in all examples.^[40]

The [ol] in Seezo’s reconstructed inclusive form (Figure 3) could be cognate with the /-(w)ol/ plural we find in MA and the /-el/ plural that we find in MA’s 1PL and 2PL pronouns. It is perhaps surprising to find a possible PL marker associated with the inclusive form within a clusivity set and not with the exclusive form. If the form is related to Seezo’s 1PL, then it must have been an internal development and not an externally induced change nor a part of the pronoun that is reconstructable for the whole Mao group. It could also be the case that the MA /-el/ plural form could be cognate with Seezo’s 3PL /hél ∼ jél/ forms. Finally, it’s worth noting that Bender has mentioned an /-el/ form in Dizoid languages (along with the MA form) as a plural “fragment” in Omotic (Bender 2000:213).

The 2PL in Mao is clearly innovative when one considers the rest of Omotic. We don’t find the expected /int/ or /it/ forms that Bender notes for other branches of Omotic (2000ː223). Bender offered /*nam/ for Proto-Mao’s 2PL, and given that two of the Mao languages exhibit this form (with polar tonal differences), the choice is reasonable (given lack of counterevidence) and even more so, when we come to Ganza’s subject enclitic /=ma/. The labial-velar approximant [w] in MA is perhaps a further weakening of the bilabial, intervocalically.

4.1.2 The *ha enclitic host and paradigmatic intrusion

Today’s patterns in Ganza (Section 2.2.2) suggest that these enclitic forms in Proto-Mao would have had a variety of hosts, not only for interrogative or affirmative utterances, but also as a means for establishing focus on different elements (as seen in other Omotic languages, Section 3). Common hosts were likely NPs, pronouns, modal markers (e.g. Ganza’s /wǎ/ today) and, most importantly, the /*ha/ affirmative form, derived from a proximal demonstrative / locative noun ’here’ that has been reconstructed for much of Omotic (Bender 2000:206).^[41]

In MA and Ganza, the /*ha/ is reflected as an affirmative marker in terms of distribution and function. In MA, the affirmative marker /ha-/ is optional (for those person subjects where it has not fused with the realis subject markers)^[42] in declarative and hearsay verbal constructions. The /ha-/ affirmative is required in polar interrogatives, at least in those where the expected answer is affirmative (see Ahland 2012:474–475). MA’s /ha-/ is prohibited in content interrogatives, imperatives, jussives and negatives (Table 10).

Unlike MA, Ganza’s affirmative marker is not fused with any of the subject enclitics. After careful examination of my own fieldnotes, Hofmeister’s texts and manuscript, and Smolders’ fieldnotes, I have been able to compile a distributional chart for Ganza’s affirmative marker; its distribution is strikingly similar to MA (Table 11).

Without a full grammar of the Ganza language, there simply are not enough data to be sure about when the affirmative marker is optional versus required. It is possible, though, to observe where it is and is not attested in the available data, and this is what Table 11 offers.

Rather than assuming that MA and Ganza each developed an affirmative marker independently, it is more reasonable to posit that this was part of the Proto-Mao state and the distribution today in these languages is a retention. But to do this, then raises the question about where this /*ha/ would have gone in Seezo and Hoozo.

There is evidence the /*ha/ form may have intruded into Seezo as well.^[43] It appears that Seezo’s pronominal inventory bears the mark of an intrusive [ha] form. First, the 1SG pronoun /hàː/ and the corresponding enclitic /hà=/ are not reconstructable for 1SG for the Mao languages. The form is clearly innovative and, given Ganza and MA, a reasonable hypothesis would be to consider the common historical host for the old subject markers. The 3SG pronoun in Seezo also shows an unexpected intrusionː /hán/. Neither of these Seezo pronominals can be reconstructed as part of the pronominal inventory directly. But taken together with MA and Ganza data, intrusion of a historical /*ha/ host seems reasonable.

