Yakov Testelets, Russian State University for Humanities, Moscow

Are there Strong and Weak Pronouns in Russian?

It is usually assumed that most pronouns are proforms for noun phrases, i. e. that the syntactic distribution of pronouns and full NPs is similar, if not identical. It is obvious, however, that pronominal clitics show systematic and predictable differences from the distribution of full NPs and stressed pronouns. As is well known, standard Russian lacks pronominal clitics, and the syntax of Russian pronouns (apart from their binding properties) has not been subject to theoretical analysis, cf. (Franks and King 2000). Things became different since Anna Cardinaletti and Michal Starke proposed what they believe to be a universal tripartite typology of pronouns (and apparently not only pronouns): clitics vs. weak elements vs. strong elements in their influential paper (Cardinaletti and Starke 1999a). The starting point of my article will be to check whether the predictions of Cardinaletti and Starke’s theory are borne out by Russian data (section 1). I will show that there is no reason to split the class of the stressed pronouns in Russian in the way that Cardinaletti and Starke propose. Some crucial facts on which their typology of pronouns is based are better explained if a class of positions is determined in which stressed pronouns are not normally available; I will try to demonstrate that such a difference in the distribution of full NPs and pronouns has theoretical significance (section 2).

The main points of Cardinaletti and Starke’s theory can be outlined as follows. The grammatical feature that takes the three above-mentioned values is structural deficiency, i. e. the deficiency in syntactic structure that underlies pronominal elements. Clitics are structurally deficient with respect to weak pronouns, which are in turn deficient with respect to strong pronouns. The former two classes are deficient, the latter (i. e. strong pronouns) is not:

Cardinaletti and Starke claim that deficient and strong pronouns show overt or covert distinctions in any language. The only Slavic language they include in their study is Slovak, in which they find a covert distinction between strong and weak pronouns and overtly distinguished pronominal clitics. It is expected therefore that in Russian the two classes must also be found.

The deficiency characteristics involved in Cardinaletti and Starke’s study are pervasive in the linguistic structure in the sense that they involve syntactic distribution, morphology, semantics and prosody.

The distributional properties can be summarized as follows. Only strong pronouns can be coordinated, cf. the Italian strong 3rd person feminine pronouns lei (singular) and loro (plural) vs. weak feminine pronouns essa (singular) and esse (plural):

As has been known since (Kayne 1975), clitics cannot be coordinated. This property is common to both classes of deficient pronouns. Essa ‘she’ and esse ‘they (fem.)’ in Italian are stressed, but weak pronouns, whereas lei and loro with the same meanings are strong, and only the latter can be coordinated.

Another syntactic difference is that only strong pronouns can be modified by adverbs:

Also since (Kayne 1975), it has been well known that clitics cannot be modified. Again, this property is shared by both classes of deficient pronouns.

The next relevant syntactic feature is that deficient pronouns occur in S-structure in derived positions but cannot occur in base position (4a), dislocation (4b), cleft (4c), and isolation (4d):

That the dative clitic gli ‘to him’ has undergone head movement in (5), can be seen from the fact that it moves together with the verb: the clitic is “transported” by the verb over the subject in the construction of conditional inversion.

The most relevant semantic feature is humanness, viz. strong pronouns denote human referents only:

Very often, — in fact, judging from Cardinaletti and Starke’s material, almost in all languages, — strong and weak pronouns are homophonous, and the only method to uncover the difference is to correlate their distributional characteristics with semantics. Cardinaletti and Starke dedicate a large portion of their paper to show that in all languages in their sample personal pronouns, when coordinated or modified, can denote human referents only. Pairs of strong and weak pronouns that are different in form like essa and lei in Italian are rare; in most languages the fact that the difference exists at all, comes to light in that a pronoun can denote only humans in a “strong” distributional context, for instance, when it is coordinated:

Cardinaletti and Starke claim therefore that there are in fact two pronouns, and not a single one, realized as elle in French, sie in German, ich in Slovak, and so on — one strong and one weak.

