Abstract
The paper offers a synchronic and diachronic account of markedness of the oppositional tense-aspect stems in Classical Armenian. The synchronic part explores the correspondence between markedness and productivity of verb classes as attested in the Armenian Bible translation, as well as the correspondence between markedness and token frequency of a selection of fifty most frequent verbs in the same text. The default pattern, characterized by an unmarked present and marked aorist stems, constitutes some two-thirds of the entire dataset but is less common in the most frequent verbs. By contrast, the two patterns with the unmarked aorist are significantly better represented and the token frequency of their aorist stems is typically higher for such verbs. This evidence is discussed in the context of a typological generalization predicting the lower markedness of more frequently used forms. An outlook on the historical grammar of Classical Armenian suggests that the attested system reflects a transition from aspect- to tense-oriented marking of stems.
1 Introduction
The paper is dedicated to the assymmetries in the morphological markedness of tense-aspect stems of Classical Armenian (CA) in the synchronic and diachronic perspectives.
Cross-linguistic studies offer rich evidence on the correspondence between morphological markedness, frequency of use, and diachronic stability (e.g., Greenberg 1966; Bybee 1985; Croft 2003; Haspelmath 2006 with further references). Thus, Croft (2003: 87–121) works with the notion of “typological markedness” that specifies categorial values to relate the morphological markedness to frequency. According to Croft, (a) the typologically marked value of a grammatical category tends to be expressed by at least as many morphemes as the unmarked value of that category (p. 92), and (b) the tokens of a typologically unmarked value tend to occur at least as frequently as those of the marked value (p. 110). Haspelmath (2006) argues in favour of a direct correspondence between overt coding and lower frequency of use without recourse to semantic markedness or similar notions. Within the latter approach, the aforementioned generalizations can be re-formulated as follows: an equally or less frequently used token representing a grammatical value will have at least as many morphemes as an equally or more frequently used token representing another value in the same category. It follows that the more frequently used token is not expected to be expressed by more morphemes compared to the expression of the opposed value. The paper aims to specify to what extent these generalizations hold true with respect to the CA evidence. In what follows, the term “(un)marked” refers to morphological markedness in terms of the relative number of morphemes that encode oppositional grammatical values, the marked value being expressed by more morphological material.
The CA verbal sytem is built on the opposition of two stems—the “present” and the “aorist”, from which all the synthetic personal verb forms are derived (Meillet 1913; Jensen 1959; Klingenschmitt 1982; Bubenik 1997). Either of the two stems (a), or both (b), or none (c) can be marked by a suffix:
-
prs. tes-an-em, aor. tes-i ‘I see/saw’; prs. gr-em, aor. gr-ecʽ-i ‘I write/wrote’;
-
prs. moṙ-an-am, aor. moṙ-acʽ-ay ‘I forget/forgot’;
-
prs. ber-em, aor. ber-i ‘I bring/brought’.
The grammatical function of this opposition is traditionally associated with encoding of aspect rather than tense (Meillet 1910–1911). Among other things, this interpretation is justified by the distribution of the stems across the verbal paradigm: both stems occur in forms expressing the past and future values of tense. Thus, the marked present stem tes-an-, used in the present tes-an-em ‘I see’ and imperfect tes-an-ei ‘I saw’, encodes the imperfective aspect rather than the present tense. Therefore, the unmarked aorist stem tes- of this verb does not contradict the typological universal that the past tense tends to be encoded by at least as many morphemes as the present, cf. the English past tense forms in -ed opposed to the unmarked present; cf. I walk vs. I walk-ed (cf. Greenberg 1966: 47–49; Bybee 1994: 248; Haspelmath 2008: 8). Altogether, it has never been properly explained why CA verb classes differ in their stem-marking patterns. The question arises whether the synchronic distribution of those patterns can be clarified by their frequency of use in line with the aforementioned typological generalizations. Is there any evidence to support the claim that verbs with the unmarked aorist stem were more frequently used to express grammatical values associated with that stem? Another question is whether the frequency-based explanation can help to better understand the nature of diachronic changes in the stems that can be reconstructed based on the comparative grammar of Indo-European languages. Some preliminary observations put forth in the present paper are intended to contribute to this discussion without pretence of an exhaustive analysis.
§ 2 offers an overview of the CA verb classes and a discussion of their type frequency in the Classical Armenian translation of the Bible. § 3 provides an account of the type frequency of the 50 most frequent verbs of the biblical corpus as well as the relative token frequency of their present and aorist stems. A snapshot of the correspondence between markedness and frequency, as reflected in the most frequent segment of the verbal lexicon, is then discussed in the diachronic perspective in § 4.
2 Markedness and productivity
CA has three synthetic tenses derived from the present stem (the present, the imperfect and the present subjunctive) and two tenses derived from the aorist stem (aorist and aorist subjunctive), see Table 1.
The present stem can be suffixless or contain one of the five suffixes: -n-, -an-, -čʽ-, -nčʽ- and -ančʽ-. The thematic vowels that follow the stem in the present indicative (-e-, -i-, -u-, -a-, -o-) belong to the endings and not to stems as is clear from the comparison of the inflections of the present indicative and present subjunctive, both derived from the present stem; cf. prs. ber-em ‘I bring’ next to prs.sbj. ber-icʽ-em ‘I will be bringing’. These vowels therefore do not constitute part of the stem markers. The aorist stems can be suffixless, or contain one of the four suffixes -cʽ-, -i- (with its allomorph -e- occurring in front of the following -a- of endings), -acʽ- (with a word-final allomorph -a-) and -ecʽ- (with word-final allomorphs -eacʽ- and -ea-).
Table 1
Distribution of tense-aspect stems in the CA verbal paradigm
|
Aspect |
|||
|---|---|---|---|
|
Imperfective |
Perfective |
||
|
Tense |
Present |
Present 3sg. gorc-em ‘I work’ |
|
|
Past |
Imperfect 3sg. gorc-ei ‘I was working’ |
Aorist 3sg. gorc-ecʽ-i ‘I worked’ |
|
|
Future |
Present subjunctive 3sg. gorc-icʽ-em ‘I will be working’ |
Aorist subjunctive 3sg. gorc-es-icʽ-Ø ‘I will work’ |
The aorist indicative forms derived from a suffixless stem beginning with a consonant are additionally marked by a prefix e- (augment), if the form would otherwise be monosyllabic. Typically, such are the 3sg. forms of the active voice with zero endings; cf. 3sg. aor. e-ber-Ø of berel ‘to bring’ as opposed to 3sg. aor. ancʽ-Ø (next to the post-classical ē-ancʽ-Ø) of ancʽanel ‘to pass’ beginning with a vowel. Still in line with the general rule are augmented forms like 1sg. aor. etu of tal ‘to give’. The grammatical function of the augment is controversial: does it encode the perfective aspect as part of the aorist stem, or does it encode the past tense? Note that under the latter interpretation, it is insignificant that the augment does not occur in the imperfect, which has no forms that would be monosyllabic without the augment. Analyzing the augment as a tense-coding morpheme conforms to the aforementioned universal according to which the aorist tends to be coded by at least as many overt morphemes as the present, because the past value of the tense category receives an additional marker in the pivotal form of the aorist paradigm; cf. 3sg. prs. ber-ē ‘s/he brings’ next to 3sg. aor. e-ber-Ø ‘s/he brought’. With respect to its uncertain grammatical status, the augment is not considered as part of stem-marking patterns in the present paper.
The combinations of present and aorist stems constitute verb classes of varying productivity. These classes can be arranged into four groups depending on the markedness of stems, see Table 2.
