Collective markers can also often appear with stative verbs designating location 3, which includes predicates of static physical configurations such as in They stood together. I consider a true collective marker to be a form that can be used to designate joint action, and not just common location; the purely locational markers lack the semantics of collectivity as described below.
The notion of joint action, although intuitively straightforward, is semantically complex. Four factors can be identified as relevant to a collective interpretation: Spatial proximity, temporal proximity, sameness of action/state, and cooperation of the participants. The first two factors operate in conjunction. Particularly strong facilitating factors for collective coding are the speaker's simultaneous perception of both participants, in relative proximity to one another, and engaged in the verbal action or state. Spatio-temporal proximity is not in itself sufficient, since the verbal relations must also be of the same type (cf. the infelicitousness of ?John came and Mary left together). A further facilitating factor, therefore, is a perceptual similarity in the actions or states. The fourth factor, which applies only to sentient participants, entails a unitary purpose on the part of the participants (even if the purpose is simply social interaction).
I have referred to these factors as "facilitating" because each of them can be suspended to a certain degree without obviating the possibility of collective marking. For example, collective markers can be used for situations in which the participants are carrying out perceptually dissimilar actions, as in Mary and her mother made a dress together, used to describe a situation in which Mary did the cutting and measuring and her mother operated the machine. What is relevant is that the action be viewed at some level as one action, as signalled by the choice of a verb at an appropriate level of granularity embracing the different physical actions. Similarly, the notion of cooperation is not necessary even for sentient participants, since collective markers cross-linguistically can be used for situations in which the participants are not aware of each other's actions, and hence cannot be cooperating.4
I suggest that what defines collectivity is the imposition of a unitary "frame" on the actions or states such that they are viewed as a single, but internally complex verbal event. The perceptual factors referred to above can contribute to this unitary conception, but are neither necessary nor sufficient for defining it. We will now observe the related linguistic domain of reciprocal semantics. In Section 3 I will discuss the parallel between collective and reciprocal marking.
In an article dealing with reciprocal constructions, Lichtenberk (1985) pointed out that in many languages, a single form serves the function of marking both reciprocal and collective events. The Ainu prefix u- cited at the beginning of this paper, for example, is used for reciprocal as well as collective events, as in the verbs u-ekuba `bite each other' (ekuba 'bite') and u-erangara 'greet each other' (erangara 'greet'). Languages with independent cases of polysemous markers of this type include Shona, Palauan, Yindjibarndi, Basque, Japanese, Kusaiean and a large number of others.
The collective/reciprocal polysemy is motivated by a similarity in the semantic configurations associated with these two categories, as Lichtenberk notes. For one thing, in both the collective and the reciprocal situation types more than one entity is carrying out the same type of action; secondly, each entity plays two roles in the event. In the most basic type of reciprocal situation, there are two participants, and each is at the same time an agent (or other event-initiating participant) and a patient (or other type of affected entity). For example, in a sentence designating the event 'the girls see each other', each girl is both a seer and a 'seen' entity.
In the case of collective situations, the acting entities can be seen as participants that each play two roles, although these roles typically do not receive separate grammatical instantiation: Each is at the same time an agent and a 'companion' of the other entities (or a 'co-actant'). The fact of the presence of two roles is a prominent part of the semantics of reciprocality, but not particularly prominent for the collective, for which the dual roles simply follow from the fact that the participants take part in the same action.
The similarity between the collective and the reciprocal goes deeper, however. Haiman (1983) noticed that a number of languages formally distinguish two modes of marking reciprocal semantics, which are complementary in function. One reciprocal form occurs with most transitive verbs, forming predicates such as 'see each other' and 'hit each other'. I refer to such forms as heavy reciprocal markers for reasons discussed below. The other form, termed a light reciprocal marker, is restricted in its reciprocal function 5 to a relatively small lexical class, specifically verbs that either necessarily or typically involve reciprocal action. Light markers, for example, often appear on verbs designating actions like 'meet' and 'fight'. I refer to such verbs as naturally reciprocal verbs. Reciprocality for such verbs is natural in that it is the predictable or expected case. The precise verbs which take light marking vary somewhat across languages, in accordance with linguistic convention; but very similar verb meanings recur cross-linguistically with light marking, e.g. Turkish bul-us- 'meet', dv-s- 'fight', Latin osculo-r 'kiss', amplecto-r 'embrace', Bahasa Indonesia ber-kelahi 'quarrel', ber-tjakap 'converse'.