While there is no evidence of /*ha/ intruding into the personal pronoun paradigm in Hoozo (see Table 7), the interrogative pronoun ‘who’ is /hà/ as seen in multiple case forms of the word as well as the one-word utterance /hàgaba/ ‘who?’ (Getachew Kassa 2015:136).^[44]

Apart from the Mao group, a suspicious [ha] form can be found in pronominal subject marking paradigms in the Omotic-Dizoid languages. In Sheko (a Dizoid-Omotic language, Hayward 2000; Bender 2003), 2SG and 3SG subjects are marked with proclitics /ha=/ and /há=/, respectively (Hellenthal 2010:323). Bender does not reconstruct these [ha] forms as part of the pronominal paradigms or as part of the subject marking system (2000:223). In Dizin (closely related to Sheko), there are two forms which could be a further simplification of the [ha] form seen in Sheko: possessive markers and subject proclitics show the forms /à=/ and /á=/ for 2SG and 3SG, respectively (Beachy 2005:245).

Apart from the pronominal systems, the form /ha/ is found throughout Omotic as the proximal demonstrative ‘this’ and as the locative noun/deictic ‘here’: in Macro-Ometo (NW Ometo and SE Ometo) (Bender 2000:79); in the Gimira-Yem-Kefoid languages (Bender 2000:138); and in the Dizoid languages (Bender 2000:145). Bender, in fact, reconstructs /*ha/ (with variants /*har-/ in C’ara and /*han-/ in Kefoid) for the proximal demonstrative and ‘here’ form in his TN group of Omotic (2000:206)--that’s all of Omotic except for Mao, Dizoid and Aroid, according to Bender’s classification (2000:206). It is clear that an Omotic source for MA’s and Ganza’s affirmative marker and the [ha] forms is likely, given the many correspondences of /ha/ as a demonstrative and locative noun (and as a pronominal form today).

4.2.1 Phonological reduction in 1PL and 2PL forms

After MA broke away from Proto-Omotic, the 1st and 2nd person enclitics simplified in shape. For the 1PL forms, this involved the loss of the final vowel with the tones maintained. For the 2PL, /*nam/ likely reduced to an intermediate /=mà/ (as is maintained in Ganza’s 2PL enclitic today, see Table 7) before the final vowel was lost and the bilabial nasal weakened to the approximant [w]--again, the tone is maintained (Table 12).^[52]

These host + enclitic composite forms could move throughout the preverbal domain in realis verb constructions, as has already been established for Proto-Mao. Only the enclitic forms would have been used in irrealis verbal constructions; in those constructions, the verb itself would have been the host.

4.2.2 Development of dual in 1st person and spread to 2nd person

Another important development is the reanalysis of the Proto-Mao 1PL inclusive versus exclusive distinction into one of dual versus plural. The comparative consonant and tonal pattern for the 1PL reconstructions suggests that the H-tone form (i.e. the form associated with inclusive in Proto-Mao and in Seezo) is the one that was reanalyzed as a dual.^[53] It’s clear that the dual is an innovation in MA and is not a pattern that can be reconstructed for any Omotic subgroup.

In MA, the reanalysis of the 1PL inclusive occurs in the first stage of innovation. There is no formal change--only a meaningful reanalysis to dual number, leaving the historical exclusive as the sole plural form in 1st person (Table 13).

Cysouw notes that crosslinguistically innovation of a dual opposition typically spreads across the person paradigm:

When there are grammaticalized dual forms in a pronominal paradigm, then it is extremely rare for there to be no dual involving all persons. Dual forms show up across the paradigm, or not at all. There are only very few examples that have a dual only in a part of the pronominal paradigm. Among these few examples, a dual in in the first person is indeed slightly more frequent then other duals, but the total amount of cases is too low to allow for any significant generalizations. (Cysouw 2003:210)

In MA, this tendency for dual to spread across persons holds. In stage 2 (Table 12), the 2PL could then have subdivided, following the same tonal pattern observed in 1st personː H tone for DU and L tone for PL. What’s interesting here is that because there was no pre-existing morphological material which could serve as the dual form, the labial-velar approximant root form appears to have been copied from the plural form for the new dual new structure (with the H tone for dual replacing the plural-indicating L tone)--perhaps an analogical extension of the tonal pattern found in 1DU and 1PL.