Traditional grammars of CA describe verb classes without details concerning their type frequency. Consequently, there is no readily available data on the relative productivity of the markedness patterns listed in Table 2. By way of partly filling in this descriptive gap, Table 3 shows the distribution of the 1777 verbs attested in the online concordance of the Armenian Bible (
Table 2
Markedness patterns of CA verb classes
|
Markedness pattern |
Present stem |
Aorist stem |
|
|---|---|---|---|
|
A |
directed |
zero |
marked |
|
B |
marked |
zero |
|
|
C |
nondirected |
marked |
marked |
|
D |
zero |
zero |
|
Table 3
Distribution of verb types across verbal classes (Bible)2
|
Stems |
Aorist |
|||||
|---|---|---|---|---|---|---|
|
zero |
-cʽ- |
-i- |
-acʽ- |
-ec‛- |
||
|
Present |
zero |
28–34 |
0–1 |
64–103 |
110 |
|
|
-n- |
5–14 |
6 |
7–12 |
|||
|
-an- |
331 |
3 |
170–209 |
|||
|
-č‛- |
7–9 |
|||||
|
-nčʽ- |
3 |
|||||
|
-ač‛- |
0–1 |
|||||
|
-ančʽ- |
1 |
Pattern A (ca. 65.5–68 %). Here belongs the most productive verb class of the e/i-conjugation with an aorist ec‛-stem, cf. prs. gr-em, aor. gr-ecʽ-i ‘to write’. The pattern is further represented by verbs of the a-conjugation with the aorist acʽ-stem. Next to 61 verbs of the latter type, for which the present stem is attested (cf. prs. eṙ-am, aor. eṙ-acʽ-i ‘to boil’), one finds 39 verbs without an attested present stem (cf. prs. xanmt-am/xanmt-an-am, aor. anmt-acʽ-ay ‘to be/become thoughtless’), and they may instantiate patterns A or C. This morphological ambiguity constitutes a major area of overlap between the directed and nondirected patterns. Besides, the pattern includes three verbs of the e/i-conjugation with the aorist acʽ-stem: as-em, aor. as-acʽ-i ‘to say’, git-em, aor. git-acʽ-i ‘to know’ and karem, aor. kar-acʽ-i ‘to be able’. From a purely synchronic perspective, čanačʽ-em, aor. can-e-ay ‘to realize’ is ambiguous since the boundary between the root and suffix is not transparent in the present stem; this allows for the analysis prs. čanačʽ- / aor. can-e- (pattern A) or prs. čan-ačʽ- / aor. can-e- (pattern C). Under the latter analysis, the isolated present suffix -ačʽ- can be compared to the isolated but unambiguously segmentable suffix -ančʽ- of meł-ančʽ-em, aor. meł-ay ‘to sin’ of pattern B.
Pattern B (ca. 19–20 %). The majority type here is the verb class with the present an-stem, cf. prs. el-an-em, aor. el-i ‘to go out’. Most of these verbs are morphological causatives built with the derivational suffix -oycʽ/ucʽ-, cf. prs. us-ucʽ-an-em, aor. us-ucʽ-i ‘to teach’ (261 verbs). Here also belongs the suppletive verb prs. hark-an-em, aor. har-i ‘to hit’. Meł-ančʽ-em, aor. meł-ay is an isolated case with the present suffix -ančʽ-. Another 5–8 verbs of the e/i- and u-conjugations have the present n-stem (cf. prs. aṙ-n-um, aor. aṙ-i ‘to take’), including three verbs for which the aorist stem is not attested and which can therefore be counted as instances of patterns B or C: ayt-n-um ‘to swell’, zbał-n-um ‘to be occupied’, zbaws-n-um ‘to take a rest’. One may arguably attribute here six more suppletive verbs of the e/i- and a-conjugations like prs. aṙ(-)n-em, aor. arar-i ‘to make’ see below on pattern D. They have a root aorist stem and, depending on whether the suffix is segmented in the present stem, they can be ascribed to patterns B or D, respectively.
Pattern C (ca. 11–13.5 %). This is the most heterogenous pattern as far as verb classes are concerned. The majority class has the present an-stem and aorist ac‛-stem; this includes a productive type of inchoatives derived from nouns and adjectives, cf. adj. čʽor ‘dry’ → prs. čʽor-an-am, aor. čʽor-acʽ-ay ‘to become dry’. The remaining verbs belong to six verb classes and include verbs with the aorist i-stem and present n-stem of the u-conjugation (7–12 verbs including two verbs for which the present stem is not attested, cf. prs. xost-n-um/xost-čʽ-im, aor. ost-e-ay ‘to jump’, and the three verbs for which the aorist stem is not attested, see above on pattern B) and of the e/i-conjugation (only prs. yaṙ-n-em, aor. yar-e-ay ‘to rise’); the aorist i-stem and present čʽ-stem (7–9, cf. prs. hang-čʽ-im, aor. hang-e-ay ‘to rest’); the present nčʽ-stem and aorist i-stem (prs. erk-nčʽ-im, aor. erk-e-ay ‘to fear’, prs. kor-nčʽ-im, aor. kor-e-ay ‘to disappear’, prs. mart-nčʽ-im, aor. mart-e-ay ‘to fight’); the present an-stem and aorist i-stem (prs. tʽṙ-an-im, aor. tʽṙ-e-ay ‘to fly’, prs. yancʽ-an-em, aor. yancʽ-e-ay and prs. zancʽ-an-em, aor. zancʽ-e-ay ‘to transgress’); and the present n-stem and aorist cʽ-stem (6 verbs, cf. prs. l-n-um, aor. l-cʽ-i ‘to fill’). One may also arguably attribute here čanačʽem (see above on pattern A).
Pattern D (ca. 1.5–2 %). Besides 12 verbs of the u-conjugation (cf. prs. tʽoł-um, aor. tʽoł-i ‘to let’) and 6 verbs of the e/i-conjugation (cf. prs. ber-em, aor. ber-i ‘to bring’), this pattern includes most suppletive verbs including 4–8 verbs of the e/i-conjugation (cf. prs. aṙ(-)n-em, aor. arar-i ‘to make’ and prs. u(-)n-im, aor. kal-ay ‘to have’) and 4–6 verbs of the a-conjugation (cf. prs. ertʽ-am, aor. čʽog-ay ‘to go’ and prs. baṙ(-)n-am, aor. barj-i ‘to lift’). The morphological uncertainty concerns the segmentation of the nasal suffix -n-, which is ambiguous in the case of suppletive verbs like aṙnem because the aorist stem does not justify the synchronic segmentation of the root. Cases like prs. spaṙn-am, aor. spaṙn-acʽ-ay ‘to threaten’ show that the etymological marked present n-stem could be reanalyzed as an unmarked one, and it is unclear to which extent the same applies to suppletive verbs. Such verbs therefore are transitional between patterns A and D.
This evidence reveals that pattern A characterizes some two thirds of the total number of verbs found in the bibilical corpus. According to the hypothesis put forth in the introduction, the markedness pattern is conditioned by the relative token frequency of the two stems. If that hypothesis is valid, the evidence may be interpreted in two ways. Firstly, one may assume that the attested distribution of type frequencies per markedness pattern reflects the relative frequency of use of the two stems for individual verbs and that most verbs are more often used in imperfective contexts (a lexeme-level marking of aspect). Alternatively, one can postulate a hybrid system in which the lexeme-level strategy is combined with a system where the perfective value is invariably marked and the imperfective one is invariably unmarked due to the overarching greater frequency of the imperfective value across the CA verbs irrespective of the relative frequency of stems for individual verbs (a category-level marking of aspect). Within the latter solution one may assume that not all verbs following pattern A were mostly used in imperfective contexts; cf., for example, such telic verbs as aṙakʽem, aor. aṙakʽ-ecʽ-i ‘I send/sent’. The two coding strategies may then be accounted for in diachronic terms, the lexeme-level type being older and synchronically less productive. By contrast, the category-level marking of the perfective aspect can be viewed as an innovation comparable (and possibly related) to the typologically common lexeme-independent markedness of the past value in the tense category. Note that the morphological boundary between the coding of tense and aspect is unstable in CA; it concerns not only to the issue of the grammatical analysis of the augment mentioned in § 1, but also to the ecʽ-aorist of the most productive verb class within pattern A. The 3sg. active form of the ecʽ-aorist has a zero ending, which goes against the universal tendency to code the past with at least as many morphemes as the present. However, the accented aorist suffix retains the diphthong which is monophthongized in the remaining forms in an unaccented position, cf. 3sg. gr-eacʽ-Ø next to 1sg. gr-ecʽ-i (see Table 4). One may tentatively suggest that the phonologically “heavier” form of the stem compensates for the lack of overt coding of the past tense. This suggestion is admittedly speculative and amounts to an expectation that, had the phonological alternation related to the position of stress not applied, the pivotal 3sg. form of the aorist paradigm would have been recharacterized by an overt marker. If the claim is correct however, the productive lexeme-independent marking of the perfective aspect might have been at least in part motivated by a requirement to obligatorily mark the past tense in the pivotal 3sg. form of the aorist paradigm.