One more semantic feature is discourse prominence: deficient pronouns must have an antecedent prominent in the discourse. It is not the case, as it may seem, that deficient pronouns can never be contrastively focussed or accompany ostension. Deficient elements are permissible with contrastive stress and ostension only if they refer to an entity which is already prominent in the discourse, as is the case in (10c). The arrow in (10) denotes ostensive use of the pronoun, for example when it is accompanied by an indicating gesture. Ostensive use is impossible when a pronoun is not discourse-prominent (10a):

In prosody, deficient pronouns are characterized by restructuring: unlike strong pronouns, they may undergo liaison, reduction, etc.:

(11) a. I saw’ya in the garden.

b. *I saw’ya and John in the garden.

c. I saw you and John in the garden.

Only deficient pronouns may be destressed and contracted, which, according to Cardinaletti and Starke, is the strongest evidence for the existence of otherwise covert homonymy of strong and deficient pronouns in English. 1st and 2nd person pronouns always denote humans, which means that they may be either strong or weak. In (11c) you is strong because it is coordinated and therefore cannot undergo restructuring, cf. (11b); restructuring is possible only when the pronoun is weak (11a).

One more prosodic feature is stress: clitics cannot bear lexical word stress, whereas strong and weak pronouns can.

After a detailed survey of the relevant features Cardinaletti and Starke conclude: “Out of all the possible combinations of strong and deficient (personal) pronouns inside a tripartite paradigm, only one is attested: one strong and two deficient. It is never the case that a tripartition stems from there being two strong and one deficient pronoun, etc. Similarly, it is never the case (to our knowledge) that there is more than three classes, with two types of strong and two types of deficient pronouns, etc.” (op. cit., 167).

In Cardinaletti and Starke’s analysis, deficient pronouns correspond to less syntactic structure than strong pronouns. They split the traditional Determiner Phrase into CP (C⁰ is the locus of referential features and Case) and IP (I⁰ is the locus of agreement-type information, i. e. phi-features), thus suggesting the same structure for extended noun projections and for extended verb projections in a sense very close to that of (Grimshaw 1991). They also postulate the S⁰ node, which denotes a functional category between C⁰ and I⁰ containing polarity, focus and prosody-related features. Pronouns that lack this level are prosodically deficient, i. e. clitics. Strong elements have a CP (the topmost part of the traditional DP, (12a)), but deficient elements (clitics and weak pronouns) lack it. In other words, weak pronouns are “peeled” strong pronouns (12b), and clitics are “peeled” weak pronouns (12c). Strong elements may stay in situ; deficient elements must move in order to compensate for the deficiency of their structure and of the features that correspond to the missing components of structure.

Demonstrative pronouns stand apart from the tripartite classification: they have a special morphological marker (like th- in English); they must be disjoint from any c-commanding antecedent (Principle C), whereas personal pronouns must be disjoint only from local antecedents (Principle B); demonstratives may refer to non-human entities in contexts requiring strong forms like coordination:

(13)  Which flowers? Those, and those.

So when demonstrative pronouns have antecedents, they behave like long-distance pronominals and full NPs, i. e. they cannot be bound by their antecedents.

According to Cardinaletti and Starke’s theory, in Russian, like in English, a covert distinction between homophonous weak and strong personal pronouns must exist. However, the expected correlation between semantic and distributional characteristics, i. e. the inability of pronouns with nonhuman referents to be coordinated or modified, is poorly supported by Russian data (counterexamples are found in other languages too, cf. (Diesing 1999) on English; (Holmberg 1999) on Swedish etc.).

At first glance, the correlation holds true in Russian. The 3rd person pronouns on ‘he, it (M)’ and ona ‘she, it (F)’ can be coordinated when they denote human referents (14a), but are reluctant to coordinate when they denote non-human referents belonging to the masculine and the feminine genders respectively (14b):

In fact, nonhuman personal pronouns can, sometimes marginally, but very often quite normally occur as leftmost conjuncts (15a, 16a, 17–18), although they are excluded as non-leftmost ones (15b, 16b, 17–18). Pronouns with human reference are coordinable in both positions (16cd):

Moreover, even this weaker restriction for nonhuman pronouns, contrary to Cardinaletti and Starke, cannot be accounted for by the structure of pronominal constructions. Cardinaletti and Starke explain the fact that deficient pronouns are non-coordinable by recourse to the principle that only CPs can be coordinated: recall that deficient pronouns lack the CP level (12bc), and therefore cannot undergo coordination.