The following examples show the use of the two reciprocal markers in Turkish (glossed HR and LR for heavy and light reciprocal marker, respectively).
(1) a. Birbiri-ni gr-d-ler. HR-ACC see-PAST-3PL 'They saw each other.' (*'They met.') b. Gr-s-t-ler. See-LR-PAST-3PL 'They met.'(*'They saw each other.') (cited in Haiman 1983)
Other languages with two reciprocal forms include Hungarian (egyms vs. -ked-/-kod-), Latin (inter se vs. -r), Old Norse (hvrr annan vs. -sk), Bahasa Indonesia (silih vs. ber-), French (se...l'un l'autre vs. se), Russian (drug druga vs. -sja), and Manam (e-... vs. e-).
A related formal distinction is found in some languages between overt reciprocal marking and zero-marking of reciprocal semantics. English is a language of this type, cf. they slugged each other vs. they fought ; they hugged each other vs. they embraced. Another such language is Twi, which has zero marking on naturally reciprocal actions/states but an overt reciprocal marker for other verbs.
It is notable that in all of these languages, the two modes of expression for reciprocal semantics show similar formal and functional distinctions. The first generalization, from which the terms heavy and light marking derive, is that in each language, one form has a greater degree of formal substance than the other, for example in terms of segmental content and degree of phonological independence from the verb root. In some cases the heavy form is composed of the light form and some other element (French, Manam), which means that it has a greater number of morpheme boundaries than the light marker, another indication of greater formal substance. In the English type of language, the light form is simply zero, which of course embodies the minimum possible formal substance.
The second generalization is that heavy and light reciprocal forms show robust distributional parallels within and across languages. The light forms are the normal forms to use with naturally reciprocal verbs, and (in reciprocal readings) are restricted to those verbs, while the heavy forms are productive with transitive verbs in general. Moreover, where there is a possibility of using either form with a given verb root, there is a meaning distinction between the two forms, as in the case of the Turkish examples cited above. The heavy form is used to form a reciprocal construction which preserves the literal meaning of the root verb ('see each other'), while the light form occurs with an action of naturally reciprocal semantics ('meet; see one another socially'). These and many similar examples from other languages (cf. Kemmer In press) indicate that the two forms express semantically different reciprocal configurations.
One aspect of the difference has to do with the possibilities for temporal separation of the participants' actions. Heavy forms allow such a separation, but light forms do not. (Contrast the English examples John and Mary kissed each other, one after the other vs. *John and Mary kissed, one after the other). Similarly with relative degree of spatial separation of actions.Lack of spatial and temporal separatability are indications that light marking codes a lower degree of conceptual separation of the relations holding between the participants than heavy marking. Some support for this analysis is the fact that light marking, but not heavy marking, recurs cross-linguistically on predicates expressing stative configurations of objects (e.g. Manam e-maranati 'stick together', Latin copulo-r 'be linked'). Stative locational configurations, which involve minimal directionality and spatio-temporal separation, clearly involve a minimal conceptual separation of the participant relations.
Light marking signals a construal in which there is some internal complexity (because the multiple participants are engaged in discernible verbal relations). But superimposed on that complexity is a conception of an essentially unitary event. In this respect naturally reciprocal actions resemble collectives. Yet naturally reciprocal actions involve even greater conceptual fusion of their component events than collectives in general. The lack of expectation of separateness inherent to natural reciprocality is itself, in my view, an aspect of conceptual fusion.
