Printer Friendly

Towards a typology of reciprocal constructions: focus on German and Japanese (1).

"L'amour ainsi exige de l'autre ne saurait rien demander : il est pur engagement sans reciprocite'"

(Sartre, L'Etre et le Neant, Paris: Gallimard, 1943: 424)


The aim of this article is twofold." (a) to offer a detailed analysis of reciprocal constructions in German and Japanese against the background of a preliminary typology of such constructions developed in analogy to the one proposed by L. Faltz (1985) for reflexivity and (b) to examine the implications of these analyses for supporting and possibly refining this typology. Reciprocity is defined as the grammatical encoding of symmetric relations, and four different types of reciprocal constructions are distinguished. Languages typically have more than one strategy at their disposal, which can often be combined but may also contrast and differ in their interpretation. It is shown that symmetric predicates play an important role in reciprocal structures and provide in fact one of their historical foundations. Combinations of quantifiers (and alterity expressions like other), another important source for the development of reciprocal markers, are shown to manifest considerable variation across languages, which is not compatible with the view that they can simply be subsumed under the category "anaphor." Our typological approach provides some new perspectives and perhaps even solutions for some traditional puzzles in the analysis of reciprocal constructions in German and Japanese.

1. Introduction

Even in reference grammars of major languages, reciprocal constructions are not given much space or attention. Nor did such constructions seem to have posed a major challenge for syntactic or semantic studies during the second half of the last century. It was only during the last ten years or so that they have found a more prominent place both in theoretical discussions (Heim et al. 1991; Dalrymple et al. 1998; Philip 2000; Plank forthcoming, etc.), in descriptive analyses of individual languages (cf. Frajzyngier 2000), and in crosslinguistic studies. Among the latter we find Kemmer (1993), Frajzyngier (2000), Siloni (2002), Maslova and Nedjalkov (2004), as well as a large survey initiated and coordinated by Nedjalkov (forthcoming). In spite of providing a wealth of interesting data, observations and analyses for individual languages, however, these crosslinguistic studies do not give us an integrated and reasonably adequate typology of reciprocal constructions. (2) There is still much work to be done before we can identify the major parameters and limits of variation in this domain, as well as the major implicational connections between variant properties. It is to this goal that our article is meant to contribute. We will examine the salient facts of German and Japanese against the background of what is known about reciprocal constructions in other languages and assess these language-specific observations as to their typological implications. (3)

2. Identifying reciprocal constructions

The core area of reciprocal constructions is easy to identify in terms of prototypical examples, but there are problems at the periphery, and the task of providing explicit criteria for their identification across languages and their delimitation from other constructions is certainly not a trivial one. In contrast to most other studies, we will therefore try to formulate such criteria. To begin with, we draw a strict distinction between SYMMETP, JC as a semantic property and RECIPROCAL as a syntactic property. The class of symmetric predicates and the set of reciprocal constructions can now be defined as follows:

(i) SYMMETRIC predicates are basic predicates with at least two argument (valency) positions which denote binary (or ternary) relations R among members of a set A with the following semantic property: [for all]x, y [member of] A (x [not equal to] y [right arrow] R(x, y)), that is, for specific substitutions of values a and b (a, b [member of] A) for the variables x and y: aRb [left and right arrow] bRa.(4)

(ii) RECIPROCAL constructions are grammatical means for the expression of symmetrical relations for any n-ary predicate and for at least one set of arguments A, with [absolute value of A] [greater than or equal to] 2; (5) it is a typical feature of such constructions that one of the arguments denotes a set A as specified above and that the basic argument structure of the relevant predicate is reduced or changed in such a way that not all argument positions are filled by referential expressions.

Let us add a few comments to these definitions. Only those predicates are characterized as symmetric which have the relevant bi-directional converse implication in their basic (transitive or ditransitive, etc.) use (cf. [1d]). According to this definition, not only the verbs in (1a) are symmetric, but also the adjectives and noun phrases in (1b) and the preposition in


(1) a. meet, marry, exchange, agree with, argue with, make love to, dance with, adjoin, fight with, date, resemble, join, compete with, speak with, separate y from z, etc.

b. be similar to, be different from, be parallel to, be analogous to, be equivalent to, be adjacent to, be engaged to ... be a relative of, be a friend of, (6) be the opposite of, be a partner of, be the mirror image of, be the counterpart of, etc.

c. be with

d. John married Annie. [right arrow] Annie married John.

In addition to the examples mentioned, we might also want to regard the verbs in (2) as instances of symmetric predicates since they are prototypically and frequently used to describe symmetric situations. Note, however, that the basic use of such predicates may denote a certain asymmetry of power, control, initiative, or involvement:

(2) embrace, divorce, greet, hug, kiss, split up with, share y with z, collide with, etc.

The distinction between such semantic and pragmatic symmetry is not an easy one to draw, however, since even the clear cases of semantic symmetry may express an asymmetry of control, initiative, or perspective in their basic use (cf. [3]). Perfect symmetry is only expressed by reciprocal constructions and this symmetry is often iconically signalled by the fact that the relevant semantic arguments are encoded by the same grammatical relation (cf. [4]): (7)

(3) a. Nowadays women are dating men rather than the other way round.

b. John met Mary at the station.

c. They swapped one car for the other.

(4) a. They are dating.

b. John and Mary met at the station.

c. They swapped cars.

It follows from our definition that the examples in (5) are instances of reciprocal constructions whereas those in (6) are not, since the symmetry is either spelled out in a sequence of clauses or expressed by lexical means: (8)

(5) a. All tenants know each other.

b. Frankie and Johnny were lovers.

c. Fred and Mary met at the station.

(6) a. John loves Mary and she loves him, too.

b. John tends to avoid Bill and vice versa.

c. Madonna loves the British and they love her back.

In defining reciprocal constructions with the help of the more basic symmetric predicates, our typology makes certain predictions: We assume that (a) such basic symmetric predicates can be found in all languages of the world, (b) that they will receive parsimonious coding when they are used in reciprocal constructions, and (c) that they will play an important role in the historical development of reciprocal morphology. All of these predictions of course need to be tested in a wider variety of languages. As far as the first of these claims is concerned, our evidence is scanty so far, but we noted with great interest that even languages in which reciprocal verbs are derived from transitive ones through the addition of a suffix in a less than fully productive derivational process, such as Turkish, still have basic symmetric predicates.

Our second claim can be illustrated by the symmetric predicates of English listed above. With plural subjects these predicates invariably allow a symmetric interpretation even if they are not followed by the reciprocal "anaphor" each other or one another. The following example (7a) could be the plot of an unhappy love story and (7b) is an analogous example of adjectival predicates: (9)

(7) a. They met/dated/danced/embraced/kissed/agreed/married/ quarrelled/split up/divorced ...

b. John and Peter are similar/different/completely unlike/ acquainted/very close ...

Note furthermore that it is the symmetric preposition with that tends to be omissible in connection with reciprocal constructions of certain verbs. Moreover, comitative prepositions like with are frequently used to derive symmetric nouns from basic nouns (cf. German Mitarbeiter 'collaborator', Mitstreiter 'brother-in-arms', Mitreisender 'fellow passenger', etc.). In Japanese the comitative marker -to also plays an important role in reciprocal sentences. Finally, as far as our third prediction is concerned, we will briefly mention at this point that reciprocal verbs may be derived in Japanese with the help of the symmetric predicate au 'meet, fit' and that -banji/-wanji 'exchange' is used for the same purpose in Nyulnyualan (cf. McGregor 2000). In Mandarin the pair of directional opposites 'come' and 'go', which create a symmetric effect in combination, are used in a productive process of deriving reciprocal verbs (Liu 2000):

(8) Tamen da-lai-da-qu. 3PL beat-come-beat-go 'They beat each other.'

In delimiting our field of inquire and in defining the third of comparison for our crosslinguistic study in terms of a semantic property (SYMMETRY, AUTO-CONVERSENESS), we are only looking at the core area of reciprocal constructions and are probably excluding a periphery with specific syntactic and semantic properties. Still, such an approach seems to us preferable to a definition in terms of a prototype or in terms of a reciprocal expression in English (each other).

3. Types of reciprocal constructions

Since there is no generally accepted, reasonably comprehensive typology of reciprocal constructions available, we will draw a preliminary distinction between different types in analogy to the well-known typology formulated by Faltz (1985) for reflexives:


As is shown in Figure 1, Faltz draws a first distinction between verbal and nominal strategies. The nominal strategies are further divided into a pronominal one and genuine nominal ones, which are further subdivided into head reflexives and adjunct reflexives. Since the latter two are based on intensifiers like English -self, they are collapsed into one type above.

