Referents of animal communication

SHORT C O M M U N I C A T I O N S out providing an explicit definition. Similarly, Abele & Gilchrist (1977) discuss the adaptive significance of male homosexual rape among Acanthocephalan worms without defining the term. Finally, Thorn_hiU (1980) has redefined the term from its commonly accepted definition to fit the needs of his specific research interests. We argue that to apply a human label to the behaviour of nonhumans does not necessarily make the events the same. Indeed, to use such a label may imply false similarities and mislead about the motivation (proximal causes) and functions (ultimate causes) of the animal's behaviour. For example, the single defining characteristic of rape, as it is commonly defined, is that a sexual act is forced upon an individual against its will. Thus, by using rape to describe non-human mating, one implies that the raped individual's motivation and/or preference for a given sexual partner are known. These are often difficult criteria to demonstrate in non-humans (and humans as well) because behavioural indices of unwillingness such as struggle, resistance or escape behaviour are not sufficient in and of themselves as evidence of motivational state. Behavioural assessments of unwillingness may be incorrect because individuals may show struggle and resistance as a normal part of their sexual behaviour, presumably even when they are willing to mate (e.g. northern elephant seals, Cox & LeBoeuf 1977; mink, marten and sable, Ford & Beach 1951). Other evidence of motivation, such as experimental preference tests or intraspecific comparisons of the behaviour in different contexts, must be presented. Frequently the term rape has been applied to non-humans without such evidence. The redefinition of commonly used words with generally accepted definitions, such as rape, to fit the needs of one's particular research area can only result in confusion and misunderstanding. For example, Thornhill's (1980) definitional requirement that rape results in the enhanced fitness of the rapist departs from the common definition where no such requirement is made. Thus, Thornhill requires that rape be defined by the functional outcome (or adaptive significance) to the rapist as well as by the immediate causes or motivation of the raped individual. Clearly, confusion would result when animal behaviourists using Thornhill's definition attempt to discuss rape with sociologists using the common definition. We also feel that there are serious connotative problems associated with using the term rape to describe nonh u m a n behaviour. Rape is an emotionally charged word that carries with it a wide range of social and ethical implications. By using the term to describe non-human behaviour, we are forcing certain h u m a n cultural standards on non-humans. We assume that scientists who apply the term to non-human behaviour do not intend these connotations, yet they cannot be avoided. What then is the solution? First, we argue that the term rape should not be applied to non-human behaviour because the term is sensationalistic because o f its connotations. The loose application of such words as rape is imprecise and does not further the goals of science. Second, we suggest that the warnings previously issued by Beach (1978) be kept in mind when making interspecific behavioural comparisons: detailed intraspeeific comparisons should always precede interspeeifie ones; the adaptiveness and function of behaviours should not be assumed to be the same across species; and where possible, the same definitions of behaviour should be used to lessen confusion. Finally, since we believe that these non-human behaviours are real and worthy of study, we recommend an alternative term be used to describe them.

1273

We suggest that the term 'resisted mating' be used to provide a purely descriptive characterization of the behaviour of a mating pair when one partner shows active resistance, struggle and/or escape behaviour. It implies nothing about the motivation of the resisting animal and can be used regardless of whether or not the individual is actually willing to mate. If it is possible to assess the proximal motivation, through the means suggested earlier, then other more specific terms could be used such as 'forced mating' (unwilling and resisted mating) or 'willing resistance' (willing but resisted mating). To ensure the broadest useful application of these terms, we urge that the adaptive significance of these acts not be pro-. scribed as part of their definitions. We believe that by avoiding the term rape and using the terms suggested here, confusion can be lessened and broadly comparative research can be facilitated. DANIEL Q. ESTEP KATHERINE E. i . BRUCE

Department of Psychology, Universit)~ of Georgia, Athens, GA 30602. References Abele, L. G. & Gilchrist, S. 1977. Homosexual rape and sexual selection in Acanthocephalan worms. Science, N. Y., 197, 81-83. Alcock, J. 1979. Animal Behavior: An Evolutionary Approach, 2rid edn. Sunderland, Mass.: Sinauer Associates. Beach, F. A. 1978. Sociobiology and interspecific comparisons of behavior. In: Sociobiology and Human Nature (Ed. by M. S. Gregory, A. Silvers & D. Sutch), pp. 116-135. San Francisco: Jossey-Bass Publishers. Beach, F. A. 1979. Animal models and psychological inference. In: Human Sexuality: A Comparative and Developmental Perspective (Ed. by H. A. Katcbadourian), pp. 98-112. Berkeley: University of California Press. Cox, C. R. & LeBoeuf, B. J. 1977. Female incitation of male competition: a mechanism in sexual selection. Am. Nat., 111, 317-335. Ford, C. S & Beach, F. A. 1951. Patterns of Sexual Behavior. New York: Harper & Row. LeBoeuf, B. J. 1978. Sex and evolution. In: Sex and Behavior (Ed. by T. E. McGill, D. A. Dewsbury & B. D. Sachs), pp. 3-33. New York: Plenum Press. ThornhiU, R. 1980. Rape in Panorpa scorpionfiies and a general rape hypothesis. Anim. Behav., 28, 52--59.

