1
Reciprocals Cross-linguistically, Freie Universität Berlin, 30.11.-2.12.2007
Four kinds of concepts that we (might) need for
comparing reciprocal constructions
MARTIN HASPELMATH
Max-Planck-Institut für evolutionäre Anthropologie
1. The Boasian commitment and the Dryerian observation
(the Boasian commitment; Boas 1911):
Latin, or English, or the latest version of GB/LFG/RRG/LDG/BLT/... are not
necessarily the best model of universal grammar. Hence, we should avoid
Eurocentrism, RRG-centrism, etc. and describe a language in a bottom-up
fashion, "in its own terms".
(the Dryerian observation; Dryer 1997):
Languages differ not only in the inventory and combinatory possibilities of
categories, but also in the criteria for defining categories. (E.g. nouns in English
inflect for number, nouns in Choctaw inflect for possessor, nouns in Vietnamese do not inflect
at all, ...) Hence, categories cannot be equated across languages. (Dryer 1997)
• Consequence No. 1: Linguists describing a language have to create their own
descriptive categories and cannot take them "off the shelf" (cf. also Lazard 2006).
• Consequence No. 2: Category-assignment controversies ("Is phenomenon X in
my language a Y or a Z?") are pointless (Haspelmath 2007a).
Question:
How is typological comparison possible, if languages are so different that
they all have different categories?
Proposed answer:
Cross-linguistic comparison is not category-based, but is based on (somewhat
arbitrary) typologists' comparative concepts.
Four kinds of concepts:
• *(language-specific) descriptive categories (form-based categories like the
English Perfect, the German Stative Passive, the Japanese Nominal Adjective)
• universal categories: form-based categories that are claimed to exist in all
languages (universally instantiated categories) or that could exist in any language
(universally available categories; ="cross-linguistic categories"; Newmeyer 2007).
• *comparative concepts: form-based concepts that are created by typologists
for the purpose of cross-linguistic comparison; these concepts need not
correspond to any descriptive categories.
• conceptual-semantic categories: universal categories of meaning (e.g.
Wierzbickian semantic primes, Jackendovian decompositional constants, etc.)
Claim: universal categories are not needed.
Suggestion: Conceptual-semantic categories may not be needed either.
2
2. A note on terminology
Linguistic terminology is very often multiply ambiguous/vague, as linguists
often fail to distinguish properly between the four kinds of concepts. Terms
like adjective and dative can have (at least) four different meanings:
(i) Adjective/Dative as descriptive categories e.g. in a description of German
(ii) adjective ("[+N, +V]")/dative ("[+dative]") as categories of UG
(iii) adjective and dative as comparative concepts created by typologists
(iv) adjective (='property word') and dative (cf. Fillmore 1968) as semantic
categories
Some principles of linguistic terminology:
The more different terms we have, the better. We will never have as many
terms as we have different concepts. Learning a new term is easier than
learning a new concept. We learn new concepts all the time, so the reason we
have so few new terms is not that it would be too difficult to remember them
all. The reason is that speakers are generally reluctant to accept new words
("neophobia").
3. Conceptual-semantic categories vs. form-based concepts
Linguists often try to distinguish between semantic and the corresponding
form-based concepts:
proposition
event
participant
question
time
transitive agent
hearer-identifiable
sentence
clause
argument
interrogative
tense
ergative
definite
Thus, it makes sense to also distinguish carefully between reciprocal
constructions (a form-based concept) and symmetric (König & Kokutani
2006) or mutual situations (Haspelmath 2007b).
Nedjalkov 2007: a reciprocal construction expresses a reciprocal situation;
participant occurring in a reciprocal situation: reciprocant.
König & Kokutani (2006:272-273):
...we draw a strict distinction between symmetric 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: ∀x, y ∈ A (x ≠ y ––> R(x, y)), that is, for specific substitutions of values
a and b (a, b ∈ A) for the variables x and y: aRb <––> bRa.
(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 |A | ≥ 2; 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.
3
Haspelmath 2007b: mutual situations, with mutuants as the participants.
(1) German (Wiemer & Nedjalkov 2007)
a. Hans schlägt Paul, und Paul schlägt Hans.
'Hans hits Paul, and Paul hits Hans.'
b. Hans und Paul schlagen sich.
'Hans and Paul hit each other.'
c. Hans schlägt sich mit Paul.
'Hans and Paul hit each other.' (Lit. 'Hans hits each other with Paul.')
– Two mutuants in all of (1a-c).