Table 4
Markedness of tense and aspect values in the ecʽ-aorist
|
Present |
Aorist |
|||
|---|---|---|---|---|
|
Active |
Mediopassive |
Active |
Mediopassive |
|
|
1sg. |
gr-em |
gr-im |
gr-ecʽ-i |
gr-ecʽ-ay |
|
2sg. |
gr-es |
gr-is |
gr-ecʽ-er |
gr-ecʽ-ar |
|
3sg. |
gr-ē |
gr-i |
gr-eacʽ-Ø |
gr-ecʽ-aw |
The two aforementioned approaches to the interpretation of the relative productivity of the four markedness patterns will be further discussed based on the evidence of the relative frequency of stems in a selection of individual verbs.
3 Markedness and frequency
This section examines a selection of the 50 most frequent verbs of the biblical corpus in order to establish: (a) the relative type frequency of the markedness patterns in the most frequent segment of the lexicon in comparison to that of all the verbs attested in the Bible (Table 5); and (b) the correspondence between the markedness patterns and the relative token frequency of the present and aorist stems (Table 7–9); see Supplement 2 at
Table 5
Distribution of the 50 most frequent verbs of the Bible over the markedness patterns
|
Lemma |
Token fq |
Pattern |
Lemma |
Token fq |
Pattern |
|---|---|---|---|---|---|
|
em ‘to be’ |
13768 |
– |
xndrel ‘to ask’ |
847 |
A |
|
asel ‘to say’ |
8732 |
A |
kočʽel ‘to call’ |
842 |
A |
|
linel ‘to become’ |
6739 |
B/D |
hasanel ‘to reach’ |
806 |
B |
|
aṙnel ‘to make’ |
4556 |
B/D |
ankanel ‘to fall; to twine’ |
804 |
B |
|
tal ‘to give’ |
3751 |
D |
awrhnel ‘to bless’ |
731 |
A |
|
gal ‘to come’ |
2645 |
D |
bnakel ‘to dwell’ |
707 |
A |
|
aṙnul ‘to take’ |
2238 |
B |
ancʽanel ‘to pass’ |
693 |
B |
|
tesanel ‘to see’ |
2032 |
B |
tʽołul ‘to leave’ |
674 |
D |
|
elanel ‘to go out’ |
1987 |
B |
pahel ‘to protect’ |
662 |
A |
|
lsel ‘to listen’ |
1896 |
D |
arkanel ‘to throw’ |
650 |
B |
|
xawsel ‘to talk’ |
1715 |
A |
acel ‘to lead’ |
633 |
D |
|
kal ‘to stay’ |
1601 |
A |
gorcel ‘to work’ |
631 |
A |
|
daṙnal ‘to turn; return’ |
1275 |
B/D |
patmel ‘to tell’ |
621 |
A |
|
mtanel ‘to enter’ |
1254 |
B |
gol ‘to exist’ |
615 |
– |
|
ertʽal ‘to go’ |
1235 |
D |
matucʽanel ‘to offer’ |
606 |
B |
|
dnel ‘to put’ |
1203 |
B |
nstel ‘to sit’ |
605 |
D |
|
unel ‘to have’ |
1165 |
B/D |
žołovel ‘to gether’ |
594 |
A |
|
gitel ‘to know’ |
1164 |
A |
cnanel ‘to give birth’ |
587 |
B |
|
utel ‘to eat’ |
1005 |
D |
yaṙnel ‘to get up’ |
584 |
C |
|
meṙanel ‘to die’ |
1003 |
B |
šinel ‘to build’ |
578 |
A |
|
harkanel ‘to hit’ |
951 |
B |
grel ‘to write’ |
576 |
A |
|
aṙakʽel ‘to send’ |
931 |
A |
kamel ‘to want’ |
563 |
A |
|
gtanel ‘to find’ |
877 |
B |
sirel ‘to love’ |
544 |
A |
|
hanel ‘to draw’ |
855 |
D |
spananel ‘to kill’ |
494 |
B |
|
gnal ‘to go’ |
851 |
A |
karel ‘to be able’ |
486 |
A |
Regarding the type frequency, the distribution of the markedness patterns is significantly different in the 50 most frequent verbs as compared to the overall verbal lexicon of the Bible, see (Table 6). Although pattern A is still the majority type, it is less common among the most frequent verbs. Patterns B and D, both with an unmarked aorist, are, on the contrary, relatively more common and host around half of the most frequent verbs—this is especially noticeable for D. Pattern C is underrepresented and instantiated by only one verb with the aorist i-stem, yaṙnem.
Table 6
Distribution of the stem marking patterns in the Bible
|
Pattern |
50 most frequent verbs of the Bible |
All verbs of the Bible |
|---|---|---|
|
A |
40 % |
65.5–68 % |
|
B |
28–36 % |
19–20 % |
|
D |
18–28 % |
1.5–2 % |
|
C |
2 % |
11–13.5 % |
According to the data in the Tables 7–9 below, the patterns have different etymological profiles. The verbs within patterns B, B/D, C, and D, except for a few dubious cases and a derived causative, have a sound Proto-Indo-European (PIE) etymology. By contrast, only half of pattern A verbs have a reliable PIE etymology within the most frequent segment, while the other verbs are either Iranian loanwords or do not have an established etymology. This distribution supports the hypothesis that patterns B and D represent an archaic layer of the lexicon containing synchronically unproductive morphology retained due to higher frequency of use. Moreover, it is probably due to that morphological inertia that respective verbs bear a trace of the lexeme-level coding of aspect typical for PIE (see § 4 below).