Analogously, a typology of reciprocal constructions can be developed by first of all distinguishing verbal strategies from nominal ones. There are, however, good reasons for drawing an additional distinction by opposing a synthetic strategy, that is, the use of a prefix or suffix as reciprocal marker, to a compound strategy, which is characterized by the development of verbs to markers of reciprocity. Nominal strategies can be further subdivided into a pronominal strategy, that is, the use of pronouns or referentially dependent nouns as markers of reciprocity and a quantificational one, namely, the use of bi-partite quantifiers as a kind of reciprocal anaphor. Before these strategies are discussed one by one, an overview and an example of each type will be given:


(i) The synthetic strategy (Swahili, Ashton 1956)

(9) a. Ali na Fatuma wa-na- pend-an -a.

Ali and Fatuma 3.PL-PRES-love-RECIP-final vowel

'Ali and Fatuma like each other.'

(ii) The compound strategy (Mandarin, cf. [8])

(9) b. Tamen da-lai-da-qu.

3PL beat-come-beat-go

'They beat each other.'

(iii) The pronominal strategy (German)

(9) c. Seit-dem meiden sich die beiden since then avoid REFL/RECIP DEF.ART.PL two Professoren.


'The two professors have avoided each other ever since.'

(iv) The quantificational strategy (English)

(9) d. John and Peter hate each other.

More than one strategy may be available in a specific language and--as is shown in Haiman (1983) and Kemmer (1993)--the choice between them is far from being arbitrary. In some languages (e.g. Romance, Tuvaluan) two strategies can be combined. In the following sections we would like to discuss these four strategies one by one. Moreover, we will also discuss the question whether there are any implicational connections or even correlations between the choice of one of these strategies and other variant properties found in the relevant language.

In addition to these distinctions, others can be drawn and are sometimes drawn, based on questions such as the following: (i) Is the same marker used for both reflexivity and reciprocity (cf. Maslova and Nedjalkov 2004)? (ii) How many strategies are available in a language (Kemmer 1993)? (iii) Is the formation of reciprocal verbs based on a morphological or on a syntactic strategy (Siloni 2002)?

3.1. The synthetic strategy

Languages using this strategy have derivational or inflectional affixes as reciprocal markers. Moreover, clitics like Russian -sja, which follows tense and person markers, and Swedish -s are also assigned to this type. Reciprocal verbs are derived from more basic verbal roots by the addition of such a reciprocal affix. In some languages the relevant affixes are also used for the encoding of reflexivity. The set of verbs derivable in this way very often includes as a core class the set of symmetric predicates ('marry', 'meet', 'resemble', 'adjoin', 'argue with', 'fight with', 'compete with', 'collide with', 'share', etc.) as well as the verbs stereotypically used to express symmetric situations ('kiss', 'embrace', 'play with', etc.). Moreover, this strategy of deriving reciprocal verbs exhibits varying degrees of productivity across languages.

Synthetic strategies for the formation of reciprocal verbs are found in Amharic, Bantu, Evenki, Finnish, Hebrew, Hungarian, Manam, Meso-American languages, Russian, (10) Somali, Swedish, Turkic, Tuvaluan, and Yimas, to mention just a few examples from different language families. The productivity of these synthetic strategies may differ quite strikingly between languages. In Finnish there are only about a dozen reciprocal verbs with the affix -ele- or -ile-, many of which cannot even be paired with a basic verb of the appropriate meaning:


(10) rakastella 'make love' (<rakastaa 'love'); kiistella 'argue' (< kiistaa 'dispute, deny, challenge'); taistella 'fight'; tapella 'fight'; keskustella 'discuss' (no basic verb);

In Turkish, the number of such derived reciprocal verbs (the relevant affix being -(i)s-, -(i)s-, -(u)s-) is much higher and includes also clear cases of nonsymmetric predicates as derivational bases:


(11) anlas 'understand each other' (< anla- 'understand'); dovus 'fight each other' (< dov- 'beat'); sevis 'to love one another' (< sev-love'); benzes 'resemble one another' (< benze- 'resemble'); vurus-mak 'beat each other' (vurmak- 'beat')

In contrast to both of these languages, the derivational processes through which reciprocal verbs can be derived seem to be fully productive in the Papuan language Yimas (Foley 1991: 284), in some Turkic languages (cf. Nedjalkov 2002), in many Pama-Nyungan languages (Dixon 1980) and in Swahili, as well as in other Bantu languages (Schadeberg 1992: 9).

Reciprocal affixes typically, and perhaps invariably, have other possible interpretations in addition to the reciprocal one. Among the most frequent additional interpretations we find the reflexive ('to V oneself'), the sociative or collective (to do something together), the frequentative/ repetitive (e.g. 'to run' vs. 'to run around'), the competitive meaning ('they competed in doing'), as well as a few other meanings (change of body position, anticausative, modal passive, etc.) that are typically associated with middle markers (cf. Nedjalkov forthcoming). Thus, if we were to postulate a general, vague meaning rather than polysemy for these affixes we could roughly describe it as "combined or repeated action by a plurality of actors, affecting a plurality of entities" (cf. Kemmer 1993; Lichtenberk 2000). The frequent identity between reflexive and reciprocal constructions is motivated by the fact that in both the same entities occur in different roles, namely, both as agents and as patients.

3.2. The compound strategy

The compound strategy resembles the synthetic one in so far as reciprocity is encoded by a verb, but in contrast to the latter, it is the verb itself rather than a verbal affix that signals the autoconverse or symmetric relationship. In addition to Mandarin and Nyulnyalan, we can also analyze Basque as using such a strategy. The reciprocal marker elkar in Basque can be assumed to derive from the verb elkartu 'meet, come together, put together'. (11)

Basque (Amaia Zaballa-Zarosa pers. comm.):

(12) Soldadu-ek elkar -?? hil zuten.

soldiers-ERG.PL meet/each.other-ABS kill AUX.3PL.PAST

'The soldiers killed each other.'

In Tuvaluan the verb fakatau 'compete, exchange' is used similarly (Besnier 2000) and in other Oceanic languages we find expressions with the basic meaning 'return', 'again'. To this category we can also assign all the serial verb constructions and reduplications frequently used to express reciprocity.

3.3. The pronominal strategy

The reciprocal markers used in the third strategy are free forms or clitics that are typically also used as reflexive markers. In other words, the relevant class of expressions includes the "reflexive pronouns" (pure pronominal reflexives) of traditional grammar, whose meaning and use typically also cover the so-called middle domain (cf. Kemmer 1993). As is well known, this strategy is found in nearly all languages of the European continent (Germanic other than English, Romance, West Slavic languages, Breton, etc.), but also in languages like Somali (cf. Saeed 1999: 78 ft.). Disambiguation between the reflexive and the reciprocal interpretation of a sentence is typically achieved by number marking. Sentences with plural subjects tend to have a reciprocal interpretation in contrast to those with a singular subject, which are only compatible with a reflexive interpretation: (12)

Somali (Saeed 1999: 78):

(13) a. Way (waa + ay) is dishay.

DECL + she REFL killed

'She killed herself.'

b. Way (waa + ay) is arkeen. (13)

DECL + they RECIP saw

'They saw each other/They saw themselves/She saw herself.'

Being syntactic in nature, this strategy is generally not restricted in its range of applicability ('productivity'), although it may compete with and thus be delimited by another strategy also available in the relevant language. In those languages (Romance, South Slavic) where the relevant pronouns are clitics, their morphological status is in between that of a free form and that of an affix.


(14) Paul et sa femme ne s'entendent plus du tout.

'Paul and his wife don't get along anymore.'

The terms "nominal" and "pronominal" strategy imply that the verbs in the relevant constructions are transitive ones and that the reciprocal markers are referentially dependent anaphors. Siloni (2002) rejects this assumption for French with convincing arguments, which cannot easily be applied to languages like German, however. The obvious alternative to distinguishing a pronominal strategy from the other types would be to group the cases discussed so far together with the synthetic strategy. We will return to this problem below and simply acknowledge here that there is a problem yet to be solved within a more comprehensive typological study.