The phenomenon described here is an example of iconic motivation of linguistic form (cf. Haiman 1983). The internal configuration of the two reciprocal categories is mirrored by the form of the reciprocal markers coding those categories. Heavy reciprocal markers code a configuration with more semantic "substance" in that the focus is on the individual verbal relations. Conversely, light markers are not only formally but also semantically less substantial, given that the focus is on one, albeit complex relation. Relatedly, the form of each [verb + reciprocal marker] construction corresponds, in degree of separation of its component elements, to the semantic configuration that it codes. Verbs expressing natural reciprocality show an interesting peculiarity, not only in languages that formally distinguish two modes of reciprocal marking, but also in those that collapse both types under one form. Naturally reciprocal verbs are very often deponents, i.e. verbs that lack a non-reciprocal form (e.g. Changana ku ringa-na 'be similar', Latin lucto-r 'wrestle'). The prevalence of deponents is another indication that these types of action are easily viewed as inherently reciprocal, distinct from similar types of actions that occur non-reciprocally.
(2) Nunggubuyu -nyji-: -yardi-nyji- 'to touch each other, be closely bunched'; -wuruma-nyji- 'to buzz around each other, swarm'; -a:ru-nyji- 'to abandon one another, disperse' Lushai in- : in-khawm 'assemble, come together'; in-fin 'merge into one, gather together' Rumanian se : se impreuna 'gather'; se aduna 'collect, come together' Kapampangan mi- : mi-lupung 'gather'; mi-lukub 'nestle'; mi-salbag scatter'; mi-tumpaktumpak 'be piled on top of one another' Hungarian -ked-/-kod- : tle-ked- 'throng' Turkish -Is- : u-us 'fly together' Old Norse -sk: flykkja-sk 'flock together'Verbs of this type in a given language form a lexical class designating actions or states that are necessarily or very frequently carried out by many participants. For example, to 'throng' is necessarily a collective action, as are 'assemble', 'swarm', 'flock (together)', 'disperse' and 'scatter'. Flying together (in a flock or swarm) is something that birds and insects are often observed to do. Baby birds often nestle together in a nest. Other languages having verbs of this type marked by a light marker include Squamish, Lingala, German, and Latin.We can refer to verbs of this lexical type as naturally collective verbs. The parallel with the semantic class of naturally reciprocal verbs is easily observed. In both cases, the verb class is one of a lexically restricted type comprising actions or states in which a particular participant configuration is expected or required. In this configuration, the component actions or states are minimally distinguished from one another, in that they either may not be performed at all individually, or else individual action is the unusual case.
Examples of naturally collective verbs are found in various types of languages: In some, the marker is the ordinary way of indicating collective semantics, and also functions as the reciprocal marker (e.g. Nunggubuyu). In Kapampangan, the marker mi- is the reciprocal marker, but there is no explicit way of indicating collectivity except with this class of verbs. In a third type, the marker is the same as that used for naturally reciprocal verbs, but there is a distinct way of indicating collectivity with the majority of verbs in the language.
The latter type of language is of particular interest. Turkish beraber, Hungarian egytt, Latin una, and German zusammen, 'together', are all productive collective markers that are distinct from the markers found on the naturally collective verbs (cf. Latin colligo-r 'gather'; German sich zusammenscharen 'flock together', and examples in (2)). Here, we can see the same formal relation as that found with the reciprocal/naturally reciprocal distinction: in each case, the ordinary collective marker, being a free form and containing greater segmental substance, is phonologically heavier than the marker found on the naturally collective verbs. Similarly in languages like English, in which the form used for naturally collective verbs is the (zero-marked) intransitive verb, as opposed to the overt collective marker together. Formal substance, again, iconically reflects the degree of semantic substance associated with the form. Heavy reciprocal and collective markers, which are used principally with roots that are unspecified as to separateness of action, introduce a greater degree of separation of actions than their corresponding light markers. For heavy reciprocal marking, there is unexpected reciprocality and hence unexpected separateness of the verbal actions. Heavy collective marking indicates some lack of separateness, but significantly, an unexpected lack of separateness that is not inherent to the verb. In contrast, light marking in both the reciprocal and collective cases merely reiterates the expected lack of separation that is part of the meaning of the root forms it occurs with. Viewed from the perspective of the verbal complex of root plus marking, we can see that the degree of formal separation between the elements of each pair of constructions (heavy vs. light) corresponds to the relative degree of conceptual separation associated with the associated meaning of the verb complex (more vs. less).