Cysouw’s typological work suggests that an inclusive/exclusive paradigm (without dual) is a possible source for the type of pronominal paradigm we find in MA: SG, DU, PL forms across three persons but with no inclusive/exclusive distinction --a paradigmatic pattern Cysouw calls ’dual-unified-we’ (Cysouw 2003:278). Citing data from the Miwok languages (Plains Miwok, Northern Sierra Miwok, and Bodega Miwok), Cysouw suggests the older 1PL inclusive is cognate with the new 1DU, while the older 1PL exclusive appears to be related to the 1PL in the resulting paradigm (Cysouw 2003:277–278). Interestingly, the older 1PL inclusive in Miwok was “probably only prototypically used for the speaker-addressee dyad 1 + 2” and was a general inclusive rather than a strict minimal inclusive (Cysouw 2003:277). So, even in cases without a minimal inclusive, deriving a dual from an inclusive is not without apparent precedent.

Before continuing, it’s important to note that the phonological reduction of pronominal enclitics and the innovation of dual in these two stages provides a unifying account of an otherwise perplexing pronominal/subject-marking paradigm (as originally seen in Table 2, above). While the innovated host + enclitics in Table 13, are the same morphological shapes found on MA realis verbs today, the phonological reductions of the reconstructed pronominals and their tones (without the /ha/ host) match the forms found on irrealis non-future verbs (see shaded rows in Table 14).

In MA realis verb constructions, the enclitics would have needed a host as they moved within their preverbal domain (in the same way as we see in Ganza today). But in irrealis constructions, the verb itself would have been the host: in the non-future, the enclitics attach to the verb itself and in the future, they attach to the verbal word, but are positioned after the future suffix. It is after this future suffix that the 2nd person non-singular forms weakened even further, losing all segmental material and maintaining only the H and L tones for dual and plural, respectively.^[54]

4.2.3 Fusion of the /ha/ host with 1st and 2nd person non-singular forms

The fusion of the affirmative marking /ha/ host with the 1st and 2nd person dual and plural subject-marking enclitics is crucial to the understanding of pronominal innovations in MA. The fusion of the affirmative /ha/ host with these enclitic forms (and the later impact on the shape of new free pronouns, discussed below) strikingly sets the MA pronominals apart from the rest of Mao and Omotic. The [ha] shape was a major reason for Bender’s argument that MA pronouns were the result of borrowing from Nilo-Saharan (see Bender 2000:184 and also Section 5.3, below).

Questions central to the fusion of the affirmative host with these particular enclitics include the following: 1) Why is this fusion not seen on irrealis verbal subject marking in the corresponding persons? and 2) What motivates the fusion on the 1st and 2nd person non-singular forms (and thus does not trigger the same for other persons)?

First, as perhaps is obvious, the fusion could only apply in the preverbal realis domain. This is the domain where the /ha/ affirmative form was hosting the enclitics. In cases where the verbal word was serving as the host for the enclitics (e.g. the irrealis verb), the subject markers remain unaffected. Second, the motivation for this fusion could have been due to phonological necessity: these particular 1st and 2nd person non-singular enclitics have weakened to single consonants, as this was unfolding, they would have become less easily hosted by elements ending in consonants because they were all consonants themselves. MA’s specific phonotactic requirements of complex codas would have prevented any of these (/ń/, /m̀/, /ẃ/, and /ẁ/) forms from serving as the second C of a CC coda.^[55]

The fusion of the /ha/ affirmative host with the enclitics is represented in Table 15.^[56]

Presumably, these fused forms would have maintained their freedom to move within the preverbal domain (as is still the case in Ganza).