The label "pronominal strategy" is probably less problematic for structures in Chadic languages (Frajzyngier 2000:186 ft.), where reciprocal markers just like reflexive markers (cf. Heine 2000) derive from expressions for body parts or, more typically, from the expression for 'body'. In view of the nondistinctness of reflexive and reciprocal markers found in many languages in different manifestations, it should not come as a surprise that the same expression ('body') is the source for both uses. Much more interesting are those cases where different sources underlie reflexive and reciprocal markers, respectively. In these cases it is invariably the more general expression 'body' that gives rise to reciprocal markers, whereas the more salient body part ('head') leads to reflexivity.

Analogously, those structures where the reciprocal marker derives from a symmetric nominal like 'comrade', 'fellow', 'mate', 'companion', 'friend' can also be assigned to the pronominal subtype. Welsh provides a well-known example (ei gilydd 'his fellow' for third person subjects) and Heine and Mayashita (forthcoming) show that this type is very wide-spread in African languages.

3.4. The quantificational strategy

The reciprocal markers involved in the fourth strategy differ strikingly from the ones used in the other three in having more semantic substance and in manifesting an NP-like behavior. The relevant expressions (English each other, one another; Russian drug druga; Punjabi ikk duuje; Dutch elkaar; Turkish birbiri; etc.) are generally analyzed as anaphors in generative studies, that is, as variables that have to be bound in their governing category. The verbs combining with such expressions are clearly transitive or ditransitive:

(15) Paul and Mary do not respect one another.


(16) a. Oni casto vid'at drug drug-a.

they often see.3PL one another-GEN

'They often see each other.'


(16) b. Ci ha presentati gli un-i a-gli

we.DAT he.has introduced the one-PL.MASC to-the altr-i.


'He introduced us to each other.'

As is shown by the preceding two examples, reciprocal anaphors inflect for categories such as person, gender, number, and case in some languages, but are invariant in others.

Historically, such reciprocal markers derive from bipartite quantifiers or demonstratives. More often than not, the first constituent of these expressions is a quantifier and the second an "alterity" expression with the basic meaning 'second, other'. (14) Combinations of two alterity expressions are also possible, however:


(17) a. Matti ja Liisa pita-vat toinen toise-sta-an.

Matti and Liisa like-3PL.PRES other other-ELA-3POSS

'Matti and Liisa like each other.'

Starting out as such combinations of two quantifiers, the relevant expressions are typically subjected to processes of fusion and univerbation (cf. Dutch elkaar/mekaar) or reduction, that is, processes resulting in a simplex, noncomposite expression. The reduction of a basically bipartite expression to a simplex one can be illustrated with an example from Finnish. The content of (17a) can also be expressed by a simpler structure, which is generally regarded as being more grammaticalized and more common. Note that the simplex reciprocal anaphor is marked for the plural in Finnish, in contrast to the bipartite one:

(17) b. Matti ja Liisa pita-vat tois -i -sta -an.

Matti and Liisa like -3PL other-PL-ELA-3POSS

'Matti and Liisa like each other.'

3.5. A hierarchy

If these four strategies of forming reciprocal constructions are compared, some interesting observations and generalizations emerge. Note, first of all, that many languages have more than one strategy. Both the synthetic strategy and quantificational strategy are found in Russian, Turkish, Hungarian, Finnish, and Greek, both the pronominal strategy and the quantificational strategy occur in German, (15) French, and Italian, both the compound strategy and the quantificational strategy are found in Japanese, whereas only the quantificational strategy is found in Modem English and as many as three strategies have been reported for Tuvaluvan (Besnier 2000:215 f.).

Moreover, additional comparative observations can be made if the four strategies are ordered on a scale as follows:

Reciprocal constructions:

(18) synthetic (I) < compound (II) < pronominal (III) < quantificational (IV)

(a) If we move on the scale from right to left we find an increase in bonding and a loss of independence. (16) Analogously, we find an increase in morphological substance and complexity as we move from left to right.

(b) Inherently symmetric predicates and those denoting stereotypically symmetric situations tend to be used with strategy I or II, rather than with the more emphatic and less economical strategies III and IV. If neither strategy I nor II is available, a zero strategy may be used (cf. [7]), but, as pointed out above, this may be a matter of pragmatics rather than grammar. (c) In moving from left to right we find fewer combinatorial restrictions as far as verbs and syntactic environments are concerned. (d) Finally, the range of possible meanings seems to increase as we move from right to left on the hierarchy. The reciprocal anaphors or quantifiers seem to have no other use in many languages, whereas polysemy is the standard situation for reciprocal affixes and reciprocal pronouns. The most wide-spread of these recurrent meanings were listed above. (17)

These comparative observations about the four strategies of encoding reciprocity and the hierarchy derived from them are, of course, hypotheses to be tested against a richer body of data, rather than well-established facts. Such hypotheses are an essential and valuable part of any attempt to formulate a typology, as long as they meet the requirement of being falsifiable by additional data. Even though a detailed demonstration of the facts and observations underlying our hierarchy cannot be given at this point, (18) we would like to cite at least some corroborating evidence. The observation made in (a) hardly needs any supporting evidence: Reciprocal anaphors are free forms and affixes are employed in strategy I. Among the pronominal elements of strategy III we find both strong pronouns, that is, independent, focusable elements as Polish siebie and the clitic pronouns of the Romance and Slavic languages such as French se or Polish sic, which invariably lean on a following or, in some cases, on a preceding verb. Our second observation (b) simply repeats an observation made for a wide variety of languages. Among the reciprocal verbs derived from basic stems with the help of the very unproductive affixes -(V)kVd(ik)-/-(VkV)z(ik)- in Hungarian, for example (cf. Ratkosi 2002), we only find a few symmetric predicates ('kiss', 'meet', 'quarrel', 'correspond', 'converse', 'compete', 'make love to') and analogous data were given above for Finnish and Turkish. The third observation (c) relates to the fact that the synthetic strategy may be of very limited productivity, whereas the fourth strategy can even be used with the symmetric predicates, albeit with a specific semantic effect. As was pointed out in Kemmer (1993: 111), the following two sentences are not completely equivalent. Sentence (19a) suggests a single unitary action, whereas (19b) suggests a sequence of actions:

(19) a. They kissed.

b. They kissed each other.

Our remark about productivity has to be taken with a great deal of caution, based as it is on a very restricted sample of languages. As already noted above, the synthetic strategy seems to be quite productive in Austronesian, Papuan, and Bantu languages, whereas the quantificational strategy is used very parsimoniously and for special semantic effect in these languages. (19)

Moreover, all reciprocal markers that are licensed in genitive (possessive) phrases are instances of strategy IV. And those languages whose exponents of the pronominal strategy are clitics have to resort to the quantificational strategy (e.g. Romance), or to strong pronouns (e.g. Polish (20)) with prepositional phrases:

(20) They admire each other's house.


(21) a. Ils sont tres tiers l'un de l'autre.

they are very proud one of the other 'They are very proud of each other.'

b. Ils se sont marirs.

they RECIP/REFL are married 'They got married.'

The wider range of possible interpretations for the shorter and more economical reciprocal markers is simply a specific manifestation of Zipf's Economy Principles (Zipf 1949). It is interesting to note in this context that the anaphors of strategy IV only have one single interpretation, whereas the exponents of the three other strategies may have several.

So far, very little has been reported on implicational connections between variant properties of reciprocal constructions. Some of the most detailed and fine-grained observations in this domain are found in Siloni (2002), but the relevant connections are only demonstrated and discussed for six languages. Her observations are thus more interesting suggestions for further study than generalizations based on a representative sample of facts.

It is against the background of these preliminary typological findings that we will take a closer look at reciprocal constructions in German and Japanese.

4. Reciprocal constructions in German

4.1. Reciprocal expressions

The most important grammatical devices for the expression of reciprocity in German are the so-called "reflexive pronoun" sich and the reciprocal "anaphor" einander. As is well known, the distinction between reflexive pronouns and personal pronouns is only drawn in German (as in most Germanic and Romance languages) for the third person. Sich can be used both in the position of a direct (accusative) object and in that of a dative object, though not as a genitive object. This expression is invariant, that is, it does not inflect for the so-called [phi]-features, and only permits a reciprocal interpretation with plural subjects. (21)

(22) a. Die beiden Professoren meiden sich.

the two professors avoid REFL/RECIP 'The two professors avoid each other.'

b. Wir sollten uns ofters sehen.

we should us often.COMPAR see

'We should see each other more often.'