In fact, coordination of any constituent as a non-leftmost conjunct containing a personal pronoun with a nonhuman referent is infelicitous, cf. coordination of converb clauses (19), finite clauses (21a), VPs (22), and PPs (23–24):

The contrast in (21) shows that the pronoun ego ‘it.M.ACC’ can be used if it belongs to the first conjunct (an infinitival clause in (21b)), but cannot if it belongs to the second conjunct (a finite clause in (21a)).

Whatever factor makes it hard or impossible to coordinate a non-leftmost conjunct containing a nonhuman personal pronoun, it is obvious that this must be a non-local factor, and not the structure of the pronoun itself, — contrary to what Cardinaletti and Starke hypothesize. It seems obvious from the above examples that the restriction on coordinating nonhuman pronouns may be due to a particular instance of a more general pragmatic principle. This principle demands that coordination of anything containing nonhuman pronouns be avoided, although sometimes a leftmost conjunct containing a nonhuman pronoun is permissible.

The general tendency in the use of anaphoric pronouns in coordinate structures seems to be the following: the antecedent must be contained in one of the preceding conjuncts (27–28). If there is no preceding conjunct at all, i. e. if the pronoun is contained in a leftmost conjunct, its antecedent must be found outside the coordinate construction. This pragmatically motivated tendency seems to be more or less consistently neglected with human antecedents, but kept more strictly with non-human antecedents:

The antecedent in (28) normally cannot be found outside the coordinate construction which is noted in the referential indexes.

It can be seen further that Russian personal pronouns with nonhuman reference can be modified by adverbs contrary to what the theory predicts, cf. (27) and the following examples:

The only isolated context where nonhuman personal pronouns are impossible is answers to wh-questions (34b); however, in the same context human pronouns also sound infelicitous (34a):

In the absence of reliable correlation between semantics and distribution of the two classes thought to be homophonous, there is no reason to split the class of personal pronouns. We cannot even say that only strong pronouns exist in Russian, because all personal pronouns can denote non-human referents, and the restrictions on humanness, as we have seen, cannot be summarized in terms of grammatical classes. We cannot say either that all pronouns in Russian are weak, because they all in principle can be coordinated, modified, used in isolation, cleft etc., and the restrictions that can be observed do not imply any grammatical classification of pronouns.

We may conclude therefore that there is (as seems to be the case in many languages) one single syntactic class of personal pronouns in Russian.

Is there any structural difference that underlies the distinction of strong vs. weak pronouns? It seems that some positions are available, or more easily available, for pronouns with human referents only. There are also other positions in which pronouns are either unavailable at all, or available also under special conditions like discourse prominence, cf. (10a) vs. (10c).

There are “strong” positions that require that a pronoun be human and non-reduced phonologically, like coordination, at least in a non-leftmost conjunct, and modification, as is the case in Romance and Germanic languages, cf. (3). In determining what is a strong position, an interplay of semantic, pragmatic and structural characteristics must be taken into account.

There are also “weak” positions that require discourse prominence and allow phonological reduction. In some languages like Italian, there are pronouns for which only weak positions are available. Finally, there are positions available for clitics only. Unlike strong and weak pronouns, clitics obviously constitute a separate grammatical class.

Now we turn to weak positions in Russian in which pronouns (not only personal ones) are either impossible or possible only if they have referents prominent in the discourse. In Cardinaletti and Starke’s data, positions that might be called “weak” are also found: weak pronouns are those that start in th-marked positions, but then undergo movement motivated by the lack of case characteristics located in the CP level that they lack. So they can be overtly realized only in weak, i. e. non-th-marked, positions, cf. (4a).

We could not find any testable correlation between the “weak” properties of Russian pronouns — such as referent’s discourse prominence and the ability to undergo phonological reduction, with their being in movement sites. It seems that all Russian personal pronouns, with human or non-human referents alike, may precede or follow sentence-level adverbials and may precede or follow the verb, which probably indicates that the distribution of pronouns is not obviously motivated by movement.

