Chapter Reciprocal Constructions

by Elena Maslova and Vladimir P. Nedjalkov

1. Introduction

Map 106A displays different ways in which languages encode reciprocal situations, i.e. situations like ‘They love each other’. The reciprocal situation comprises at least two simple situations (e.g., She loves him and He loves her), so one way to encode such a situation is to combine expressions for two simple situations, as seen in (1) from Cantonese:

(1) Cantonese (Matthews and Yip 1994: 87)













‘He and I respect each other.’

Apparently, this strategy of encoding reciprocated situations is possible in all languages, although the extent to which such a pattern is conventionalized in grammar and common in actual discourse differs from language to language. However, most languages also have simple or complex reciprocal markers (verbal affixes, pronouns, particles, adverbs, etc.), which, if combined with a verb within one clause, signal that the clause describes a reciprocal situation and not just the situation denoted by this verb, without repeating the verb for each simple situation. For example, English has two pronominal reciprocal markers, each other and one another; and in Kolyma Yukaghir (eastern Siberia), there is one reciprocal marker which is prefixed to the verb stem:

(2) Kolyma Yukaghir (own fieldwork)










‘I saw you.’








‘We saw each other.’

English each other and Kolyma Yukaghir n’e- are unambiguous, i.e. they always express reciprocal meaning. In other languages, the reciprocal meaning may constitute one of several distinct functions of a grammatical construction. In Imbabura Quechua, for instance, the primary reciprocal marker is ambiguous between reciprocal and reflexive meanings, as seen in (3a); in addition, the reciprocal situation can be expressed by a suffix with collective meaning (‘jointly, together’), as in (3b).

(3) Imbabura Quechua (Cole 1982: 91)








‘The children saw each other.’


Or: ‘The children saw themselves.’








‘We hit jointly.’ Or ‘We hit each other.’

This chapter focuses on one type of polysemy pattern that is associated with reciprocal constructions cross-linguistically, namely, reciprocal-reflexive polysemy as in (3a).

2. Definition of values

Values of Map 106A. Reciprocal Constructions
Go to map
There are no non-iconic reciprocal constructions. 16
All reciprocal constructions are formally distinct from reflexive constructions. 99
There are both reflexive and non-reflexive reciprocal constructions. 16
The reciprocal and reflexive constructions are formally identical. 44
Total: 175

2.1. The existence of non-iconic reciprocal constructions

In some languages, the encoding of a reciprocal situation always involves repetition of the main verb. This type of reciprocal encoding can be referred to as iconic, since the complex structure of the reciprocal situation is straightforwardly reflected in the structure of the grammatical construction. The iconic encoding can follow a conventionalized pattern distinct from the regular combination of two clauses. For example, the reciprocal construction in Amele (Madang, Trans-New Guinea; Papua New Guinea) involves 3rd person singular suffixes on each of the two forms of the verb, independently of the actual person of the reciprocal participants.

(4) Amele (Roberts 1987: 307)









‘We (two) despise each other’

A particular case of iconic reciprocal encoding is reduplication of the verb stem, as in Godié (Gur, Niger-Congo; Côte d'Ivoire).

(5) Godié (Marchese 1986a: 231)





‘They love each other.’

Languages that can express reciprocal meaning without repeating the verb, i.e. just by adding a reciprocal marker, show considerable variation in the range of contexts in which their reciprocal markers can be used. Most importantly, some reciprocal markers are compatible only with certain classes of verbs and can express a reciprocal relation only between certain participants of the situations signified by these verbs. For example, the reciprocal construction of West Greenlandic is available for transitive verbs only, and encodes a reciprocal relation between the core participants of the situation (“actor” and “patient”), as in the following example:

(6) West Greenlandic (Fortescue 2007)





‘You are going to kill each other/yourselves’

Some reciprocal constructions are constrained to very small lexical classes of verbs; for example, Modern Hebrew has a reflexive/reciprocal construction which is used in its reciprocal meaning with ca. 10-15 verbs.

Secondly, there are languages where the reciprocal participants must be referred to by a single noun phrase (rather than by a conjunction of two or more noun phrases, as in Mary and John). This is the case in Ngiyambaa (Pama-Nyungan; Australia), where the reciprocal construction is marked by the verbal suffix -la.

(7) Ngiyambaa (Donaldson 1980: 166)





‘The dogs are biting each other.’

This construction cannot be used, however, if the reciprocal participants are denoted by two different noun phrases, as in The dogs and the dingoes are biting each other. Such a meaning can only be conveyed by a conjunction of two clauses, i.e. an iconic construction (Donaldson 1980: 163).

Thus, the availability of non-iconic reciprocal encoding in a given language can be viewed as a matter of degree, rather than a simple binary parameter. Since this study focuses on the polysemy of reciprocal markers (see §2.2), we have decided to take into consideration all non-iconic constructions, regardless of any language-specific constraints. A language is assigned the feature value 1 (no non-iconic constructions) only if it has no such constructions at all. As a result, some languages can be assigned identical feature values (2, 3, or 4) in spite of considerable differences in the range of lexical or grammatical contexts in which specific non-iconic reciprocal constructions are available. The distinction between the values 1, on the one hand, and 2-4, on the other, is therefore of limited theoretical significance. Besides, since descriptive grammars differ in the amount of attention paid to lexically constrained or otherwise marginal constructions, the feature value 1 may have been mistakenly assigned to some languages that do have reciprocal markers. As a result, the number of languages without non-iconic reciprocal constructions may well have been overestimated.