In combination with predicates that denote typically other-directed situations, this reciprocal interpretation is the preferred option, whereas a stereotypically self-directed predicate such as verteidigen 'defend', schutzen 'protect', or vorbereiten 'prepare' makes a reflexive distributive interpretation of sich more probable even with plural subjects (cf. Konig and Vezzosi 2004). Such cases of default interpretation or ambiguity may be overridden or disambiguated by adding one of the three adverbs gegenseitig, wechselseitig 'mutually', or untereinander 'among themselves'. (22)

(23) a. Die beiden Studenten haben sich nut verteidigt.

the two students have REFL/RECIP only defended 'The two students only defended themselves.'

b. Die beiden Studenten haben sich gegenseitig the two students have REFL/RECIP mutually verteidigt.


'The two students defended each other.'

In terms of the typology formulated above, German reciprocals with sich can be regarded as an instance of the pronominal strategy. What is not perfectly clear, however, is whether sich can really be analyzed as a purely grammatical marker and whether the verb it follows is thereby changed into an intransitive one, as argued by Siloni (2002). Neither of the arguments that can be adduced for the intransitive character of French reciprocal verbs with the reflexive clitic se (23) carries over to German. Moreover, the fact that reciprocal sich can be coordinated with NPs argues for an analysis as referentially dependent anaphor:

(24) Die beiden Angeklagten beschuldigten sich gegenseitig und dazu noch ihre Nachbarn.

'The two defendants accused each other as well as their neighbors.'

As is shown by the preceding example, however, such a coordination requires the presence of the reciprocal adverb gegenseitig, which is also required in those cases where the reciprocal marker carries a contrastive focal stress:

(25) Sich gegenseitig haben sie beschuldigt, nicht ihre Nachbarn. RECIP mutually have they accused not their neighbors 'They accused each other, rather than their neighbors.'

On the other hand, the fact that that a focal stress on sich clearly excludes a reciprocal interpretation argues for Siloni's (2002) view that the verbs in sentences like (22)-(23) have lost their second argument position and that sich is a purely grammatical marker.

The reciprocal marker einander is a clear instance of the quantificational strategy (IV) distinguished above from three others. The internal make-up of this expression is still sufficiently transparent to show that it is the result of combining an existential quantifier (ein-) with an alterity expression (ander-). A plausible reconstruction of this process of fusion and univerbation is presented in Plank (forthcoming). The starting points of this process can still be reconstructed by looking at those contexts which apparently prevented such a development, that is, the fusion of the quantifier with a negation, the syntactic position as genitive (possessive) modifier, and the occurrence of two prepositions:

(26) a. Kein Mensch kennt den andern. (H. Hesse)

'No human being knows the other.'

b. Ein jeder trage des Anderen Last. (Bible)

'Everybody may carry the burden of the other.'

c. Karl ubersetzt Texte von einer Sprache in die andere.

'Charles translates texts from one language into the other.'

The most remarkable aspect of this process of grammaticalization was that it occurred preferably and perhaps exclusively around an intervening preposition (cf. Plank forthcoming). In prepositional phrases einander is the only possible marker of reciprocity and thus contrasts with sich, which can only have a reflexive interpretation in such contexts:

(27) a. Die Spieler reden nicht mehr fiber sich.

'The players do not talk about themselves any longer.'

b. Die Spieler reden nicht mehr ubereinander.

'The players don't talk about each other any longer.'

If einander developed historically from a bi-partite quantificational expression, which is not implausible, the first part ein- must have been attracted by the second part and floated into a position inside the PP. Plank (forthcoming) reconstructs this development as the result of three stages of quantifier floating, which can roughly be illustrated by the following English examples: (24)

(28) a. One/each ... earl fought with (the/an) other.

b. The earls one/each fought ... with (the/an) other.

c. The earls fought one/each ... with (the/an) other.

d. The earls fought with each other/one another.

Our prediction that symmetric predicates as well as those denoting stereotypically symmetric situations will receive parsimonious coding when they are used in reciprocal constructions is partly confirmed by the relevant data from German. Such verbs do not always require reciprocal markers, but may occur without einander or sich, provided they have a plural subject. In contrast to English, however, only a few transitive verbs allow such an empty object position. Heiraten 'marry' is a case in point. The other two examples given in (29a) are remarkable in so far as they may drop a prepositional phrase with mit and a nonreferential sich in the reciprocal construction as opposed to the basic use with singular subjects (29b):

(29) a. Sie heirateten. Sie tanzten (miteinander). Sie stritten (miteinander).

'They married. They danced. They quarrelled.'

b. Er heiratete sie. Er tanzte mit ihr. Er stritt sich mit mir.

'He married her. He danced with her. He argued with me.'

There are, however, a wide variety of two-place predicates whose second argument is introduced by the comitative preposition mit 'with'. The so-called "reflexive pronoun" sich that accompanies these verbs is an instance of the purely grammatical marker that accompanies inherently reflexive verbs. When these verbs combine with a plural subject the relevant prepositional phrase (miteinander 'with each other') is generally omissible:

(30) a. Sic unterhielten/zankten/versohnten/einigten/verstanden ... sich (miteinander).

'They conversed/quibbled/made up/agreed/got along ...'

It is also within this group of symmetric predicates that we find the possibility of "discontinuous reciprocals" (cf. Dimitriadis 2002). Strictly speaking, this term should not be used for symmetric predicates such as the ones in (30), which are subeategorized for a prepositional phrase introduced by the comitative preposition mit in addition to a subject:

(30) b. Karl unterhielt/einigte sich mit Mafia.

'Charles chatted/came to an agreement with Maria.'

Examples such as (30a) are simply instances of symmetric predicates, which do not lose that interpretation, of course, if they occur with plural subjects. The only remarkable cases are those that are basically subcategorized for direct objects such as duzen 'tutoyer, address informally', bekampfen 'fight', beraten 'advise', umarmen 'hug', treffen 'meet', kassen 'kiss', etc. Even though judgments vary, most speakers of German can also use these verbs with singular subjects and prepositional complements introduced by mit:

(31) a. Karl berat/duzt seinen Chef.

'Charles advises his boss.'

b. Karl berat/duzt sich mit seinem Chef.

'Charles consults with his boss.'

4.2. The distribution of sich vs. einander

One of the most remarkable facts about reciprocity in German is the distribution of the two reciprocal markers sich and einander. There is a great potential for overlap, but such extensive overlap is not found in corpora of authentic data and these two expressions overlap only partially in their distribution, especially in quantitative terms. As was already mentioned, sich is excluded as a genitive complement and requires a subject as antecedent (cf. [25]). Accordingly, einander is the only possible option after verbs which govern the genitive, like bedurfen, as well as in cases where the antecedent is an object:

(32) Die Menschen in New York bedurfen einander.

'People in New York need each other.'

(33) Die Gastgeberin stellte die Gaste einander vor.

'The hostess introduced the guests to each other.'

These restrictions are not particularly remarkable, given that sich is also excluded as genitive complement in its reflexive interpretation and that pure pronominal reflexives are usually subject-oriented. What is really remarkable is the division of labor between sich and einander in object position (dative and accusative) and in prepositional phrases. As noted in many studies and as already demonstrated in connection with (27), the only possible interpretation of sich in a prepositional phrase is a reflexive one, so that sich and einander are in clear opposition in such contexts. A further example of this semantic contrast is provided by (34):

(34) a. Die Pfadfinder passten gut auf sich auf.

'The boy scouts took good care of themselves.'

b. Die Pfadfinder passten gut aufeinander auf.

'The boy scouts took good care of each other.'

In the position of direct and indirect object einander would certainly not be excluded in acceptability judgments by native speakers. The interesting thing is that it rarely occurs in these functions in actual performance. Plank (forthcoming) totally excludes einander in object positions for Bavarian, a southern dialect of German, and also considers its occurrence in those positions extremely rare, if not nonexistent, in Standard German. Our survey of current usage based on authentic data from the Mannheim Corpus COSMAS, which comprises some 1600 million word forms, gives a somewhat modified picture, but confirms the basic intuitions formulated by Plank. Einander is rare outside of prepositional phrases, especially in conversational data. But it does occur as dative or accusative object in the written language. There is a clear tendency for einander to occur as object in those cases where the preferred interpretation of sich in the same position would be a reflexive one. Such an explanation is not possible for all cases, however. Moreover, there are a few verbs which only combine with einander and totally exclude sich as a possible object. Interestingly, these are verbs like jagen and folgen which encapsulate the same semantic content as the preposition nach:

(35) a. Grosse Intervallsprunge statt kleiner Schritte folgen einander/ *sich.