The cross-linguistically common tendency to keep the more frequently used grammatical values formally unmarked predicts that pattern B is likely to be retained in verbs with relatively greater frequency of forms derived from the aorist stem. The evidence of the 50 most frequent verbs of the biblical corpus fully supports this prediction. All verbs that unambiguously belong to pattern B use forms with the aorist stem at a higher rate. The ratio of the present/aorist stems ranges from 0.27 to 0.70 (see Table 7).3
Table 7
Relative frequency of pattern B verb stems
|
Lemma |
Prs. stem |
Fq |
Aor. stem |
Fq |
Ratio |
Origin |
|---|---|---|---|---|---|---|
|
aṙnul |
aṙ-n- |
519 |
aṙ- |
1713 |
0.30 |
PIE (EDAIL 112–113) |
|
tesanel |
tes-an- |
625 |
tes- |
1407 |
0.44 |
PIE (LIV2 109–112) |
|
elanel |
el-an- |
598 |
el- |
1389 |
0.43 |
PIE (EDAIL 248–249) |
|
mtanel |
mt-an- |
461 |
mt- |
793 |
0.58 |
PIE (Martzloff, fthc.) |
|
dnel |
d-n- |
258 |
d- |
945 |
0.27 |
PIE (EDAIL 240–241) |
|
meṙanel |
meṙ-an- |
267 |
meṙ- |
736 |
0.36 |
PIE (EDAIL 463) |
|
harkanel |
hark-an- |
302 |
har- |
649 |
0.46 |
PIE (? LIV2 301 & 473) |
|
gtanel |
gt-an- |
275 |
gt- |
602 |
0.45 |
PIE (EDAIL 216) |
|
hasanel |
has-an- |
253 |
has- |
553 |
0.45 |
PIE (EDAIL 90–91) |
|
ankanel |
ank-an- |
192 |
ank- |
612 |
0.31 |
PIE (EDAIL 90–91) |
|
ancʽanel |
ancʽ-an- |
228 |
ancʽ- |
465 |
0.49 |
PIE (Olsen 1999: 88) |
|
arkanel |
ark-an- |
232 |
ark- |
418 |
0.55 |
? (Klingenschmitt 1982: 204–205) |
|
matucʽanel4 |
matucʽ-an- |
238 |
matucʽ- |
368 |
0.64 |
Derived causative. |
|
cnanel |
cn-an- |
95 |
cn- |
492 |
0.19 |
PIE (EDAIL 342–343) |
|
spananel |
span-an- |
204 |
span- |
290 |
0.70 |
? (Klingenschmitt 1982: 226–227) |
All verbs withing the morphologically ambiguous pattern B/D and pattern D belong to synchronically unproductive classes and include many irregular and suppletive verbs. With respect to the relative frequency of stems, three of the four verbs classified as B/D, aṙnel, daṙnal, and linel, align with pattern B, while unel ‘to have’ has a significantly more frequent present stem (see Table 8). One may argue that the present stem was considered as containing a nasal suffix in the former cases but as a root in the latter one, cf. the root noun oyn(kʽ) ‘habit; valor’ also attested as the second member of the compound ołǰ-oyn ‘salute!’ lit. ‘have health!’ (de Lamberterie 2020: 63–64).
Table 8
Relative frequency of pattern B/D, C, and D verb stems
|
Lemma |
Prs. stem |
Fq |
Aor. stem |
Fq |
Ratio |
Origin |
|---|---|---|---|---|---|---|
|
Prs. < aor. |
||||||
|
linel |
li(-)n- |
1507 |
eł-, l(i)- |
5232 |
0.20 |
PIE (Klingenschmitt 1982: 164) |
|
aṙnel |
aṙ(-)n- |
1377 |
arar-, ara(s)- |
4556 |
0.30 |
PIE (EDAIL 112) |
|
tal |
t(-)a- |
16925 |
et-, tu- |
2059 |
0.82 |
PIE (EDAIL 595) |
|
gal |
g(-)a- |
670 |
(e-)k- |
1975 |
0.34 |
PIE (EDAIL 196) |
|
lsel |
ls- |
513 |
lu- |
1383 |
0.37 |
PIE (EDAIL 313–315) |
|
daṙnal |
daṙ(-)n- |
222 |
darj- |
1053–7596 |
0.21–0.29 |
PIE (EDAIL 234) |
|
utel |
ut- |
408 |
ker- |
597 |
0.68 |
PIE (EDAIL 644) |
|
hanel |
han- |
285–315 |
han- |
540–570 |
0.50–0.58 |
PIE (EDAIL 389) |
|
tʽołul |
tʽoł- |
135 |
tʽoł- |
539 |
0.25 |
PIE (EDAIL 90–91) |
|
acel |
ac- |
217–250 |
ac- |
383–416 |
0.52–0.65 |
PIE (EDAIL 16) |
|
yaṙnel |
yaṙ-n- |
87 |
yar-i- |
497 |
0.17 |
PIE (LIV2 299–301) |
|
Prs. ≈ aor. |
||||||
|
nstel |
nst- |
270–312 |
nst- |
293–3357 |
0.80–1.0 |
PIE (EDAIL 505–506) |
|
Prs. > aor. |
||||||
|
ertʽal |
ertʽ- |
1102 |
čʽog- |
133 |
8.3 |
PIE (EDAIL 263) |
|
unel |
u(-)n- |
850 |
kal- |
315 |
2.6 |
PIE (EDAIL 636–637) |
The aorist stem is also more frequently attested for most verbs with the nondirected patterns C and D. Two exceptions are verbs ertʽal and nstel. The stem frequency of nstel depends on how one analyses the derivational basis of the past participle of this verb. If all potential participial forms are counted towards the aorist stem, the ratio is 0.80, which is comparable to that of tal. The pair ertʽ- / čʽog- is thus isolated among the suppletive verbs regarding the exceptionally high frequency of the present stem. This can be partly explained by the irregular paradigm in which the stem ertʽ- is used for the aorist subjunctive (242 occurrences) and the imperative (239 occurrences). However, the aorist subjunctive and imperative forms aside, the frequency of the present stem is still much higher with a present/aorist stem ratio of 4.7.
The verbs with the productive pattern A are more heterogeneous with respect to the relative frequency of stems (see Table 9). Only 11 of 19 verbs have a more frequently attested unmarked present stem.
Table 9
Relative frequency of stems of the verbs with the patterns A
|
Lemma |
Prs. stem |
Fq |
Aor. stem |
Fq |
Ratio |
Origin |
|---|---|---|---|---|---|---|
|
Prs. > aor. |
||||||
|
asel |
as- |
7335 |
as-acʽ- |
1397 |
5.2 |
PIE (EDAIL 118) |
|
kal |
k(-)a- |
908 |
k(-)acʽ- |
693 |
1.3 |
PIE (LIV2 205) |
|
gitel |
git- |
791 |
git-acʽ- |
373 |
2.1 |
PIE (EDAIL 211) |
|
xndrel |
xndr- |
454 |
xndr-ecʽ- |
393 |
1.1 |
? (Djahukyan 2010: 335–336) |
|
awrhnel |
awrhn- |
373 |
awrhn-ecʽ- |
358 |
1.0 |
Iran. (Djahukyan 2010: 104) |
|
bnakel |
bnak- |
405 |
bnak-ecʽ- |
302 |
1.3 |
Iran. (Djahukyan 2010: 131) |
|
gorcel |
gorc- |
393 |
gorc-ecʽ- |
238 |
1.6 |
PIE (EDAIL 226–227) |
|
grel |
gr- |
331 |
gr-ecʽ- |
245 |
1.3 |
PIE (Djahukyan 2010: 162) |
|
kamel |
kam- |
408 |
kam-ecʽ- |
156 |
2.6 |
? Iran. (Djahukyan 2010: 379) |
|
sirel |
sir- |
341 |
sir-ecʽ- |
203 |
1.7 |
PIE (Djahukyan 2010: 677–678) |
|
karel |
kar- |
327 |
kar-acʽ- |
159 |
2.0 |
? PIE (Klingenschmitt 1982: 138–139) |
|
Prs. < aor. |
||||||
|
xawsel |
xaws- |
582 |
xaws-ecʽ- |
1133 |
0.5 |
? Iran (Djahukyan 2010: 324–325) |
|
aṙakʽel |
aṙakʽ- |
176 |
aṙakʽ-ecʽ- |
761 |
0.23 |
? (Djahukyan 2010: 106) |
|
gnal |
gn- |
305 |
gn-acʽ- |
546 |
0.5 |
PIE (LIV2 196) |
|
kočʽel |
kočʽ- |
185 |
kočʽ-ecʽ- |
657 |
0.28 |
PIE (EDAIL 371) |
|
pahel |
pah- |
330 |
pah-ecʽ- |
332 |
0.99 |
Iran. (Djahukyan 2010: 617) |
|
patmel |
patm- |
158 |
patm-ecʽ- |
463 |
0.34 |
? Iran. (Djahukyan 2010: 617) |
|
žołovel |
žołov- |
177 |
žołov-ecʽ- |
399 |
0.44 |
? (Djahukyan 2010: 282) |
|
šinel |
šin- |
202 |
šin-ecʽ- |
376 |
0.53 |
Iran. (Djahukyan 2010: 586) |
Under the assumption that the most frequent verbs retain traces of the lexeme-level marking of aspect in stems, the inherited verbs with an unmarked and more frequent present stem may reflect archaic morphology if indeed their unmarked present stems prove to be inherited. By contrast, the verbs with an unmarked and less frequent present stem most likely reflect innovations created after the drift of the stem-marking from a lexeme to a category-level strategy. This shift assigned pattern A to verbs irrespective of their lexical aspectual features and relative frequency of stems. The process must have been completed by the time of the influx of Middle Iranian loanwords in the first centuries CE; cf. the potential Iranian loanwords in Table 9.
