Nominal-type word-stem morphology

When establishing nominal-type word stems, it might occur to one to follow the three-vowel split system observed for verbs in the Finnish and apparently the Estonian Schools of Mordvin studies (cf. Ravila 1929: 104-105; Pall 1996: 22; Bartens 1999: 122; Hamari 2007: 66). The verbs, it is maintained, can be divided into three groups on the basis of which vowel precedes the ms segment in the in nitive: a (pala%ms ‘‘to kiss’’), o (vano%ms

‘‘to watch’’) or e (ńiĺe%ms ‘‘to swallow’’). The problem with this three-way split is that, while a%stem verbs always retain their vowel in IND.PRETI.PRED-3SG marking, the two mid-vowel-stem verbs given lose theirs, hence pala+ś kiss_V+IND.PRETI.PRED-3SG shows vowel retention, whereas van+ś watch_v+IND.PRETI.PRED-3SG and ńiĺ+ś swallow_v+IND.

PRETI.PRED-3SG do not. What makes this awkward from a point of concatenation is that there are also mid-vowel verb stems that retain their vowels, e.g. pid́ems ‘‘to cook’’: pid́e+ś cook_v+IND.PRETI.PRED-3SG and udoms ‘‘to sleep’’: udo+ś sleep_v+IND.PRETI.PRED-3SG.

In the most recent Erzya-Russian dictionary (ɗɪɡɹɧɶ-ɪɭɡɨɧɶ ɜɚɥɤɫ, 1993), hence-forth (ERV 1993) a vertical separator ““|”” is implemented to indicate a breaking point in the headword where in exion of the various word types can readily be conjugated or declined. In the instance of verb headwords, this means that an additional indication of the indicative preterit I third person singular will be given, and in the case of noun head-words the inde nite nominative plural; for some reason, however, adjectives and other modi ers are not systematically marked.

The role of the vertical separator in verb headwords is to indicate whether the vowel occurring before the ms in nitive marker is, in fact, a stem vowel or a linking vowel; the IND.PRETI.PRED-3SG marker %ś has no af x-initial vowel (see also Tsypkaikina 2000: 156).

On the basis of the above, we can hypothesize two verb-stem types, i.e. verbs that retain their stem vowels in the IND.PRETI.PRED-3SG slot, and those that do not. Thus the verbs palams ‘‘to kiss’’, pid́ems ‘‘to cook’’ and udoms ‘‘to sleep’’ belong to one group, and vanoms ‘‘to watch’’ and ńiĺems ‘‘to swallow’’ to the other. A second hypothesis we can make is that the in nitive marker is, in fact, -Oms with an archiphoneme O to indicate that a vowel must always be present; in the literary language that is (cf. Evsev'ev 1963:

286). Evsev'ev indicates that there are certain verb stems that lose their mid-vowel in dialect representation of the in nitive, e.g. moĺems ~ moĺmeks ‘‘to go’’ and kadoms ~ kadmoks ‘‘to leave (VT)’’, while others do not: udoms ~ udomks ‘‘to sleep’’. (Trosterud (2006: 250) offers a phonetic solution to stem-vowel deletion before consonant-initial suf xes. He recognizes an important role played by consonant clusters but does not see the correlation to the stem-vowel versus linking-vowel dichotomy.)

The dichotomy ““stem vowel versus linking vowel”” can be further developed upon perusal of the Erzya-Russian dictionary. There are, in fact, three types of verbs to be attested, i.e. sod|oms: %ś ‘‘to tie’’ (verbs with linking vowels), soda|ms: %ś ‘‘to know; to rec-ognize’’ (verbs with stem vowels), and kundato|ms: %tś ‘‘to become tied (of the tongue)’’

(verbs with stem vowels and additional T morphology). The third verb type, originally brought to my attention by Salo (cf. Salo, forthcoming), takes an additional T before the

IND.PRETI.PRED-3SG marker. A parallel to this morphological variation between m and T can be observed in the attestation of two noun forms in Kozlovka dialect: utom ‘‘store-house’’: utotso ‘‘in the storehouse’’ and kaštom ‘‘oven’’: kaštotso ‘‘in the oven’’ (cf. Bubrikh 1930: 33).