'Big intervals rather than small steps follow one another.'

b. Die Termine jagen einander/?sich.

'Appointments follow each other very closely.' (25)

5. Reciprocal constructions in Japanese

5.1. Derivational strategies

As in the other languages discussed above, there is also a small set of symmetric predicates in Japanese and these exhibit special properties in reciprocal constructions. A representative subset of these verbs is listed in (36):

(36) au 'meet, fit, get along with', tatakau 'fight', kekkonsuru 'marry', rikonsuru 'divorce', surechigau 'pass, cross paths with', kaiwasuru 'converse', odoru 'dance (with)' (26)

Among these expressions, the verb au, which belongs to the basic vocabulary of Japanese and may thus cover a wide range of meanings, plays a very special role: au combines with other verbs to form compound verbs with a reciprocal use and interpretation. The following examples (37)-(38) illustrate the basic use of au as a symmetric predicate:


(37) a. Taroo-ga Hanako-ni/to machi-de au.

Taroo-NOM Hanako-DAT/COM street-LOC meet

'Taroo meets Hanako on the street.'

b. Kono sen-wa bin-ni awa-nai.

this cork-TOP bottle-DAT fit-NEG

'This cork does not fit the bottle.'

(38) a. Tarooto Hanako-ga machi-de au.

Taroo and Hanako-NOM street -LOC meet

'Taroo and Hanako meet on the street.'

b. Kono sen to bin -wa awa-nai.

this cork and bottle-TOP fit-NEG

'This cork and the bottle do not fit.'

In its basic use the verb au takes two arguments, one of which is marked either by the dative or by the comitative case. In reciprocal constructions there is only one conjoined or plural argument which is marked either by the nominative affix ga or the topic marker wa. The remarkable fact about the verb au, however, is that it can be combined with other verbs to form reciprocal verbal compounds. Used as a prefix, this verb takes the phonological form ai- and combines with a number of basic verbs to form symmetric predicates like the following, as a result of a compounding process that is no longer productive:

(39) ai-taisuru 'face/confront each other', ai-kotonaru 'differ from each other', ai-hitoshii 'be equal to', ai-irenai 'be mutually exclusive', ai-zengosuru 'occur at the same time', etc.

In contrast to the preceding marginal process of deriving symmetric predicates and reciprocal verbs, the process in which au is used as a suffix (or final component) is very productive. Because of its high productivity most verbs derivable by this process are not listed in the lexicon. The following examples illustrate the derivation and use of these verbs:

(40) tasukeru 'help' > tasuke-au 'help each other' (with accusative)

a. Hanako-ga Taro-o tasuke-ta.

Hanako-NOM Taro-ACC help-PAST

'Hanako helped Taro.'

b. futari-wa tasuke-at-ta.

two.CLASS(person)-TOP help-RECIP-PAST

'These two helped each other.'

(41) izonsuru 'depend on' > izonshi-au 'depend on each other' (with dative)

a. Taro-wa Hanako-ni izonshi-tei-ta.

Taro-TOP Hanako-DAT depend.on-PROG-PAST

'Taro was dependent on Hanako.'

b. futari-wa izonshi-at-tei-ta.

two.CLASS(person)-TOP depend.on-RECIP-PROG-PAST

'These two were dependent on each other.'

The preceding examples show that reciprocal constructions wit au do not only require a plural subject but also a gap in the relevant sentence. This gap indicates that the entities denoted by the subject participate in the relevant situation not only as agents, but also as patients or recipients, etc. The gap can be filled by a reciprocal anaphor without changing the reciprocal reading, as will be shown below (cf. [48]-[49]). (27) If it is filled by another argument, the resultant sentence can only have a sociative or frequentative interpretation (cf. Nishigauchi 1992: 161).

So far, the derivational process involving the verb au has been characterized as a process of compounding. This characterization is well in line with the fact that all lexical elements derived by adding au are verbs. Given that compounds in Japanese are head-final, this is exactly the result one would predict. The preceding examples as well as the following show, however, that au has lost its basic meaning 'meet, fit' in these compounds so that its contribution to the meaning of the complex expression can simply be characterized as 'reciprocity'. The only other possible interpretation is a frequentative one, as is shown by the example (42). (28) In other words, we find the type of polysemy that is typical of reciprocal expressions:

(42) renraku-o toru 'make contact' > renraku-o tori-au

a. futari-wa renraku-o toil, at-ta.

two.CLASS(person)-TOP contact-ACC take meet-PAST

'These two (once) made contact and met.'

b. futari-wa renraku-o tori-at-ta.

two.CLASS(person)-TOP contact-ACC take-RECIP-PAST

'Both of them made (several) attempts to establish contact.' or 'These two kept in touch (with each other).'

In (42a) both toru 'take' and au 'meet' are main verbs and have their standard interpretations. In (42b), by contrast, tori-au is a complex verb with the meaning of the basic verb toru, to which au adds the reciprocal or frequentative component. In addition to the reciprocal and frequentative interpretation derived verbs with au may also have a sociative interpretation (cf. [43]). This fusion of au and other basic verbs and the concomitant semantic change show that the symmetric predicate au 'meet, fit' has developed into a kind of derivational atfix or reciprocal marker as a result of grammaticalization.

(43) yorokobu 'rejoice' > yorokobi-au 'rejoice together'

kokumin-wa minna-de kootaishi -no tanjoo-o

people -TOP all -LOC crown.prince-GEN birth -ACC yorokobi-at-ta.


'The whole nation rejoiced together at the birth of the crown prince.'

In addition to the symmetric predicate au two more lexical items are used for the derivation of symmetric or reciprocal expressions. These expressions, however, combine with nominal rather than verbal predicates. Used as a noun dooshi means 'brothers-in-arms'. But this use is archaic, in contrast to the use of this expression as suffix, where it denotes a symmetric relationship:

(44) Maya to Mai-wa tomodachi-dooshi-da.

Maya and Mai-TOP friend -mutual-PREDICATE

'Maya and Mai are (mutual) friends.'

That dooshi has lost its original meaning is particularly apparent in combinations like teki-dooshi 'mutual enemies'.

Apart from being used as an anaphor in written, formal Japanese soogo (29) can also be used as an affix, much like dooshi. Its most important use, however, is that of a prefix in the derivation of reciprocal verbal nouns. Being of Chinese origin, soogo does not combine with verbal nouns of Japanese origin. Combined with verbal nouns of the noninherited lexical stock, soogo forms reciprocal verbal nouns in a semiproductive process. In their English translation these formations correspond to a verbal noun and the reciprocal adjective 'mutual':

(45) a. fujo 'help' > soogo-fujo 'reciprocal help'

b. shien 'support' > soogo-shien 'mutual support'

c. kooryuu 'exchange' > soogo-kooryuu 'mutual exchange'

d. rikai 'understanding' > soogo-rikai 'mutual understanding'; soogo-rikai-o suru (or: soogo-rikai-suru) 'to show mutual understanding'

e. hyooka 'estimation' > soogo-hyooka 'reciprocal estimation': soogo-hyooka-o suru 'to make a mutual assessment'

f. syoonin 'approval' > soogo-syoonin 'mutual approval'

In combination with the verb suru 'do' (light verb) these verbal nouns can be used as complex predicates. Finally, the comitative marker -to also plays a special role in the encoding of symmetric relations. The use of this case marker instead of the case usually governed by the relevant verb changes an asymmetric situation to a symmetric one:

(46) dareka-ni koi-o suru-no-wa yasashii-ga, someone-DAT love-ACC do-COMP-TOP be.easy-but dareka-to koi-o suru-no-wa muzukashii.

someone-COM love-ACC do-COMP-TOP be.difficult 'To fall in love with someone is easy, but to find someone who loves you too is difficult.'

In the preceding example, the replacement of the dative case marker by the comitative one changes the meaning from a one-sided romantic involvement to a mutual one. The only verbs permissible in this construction are symmetric predicates and verbs denoting prototypically symmetric situations. With the former, group structures with singular subjects and comitative complements are equivalent to the more common structures with plural or coordinate subjects:

(47) Taroo-ga Hanako-to kenkasuru. = Taroo to

Taroo-NOM Hanako-COM quarrel Taroo and Hanako-ga kenkasuru.