However, there is one kind of weak positions: NP adjuncts and PP adjuncts. In Russian, many NP adjuncts and NPs within PP adjuncts are either unavailable for pronouns or available for those having discourse-prominent reference. We may hypothesize therefore that pronouns lack some component of structure or some structurally dependent feature, which, whatever it might be, makes it possible for non-pronominal (i. e. full) NPs to occur as NP adjuncts or inside PP adjuncts.

The list of some syntactic contexts unavailable for pronouns may be found in (Švedova (ed.) 1980: 455). Even the first look at this list enables us to see that positions in which pronouns are ungrammatical or infelicitous are mostly NP and PP adjuncts, although the reverse is not true — not all NP or PP adjuncts cannot be filled with pronouns.

Positions available for all kinds of pronouns include

1) arguments:

(36) a. da-t’ odn-u minut-u (ee)

give-INF one-ACC minute-ACC (it.ACC)

‘to give one minute (it)’

b. skuča-t’ po druz’-jam (nim)

miss-INF PR friend-PL.DAT (they.PL.DAT)

‘to miss (one’s) friends (them)’

c. vystrelil iz ružj-a (nego)

he.shot from gun-GEN (it.GEN)

‘He shot the gun (it)’

d. rešenie problem-y (ee)

solution problem-GEN (it.GEN)

‘solution of the problem’

e. razgovor s drugom (nim)

conversation with friend (him)

‘conversation with a friend (him)’

2) some PP adjuncts like those with prepositions dlja ‘for’, s ‘with’, iz-za ‘because of’, bez ‘without’, u ‘at’. Here belong most “lexical” (Yadroff and Franks 2001), or “non-prototypical” (nepervoobraznye, Es’kova 1996) prepositions like posle ‘after’, vvidu ‘because of’, vokrug ‘around’:

(37) a. zapiska dlja Saš-i (dlja nego/ kogo?)

message for S.-GEN (for him.GEN/ whom.GEN?)

‘a message for Saša (him/whom?)’

b. rabotaem s prepodavatelem (nim/kem?)

we.work with teacher (him/whom?)

‘we work with the teacher (him/whom?)’

c. otsutstvova-t’ iz-za bolezni (nee/ čego? ...)

be.absent-INF because.of illness (it/ what?)

‘to be absent because of the illness (it/what?)’

d. rabotat’ u Ivana (nego)

work-INF at Ivan (him)

‘to work at Ivan’s (his place)’

e. uexal posle prazdnika (nego)

left after feast (it)

‘he left after the feast (it)’

f. izmenit’ rešenie vvidu novyx obstojatel’stv (nix)

change decision because.of new circumstances (them)

‘to change the decision because of the new circumstances (them)’

g. side-t’ vokrug stola (nego)

sit-INF around table (it)

‘to sit around the table (it)’

3) some NP adjuncts like benefactive datives:

(38) prigotovit’ obed syn-u (emu/ komu? ...)

cook-INF dinner.ACC son-DAT (him.DAT/ who.DAT?)

‘to cook a dinner for the son (him/ whom?)’

Weak positions are those within other PP adjuncts. “Weak” prepositions are v, na (in spatial meaning), iz ‘from’; s ‘since’, po with various meanings, do ‘until’, na ‘for (temporal)’ and others:

(39) a. besedova-t’ v gostinoj (?nej/ *čem?)

talk-INF in lounge (it/ what)

‘to talk in a lounge (?it/ *what?)’

b. lekcija iz istori-i (*nee)

lecture from history (it)

‘a lecture from history (*it)’

c. poezd idet iz Moskvy (??nee; *čego?)

train comes from Moscow (??it; *what?)

‘the train comes from Moscow (??it; *what?)

d. s prošlogo mesjaca (*nego; *cego?) boleet

since last month (*it; *what?) is.ill

‘he has been ill since the last month (*it; *what)’

e. priexal na pjat’ dnej (*nix)

came for five days (*them)

‘he came for five days’

Note the contrast between argument (a) and adjunct (b) uses of “weak” prepositions in (40–43):

(40) a. Po komu oni skučajut?

PR whom they miss

‘Whom do they miss?’

b. *Po komu (znakomym) oni xodjat?