The semantics of nominal and verbal collectivity is similar. As we saw above, verbal collective markers express a construal involving an imposition of a unity on what can also be viewed as separate actions or relations of multiple participants. In the nominal case, collective markers express a semantic configuration in which the nominal entities are construed as a group. Group construal requires the imposition of a unitary conceptual frame on what could potentially be considered in isolation as individual entities. The entities are treated as a single, but internally complex participant.
Collective marking in nominals shows a number of different patterns in the languages of the world (cf. Barlow 1992, Greenberg 1991). One pattern of relevance here is found in languages which have an obligatory singular/plural distinction in most nouns, but for a subset of nouns have a three-way opposition between unmarked forms with collective interpretation and overtly-marked "singulative" forms to which singular and plural morphology can be added. It is notable that in such systems, the nouns concerned denote entities which are frequently are viewed in sets. For example, Shatt has unmarked collectives for the nouns gilis 'egg', abax 'hair', and nyix 'tooth' (Barlow 1992). Eggs typically are found in birds' nests in groups, and of course hair and teeth occur in groups rather than alone. We can think of such nouns as naturally collective nouns in the same sense as the verbs discussed above were termed naturally collective: the collective case occurs necessarily or very often 6, and thus is in some sense the norm for those entities.
English is of particular interest in that it contains a small lexical set of nouns with morphologically related pairs contrasting in collectivity. The case is slightly different from the one discussed immediately above, since here the contrast is between collective and plural rather than two types of collective. The overall parallel, however, is close, suggesting that the same motivation is at work.
Consider the following examples of English nouns that are said to have two plural forms, taken verbatim from Meiklejohn (1891), a traditional English grammar. The parenthetical explanations are Meiklejohn's.7
(3) SINGULAR PLURAL1 PLURAL2 die dies (stamps for coining) dice (cubes for gaming) cloth cloths (kinds of cloth) clothes (garments) penny pennies (taken separately) pence (taken collectively) brother brothers (by blood) brethren (of a community) shot shots (separate discharges) shot (balls, collectively) fish fishes (looked at separately) fish (taken collectively)The column headed PLURAL1 contains plurals formed with the productive plural marker -s, and the column headed PLURAL2 has "plural" forms that deviate in particular respects from ordinary plural marking, and that are semantically different from the other form.
Despite differences in the precise type of semantic divergence, it is clear from the table that the nouns in the PLURAL2 column involve collective construals, contrasting with the nouns in the PLURAL1 column which denote individual entities, either types or tokens.
The kind of group or collection involved in each example under PLURAL2 varies from case to case. In some instances, the entities comprise a set that is used for some single purpose, as in the case of dice and clothes.8 Brethren, used in reference to members of a religious community (e.g. Plymouth Brethren) designates a group constituted for a specific purpose. Pence refers to a sum of money in terms of its collective value rather than the individual coins that are or could be associated with that value. Both "plurals" of shot are metonymic in nature, but the unmarked form designates a set consisting of partially-differentiated particles while the other form refers to individual (abstract) entities. Fish can refer to something that is eaten at table as a mass rather than as a whole individual animal. Moreover, animals like fish can be thought of as game and, as such, things that can be pursued in the aggregate (cf. the contrast between ducks in a barnyard vs. to shoot duck).9
In all of these cases, we can appropriately refer to the collective nouns as naturally collective, since for each kind of entity designated, there is something that facilitates thinking of them in terms of a group rather than as individuals. In some cases, there has been enough semantic divergence that the two plural forms are simply seen as designating two different kinds of things, and in certain cases (clothes, shot) the collective form can be said to altogether lack a corresponding singular lexical form, as in the case of the deponents mentioned in Section 2.