Hanako-NOM quarrel

'Taroo and Hanako quarrel.'

5.2. Reciprocal "anaphors"

The quantificational strategy discussed above is also clearly identifiable in Japanese. Its formal exponent is (o)tagai, (30) an expression which may combine with any case marker and is therefore clearly a noun:

(48) futari-wa otagai-o nonoshiri-at-ta.

two: CLASS(person)-TOP each.other-ACC curse-RECIP-PAST

'These two curse each other.'

(49) kongo-wa otagai-ga tasuke-ao-o. each.other-NOM help-RECIP-ADH

'Let us help each other in the future.'

The etymology of (o)tagai is no longer transparent, but it is quite plausible to assume that this expression is related to the verb tagau/tagawa- 'to be contrary to/different from', which still occurs in fixed expressions like the following:

(50) kitai-ni/yosou-ni tagawa-zu


'As was expected ...'

Moreover, it is reasonably well established that the verb tagau has the same origin as the symmetric verb chigau (< tigafu) 'to differ from', which is still used today. (31) The assumption that (o)tagai is a nominalization of the verb tagau with the basic meaning 'the other' is furthermore 2supported by what we know about the historical development of reciprocal anaphors in other languages. As was pointed out above, such alterity expressions frequently develop into reciprocal markers when they occur twice and are thus found as an alternative to the combination quantifier + alterity expression ('each/the one ... the other'). And such combinations of two occurrences of otagai do indeed occur:

(51) tagai-ga tagai -o tasuke-au, sore-ga

each.other-NOM each.other-ACC help-RECIP such-NOM



'Helping each other is good.'

In Modern Japanese such double occurrences of otagai only occur if there is no other constituent filling the relevant subject or object position. (32) Just as in English or Dutch, but in contrast to German, otagai may also be used as a possessive (genitive) modifier:

(52) bokutachi-wa tagai-no ketten-o yoku shit-teiru.

1PL-TOP each.other-GEN fault-ACC well know-PROG

'We know each others' faults very well.'

Otagai does not have to occur in the same clause as its antecedent, but can also be in a lower clause. In other words, the following examples look very much like instances of long-distance binding (cf. also Nishigauchi 1992: 160):

(53) Taro to Hanako-wa(jibun-tachi-wa) otagai-o

Taro and Hanako [TOP SELF-PL-TOP each.other-ACC mi-ta to omot-ta.

see-PAST] COMP think-PAST

'Taro and Hanako thought they saw each other.'

Considered in the context of the total distribution of otagai, however, sentences like (53) and analogous examples are not just a minor problem for an analysis in terms of binding theory in some version or other. As is shown in (49) and (51), otagai may occur in subject position and thus not be bound in its governing category. These distributional facts are more or less parallel to those observable for the reflexive marker zibun, which has proved similarly problematic for binding principles. Note that in the following example zibun is part of the phrase filling the subject position, but is referentially dependent on a following genitive phrase:

(54) jibun-no shashin-ga Taro-no okorimono-dat-ta.

self-GEN photo-NOM Taro-GEN present-PRED-PAST

'His own photo was Taro's present.'

As far as we can see, neither zibun nor otagai can be easily be subsumed under binding theory. The facts discussed in this article alone argue strongly against any attempt to simply extend that theory from the domain of reflexivity to that of reciprocity or from languages like English to languages like Japanese.

(O)tagai also provides the basis for the adverb (o)-tagai-ni 'mutually', which may be used instead of the anaphors discussed above:

(55) tagai-ni (tagai-o) tasuke-au, sore-ga

each.other-adv (each.other-ACC) help-RECIP such-NOM



'Helping each other is good.'

A remarkable fact of Japanese not observable in any of the other languages discussed so far is the possible co-occurrence of the derivational and the quantificational strategy. In all of the examples listed in this section, the verbs are derived reciprocal forms with the basic symmetric predicate au added as suffix. Such a co-occurrence, however, is by no means obligatory, as is shown by the following examples. In (56a) we find the au-form of the verb with the progressive marker -teiru and thus a clear reciprocal meaning independent of the rest of the sentence, whereas in (56b) we find the progressive (i.e. -teiru) form of the verb nikumu 'hate' together with a reciprocal anaphor. There is no difference of meaning between the two sentences, but (56a) is the more frequent construction:

(56) a. futari -wa nikumi-at-teiru.

two.CLASS(person)-TOP hate-RECIP-PROG

'These two hate each other.'

b. futari -wa otagai -o nikun-deiru.

two.CLASS(person)-TOP each.other-ACC hate-PROG

'These two hate each other.'

6. Concluding remarks

The goal of this article was twofold: (a) to offer a detailed analysis of reciprocal constructions in German and Japanese against the background of a preliminary typology of such constructions developed in analogy to the one proposed by L. Faltz (1985) for reflexivity, and (b) to examine the implications of these analyses for supporting and possibly refining this typology. A major result of our investigation is that the two languages, which differ strongly in their genetic and geographic affiliations, exhibit striking similarities in the domain under investigation. In both languages, symmetric predicates, including predicates subcategorized for the comitative preposition 'with', can be shown to play a special role in reciprocal structures. In German, as in many European languages, reciprocal constructions with symmetric predicates are parsimonious in their formal coding, and in Japanese the prototypical symmetric predicates form one basis of such constructions. In both languages more than one strategy of encoding reciprocity is available and these strategies can be assigned to one of the four options distinguished in our preliminary typology. Moreover, we found that, on the one hand, such different strategies may have a different distribution and even contrast in their interpretation, as sich and einander do in German. On the other, they may also co-occur, as au and (o)tagai in Japanese. Our data also shed some interesting light on the analysis of expressions like each other in English and similar quantificational markers of reciprocity as reciprocal anaphors, that is, as expressions that are bound by an antecedent in a local domain. The variation observed above in the formal properties and syntactic behavior of the relevant expressions, for example, their possible use as subjects in Japanese, argue strongly against the simple extension of such an analysis from English to other languages. Exponents of the quantificational strategy of encoding reciprocity exhibit both different degrees of grammaticalization and striking differences in their possible syntactic functions. Even if the analysis of each other in English and its counterparts in other West Germanic languages (e.g. German einander, Dutch elkaar, etc.) as reciprocal anaphor, that is, as expressions bound in their governing category, is adequate--despite some subtle syntactic differences (cf. Everaert 2000)--the counterparts of these expressions in Japanese, Finnish, and Romance are quite different in their syntactic properties. In both German and Japanese, as well as in the other languages briefly mentioned in this article, reciprocal structures have plural or coordinate subjects as well as an argument position that is either empty (as in They married) or filled by a referentially dependent expression such as sich, einander, or otagai. Both the gaps and such anaphoras indicate that the persons identified by the subject are also involved in another participant role in addition to being agents.

Among the contrasts identified for the two languages, we would like to draw particular attention to the compounding structures with au, which do not seem to have a parallel in European languages with the possible exception of Basque and provide another piece of evidence for postulating a special compound subtype of reciprocal constructions. Another interesting difference concerns the availability of special reciprocal affixes in the nominal domain in Japanese, which is lacking in German. In view of the areal patterns described in Lichtenberk (1985, 2000) and Maslova and Nedjalkov (2004), it is less surprising that there should be no formal similarity between reciprocals and reflexives in Japanese, in contrast to German. Languages with both "reflexive and nonreflexive reciprocal constructions" (Maslova and Nedjalkov's "mixed type") are rare outside of Europe. What is remarkable about German is the fact that sich and einander are in strict opposition in prepositional phrases, though not in object positions, and that the reflexive pronoun as well as the relevant personal pronouns only allow a reflexive interpretation in such contexts. We have no complete explanation for this traditional puzzle of German grammar, but our typological approach presents these facts in a completely new light. In Romance languages the so-called reflexive pronouns (Italian si, Spanish se, French se) may have a reciprocal interpretation, in addition to their reflexive one, when they are used as clitics together with a plural subject (cf. Italian Se odiano. 'They hate each other.'). This interpretation is not possible, however, for the strong, emphatic forms of these reflexives (Italian se, Spanish si, French soi). (33) And one of the contexts requiring these strong forms is prepositional phrases. Prepositional phrases with the strong forms of the reflexive pronouns only allow a reflexive interpretation and require the use of the quantificational strategy for the expression of reciprocity (cf. Italian Sono contenti di se [stessi]. 'They are quite content with themselves.' vs. Sono contenti l'uno dell'altro.' They are quite content with each other.'). These observations are yet another example of the well-established fact that language-specific phenomena may find an illuminating explanation in the context of a crosslinguistic study.