PR whom (acquaintances) they visit

‘Whom (their acquaintances) do they visit’

(41) a. Na čem (divane) oni sidjat?

on what (sofa) they sit

‘What are they sitting on (on the sofa)?’

b. ?? Na čem (divane) oni razgovarivajut?

on what (sofa) they talk

lit.: ‘What are they talking on (on the sofa)?’

(42) a. V čem oni soznalis’?

in what they confessed

‘What did they confess to?’

b. *V čem oni zasedali?

in what they had.session

‘In what (room) did they hold session?’

(43) a. ?? Iz nego (okna) on posmotrel

from it (window) he looked.out

‘He looked out of it (the window)’

b. Iz nego (ružja) on vystrelil

from it (gun) he shot

‘He shot it (the gun)’

Other NP adjuncts like measure accusative adjunct NPs denoting time and distance also cannot be pronominal:

(44) a. govori-t’ odn-u minut-u (*ee)

speak-INF one-ACC minute-ACC (*it.ACC)

‘to speak one minute (*it)

b. Ivan nes čemodan vsju dorog-u (*ee)

Ivan carried suitcase all way-ACC (*it.ACC)

‘Ivan carried the suitcase all the way (*it)’

c. Ivan pro-begal po magazinam ves’ den’ (*ego)

Ivan PRO-run PR shop all day (*it)

‘Ivan went to shops all day (*it)’

In (44c), Yadroff and Fowler (Jadrov and Fauler 1997) see a special case of what they call a “quasi-argument”, i. e. a non-referential, i. e. not referring to any participant, but nevertheless theta-marked category.

The same is true for various instrumental NP adjuncts:

(45) a. še-l les-om (*im/ *čem?)

walked forest-INSTR (it.INSTR/ what.INSTR)

‘he walked through the forest (*it/ *what?)’

b. propadat’ celymi nedeljami (*imi; *čem?)

be.absent whole weeks-INSTR (*they.INSTR;*what.INSTR)

‘be absent for whole weeks (*for them; *for what)’

c. pokupat’ saxar mešk-ami (*imi; *čem)

buy sugar bag-INSTR (*them; *what)

‘to buy sugar in bags’

d. vstretit’sja pozdnej osen’-ju (*čem)

meet late fall-INSTR (*what?)

‘to meet in late fall (*in what)’

Weak positions are either unavailable for pronouns or available under discourse prominence:

(46) V dome xoxot, i stekla zvenjat,

in house laughter and glasses clink

V nem šinkujut, i kvasjat, i perčat,

in it.M chop and pickle and pepper

I gvozdiku kladut v marinad.

and clove put in marinade

‘There is laughter in the house, and glass clinking, in it (the house) they are chopping, and pickling, and peppering, and putting clove in the marinade’ (Pasternak, “Indian Summer”)

(47) Komnata ploxo osveščena, poetomu rabota-t’

room badly illuminated therefore work-INF

v nej neudobno

in it uncomfortably

‘The room is badly illuminated, so it is uncomfortable to work in it’

(48) Tam byla malen’kaja gostinaja.

there was small lounge

V nej razgovarivali dva čeloveka.

in it talked two men

‘There was a small lounge. Two men were talking in it’

(46–48) represent the only instance of what to our knowledge can be called a weak pronoun in Russian, or, more exactly, a pronoun in a weak position — a position that requires that the pronoun’s referent be discourse-prominent.

If we have a closer look at the two classes of prepositions: one that permits and another that does not permit pronominal complements, we see that the prepositions that can be used with pronouns, as a rule, have more definite meanings and are less idiomatic that prepositions that normally do not take pronominal complements as adjuncts. They former can be regarded either as separate predicates with thematic potential of their own, or as derivative morphemes that extend the thematic grid of the verbal predicate changing its meaning (like causative, applicative and benefactive derivations do). In other words, in order to license a pronominal complement in an adjunct use, a preposition must be a predicative expression, contributing to the thematic structure. “Light” prepositions that cannot theta-mark their complements either do not occur with pronouns, or take them as complements only when they are arguments of the lexical predicate and theta-marked by it, cf. (40–43).