The results of the /ha/ fusion are important. First, the fusion of the /ha/ host in these cases renders the realis subject-marking paradigm independent from the irrealis subject-marking paradigm in MA.^[57] And most important for our focus here, this fusion is the initial intrusion of the /ha/ host into the MA pronominal system. We now have obligatory (i.e. highly frequent and salient) markers for all realis constructions which involve a combination of subject markers fused with their erstwhile host and others (like 1SG and 2SG) that are not fused. These moveable, preverbal, obligatory subject markers also show a dual versus plural distinction in 1st and 2nd person. But of course, there was a problem: the old free/full pronouns would not have matched in shape with the intruded /ha/ nor with the innovated dual versus plural distinction. This would have been an issue for expressing persons in a number of important environments: as grammatical objects, establishing emphasis, contrastive focus and for responding to questions of ’who’ (just to name a few).

4.2.4 New free pronouns through augmentation

An examination of synchronic paradigms in MA may initially lead one to assume that the bound subject marking on realis and irrealis verbs is the result of reductions from the free pronouns. The problem with such an analysis, of course, is that the free pronouns contain various elements which are not reconstructable as part of the Mao pronoun system: most strikingly the [ha] shape on four pronouns and the [el] shape on first and second person plural pronouns.

Certainly, there is a strong case to be made in the historical literature for the idea that bound morphological forms develop from a further grammaticalization and reduction of free pronouns. This pathway of reduction from free, full pronoun to bound (or otherwise reduced) form (including agreement) has been well-established in the literature: Meillet 1912, Givón 1976, Hopper and Traugott 1993, Ariel 2000--among the many prominent works.

But the internal facts in MA do not support such an analysis in this particular instance. If today’s realis prefixes are reduced from the pronouns themselves, how did the [ha] form, which cannot be reconstructed as part of the pronouns, enter into the paradigm? Why do the 1PL and 2PL pronouns transparently exhibit internal morphology, such as the /-el/ PL marker? The answer to these questions lies in how today’s free pronouns (1DU, 1PL, 2DU, and 2PL) developed via augmentation (with necessary morphology) from copies of the newly fused obligatory subject-marking forms (the erstwhile /ha/ + enclitic).

There is growing evidence that the relationship between free and reduced/bound pronominals is not always unidirectional (i.e. less grammatical to more grammatical). More recent work has established that there may be developmental cycles whereby free pronouns can produce new reduced/bound pronominals and where the reduced (often obligatory and bound) pronominals can spark the development of new pronouns as well (see Harvey 2003, Mushin and Simpson 2008, and Norde 2009:204-207). Mushin and Simpson conclude:

This leads to the suggestion that, if a language has different forms for free and reduced pronouns, all else being equal, the reduced pronouns should be taken as representing the earlier stage of the language, and the free pronouns as being the result of paradigm extension, or borrowing, or innovation on the basis of the reduced pronouns with additional material (Mushin and Simpson 2008:591)

The developments in MA provide evidence that while the bound pronominal forms very likely derived originally from the free pronouns in the Proto-Mao state, innovations within the obligatory bound pronominal forms (fused host + enclitics) would have created a need for new pronouns to develop through augmentation based off of these obligatory forms.

First, given that the subject marking enclitics were obligatory in all clauses, as the subject marking device, and given that these most frequent forms had developed innovations related to dual number in both 1st and 2nd person, updates in the free / full pronoun paradigm would have been needed so that all relevant person/number categories could be expressed. These enclitics with their hosts, of course, would have been already moving freely preverbally, but they were limited functionally as subject markers and were not used for many of the discourse functions for which free pronouns are required. Speakers of MA, appear to have done just as Mushin and Simpson suggest: they appear to have constructed new pronouns on the basis of the fused host + enclitic 1DU/PL and 2DU/PL forms. In this case, the additional material involved the toneless terminal vowel /-e/ that marks all nominals (including the other MA pronouns) and the addition of the /-el/ PL suffix to the 1PL and 2PL forms (Table 16).