Free University, Berlin

University of Kyoto

Received 20 January 2004

Revised version received

18 November 2004</p> <pre> Appendix. Abbreviations ABS absolutive ACC accusative ADH adhortative ART article AUX auxiliary CLASS classifier COM comitative COMP complementizer COMPAR comparative DAT dative DECL declarative DEF definite ECM exceptional case marking ELA elative ERG

ergative F feminine GEN genitive INDEF indefinite LOC locative MASC masculine NEG negation NOM nominative PAST past tense PL plural POSS possessive PRED predicate PRES present tense PROG progressive RECIP reciprocal REFL

reflexive SG singular SOC sociative TOP topic </pre> <p>References

Ashton, E. (1956). Swahili Grammar. London: Longman.

Besnier, Niko (2000). Tuvaluan. A Polynesian Language of the Central Pacific. London: Routledge.

Dalrymple, Mary; Kanazawa, Makoto; Kim, Yookyung; Mchombo, Sam; and Peters, Stanley (1998). Reciprocal expressions and the concept of reciprocity. Linguistics and Philosophy 21, 159-210.

Dimitriadis, Alexis (2002). Discontinuous reciprocals and symmetric events. Paper presented at the Anaphora Typology Workshop in Utrecht, October 2002.

Dixon, Robert M. W. (1980). The Languages of Australia. Cambridge: Cambridge University Press.

Everaert, Martin (2000). Types of anaphoric expressions: reflexives and reciprocals. In Reciprocals: Form and Function, Z. Frajzyngier and T. S. Curl (eds.), 63-83. Amsterdam: John Benjamins.

Faltz, Leonard (1985). Reflexivization. A Study in Universal Syntax. New York: Garland.

Foley, William A. (1991). The Yimas Language of New Guinea. Stanford: Stanford University Press.

Frajzyngier, Zygmunt (2000). Coding of the reciprocal function: two solutions. In Reciprocals: Form and Function, Z. Frajzyngier and T. S. Curl (eds.), 179-194. Amsterdam: John Benjamins.

Goldberg, Adele (1995). Constructions. Chicago and London: University of Chicago Press.

Haiman, John (1983). Iconic and economic motivation. Language 59, 781-819.

Heim, Irene; Lasnick, Howard; and May, Robert (1991). Reciprocity and plurality. Linguistic Inquiry 22, 63-101.

Heine, Bernd (2000). Polysemy involving reciprocal and reflexive markers in African languages. In Reciprocals: Form and Function, Z. Frajzyngier and T. S. Curl (eds.), 1-30. Amsterdam: John Benjamins.

Heine, Bernd; and Mayashita, Hiroyuki (forthcoming). The intersection between reflexives and reciprocals: A grammaticalization perspective. In Reflexivity and Reciprocity: Crosslinguistic Explorations, Ekkehard Konig and Volker Gast (eds.). Berlin and New York: Mouton de Gruyter.

Kemmer, Suzanne (1993). The Middle Voice: A Typological and Diachronic Study. Amsterdam and Philadelphia: John Benjamins.

Konig, Ekkehard; and Vezzosi, Letizia (2004). The role of predicate meaning in the development of reflexivity. In What Makes Grammaticalization? A Look from Its Fringes and Its Components, Walter Bisang, Nikolaus Himmelmann, and Bjorn Wiemer (eds.), 213-244. Berlin and New York: Mouton de Gruyter.

Levin, Beth (1993). English Verb Classes and Alternations. Chicago: University of Chicago Press.

Lichtenberk, Frantisek (1985). Multiple uses of reciprocal constructions. Australian Journal of Linguistics 5, 19-41.

--(2000). Reciprocals without reflexives. In Reciprocals: Form and Function, Z. Frajzyngier and T. S. Curl (eds.), 31-62. Amsterdam: John Benjamins.

Liu, Meichun (2000). Reciprocal marking with deictic verbs "come" and "go" in Mandarin. In Reciprocals." Form and Function, Z. Frajzyngier and T. S. Curl (eds.), 123-132. Amsterdam: John Benjamins.

Maslova, Elena; and Nedjalkov, Vladimir P. (2004). Reciprocal constructions. In Worm Atlas of Language Structures, Martin Haspelmath, Matthew Dryer, David Gil, and Bernard Comrie (eds.), 430-433. Oxford: Oxford University Press.

Matthews, Stephen; and Yip, Virginia (1994). Cantonese: A Comprehensive Grammar. London: Routledge.

McGregor, William (2000). Reflexive and reciprocal constructions. In Reciprocals: Form and Function, Z. Frajzyngier and T. S. Curl (eds.), 85-122. Amsterdam: John Benjamins.

Nedjalkov, Vladimir P. (2002). Karachay-Balkar reciprocals. Turkic Languages 6(1), 19-80.--(ed.) (forthcoming). Typology of Reciprocal Constructions. 2 vols. Munich: Lincom.

Nishigauchi, Taisuke (1992). Syntax of reciprocals in Japanese. Journal of East Asian Linguistics 1(2), 157-196.

Philip, Wilham (2000). Adult and child understanding of simple reciprocal sentences. Language 76, 1-27.

Plank, Frans (forthcoming). Thoughts on the origin, progress and pro status of reciprocal forms, occasioned by those of Bavarian. In Reflexivity and Reciprocity: Cross-linguistic Explorations, Ekkehard Konig and Volker Gast (eds.). Berlin and New York: Mouton de Gruyter.

Rakosi, Gyorgy (2002). Hungarian reciprocals. Paper presented at the Anaphora Typology Workshop in Utrecht, November 2002.

Saeed, John (1999). Somali. Amsterdam: John Benjamins.

Schadeberg, Thilo (1992). A Sketch of Swahili Morphology. Cologne: Rudiger Koppe.

Siloni, Tal (2002). Reciprocal verbs. In Proceedings of the Israel Association of Theoretical Linguistics 17, Yehuda N. Falk (ed.). [].

Wiemer, Bjorn (1999). The light and the heavy form of the Polish reflexive pronoun and their role in diathesis. In Beitrage der Europaischen Slavischen Linguistik, K. Bottger, M. Giger, and Bjorn Wiemer (eds.), 300-313. Munich: Sagner.

Zipf, George K. (1949). Human Behavior and the Principle of Least Effort. Cambridge, MA: Addison-Wesley.


(1.) The authors are indebted both to the DFG (Ko 497/8-1) and the Japan Society for the Promotion of Science for supporting their research. Moreover, we would like to thank Nick Evans, Osten Dahl, Berthold Frommann, Martin Haspelmath, Daniel Hole, Claire Moyse-Faurie, as well as two anonymous referees of this journal for reading and criticizing an earlier version of this paper. Pia-Maria Paivio has contributed data from Finnish. Correspondence address: Ekkehard Konig, Institut fur Englische Philologie (WE1), FB Neuphil. WE 1, Freie Universitat Berlin, GoBlerstr. 2-4, 14195 Berlin, Germany. E-mail:

(2.) We will be using the term construction somewhat loosely (rather than in the sense of Goldberg 1995) and not take any stance on the issue whether the term structure or the term construction is more appropriate for the characterization of complex and composite forms in natural languages, even though our sympathies lie with the former view.

(3.) This double perspective is a consequence of (a) our interest in typological generalizations and (b) of the title of the conference--as well as of this issue of the journal-where this paper was first presented.

(4.) Analogously, we could define symmetric predicates in terms of the sense relations usually distinguished in structural semantics as "auto-converse terms." Note that our terminology differs from the more common one which reserves the term symmetric for stative predicates such as resemble, or be parallel to and speaks of "inherent reciprocity" in the case of activity verbs like discuss or fight with. Strictly speaking, there are no symmetric predicates in natural languages, in contrast to formal languages (O. Dahl pers. comm.). So all we can say is that there are certain predicates that can be used in such a way that they hold symmetrically without reciprocal markers (e.g. John met Mary. [right arrow] Mary met John.).