As for bare NP adjuncts, the idea of empty prepositions recently suggested by Yadroff and Fowler (Jadrov and Fauler 1997) for Russian accusatives of measure (44ab) seems to suggest a promising explanation. Yadroff and Fowler argue for empty prepositions, pointing out that some overt prepositions are interchangeable with null:

(49) a. pis’mo (k) materi

letter (to) mother.DAT

‘a letter to Mother’

b. čto kasaetsja (do) menja

what concerns (until) me

‘as regards me’

Most empty prepositions like that of the accusative of measure suggested by Yadroff and Fowler do not theta-mark their complements, and therefore the complement cannot be expressed by a pronoun. The benefactive empty preposition (38), on the contrary, produces a benefactive derivation (overtly marked in many languages, see (Baker 1988; Shibatani 1996)); therefore it theta-marks its complement, making a pronoun possible.

The factors involved in (36–48) may be summarized as follows:

(50) Pronouns, to be licensed, must be theta-marked, or their referents must be prominent in the discourse.

This mixture of structural and pragmatic factors is quite typical for the syntax of pronouns. We may expect that further study will clarify the problem of their interaction. English examples with PP adjuncts like (51) a. We stopped at the railway station; b. ?? What did we stop at?; cf. an argument c. What did we look at?; (52) a. I walked with John; b. Who did I walk with?; (53) a. ?He stayed at home because of it (illness); b. ??What did he stay at home because of?; c. His illness was serious, and because of it he stayed at home show that the same two factors: the class of the preposition (51–52) and discourse prominence (53) may determine the “weakness” of adjuncts in other languages.

Baker, Mark. 1988. Incorporation: a theory of grammatical function changing. Chicago: University of Chicago Press.

Cardinaletti, Anna, and Michal Starke. 1999a. The typology of structural deficiency: A case study of the three classes of pronouns. In Clitics in the Languages of Europe. Empirical Approaches to Language Typology, ed. by Riemsdijk, Henk van, 145–233. B. – N.-Y.: Mouton de Gruyter.

Cardinaletti, Anna, and Michal Starke. 1999b. Responses and Demonstratives. Ibidem, 273–290.

Diesing, Molly. 1999. Comments on Cardinaletti and Starke: “The typology of structural deficiency”. Ibidem, 243–247.

Es’kova, N.A. 1996. Pervoobraznye i nepervoobraznye predlogi. Formal’nyj aspekt. In: Rusistika. Slavistika. Indoevropeistika. Sbornik k 60-letiju Andreja Anatol’eviča Zaliznjaka, ed. by Gippius A.A., T.M. Nikolajeva, V.N. Toporov, 458–464. M.: Indrik.

Fowler, George, and Michael Yadroff. 1993. The Argument Status of Accusative Measure Nominals in Russian. Journal of Slavic Linguistics 1(2): 251–279.

Franks, Steven, and Tracy Halloway King. 2000. A Handbook of Slavic Clitics. New York, Oxford University Press.

Grimshaw, Jane. 1991. Extended Projections. Brandeis University: unpublished manuscript.

Holmberg, Anders. 1999. Comments on Cardinaletti and Starke: “The typology of structural deficiency”. In Clitics in the Languages of Europe. Empirical Approaches to Language Typology, ed. by Riemsdijk, Henk van, 263–266. B. – N.-Y.: Mouton de Gruyter.

Jadrov, M., and Dz. Fauler. 1997. Pristavka pro- i argumentnyj status imennyx grupp. In Glagol’naja prefiksacija v russkom jazyke, ed. by Krongauz, M., and Pajar D, 164–185. M.: Russkie slovari.

Kayne, Richard S. 1975. French Syntax. The Transformational Cycle. Cambridge, Mass.: MIT Press.

Shibatani, Masayoshi. 1996. Applicatives and benefactives: a cognitive account. In: Grammatical constructions, their form and meaning, ed. by Shibatani, Masayoshi, and Sandra A. Thompson. Oxford: Oxford University Press.

Švedova, Natalija Ju. (ed.). 1980. Russkaja grammatika. II. M.: Nauka.

Yadroff, Michael, and Steven Franks. 2001. The Origin of Prepositions. In Current Issues in Formal Slavic Linguistics, ed. by Zybatow, Gerhild, Uwe Junghanns, Grit Mehlhorn, Luka Szucsich, 69–79. Frankfurt a. Main etc.: Peter Lang.