The third verb type is not adhered to by all speakers of the language, such that Imaikina (originally from an Insar or Western dialect background, but with a lifetime in university-level Mordvin studies) indicates two separate re exes for the verb satoms

‘‘to be enough’’: satotś (2008: 96) and satś (ibid. 282) ‘‘suf ce_V.IND.PRETI.PRED-3SG’’, whereas the Russian-language treatise of Erzya verbs published by Mészáros ignores it altogether (cf. 1999: 116-120).

The Erzya literary language attests to a system of three verb types as depicted in ERV (1993). Therefore, the three verb types described by the Finnish School, especially the o and e verb-type descriptions of the Erzya, have little to do with the synchronic state of the language, although they may offer partial insight into language history.

In nouns the vertical separator ““|”” serves to mark the removal of the Cyrillic soft sign before the plural marker T, on the one hand, and some instances of mid vowels, on the other, whereas low stem vowels are never elided in conjugation or declension, and therefore a correlation between verb and nominal-type stems might be posited (cf. Zaicz 1998: 188––189).

Let us then address the nominal stem type, if we can, according to the same cri-teria as were used with verb stems. Nominals can also be divided into three types. The stem types do not directly parallel those of the verbs, though. The nominal stem types are based almost entirely upon the phonetic qualities of the inde nite nominative singu-lar form. The  rst split is made on the basis of whether the headword ends in a vowel or a consonant. The merits of such a split will be seen in the number of af xes, described below, that require the presence of a vowel between the consonants of an immediately preceding stem and the  rst consonant of the af x. Let us observe the variation in the nominative de nite singular marker allomorphs -oś, %eś and -ś <= -Oś: oš+oś ‘‘town_

N+NOM.DEF.SG’’ and keĺ+eś ‘‘tongue_ N+NOM.DEF.SG’’ versus kudo+ś ‘‘house_ N+NOM.DEF.SG’’

and veĺe+ś ‘‘village_ N+NOM.DEF.SG’’. The vowel- nal versus consonant- nal division of stems is then followed by two more, one concerning the consonant- nal and the other the vowel- nal headwords.

In consonant- nal nominal-type headwords the presence of a word  nal s(h)ibi-lant can bring about synonymous variation in the declension tables. If the

headword- nal consonant is a s(h)ibilant, then there is a tendency for a linking vowel to occur between the stem- nal s(h)ibilant and an af x-initial s(h)ibilant, e.g. in the illative, the word karks ‘‘belt’’ gains what here will be termed an optional stem vowel in o before the illative marker %s is added, thus karks+(o)s belt_N+ILL. This, it must be stressed, is a tendency that affects all stem- nal s(h)ibilants c, s, z, ć, ś, ź, č, š, ž in combination with the af x-initial s of the illative, inessive and elative, as well as the š of the comparative.

Therefore synonymous variation can be observed, such as that found in the inde nite il-lative declension for potmaks ‘‘bottom’’ with evidence in favor of the linking-vowel strat-egy potmaksos 184 occurrences and potmakss 28, but also the adposition veĺkses 255 and veĺkss 211 both interpreted as ‘‘over/above_POP.ILL’’. Hence phonological variation in the presence versus absence of a stem vowel, because of its seemingly non-semantic character, can automatically be ascribed to all stems ending in s(h)ibilants. This non-semantic variation might be dealt with in two manners: it might be simply described as morphophonemic variation, or if we choose to inspect its discourse-level variation, we might plot it in the continuation lexicon of concatenation, so that it can be automatically parsed. Similar synonymous variation can also be observed in the combination of other consonant- nal stems followed by translative case marker ks. Both subtypes are open to native and loanword stems.

In the vowel- nal set of all nominal-type headwords special attention must be given to a subset with headword- nal mid vowels. The subset affected comprises those which alternately exhibit a loss of that stem- nal mid vowel before certain af xes be-ginning with voiceless alveolar consonants, especially the plural marker in %T and the spatial cases %s illative, -sO inessive and -stO elative. In the table below we will observe

the three different stem types NOUNS1 in (a, b) (consonant- nal stems), NOUNS1S in (c, d, whereas this subset of NOUNS1 is entirely predictable), NOUNS2 in (e, f) (nominal stem type with synonymous stem-vowel variation) and NOUNS3 in (g, h, i) (nominal stem type with stem vowel retention).