– Reciprocal construction only in (1b-c) (mutual situations can also be expressed
by constructions that are not dedicated to the expression of mutual situations)
– Mutuants are expressed in a single argument (as in 1b): simple reciprocal
construction (Nedjalkov 2007, §7)
– Mutuants are expressed in two arguments (as in 1c): discontinuous
reciprocal construction (cf. also Dimitriadis 2007); also transitives: Aisha
married Pedro; Lisi resembles Hans.
– Usually the set of mutuants is expressed as a single argument occupying
one of the two syntactic positions in which the mutuants are in the
corresponding non-reciprocal clause pair. This argument is called the
reciprocator (Haspelmath 2007b:2092). The other syntactic position is called
the reciprocee.
– The reciprocee position is either unfilled (as in 2), or filled by a reciprocal
anaphor (as in 1b).
(2) Cashinahua (Panoan; Camargo 2007: 1869)
Paku
inun haidu dɨtɨ-nanan-ai-bu.
Paco
and Jairo hit-REC-PROGR-PL
'Paco and Jairo are hitting each other.'
4. Comparative concepts and generalizations based on them
Some well-known examples:
(i)
ergative case (= a case of A, when A ≠ S, P)
Generalization: An ergative case always has an overt marker (Dixon 1979).
(ii)
subject (= the argument with the greatest number of "subject properties",
Keenan 1976)
Generalization: The subject normally precedes the object (Greenberg 1963).
(iii) future tense (= a tense form that has 'future time reference, E > S' as a
prominent meaning)
Generalization: Future tenses tend to be more analytic than past tenses
(Dahl 1985).
4
(iv) reflexive pronoun (= a pronoun that is used for local coreference, but not
for local disjoint reference)
Generalization: Reflexive pronouns are at least as long as disjointreference pronouns (Faltz 1985, Haspelmath 2005).
(v)
wh-movement (= movement of a wh-word to a peripheral position in
the clause)
Generalization: Wh-movement is always to the left.
(vi) wh-word (= a word that is used (possibly among other uses) in
parametric questions (= special questions, content questions) to
represent the questioned content
(vii) affix (= a morpheme that can never occur on its own)
Generalization: Affixes show a strong tendency to be postposed
(as "suffixes") rather than preposed (as "prefixes").
Comparative concepts cannot be defined in terms of descriptive categories
(because these are language-specific), so they must be defined in terms of
– conceptual-semantic categories,
– other comparative concepts, and/or
– highly general relational concepts ("precedes", "is part of", "expresses", etc.).
Comparative concepts may, but need not, correspond closely to linguistic
categories:
ergative case: Defined by use in A function, but in individual languages it may
have a broader use (A + possessor: Eskimo Relative case; A + instrument, etc.)
reflexive pronoun: Defined by possibility of local coreference and
impossibility of local disjoint reference, but in different languages the
conditions of use may be quite different (e.g. English himself vs. German sich,
cf. Reinhart & Reuland 1993)
wh-word: Defined by use in content questions, but in individual languages it
may have a broader use ('who, someone', 'what, something', cf. Durie's
epistememe, Wierzbicka's ignorative)
5. Some comparative concepts for reciprocal typology and
generalizations based on them
(i)
discontinuous reciprocal construction (= reciprocal construction in which
the mutuants are expressed in two arguments, contrasting with simple
reciprocal construction)
Generalization: Only verb-marked reciprocals allow a discontinuous
reciprocal construction. (Universal 4, Haspelmath 2007b)
(Cf. Nedjalkov 2007:§7.1, Dimitriadis 2007)
5
(ii)
reciprocator (= in simple reciprocal constructions, the argument position
that has the single argument); reciprocee (= the argument position that is
not filled by a noun phrase expressing the set of mutuants)
Generalization: In verb-marked reciprocals, the reciprocator is always the
subject, and the reciprocee can only be the direct object, the indirect object, the
possessor of a co-argument, or an adverbial. (Universal 11)
(3) a. I love you. You love me.
—> *Ø Love-REC us.
(OK: We love-REC Ø.)
(iii) reciprocal anaphor (= a noun phrase that does not have independent
reference and gets it by cross-coreference from the reciprocator, i.e. the
antecedent)
Generalization: If the antecedent and the reciprocal anaphor are coarguments
of the same predicate, all languages with reciprocal anaphors allow the
construction (unless it is pre-empted by some even more grammaticalized
construction). The less local the relationship between the antecedent and the
recipient is, the less likely it is that it is acceptable. (Universal 8)
Cf. Nedjalkov's (2007:§12.4) implicational scale:
(4) coargument > possessor of coargument > subject of complement clause
> nonsubject of complement clause
(iv) verb-marked reciprocal construction (= a reciprocal construction that is
marked by an element associated with the verb, especially affixed to the
verb, that is clearly not a noun phrase or bound pronoun)
If a language has verb-marked reciprocals based on intransitive verbs, it also
has verb-marked reciprocals based on transitive verbs. (Universal 10)
See Nedjalkov (2007: §12.1.1.2).