2.2. Reciprocal-reflexive polysemy

Non-iconic reciprocal constructions fall into two types depending on whether or not they can also express reflexive meaning. This is a cross-linguistically significant parameter, since reciprocal and reflexive situations are clearly different (cf. ‘They like each other’ and ‘They like themselves’), yet they share an essential feature which sets them apart from most other situation types: each participant plays two distinct roles in the same situation (e.g., respects somebody and is respected by somebody, as in (1)). The grammar of a language can reflect this similarity and use the same construction for reciprocal and reflexive situations, as illustrated by examples (3a), (6), and (8) from Wari' (Chapacura-Wanham; Rondonia, Brazil).

(8) Wari' (Everett and Kern 1997: 191)



xujuhu’ ?




‘Why are you angry with each other?’ Or:

‘Why are you angry with yourselves?’

Alternatively, reciprocal and reflexive constructions can be formally distinct (as in English, Kolyma Yukaghir (2b) or Ngiyambaa (7)). This parameter defines three language types: a language can have a reflexive reciprocal construction (as Wari', feature value 4), a non-reflexive reciprocal construction (as in English, feature value 2), or both (mixed type, as in German, feature value 3).

If a language has a reflexive reciprocal construction, it does not mean that all or even most clauses with this marker are ambiguous. In Lithuanian, for instance, ambiguous reciprocal-reflexive clauses are impossible; although the language has a single marker that can express both meanings, only one meaning is possible for any given verb:

(9) Lithuanian (Geniušienė 2007)












‘Peter and Ann kiss each other.’








‘Ann is rocking [herself].’

Nevertheless, such constructions are classified here as reflexive reciprocals, along with those constructions that allow actual reflexive-reciprocal ambiguity. The reason for this decision is very simple: the semantics of a potentially ambiguous clause as a whole is bound to play an important role in its interpretation as either reflexive or reciprocal in any language, but the role of this factor in a non-native language is very difficult to assess, and this issue is rarely addressed in grammars.

3. Geographical distribution

Most languages in the sample for this study have non-iconic reciprocal constructions. The sample may somewhat magnify this tendency, since it was designed to include enough languages with non-iconic constructions to explore the distribution along the second parameter. However, this distortion is not likely to be very significant, since the WALS sample of 100 languages also contains a relatively small number of languages without non-iconic constructions (ca. 13%); this type is not attested in South America, in Europe and in northern and Central Asia and in Australia.

Turning now to languages that have non-iconic reciprocals, the map shows a clear distinction between two large areas, Eurasia and the rest of the world. Non-reflexive reciprocals are overwhelmingly dominant in Eurasia: the sample contains no Eurasian languages that have only reflexive reciprocals (value 4) (this does not mean, of course, that there are no such languages, simply that this type is likely to be less frequent here). This large “non-reflexive” area stretches to the Pacific (excluding Australia), although the dominance pattern is less straightforward in the Pacific region (ca. 75% of languages with reciprocals have non-reflexive reciprocal markers). The rest of the world has no dominant type, i.e., reflexive and non-reflexive reciprocals are more or less evenly distributed across major areas. This situation appears to reflect the diachronic instability of reciprocal types.

Another interesting pattern distinguishes Europe and the rest of the world: the mixed type (value 3) is common in Europe (about half of the European languages in the sample) and very infrequent elsewhere: the 100-language sample contains only one such language outside Europe, Hixkaryana (Carib). This pattern may seem to be better accounted for in terms of common genetic origin, since almost all mixed-type European languages are Indo-European; on the other hand, Hindi (a non-European Indo-European language) has only a non-reflexive reciprocal, which means that common genetic origin cannot be the only determinant factor. It may be the case that the mixed reciprocal encoding should be considered as one of the several “exotic” traits of so-called “Standard Average European” (Dahl 1990, Haspelmath 2001).

The reflexive-reciprocal polysemy pattern has been investigated within the frameworks of empirical typology (Geniušienė 1987), cognitive linguistics (Kemmer 1993), and grammaticalization theory (Heine 1999). The major theoretical issue within functionally oriented approaches has been the semantic affinity between reflexives and reciprocals which motivates cross-linguistically recurrent extension of prototypically reflexive constructions to cover reciprocal situations, and the resulting semantic bleaching of the reflexive markers. Non-reflexive polysemous reciprocals have been discussed in Lichtenberk (1985, 2000a); Kemmer (1996); Maslova (1999). Detailed descriptions of various types of polysemy patterns involving reciprocals (including reflexive reciprocals) in a variety of languages can be found in (Nedjalkov et al. forthcoming). Within generative linguistics, reflexives and reciprocals used to be considered as instances of essentially the same syntactic phenomenon (anaphor); this approach has recently been challenged on the basis of both theoretical and empirical observations (see Everaert 1999 for an overview and discussion).