It is with respect to the synchronic evidence on the correspondence between markedness and frequency, surveyed in §§ 2 and 3, that one should address the controversies of the Proto-Armenian (PA) changes in the tense-aspect morphology. Some preliminary observations are offered in § 4.
4 Markedness and diachronic stability
The evolution of the CA tense-aspect system covers a time span of over 3,000 years after the split of the Armenian branch from common PIE, presumably in the beginning of the third millennium BCE (see Kassian et al. 2021).
The encoding of aspect by stems is characteristic of ancient Indo-European languages. According to the communis opinio, PIE telic verbs tended to have a suffixless aorist stem and a present stem with an imperfective affix, while the markedness of the two stems was reverse in atelic verbs (see Napoli 2007 and Bartolotta 2009 on the coding of aspect in PIE and Homeric Greek with further references).8
Armenian shows traces of many inherited features shared with the Greek and Indo-Iranian branches of the Indo-European language family, including the imperfect tense, the augment in the aorist, and the thematic aorist (see Martirosyan 2013; de Lamberterie 2013). The diachronic analysis of the CA verbal system must therefore be based primarily on the dialectal PIE verbal system of the Greco-Indo-Iranian type characterized by four synthetic tenses built with three tense-aspect stems (imperfective “present” stem, perfective “aorist” stem, and resultative “perfect” stem) and three sets of tense endings (primary “present”, secondary “aorist”, and perfect): present, aorist, imperfect, and perfect (see Clackson 2007 for an overview). CA inherited the present-aorist system but replaced the synthetic perfect by analytic formations expressed by the past participle plus auxiliary. The older resultative stem was almost completely eliminated with the exception of a few perfecto-present verbs, in which the perfect tense functioned as the present and was integrated into the PA present system (cf. PIE *u̯oi̯d- > CA gitem ‘to know’). The CA imperfect is an inner-Armenian innovation that does not continue the PIE imperfect directly. Altogether, the latter left residual traces in the CA aorist; see Viredaz (2018), Kim (2019), and Kocharov (2019), among others, on the major patterns of stem continuity.
The PIE ablaut and most of the PA morphophonological alternations of the root have been eliminated from the nominal and verbal paradigms by the time of earliest written record; cf. Arm. aor. e-bek ‘s/he broke’ (from PIE aor. *h₁e-bheg-t) next to pres. bek-an-ē ‘s/he breaks’ (replacing the PIE infixed present *bh-ne-g-ti; see LIV2 66). In many cases, the shape of the generalized root unambiguously points to a PIE or PA tense-aspect stem as its source. Within the approach discussed in §§ 2 and 3, the diachronic stability of unmarked stems is conditioned by the frequency of use. Some preliminary observations on the origin of the four major markedness patterns are provided below.
4.1 Pattern B
The source for pattern B consists of verbs with an inherited unmarked aorist stem, like the aforementioned ebek of bekanem. In the case of bekanem, the unmarked aorist stem was probably pivotal due to the lexicalized telicity characteristic of this verb of destruction in PIE as well as in CA. While the aorist stem remained virtually intact, the PIE infixed stem was replaced by a stem with the suffixe *-an-e/o-. The same mechanism triggered the renovation of a variety of inherited PIE present stems, in particular the infixed and reduplicated ones; cf. PIE aor. *u̯id-e/o- (prs. *u̯i-ne/n-d-) > aor. git/gt- → prs. gt-an- ‘to find’ (LIV2 665–667); PIE aor. *dʰeh₁- (prs. *dʰi-dʰeh₁-) > aor. (e-)d- → prs. d-n- ‘to put’ (LIV2 136–138). The CA nasal suffixes -ane- and -ne-, matching Gk. -
Based on the comparative evidence, a thematic or root aorist stem can be reconstructed for the PIE prototypes of 6 of 15 most frequent pattern B verbs listed in Table 7:
-
PIE aor. *h₂r-e/o- ‘to receive’ (Gk. aor.
ἀρόμην ; LIV2 270) > CA aṙ- (where -ṙ- has been explained by a secondary sigmatic aorist or analogy to prs. aṙ-nu-); -
PIE aor. *dek̑- ‘to perceive’ (Myc. aor. de-ko-to, Hom. Gk.
δέκτο ‘received’; LIV2 109–110; DELG2 1287) > CA tes-; -
PIE aor. *dʰeh₁- ‘to put’ (Skt. aor. ádhāt ‘put’, etc.; LIV2 136–138) > CA (e-)d-;
-
PIE aor. *mer- ‘to die’ (Skt. aor. ámr̥ta ‘died’, etc.; LIV2 439) > CA meṙ- (where -ṙ- has been explained by a secondary sigmatic aorist or analogy to unattested prs. *meṙ-nu-);
-
PIE aor. *u̯id-e/o- ‘to find’ (Skt. aor. ávidat ‘found’, etc.; LIV2 665–667) > CA git/gt-;
-
PIE aor. *g̑enh₁- ‘to be born’ (Gk. aor.
ἐγενόμην ‘was born’; LIV2 163–164) > CA cin/cn-.
The reconstruction of a thematic or root aorist stem is formally uncontroversial although not supported by external morphological cognates for another three verbs:
-
PIE aor. *mud-e/o- ‘to enter’ (Lat. mundus ‘world’; Martzloff fthc.) > CA mut/mt-;
-
PIE aor. *sh1k̑-e/o- ‘to reach’ (Gk.
ἥκω ‘to have come’; LIV2 519; DELG2 1304) > CA has-; -
PIE aor. *sngʷ-e/o- ‘to fall’ (Goth. sigqan ‘to sink’; LIV2 531–532).
The causative matucʽanem aside, the reconstruction of the aorist stems is burdened with further etymological complications for the remaining five pattern B verbs. Yet the reliable core of the evidence unambiguously points to verbs with an inherited root and thematic aorist as the principle PIE source of the pattern.
The fact that there are no securely identified Iranian, Syriac, or Greek loanwords following pattern B indicates that the pattern had ceased to be productive (except for the morphological causative) before the influx of such loanwords in the first centuries CE. The pattern remained productive, however, well into the PA period. In particular, it probably remained a vivid source of analogy at a stage when the prototype of the morphological causative in -oycʽ/ucʽ- replaced the inherited PIE causative formations and received a marked present an-stem (see Kocharov, fthc. on the origin of the causative). That morphological innovation must have followed the decay of the inherited *ei̯e/o-causative, which was probably conditioned, at least in part, by the loss of the intervocalic *-i̯- and the subsequent vowel contraction. It is therefore likely that pattern B remained productive after those sound changes had taken place. Another hint at the inner-Armenian productivity of pattern B concerns a series of sound changes that modified the aorist stems, from which new marked present stems were derived—for example, the simplification of *Cs-clusters and *sk to dental affricates: PIE aor. *leu̯k-s- > PA aor. *leu̯c- → prs. *leu̯c-an- > CA aor. lucʽ-i, prs. lucʽ-an-em ‘to kindle’ (EDAIL 311–312); PIE impf. *pr(k̑)-ske- (cf. Skt. impf. ápr̥cchat) > PA *harc- → PA *harc-an- > CA aor. harcʽ-i, prs. harcʽ-an-em ‘to ask’ (EDAIL 396). Here may also belong two of the aforementioned verbs with aorist stems in -ṙ-, aṙ- of aṙnum ‘to receive’ and meṙ- of meṙanim ‘to die’, if one derives those aorist stem from sigmatic stems *h₂er-s- (cf. Gk. aor.