(v) reciprocal deponent (= a reciprocal construction with overt reciprocal
marking that lacks a corresponding unmarked non-mutual base form; e.g.
French se bagarrer 'fight'; *bagarrer)
(vi) uniplex mutual event (= a mutual event that cannot be thought of as
consisting of several sub-events; "irreducibly symmetric predicate")
Reciprocal deponents always express uniplex mutual events. (Universal 18)
(5) weak reciprocal deponents
a. Turkish
gör-üş'meet' (≠ 'see each other')
(Haiman 1983:806)
b. Norwegian slå-ss
'fight' (≠ 'hit each other')
(Kemmer 1993:111)
gör- 'see'
slå
'hit'
6
6. Descriptive categories vs. comparative concepts
The categories that express mutual events need not express mutual events
exclusively to qualify as reciprocal constructions; i.e. they need not be fully
dedicated.
• Russian -sja (Pazelskaya 2007):
Reciprocal derivation of the type soedinit'-sja "is not a purely reciprocal one, it
is rather an instance of decausativization"
• French se (Bar-Asher 2007)
The French se-construction (Jean et Pierre se sont lavés 'Jean and Pierre washed
themselves/each other') is "fundamentally a reflexive construction"
• Japanese -aw (Yamada 2007)
The Japanese -aw construction has a broader meaning than reciprocal, also
encompassing competitive and alternating constructions.
Thus, when describing these languages "in their own terms", it is not
sufficient to say that these language-specific constructions are "reciprocal
constructions". Maybe such a characterization is very misleading from a
language-specific perspective, and maybe the notion of mutuality is not even
necessary for describing the meaning.
But from a typological point of view, these constructions count as reciprocal
constructions (= a comparative concept).
7. Problems with universal categories
7.1. Anaphors
Sevcenko 2007:
"Binomial reciprocals are reciprocal pronouns. Reciprocal pronouns belong to
the category of anaphors, on a par with reflexive pronouns."
Here it is assumed that a universal category "anaphor" exists. But this
category is usually left undefined, and it is quite unclear how to deal with
intermediate cases (e.g. Cole et al. 2007).
cont'd
"As far as we know, in all cases of "simple" pronominal reflexives, there is no
agreement in case with the antecedents... Consequently, it does not seem to be
desirable at all to explain the same case marking between the antecedent and
the first part of binomial reflexives in terms of agreement."
Here it is assumed that "anaphors" behave uniformly, and that we have
basically understood them. It is assumed that language-specific facts
should be described in terms of universal categories, and that they restrict
the possible descriptions of individual languages. This is very similar in
spirit to the Latin-based grammatical descriptions of the 16th and 17th
centuries, or the French-based grammaire générale of the 18th centuries.
7
7.2. Reciprocal verbs vs. reciprocal pronouns
Siloni 2007:
assumes that "reciprocal verb" and "reciprocal pronoun" are universal
categories, and that her theory only applies to reciprocal verbs. But how do
we identify reciprocal verbs?
E.g. by "transitivity tests": (6c) shows that s'embrasser behaves like an
intransitive verb, and hence se cannot be a direct object:
(6)
a. Marie les fait courir.
'Marie makes them run.'
(causative of intransitive)
b. Marie les leur fait embrasser.
'Marie makes them kiss them.'
(causative of transitive)
c. Marie les fait s'embrasser.
'Marie makes them kiss each other.'
But Siloni provides no general way of identifying "reciprocal verbs" and
"reciprocal pronouns". The criteria are language-specific, and they are
sometimes contradictory:
"Neither of the arguments that can be adduced for the intransitive character of
French reciprocal verbs with the reflexive clitic se carries over to German.
Moreover, the fact that reciprocal sich can be coordinated with NPs argues for an
analysis as referentially dependent anaphor (Die beiden Angeklagten beschuldigten
sich gegenseitig und dazu noch ihre Nachbarn)." (König & Kokutani 2006:285)
Note that the differentiation between reciprocal anaphors and verb-marked
reciprocals is also problematic for a typological approach that avoids the
assumption of universal categories. The definitions in 5.(iii) and 5.(iv) above
assume that "noun phrase" and "associated with the verb" are readily
definable comparative concepts, which is actually not the case. So more needs
to be said before these definitions can be applied with sufficient rigor. But
whatever is added to the definition cannot make reference to languagespecific criteria that are applicable in some languages but not in others.
8. Semantic relativism
What if there are no universal conceptual-semantic categories?