(5.) It is a well-known fact that the same grammatical devices that express strong reciprocity--that is, strict symmetry--are also used to express weaker relations, especially with sets of cardinality greater than 2, as in: The parachutists followed each other; The pirates stared at each other. An interesting analysis of such variation in the meaning of reciprocal markers and a generalization covering the meaning of numerous cases is presented in Dalrymple et al. (1998). Our definition does not cover such cases, which seem to be special in both their semantic and their syntactic properties, as will be shown below. Note that these cases often involve predicates expressing a fundamental asymmetry, such as follow, chase, or, more specifically, sprechen zu (as opposed to sprechen mit) in German or parlare a (vs. parlare con) in Italian, which roughly correspond to English talk to but also express a fundamental social asymmetry. In cases like these the reciprocal structure is either very peculiar and difficult to contextualize (Gott und Moses sprachen zueinander. 'God and Moses talked to each other.'), or it expresses a series of actions rather than a joint one (The two boys chased each other around the garden). It is an interesting task for future typological studies to examine the special properties exhibited by such peripheral cases across languages.

(6.) Note that the sentence John is friends with Bill is an analogical backformation based on John and Bill are friends.

(7.) C. Moyse has drawn my attention to the fact that in the languages of New Caledonia we typically find co-referential pronouns in subject and object position in addition to reciprocal affixes marked on the verb.

(8.) Nor would we regard the analogous expressive devices used in Cantonese (Matthews and Yip 1994: 87) as a grammatical way of encoding reciprocity:

(i) Leih mohng ngoh, ngoh mohng leih.

I stare at you, you stare at me

'We stare at each other.'

It is interesting to note in this connection that the "reciprocal construction," that is, the sentence with a plural subject (analogous to [5]), corresponding to sentences with comitative prepositions (with) would be a sentence with the adverb together: Mary is/lives with her father. [left and right arrow] Mary and her father are/live together. This is yet another example of the widespread polysemy between a reciprocal and a sociative reading of the same expressions.

(9.) These are the verbs listed under "understood reciprocal object alternation" in Levin (1993). Note, however, that the reciprocal interpretation is only the most plausible one for such constructions, most of which may also have a distributive interpretation where the object position is simply filled in by the context:

(i) The man and the woman sitting inside the car were married, but not to each other.

(ii) Fred made an interesting proposal. John and Mary agreed.

(10.) In contrast to most of the other languages, the reciprocal marker in Russian -sja is a clitic and follows the inflectional markers for tense and person.

(11.) This root can also be found in compounds like elkarrizketa 'conversation' and elkartasuna 'union, unification'. It is, however, not perfectly clear whether the reciprocal marker derives from the verb or the other way around.

(12.) The addition of a reciprocal adverb to a sentence with a singular subject and a reflexive pronoun, as in the well-known cartoon by Philippe Geluk, results in a bizarre interpretation:

(i) Je m'aime beaucoup et reciproquement.

'I love myself dearly and vice versa.'

Similarly remarkable for its rule-changing creativity is a poem written by Jakob yon Hoddis, which ends with the lines: "Da, Freunde, hub sich gro[BETA]e Not. Ich schlug mich gegenseitig tot" ('... I killed myself mutually').

(13.) Note that the feminine third personal pronoun singular is formally identical to the third person plural form.

(14.)Cf. Plank (forthcoming), who points out that all types of quantifiers (existential, dual, mid-scale, and universal) can be found in Germanic languages. The term quantifier is used by us in the sense of 'generalized quantifier' and thus also subsumes demonstratives. Neither of these terms is fully appropriate, however, since the crucial property of these expressions seems to be that they indicate how members of the set denoted by the subject are distributed over the variables marking the argument positions of the relevant predicate.

(15.) If we consider the adverb zusammen 'together', which can be combined with a verb to form a phrasal verb, as a reciprocal marker in formations like zusammensto[beta]en 'colride', zusammenarbeiten 'cooperate', zusammenprallen 'clash', zusammen kommen 'get together', etc., German could also be analyzed as having an affixal strategy.

(16.) Since we find both critic (weak) pronouns and strong pronouns within the expressive devices of strategy HI, we have an additional scalar distinction within that strategy.

(17.) We have to admit that strategy II does not fit optimally into our hierarchy.

(18.) As far as we can see, our hierarchy is consistent with the data presented in the studies cited above.

(19.) This was pointed out to us by Claire Moyse (pers. comm.).

(20.) An interesting contrast between critic (sie) and strong pronouns (siebie) can be observed in Polish. Both of these pronominal forms may have a reflexive and a reciprocal interpretation. The critic is used for verbs which make a reflexive interpretation (non-other-directed predicates like 'defend') or a reciprocal interpretation (symmetric predicates like meet) more likely. The tonic siebie is used to force the other, less typical interpretation (cf. Wiemer 1999).

(21.) It is remarkable fact of German that the generic pronoun man can also be used in reciprocal constructions, just like on in French, but unlike one in English, even though it requires predicate agreement in the singular:

(i) Wenn man ein Jahr getrennt voneinander gelebt hat, wird man automatisch geschieden.

'If two people have rived separately for a year, they are divorced automatically.'

Collective nouns in the singular, by contrast, are treated as individuals and are thus not admissible in such structures. (?Die Gruppe begru[beta]te sich. '*The group greeted each other.')

(22.) The adverb untereinander is only used for sets of more than two: Hechte fressen sich untereinander 'Pike feeds on pike'.

(23.) Notice that this view is also somewhat problematic for French when the reflexive clitic corresponds to a dative object (Ils se lisent des histories).

(24.) Unfortunately, not all four stages of this reconstruction are attested in historical records.

(25.) D. Hole pointed out to us that jagen is possible with sich, even though there are no examples in our corpus.

(26.) It is a characteristic property of these verbs that their intransitive use with conjoined (or plural) subjects (A and B verb) and their use with singular subjects and comitative complements (A verbs with B) are equivalent.

(27.) Whether the gap-filler dependency can be described in purely syntactic terms, for example, in terms of binding, is not perfectly clear since gaps can also be filled by expressions for body parts. The referents of the subject phrase would in this case be interpreted as the possessors of those body parts:

(i) John to Mary ga atama-o naguri-au-ta.

John and Mary NOM head -ACC hit-RECIP-PAST

'John and Mary hit each other on the head.'

These structures are quite similar to structures with external possessors in European languages, where a dative phrase can be added as an additional argument.

(28.) The frequentative interpretation is possible for nondurative verbs such as kisushi-au 'kiss each other', but not for durative verbs like mitsume-au 'gaze at each other'.

(29.) Soogo was borrowed from Chinese and originally a composite expression. The character for soo is also used in Japanese for the reciprocal prefix ai (< au) and that for go is also used for tagai 'each other'. In other words, soogo can be analyzed as the result of a fusion and lexicalization of two elements borrowed from Chinese, which were later reintroduced on the basis of native resources.

(30.) The honorific form with the prefix o- is used both in spoken and written Japanese, whereas the basic form tagai is relatively rare in the spoken language.

(31.) It is interesting to note in this context that the symmetric predicate chigau is semantically the opposite ("complementary term") of au. Chigat-teiru means 'be wrong, be not identical, not fit', whereas at-teiru means 'be right, be identical, to fit'. In other words, two opposite symmetric predicates developed into reciprocal markers in Japanese.

(32.) Note that the first occurrence of otagai is a subject and thus does cannot be subsumed under the category "reciprocal anaphor." Similar data can be found in Finnish.

(33.) We owe this observation to V. Gast (pers. comm.).
COPYRIGHT 2006 Walter de Gruyter GmbH & Co. KG
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2006 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Konig, Ekkehard; Kokutani, Shigehiro
Publication:Linguistics: an interdisciplinary journal of the language sciences
Geographic Code:9JAPA
Date:Mar 1, 2006
Previous Article:4. Pragmatically motivated voice systems.
Next Article:Transitivity pairs, markedness, and diachronic stability *.

Related Articles
Germany probing man found with Japan Embassy blueprints.
Japan Embassy in Germany not target of terrorist attack.
Minister Omi relates Russian appreciation of aid projects.
Introduction: operations on argument structure *.
"Two's company, more is a crowd": the linguistic encoding of multiple-participant events *.
Rethinking the coordinate-subordinate dichotomy; interpersonal grammar and the analysis of adverbial clauses in English.
Phoronyms; classifiers, class nouns, and the pseudopartitive construction.
Studies on grammaticalization.
Third language acquisition and universal grammar.
Reciprocals and reflexives; theoretical and typological explorations.

Terms of use | Privacy policy | Copyright © 2019 Farlex, Inc. | Feedback | For webmasters