Lucʽanem tr. ‘to kindle’ from PA aor. *leu̯k-s- and above all mucanem ‘to bring in’ from PA aor. *meu̯d-s- (next to mtanem ‘to enter’ from aor. *mut-e/o-) suggest that the sigmatic suffix could have a transitivizing value in PA similarly to Greek; cf.
Another issue is related to unmarked aorist stems that go back to PIE present stems, including the marked ones (see, in particular, Kim 2019 with further references). Here belong PIE prs./impf. *bʰer-e/o- > CA prs./aor. ber- ‘to bring’ (EDAIL 176), PIE prs./impf. *ki̯eu̯-e/o- > CA aor. čʽog- (the suppletive aorist of ertʽam ‘to go’; Kocharov 2020), PIE prs./impf. *pr(k̑)-ske/o- > CA aor. harcʽ- of harcʽanem ‘to ask’, perhaps PIE prs. *h₃ei̯gʰ-e- (Gk.
Based on such examples, it has been suggested that the inherited aorist and imperfect had merged in PA and the aspectual contrast had been abandoned in the past tense until it was reintroduced at a later stage when the new imperfect developed (cf. Meillet 1936: 114; Godel 1975: 126; Viredaz 2018). Note however that the formation of a new marked present stem took place after the older present stem became unmarked as a result of sound change and began to function as an inherited root aorist stem. With that, some of the root aorists were opposed to inherited marked present stems from at least dialectal PIE to CA (cf. lkʽanem mentioned above); thus, it should be stressed that the formal contrast between the two stems had never been abandoned for the entirety of the verbal lexicon throughout the PA period. There were arguably fewer verbs in which the contrast had been neutralized through sound changes than those for which it had been maintained.
In order to explain cases where the older imperfect surfaces as the CA aorist, Kim (2019: 151–155) and independently Kocharov (2019: 239–260) suggested that the outcomes of the inherited marked present stems analogically spread to the inherited root aorist stem (a salient example being PIE prs. *pr(k̑)-ske/o- > CA aor. harcʽ- ‘to ask’). Within the approach investigated in this paper, such analogical levelling could be conditioned by the pivotal status of a formally unmarked present stem modified by sound changes. By that time, pattern B was still productive, given that marked present stems were formed at a later stage (cf. PA aor., prs. *harc- → prs. *harc-an-). Therefore, the motivation for the change of a formally unmarked and pivotal present stem *harc-e- to marked *harc-an- remains largely unexplained.
One may assume a stage of PA, when the three unmarked stems *harc- ‘to ask’, *ber- ‘to bring’, and *u̯oi̯d- ‘to know’ could be used to derive both the present and the aorist forms. Starting from that stage, the structure of the underlying PIE paradigm of these verbs (e.g., the defective paradigm of PIE *bʰer-) is not relevant for explaining subsequent PA innovations. The three aforementioned verbs developed secondary stems according to patterns B, D, and A, respectively; see Table 10.
Table 10
Secondary PA tense-aspect stems
|
harcʽanem |
berem |
gitem |
|
|---|---|---|---|
|
‘to ask’ |
‘to bring’ |
‘to know’ |
|
|
PA present/aorist stem |
*harc-e- |
*ber-e- |
*u̯oi̯d-e- |
|
CA present stem |
harcʽ-an- |
ber- |
git- |
|
CA aorist stem |
harcʽ- |
ber- |
git-acʽ- |
|
Markedness pattern |
B |
D |
A |
The fact that no aorist stems *harc-ac- (CA aor. xharcʽ-acʽ-i) and *ber-ac- (CA aor. xber-acʽ-i) were formed shows that category-level markedness of stems had not yet become default when the paradigms of these verbs were being shaped. The choice of pattern might have still depended on the relative frequency of stems, itself rooted in lexical aspectual features—after the inner-Armenian sound changes had blurred the older markedness patterns. While the aorist stem of atelic verbs (like ‘to know’) was marked by the suffix *-ac-, the aorist stem of telic verbs remained unmarked. If so, the fact that the present stem of *harc- was extended with a suffix cannot be explained by formal analogy to verbs with root aorists alone (pace Kim 2019: 151–155). With that, the lack of *ber-an- suggests that the verb was underspecified for telicity and both of its stems were frequent enough to avoid recharacterization.
4.2 Pattern D
Many verbs that can be attributed to pattern D result from the blurring of morpheme bouldaries in stems of the older pattern B, e.g., linel ‘to become’ and daṙnal ‘to turn’ (see Table 5). Here also belong cases where the thematic vowels -u- and -a-, which used to mark the present stem, were reanalyzed as part of inflection. Thus, the proto-type of tołum changed from B to D after the vowel of the present u-stem had been reanalyzed as part of endings, cf. the present subjunctive endings of the u-conjugation (-ucʽ-um, -ucʽ-us, etc.), in which the -u- no longer belongs to the stem.
Cases where both stems can be derived from an unmarked stem inherited from PIE are extremely rare, cf. prs./aor. ber- of berem from PIE *bʰer-e- (Skt. bhárati, etc.; LIV2 76–77) and prs./aor. ac- of acem from PIE *h₂eg̑-e- (Skt. ájati ‘to lead’, etc.; LIV2 255–256). Neither PIE *bʰer-e- nor *h₂eg̑-e- had an aorist stem in PIE—undoubtedly due to their lexicalized atelicity. The telic aspectual pair might have been expressed by the aorist forms of suppletive telic verbs (cf. Gk. aor.
4.3 Pattern A
Pattern A verbs have the aorist acʽ- and ecʽ-stems. The ultimate source of these formations was atelic (inchoative or iterative) verbs with an unmarked present and derived imperfect *ske/o-stems (see, in particular, Clackson 1994: 75–83). Thus, aor. kacʽ-i of kam ‘to stand’ may go back to *gʷh₂-ske/o- derived from PIE *gʷeh₂- (aor. Gk.
The stem-final *-ac- (CA -acʽ-) must have been reanalyzed as an aorist suffix at a PA stage after the vocalization of interconsonantal laryngeals. The suffix was then used to derive new aorist stems from atelic verbs, which previously used their imperfect forms to express past events. That must have been the case of a stative verb git-em, aor. git-acʽ- from PIE prf.-prs. *u̯oi̯d-e- → PA aor. *u̯oi̯d-ac-. Along the same lines, the aorist suffix *-ac- spread to atelic denominal verbs in *-e-i̯e/o- and *-ā-i̯e/o- (*-eh₂-i̯e/o-) replacing the thematic vowel, which resulted in *-ei̯-ac- and *-āi̯-ac-, respectively. There is an intricate functional connection between the atelicity and verbs of manner derived from base nouns which correspond to the direct object of content of such verbs; the direct object can be easily suspended under antipassivization, cf. gorc ‘work’ → gorc-em, aor. gorc-ecʽ-i transitive ‘to do (of work)’, antipassive intr. ‘to work’ (PA prs. *u̯orj-ei̯-e-, aor. *u̯orj-ei̯-ac- next to *u̯orj-o- ‘work’ from PIE *u̯erg̑- ‘to work’, Gk.