What if the conceptual-semantic categories that speakers know and that
linguists use are just as language-specific as form-based grammatical
categories? (cf. Levinson 2003)
Possible answer: Cross-linguistic comparison is based on nonlinguistic
stimuli, e.g. video clips (cf. "Reciprocals across languages" project, Nicholas
Evans et al.). Speakers of different languages can talk about roughly the same
topics, and transmit roughly the same information. To the extent that this is
not the case, cross-linguistic comparison indeed becomes impossible.
8
References
Bar-Asher, Elitzur Avraham. 2007. "Two strategies for expressing reciprocals."
Abstract, this conference.
Boas, Franz. 1911. Introduction to The Handbook of American Indian Languages.
Camargo, Eliane. 2007. "Reciprocal, response-reciprocal and distributive
constructions in Cashinahua." In: In: Nedjalkov, Vladimir P. (ed.) Reciprocal
constructions. Amsterdam: Benjamins, 1865-1913.
Cole, Peter & Gabriella Hermon & Chang-Yong Sim & Chonghyuck Kim & Yassir
Tjung. 2007. "A Binding Theory Exempt Anaphor." In König, Ekkehard &
Volker Gast (eds.) Reciprocity and reflexivity: cross-linguistic explorations. Berlin:
Mouton de Gruyter.
Dahl, Östen. 1985. Tense and aspect systems. Oxford: Basil Blackwell.
Dimitriadis, Alexis. 2007. "The event structure of discontinuous reciprocals." In:
Johannes Dölling, Tatjana Heyde-Zybatow, and Martin Schäfer (eds.) Event
structures in linguistic form and interpretation. Berlin: Mouton de Gruyter.
Dixon, R.M.W. 1979. Ergativity. Language 55:59-138.
Dryer, Matthew. 1992. “The Greenbergian word order correlations.” Language 68: 81138.
Dryer, Matthew. 1997. "Are grammatical relations universal?" In: Bybee, Joan &
Haiman, John & Thompson, Sandra A. (eds.) Essays on language function and
language type. Amsterdam: Benjamins, 115-143.
Faltz, Leonard M. 1985. Reflexivization: A Study in Universal Syntax. New York:
Garland.
Haspelmath, Martin. 2007a. Pre-established categories don't exist—consequences for
language description and typology. Linguistic Typology 11:119-132.
Haspelmath, Martin. 2007b. "Further remarks on reciprocal constructions." In:
Nedjalkov, Vladimir P. (ed.) Reciprocal constructions. Amsterdam: Benjamins,
2087-2115.
Keenan, Edward. 1976. "Toward a universal definition of 'subject'." In: Li, Charles N.
(ed.) Subject and topic. New York: Academic Press, 303-333.
König, Ekkehard & Kokutani, Shigehiro. 2006. Towards a typology of reciprocal
constructions: focus on German and Japanese. Linguistics 44.2: 271-302.
Lazard, Gilbert. 2006. La quête des invariants interlangues: La linguistique est-elle une
science? Paris: Champion.
Levinson, Stephen C. 2003. Space in language and cognition: explorations in cognitive
diversity. Cambridge: Cambridge University Press.
Nedjalkov, Vladimir P. 2007. "Overview of the research. Definition of terms,
framework, and related issues." In: In: Nedjalkov, Vladimir P. (ed.) Reciprocal
constructions. Amsterdam: Benjamins, 3-114.
Newmeyer, Frederick J. 1998. Language form and language function. Cambridge, MA:
MIT Press.
Newmeyer, Frederick J. 2007. Linguistic typology requires crosslinguistic formal
categories. Linguistic Typology 11:133-157.
Pazelskaya, Anna. 2007. "Russian deverbal nominals with reciprocal meaning: types,
sources and derivation." Abstract, this conference.
Reinhart, Tanya & Eric Reuland. 1993. "Reflexivity." Linguistic Inquiry 24:657-720.
Sevcenko, Anca. 2007. "The syntax of split case binomial reciprocal pronouns: A
Romance perspective." Abstract, this conference.
Siloni, Tal. 2007. The syntax of reciprocal verbs: An overview. In: König, Ekkehard &
Volker Gast (eds.) Reciprocity and reflexivity: cross-linguistic explorations. Berlin:
Mouton de Gruyter.
Wiemer, Björn & Nedjalkov, Vladimir P. 2007. "Reciprocal constructions in German."
In: Nedjalkov, Vladimir P. (ed.) Reciprocal constructions. Amsterdam: Benjamins,
455-512.
Yamada, Masahiro. 2007. "Reciprocal without reciprocal relation between two
arguments." Abstract, this conference.