The tones of the erstwhile enclitics /-ń/ and /-m̀/ in 1DU and 1PL respectively spread rightward onto the toneless suffixes. In a related manner, the tones which had already associated to the vowel of the first syllable in the 2DU and 2PL form, also spread right to the toneless suffixes.^[58]

The only other change that appears to have occurred involves a sound change in the 1PL pronoun. Strengthening of intervocalic bilabial nasal through the addition of a homorganic stop (a subtype of excrescence [m] > [mb]) before a vowel is quite commonly found in languages in western Ethiopia. For instance, consider the word ‘camel:’ /gɨmɛl/ in Amharic; /dʒəmel/ in Arabic; /kambəla/ in Gumuz (Colleen Ahland, personal communication); /hambel/ in Bertha (BGLDP 2007:112); and /hàmbèlè/ in MA (Ahland 2012:252). Cross-linguistically, this sort of nasal-stop strengthening is relatively common, where [m] is fortified to [mb] before vowels. Such a change is reported for Haida (Mithun, p.c.), Maxakalí (Gudschinsky et al. 1970), and throughout Amazonia (Hyman 2007:351).^[59] With the completion of this one sound change, we find today’s free pronouns for the 1st and 2nd non-singular forms in the final column of Table 16.

4.2.5 Subject-markers collapse into prefixes

Given the comparative data from Ganza and Seezo, it seems likely that even in the Proto-Mao stage, the Mao subject marking enclitics could at times procliticize to realis verbs. Certainly, as demonstrated above, in Seezo, realis constructions today require the procliticization of subject markers to the verbs. And in Ganza, procliticization to the realis verb is not entirely infrequent in my fieldnotes. Of course, in MA today realis verbs obligatorily carry subject marking prefixes which derive from the old enclitic system.^[60]

Regardless of how widespread the procliticization of the subject markers to the verbal word was in the earliest stages, it’s clear that in MA today, the collapse into the verb is complete. As noted in Section 2.2.1, realis subject markers are obligatory prefixes and the required verbal inflection for the expression of subject.^[61]

The relative chronology of the collapse into inflectional prefixes and the development of the new free pronouns is not entirely clear. It seems to me that an increase in frequency of procliticization of the subject marking host + enclitic composites (and certainly the complete collapse of these forms into verbal prefixes) could have been a major motivation for the development of the new free pronouns. Cysouw suggests such a situation:

…free and bound forms can also arise because the original clitics became ‘real’ inflectional markers and, as a consequence, free pronouns had to be remade by reinforcing [augmenting] the clitics. Such reinforcement of reduced pro-forms is quite common cross-linguistically. A well-known example is the development of the Latin pronouns nos/vos, which were reinforced in Spanish as nos-otros/vos-otros (Cysouw 2004:10).

4.2.6 A new 3rd person pronoun series

Finally, we arrive at what may well be one of the more recent developments in the MA pronominal paradigm: the development of a new 3rd person pronoun series. A quick inspection of Table 7 again shows widespread variation in Mao 3rd person forms. As briefly noted above in the discussion around Tables 7 and 8, there are strong correspondences between each 3rd person pronominal form and language-internal demonstratives across the Mao languages. No one single system or single demonstrative source can be reconstructed for the whole group. Ultimately, it appears that a variety of demonstratives have served as the base for the 3rd person pronouns.

In the case of MA, the base is the distal demonstrative /íʃ-é/ DIST-TV--the final vowel is the toneless terminal vowel. The 3SG pronoun includes the demonstrative base with a L tone on the terminal vowel--perhaps as a means for distinguishing the 3SG pronoun from the distal demonstrative synchronically; both are still in use. The dual and plural forms include the number and terminal vowel morphology expected of other nominals (Table 17).^[62]

Bender reports an array of possibly related isomorphs of various shapesː /izi/ Macro-Ometo, /isi/ Gimira, and /iz/ Dizoid, associated with 3SG, and /ist/ Macro-Ometo, /ic/ Gimira, and /iʃ/ Dizoid, associated with 3PL (Bender 2003:223). I don’t believe there is a clear case for reconstruction of such a form across Omotic. I do suspect it is more likely that in various subgroups, a cognate demonstrative has been brought in to the pronominal paradigm, independently. In the case of MA, the transparent quality of the morphological forms suggests the incorporation of the demonstrative to be more recent, sometime after a stage where any gender that would have been present in Proto-Mao had been lost.