Once again, the morphology associated with natural collectivity reflects the low degree of conceptual differentiation of the component elements in the conception. Shot and fish do not have any overt marking to distinguish them from the singular form; thus they follow the pattern of unmarked collective referred to above. Brethren contains two archaic forms of plural marking, but the synchronic status of these as recognizable plural markers is doubtful. In any case they are integrated with the root in a way that is not found with normal plural marking. The rest of the forms in the column contain a diachronic reflex of the plural marker -s. Here too, however, integration of the form into the root has occurred. In dice and pence, the sibilants lack the phonological changes that would be associated with the plural morpheme -s, and in clothes, sound changes of other types have operated to integrate the elements and divorce the word from any singular correspondent.
In all of these cases, the relative lack of differentiation into separate morphological subcomponents iconically mirrors the semantic lack of differentiation of a group into discrete components, despite the internal complexity of the group.10
The present study is necessarily limited in scope, but serves, I believe, to demonstrate the close relation between conceptual structuring and grammatical marking. Suzanne Kemmer Department of Linguistics, 0108 University of California, San Diego La Jolla, CA 92093-0108 U.S.A.
Croft, W., 1991, Syntactic categories and grammatical relations, Chicago: The University of Chicago Press.
Greenberg, J. H., 1991, The Semitic "intensive" as verbal plurality, Semitic studies in honor of Wolf Leslau, Vol. I, ed. by Alan S. Kaye, 577- 587. Wiesbaden: Harrassowitz.
Haiman, J., 1983, Iconic and economic motivation, Language 59:781-819.
Kemmer, S., (In press), The middle voice (Typological studies in language 23), Amsterdam: Benjamins.
Lichtenberk, F., 1985, Multiple uses of reciprocal constructions, Journal of Australian Linguistics 5:19-41.
Meiklejohn, J.M.D., 1891, The English language: its grammar, history, and literature, 6th edition.
Wierzbicka, A., 1985, Oats and wheat: The fallacy of arbitrariness, J. Haiman, ed., Iconicity in syntax, 311-342, Amsterdam: Benjamins.
2There is another attested type of collective/distributive opposition in which the distributive is the marked form, and signals separate actions by multiple actors (intransitive) or on multiple objects (transitive); the unmarked form, on the other hand, signals joint action by many on an object (transitive). This pattern, explained by Greenberg (1991), will not be dealt with here.
3Stative verbs designating mental relations do not typically appear with collective marking, and thus will not be discussed.
4An example from Basque with the collective marker elkar is Konturatu gabe elkarren ondoan gesarri ziren 'They sat down together without realizing it.'
5Light reciprocal markers in some languages also have other, non- reciprocal functions, most notably middle voice (see Kemmer, In press).
6Croft (1991:55) makes a similar point, noting that some English nouns for which collectivity is the norm also show the pattern of unmarked collective/marked individual entity, e.g. grass vs. blade of grass.
7To Meiklejohn's table, I have added only the subscripts on the column headings. I also removed a few of his examples with other, less regular meaning distinctions. Meiklejohn, it should be noted, was not making any particular point about collectives, but simply giving examples of nouns with 'two plural forms with different meaning'.
8Wierzbicka (1985) demonstrates that the semantics of clothes, in contrast with clothing, makes reference to entities in sets designed to clothe a single person. Her insightful analyses of mass nouns and various classes of -s nouns in English are relevant to many of the examples discussed here.
9Fishes is somewhat archaic for reference to a plurality of individual fish. The term is appropriate for a plurality of types, for example, in the conversation of biologists. Thus the generalization of collective vs. individual (type) still holds.
10The parallel in the case of nominals is in fact not limited to those with strictly collective semantics. Other types of complex entities that lack independent subcomponents have nouns marked with -s. For example, with ashes, embers, dregs, and entrails it is difficult to pick out individual components. Ronald Langacker (p.c.) has pointed out the connection with nouns like scissors and glasses, which have an internal complexity based on functionally dependent and structurally united subparts. Both of these classes, like the natural collective and reciprocals, often lack a corresponding unmarked form. See also Wierzbicka (1985).