After the spread of *-ac-, the suffixes *-ei̯- and *-āi̯- must have become derivational morphemes available in both tense-aspect stems and void of aspectual semantics. Following the loss of the intervocalic -i̯- and the contraction of the homorganic vowels (*-ei̯e- > *-e- and *-āi̯a- > *-a-), the type of PA prs. *u̯orj-ei̯-e-, aor. *u̯orj-ei̯-ac- must have yielded prs. *u̯orj-e-, aor. *u̯orj-eac-. Note that, if the causatives in -oycʽ- and their present stem -ucʽ-an- were formed after the loss of the intervocalic -i̯-, one may argue that pattern B had still been productive when the aorist *eac-stem of pattern A was on the rise. Starting from that stage, the aorist suffix *-eac- began to analogically spread to verbs with unmarked present stem in a thematic vowel *-e- of various origins. The analogical spread of the aorist *ac- and *eac-stems onto telic verbs, the present stem of which remained unmarked, reveals the turn towards the category-level markedness strategy. That drift must have been accomplished before the influx of Middle Iranian loanwords, which were adopted to the default class with an ecʽ-aorist regardless of their lexical aspect and relative frequency of forms derived from the two tense-aspect stems.
The verbs that follow pattern A contain very heterogeneous inherited morphological material recycled into what synchronically consitutes verbal roots, including PIE root presents (PIE prs. *kʷrep-e- > CA erew-im, aor. erew-ecʽ-i ‘to appear’), *ske/o-stems (PIE prs. *slh₂-ske- > PA *alac-i̯e- > CA prs. ałačʽ-em, aor. ałačʽ-ecʽ-i ‘to pray’), *i̯e/o- and *ei̯e/o-stems including primary and denominal verbs (PIE *l(e)h₂-i̯e- > CA l-am, aor. l-acʽ-i ‘to cry’; PIE prs. *gʷen-i̯e/o- > CA ǰnǰ-em ‘to beat’, aor. ǰnǰ-ecʽ-i; PIE prs. *u̯okʷ-i̯e- > CA prs. gočʽ-em, aor. gočʽ-ecʽ-i ‘to cry’; PIE prs. *u̯org̑-ei̯e- > CA prs. gorc-em, aor. aor. gorc-ecʽ-i ‘to work’), *eh₂(-i̯e/o)-stems (PIE *gʷi̯(e)h₃- → *gʷih₃-eh₂(i̯e/o)- > CA ke-am, aor. ke-acʽ-i; PIE prs. *mVn-eh₂(-i̯e/o)- > CA prs. mn-am, aor. mn-acʽ-i), and PIE perfect stems (PIE *u̯oi̯d- > CA git-em, aor. git-acʽ-i).
The drift towards pattern A was ongoing throughout the classical and post-classical period of Armenian. Thus, one finds numerous cases of replacement of the present an-stems by a suffixless thematic stem in CA texts (cf. anicem for anicanem ‘to curse’, kcem for kcanem ‘to bite’, lkʽem for lkʽanem ‘to leave’ etc.; see Kocharov 2019: 261–262) as well as in Middle Armenian (cf. mtem for mtanem ‘to enter’, lucem for lucanem ‘to solve’; see Karst 1901: 266–267). The present stem bek- ‘to break’, attested from the tenth century (Xosrov Anjewacʽi; see NBHL 1.479), also belongs to the post-classical instantiation of the morphological drift and completes the transition from the PIE unmarked aorist stem *bʰeg- to the Middle Armenian unmarked present bek- spanning over four thousand years. By that time, the telic semantics of the verb ‘to break’ and the frequency of the perfective forms derived from the aorist stem were no longer revelant for the markedness pattern.
4.4 Pattern C
Pattern C includes classes the rise of which belongs to different chronological layers. In PIE, the paradigms in which both the present and aorist stem had contrastive suffixes were probably marked for other categories than tense and aspect. Thus, the class with the nasal affix in the present stem and the sigmatic suffix in the aorist stem, mentioned in § 4.1, was most likely originally marked for transitivity. Similarly, the class with the present *nu-stem and the aorist *eh₁-stem, attested in Greek and Armenian, seems to have originally been marked for intransitivity or valency-decreasing derivation, cf. mp. pres.
The type of im-an-am, aor. im-acʽ-ay ‘to understand’ must have developed at the stage of PA after the aorist suffix *-ac- had become productive and before the present suffix *-an- ceased to be productive. Besides a few primary verbs, here belong productive denominal verbs with inchoative semantics ‘to become X’ (where X is a base noun or adjective); see Kocharov 2019: 136–177. One may assume that the suffix *-an- was introduced in this class to mark telicity, in contrast to stative and atelic dynamic verbs of the a-conjugation such as mnam intr. ‘to remain’, perhaps from PA *mVn-eh₂(-i̯e/o)- from PIE *men- ‘to remain’ (LIV2 437), and eṙam ‘to crawl’ from PIE *h₁ers-eh₂-i̯e/o-; cf. Lat. errāre ‘to err’ (Klingenschmitt 1982: 96). Thus, the core of pattern C was shaped at a stage when lexical aspectual features had still been relevant to the markedness of stems. This nondirected type remained stable well into the Middle Armenian texts (see Karst 1901: 282) and was not replaced by pattern A—perhaps because it did not violate the markedness constraint according to which the past tends to be coded by at least as many overt morphemes as the present.
5 Conclusions
The present study clarifies, to some extent, the role of the frequency of use for formal markedness and diachronic stability of the CA tense-aspect stems.
The variety of the CA verb classes can be arranged into four markedness patterns depending on the type of formal contrast between the present and aorist stems. The biblical evidence shows that pattern A, characterized by the unmarked present stem and marked aorist stem, is by far the most common type. At the same time, pattern B, characterized by the unmarked aorist stem and marked present stem, is significantly better represented in the most frequent segment of the verbal lexicon than in the complete set of verbs attested in the Bible.
The revised evidence suggests that the higher the overall frequency of a verb, the higher the correspondence between the relative frequency of one of the two stems and its markedness. The less frequent a verb is, the more likely it belongs to the predominant pattern A, irrespective of the attested relative frequency of the two stems.
Based on our current knowledge of the PIE stem-markedness patterns and on the etymological analysis of the selection of the most frequent verbs attested in the Armenian Bible, one may assume that there was a change from a differential marking of tense-aspect stems rooted in the lexical aspectual features of individual verbs to a marking pattern with an unmarked present stem. Taking into account this large-scale morphological drift between the two markedness strategies allows us to better understand the nature of inner-Armenian analogical changes, which only in part restored the oppositional markers eliminated by sound laws. In particular, it has been demonstrated that pattern B was still productive when the present and aorist stems merged in a limited number of verbs; therefore there is no reason to believe that the stems of the two past tenses completely merged at any stage of PA.
Acknowledgments
The research is part of the project “The Lexicon-Grammar Interface in the History of the Classical Armenian Verb” funded by the Alexander von Humboldt Foundation (2021–2022).
An important disclaimer needs to be added here. Quantitative data represent a genuinely insecure parameter when dealing with limited corpora. The biblical corpus (661,668 occurrences of 11,269 words) on which the present study relies is further constrained by its genre and the influence of the Greek original on the Armenian translation. Any results on the correspondence of markedness and frequency discussed in the present paper must therefore be considered provisional. Altogether, unless proven to be wrong based on larger datasets, these results remain a valid piece of evidence. The translation of the Bible is the earliest text written in CA in the first half of the fifth century, and it is representative of the emerging CA literary language.
The cells with ranging values reflect cases when only one of the two stems is attested in the biblical corpus and the attribution to a verb class is controversial, see below.
The frequency data provided in Tables 7–9 account for the occurrences of personal verb forms as well as of the non-personal forms in -eal, -li, -locʽ, and the infinitive; verbal adjectives in -ac, -oł, etc. have been excluded. The complete lists of occurrences are provided in Supplement 2 (
The past participle matucʽeal (94×) is morphologically ambiguous since it coincides with the past participle of the base verb matčʽim ‘to approach’. Thus, depending on how many of the attested participles belong to the causative verb, the actual ratio ranges from 1.5 to 1.0.