The table illustrates possible homonymy that occurs between plural marking T and possessive cross-referential 2SG marker OT in the three stem types. No homonymy occurs in NOUNS1 stems; possessor index markers in the modern literary language always require linking vowels (see section 4.2.3.1. POSSESSOR-INDEXMARKING). Optional hom-onymy is observed in NOUNS2 stems, and total homonymy in NOUNS3 stems. In the most recent grammar (EKM 2000), NOUNS2 stems are treated as invariable NOUNS3 stems. This re ects one of the prescriptive norms proposed in the most recently printed orthographic norms ““ɗɪɡɹɧɶ ɤɟɥɶɫɷ ɫɺɪɦɚɞɨɦɚɧɶ, ɤɨɪɬɚɦɨɧɶ, ɩɭɧɤɬɭɚɰɢɹɧɶ ɥɭɜɬɧɟ”” ‘‘Orthogra-phy, Speech and Punctuation Norms in the Erzya Language’’, henceforth (EKS 1995:

34). For an extensive presentation of nominal stem variation in declension (cf. Evsev'ev 1963: 56––101; Abondolo 1987).

Table 4.1 Nominal stem types in Erzya nominative plural forms. This stem type has received attention in various grammars beginning with Evsev'ev ([1929] 1963: 82––83). As noted above at least some prescriptive grammar writers have decided to remove the NOUNS2 stem type from the agenda of Erzya morphology, even though it is extensively attested in the written language. See table below for disambiguated statistics on vowel versus consonant-stem in the expression of plural in the NOUNS2 stem type, where nine of the most frequently attested headwords are given with possessive second person singular versus inde nite nominative plural readings.

Table 4.2 Stem variation in NOUNS2 nominal stem type

The nominal stems demonstrated above will be rendered in three separate types on the basis of two parameters: (i) presence of stem- nal vowel in headword, and (ii) retention of stem- nal vowel before plural marker in -T. Although there are tendencies in the written language towards possessive second person singular versus nominative plural differentiation in some of the NOUNS2 stems, it appears that not all people in typesetting adhere to the same norms (something indicative of dialect variation). Vowel retention in some of the stems may be found in bahuvrihi type constructions, such as vaŕga kurgot

‘‘blabber-mouths (glove_N.NOM.SG mouth_N.NOM.PL)’’, or, perhaps, emphatic and dialect variation. Vowel loss occurs with the plural marker %T, but this same form or possibly a homonym is also used in the formation of adverbs, e.g. veŕev pandt ‘‘up-hill (up_ADV.

LAT hill_N.PL/DISTR: GOAL and LOC)’’. For accuracy the NOUNS2 stems could be regarded as a closed set, which for all practical purposes it is, but the following regular-expression descriptions of stem type will be helpful in the locating of plausible yet unidenti ed members.

[FTV] [ [ɩ|ɬ|ɤ|ɰ|ɱ] | ɦ ɛ | [ɜ|ɫ|ɲ] ɬ | [ɪ|ɥɶ|ɧ] ɞ | [ɫɶ|ɲ] ɤ | [ɜ|ɪ|ɥɶ|ɧ] ɝ | [ɬɶ|ɞɶ|ɪɶ|ɥɶ] ɦ | ɜ ɲ] ɟ

[BTV] [ [ɩ|ɬ|ɤ|ɰ|ɱ] | ɦ ɛ | [ɜ|ɫ|ɲ] ɬ | [ɪ|ɥ|ɧ] ɞ | [ɫ|ɲ] ɤ | [ɜ|ɪ|ɥ|ɧ] ɝ | [ɬ|ɞ|ɪ|ɥ] ɦ | ɜ ɲ] ɨ With retrospect to the consonant-cluster delimitations suggested by Trosterud (2006:

250), it should be noted that certain stems, e.g. keĺm^|e ‘‘cold; frost’’ and keĺme^|ms ‘‘to get cold; to freeze’’ do not necessarily follow identical patterns. And thus it is the nominal type stem NOUNS2 that consoant-cluster delimitations might be concentrated on.