This number includes 878 occurrences of the aorist subjunctive derived from the irregular stem ta-cʽ/ǰ-.
The biblical corpus contains 294 occurrences of the morphologically ambiguous form darjeal which functions as the past participle of daṙnal and as an adverb meaning ‘again’.
The derivational basis is ambiguous in the case of the past participle (up to 42 occurrences, if all the oblique forms belong to the paradigm of the participle and not the infinitive).
The reconstruction of the grammatical functions of the PIE tense-aspect stems remains a debated issue; see, among others, Dahl (2010) and Hollenbaugh (2021) for discussion. The present paper relies on the traditional reconstruction.
Note that the stem-final velar of aor. likʽ/lkʽ- was not palatalized by the following front vowel in the pivotal 3sg. aor. form PIE *h₁e-likʷ-e-t > CA elikʽ. In line with the approach discussed in this paper, it is unlikely that the result of the palatalization was replaced in a diachronically more stable aorist stem by analogy to the present stem. The solution therefore most probably concerns the relative chronology of the palatalization and the lowering and/or backing of *e to a non-palatalizing segment in PA (presumably in a post-tonic position). The matter requires further investigation.
Gk.
References
Bartolotta, Annamaria. 2009. Root lexical features and inflectional marking of tense in Proto-Indo-European. Journal of linguistics 45: 505–532.
Bubenik. Vit. 1997. The verbal system of Classical Armenian. In Tense and aspect in Indo-European languages. Theory, typology, diachrony, ed. John Hewson & Vit Bubenik, 67–81. Amsterdam: John Benjamins.
Bybee, Joan. 1985. Morphology. A study of the relation between meaning and form. Amsterdam: John Benjamins.
Bybee, Joan. 1994. The grammaticization of zero. Asymmetries in tense and aspect systems. In Perspectives on grammaticalization, ed. William Pagliuca, 235–254. Amsterdam: John Benjamins.
Clackson, James. 1994. The linguistic relationship between Armenian and Greek. Oxford: Blackwell.
Clackson, James. 2007. Indo-European linguistics. An introduction. Cambridge: Cambridge University Press.
Croft, William. 2003. Typology of universals2. Cambridge: Cambridge University Press.
Dahl, Eystein. 2010. Time, tense and aspect in early Vedic grammar. Leiden: Brill.
DELG2 = Chantraine, Pierre. 2009. Dictionnaire étymologique de la langue grecque. Histoire des mots2. Paris: Klincksieck.
Djahukyan, Gevorg B. 2010. Armenian etymological dictionary. Yerevan: Asoghik.
EDAIL = Martirosyan, Hrach. Etymological dictionary of the Armenian inherited lexicon. Leiden: Brill.
Godel, Robert. 1975. An introduction to the study of Classical Armenian. Wiesbaden: Reichert.
Greenberg, Joseph. 1966. Language universals, with special reference to feature hierarchies. The Hague: Mouton.
Haspelmath, Martin. 2006. Against markedness (and what to replace it with). Journal of linguistics 42: 25–70.
Haspelmath, Martin. 2008. Frequency vs. iconicity in explaining grammatical asymmetries. Cognitive linguistics 19: 1–33.
Hollenbaugh, Ian B. 2021. Tense and aspect in Indo-European. A usage-based approach to the verbal systems of the Rigveda and Homer. Ph.D. dissertation, University of California, Los Angeles.
Jensen, Hans. 1959. Altarmenische Grammatik. Heidelberg: Carl Winter.
Karst, Josef. 1901. Historische Grammatik des Kilikisch-Armenischen. Strassburg: Trübner.
Kassian, Alexei S., Mikhail Zhivlov, George Starostin, Artem A. Trofimov, Petr A. Kocharov, Anna Kuritsyna, & Mikhail N. Saenko. 2021. Rapid radiation of the inner Indo-European languages: an advanced approach to Indo-European lexicostatistics. Linguistics 59(4): 949–979.
Kim, Ronald I. 2019. PIE aorists and imperfects in Old Armenian. In Proceedings of the 30th annual UCLA Indo-European conference, Los Angeles, November 9th and 10th, 2018, ed. David M. Goldstein, Stephanie W. Jamison, & Brent Vine, 141–159. Bremen: Hempen.
Klingenschmitt, Gert. 1982. Das altarmenische Verbum. Wiesbaden: Reichert.
Kocharov, Petr. 2014. The derivational semantics of the Classical Armenian -č
῾ - stems. Acta linguistica Petropolitana 10: 202–226.Kocharov, Petr. 2019. Old Armenian nasal verbs. Archaisms and innovations. Ph.D. dissertation, Leiden University.
Kocharov, Petr. 2020. PIE *kieu-e/o- ‘move’ as a lexico-morphological isogloss. Indo-European linguistics and classical philology 24: 190–205.
Kocharov, Petr. Forthcoming. The origin of the Classical Armenian causative. In Studies in Armenian grammar and lexicon. Proceedings of the workshop on Armenian linguistics, Würzburg, 4–5 April 2022, ed. Petr Kocharov & Daniel Kölligan.
Kortlandt, Frederik. 2003. Armeniaca. Comparative notes. With an appendix on the historical phonology of Classical Armenian by Robert S.P. Beekes. Ann Arbor: Caravan Books.
de Lamberterie, Charles. 2013. Grec, phrygien, arménien. Des anciens aux modernes. Journal des savants 1: 3–69.
de Lamberterie, Charles. 2020. La version arménienne de la grammaire de Denys de Thrace. Étude linguistique. Revue des études arméniennes 39: 49–72.
LIV2 = Rix, Helmut, Martin Kümmel, Thomas Zehnder, Reiner Lipp, & Brigitte Schirmer. 2001. Lexikon der indogermanisichen Verben. Die Wurzeln und ihre Primärstammbildungen2. Wiesbaden: Reichert.
Martzloff, Vincent. Forthcoming. Mundus, chambre souterraine, cave, firmament, monde. Essai d’analyse sémantico-référentielle et comparative. In Studies in Armenian grammar and lexicon: Proceedings of the workshop on Armenian linguistics, Würzburg, 4–5 April 2022, ed. Petr Kocharov & Daniel Kölligan.
Meillet, Antoine. 1910–1911. Deux notes sur des formes verbales indo-européennes. Mémoires de la Société de Linguistique de Paris 16: 239–246.
Meillet, Antoine. 1913. Altarmenisches Elementarbuch. Heidelberg: Carl Winter.
Meillet, Antoine. 1936. Esquisse d’une grammaire comparée de l’arménien classique2. Vienne: Imprimerie des PP. Mekhitaristes.
Napoli, Maria. 2007. Telicity as a parameter of aspect in Homeric Greek. Activity and accomplishment verbs. Indogermanische Forschungen 112: 124–169.
NBHL = Awetik‛ean, Gabriēl, Xačʽatur Siwrmēlean & Mkrtičʽ Awgerean. 1836–1837. Nor baṙgirkʽ haykazean lezui [New dictionary of the Armenian language]. Venice: St. Lazar.
Olsen, Birgit A. 1999. The noun in Biblical Armenian. Origin and word-formation, with special emphasis on the Indo-European heritage. Berlin: Mouton de Gruyter.
Viredaz, Rémy. 2001–2002. Contractions et place de l’accent en arménien ancien. Revue des études arméniennes 28: 1–11.
Viredaz, Rémy. 2018. Les désinences arméniennes d’aoriste et d’imparfait. Acta linguistica Petropolitana 14: 163–268.
Willi, Andreas. 2018. Origins of the Greek verb. Cambridge: Cambridge University Press.