(Received 12 January 1981 ; revised 22 June 1981 ; MS. number: As-132) Referents of Animal Communication Learning what information is made available by signals is fundamental in studying animal communication. However, disagreement exists over the kinds of referents that signals can have. Most broadly and usefully, a 'referent' is anything becoming knowable or predictable through performance of a signal: who and where the signaller is, what it may be doing or about to do, aspects of its internal state, and things or events other than itself. In contrast, some ethologists hold that referents must be things apart from signallers. This particularly narrow view fails even to encompass much of the information in

1274

ANIMAL

BEHAVIOUR,

h u m a n speech, which conveys behavioural dispositions and feelings along with information about the 'external' world. The adjective 'external' has generated some confusion. I applied it to referents in the environment of a signaller (Smith 1977, page 74), as did Green & Marler (1979, pp. 83-84). However, Wittgenstein (1953) contrasted anything publicly observable (including behaviour) with the internal states of individuals, suggesting that because the latter cannot be directly compared among individuals, they cannot become referents. Green & Marler agreed that internal states are not referents, but confounded them with behaviour, which is public. (This led to incorrect classification of behaviourat analyses as 'affective' (Seyfarth et al. 1980), and rejection of behaviourfrom semantics.) In questioning behavioural referents Marier (1967) argued that a signal cannot represent 'a complex of physiological and behavioral states of which it itself is a part'. He cited Altmann (1967), who agreed about internal states but who, with Wittgenstein, accepted behavioural referents. Physiology and behaviour should not be lumped: only physiology is private. Behaviour is not part but product: as is an individual's identity, and Marler accepts identity referents. We say that displays have a communicator's behaviour as referents because there are probabilistic, conditional relations between signals and other observable acts. A songbird encountering a sleeping owl may utter a particular vocalization and either attack, retreat, abort approaches and withdrawals, pause and monitor, or resume its undisturbed activity. Such a call reveals the relative probabilities of each of the actions being selected. Even if uttered in diverse circumstances, its predictions remain the same (development of each event determines actual selection). The call also makes aspects of emotional and motivational states more nearly public. Owls (or cognitive representations of them) can be a referent if consistently involved in all circumstances. While I stressed that referents such as predators are entirely plausible (Smith 1977, pp. 73-74), most claims relied on insufficient observations so I treated them conservatively (p. 181). Owings & Leger (1980) have now studied ground squirrel (Spermophilus beecheyi) calls that have both behavioural and predator correlates. For example, among complexly variable chatter calls, those uttered in response to snakes and mammals differ, and snakes are approached but most callers stand and monitor a mammal. Ninety per cent of chatters elicited by terrestrial predators were distinguished from others, for instance from those uttered in intraspecific chases. As Owings & Leger suggest, without knowing sciurid cognitive procedures we must treat both behaviour and predators as referents. Seyfarth et al. (1980) studied calls of vervets (Cercopitheeus aethiops) and also observed correlations with classes of predators, although their procedures cannot reveal behavioural referents. Utterance of 'rraup' correlated with eagles, and of 141 cases one or more vervets looked upward in 52 ~ of the instances. Playback experiments done to control availability of some sources of information had problems, but elicited differential responses, e.g. looking upward occurred in 18 of 31 'eagle alarm' trials, but in fewer than half as many 'leopard call' playbacks. 'Eagle' is not the sole referent of 'rraup'. Adult males appear invulnerable to aerial predators, and do not respond to eagles or to 'rraup' calls. Having once been vulnerable and responsive juveniles, they must recognize eagles, and by uttering 'rraup' could alert relatives and others without personal risk. Yet they rarely or never

29,

4

utter it (Seyfarth et al. 1980, pages 1075, 1087): apparently an eagle is not a sufficient stimulus. More vulnerable vervets who call also monitor, avoid, or prepare to avoid eagles. If these actions are necessary correlates of calling, then they are among the referents of 'rraup'. Most animal signals do not have predators, food, or the like as referents. Having denied behavioural referents, Seyfarth et al. suggest 'social contexts'. Their model is the seven 'coo' sounds of Japanese macaques (Macaca fitscata), which Green (1975) analysed in terms of 'situations'. However, Green's interesting analysis is only a first step, and suggests referents that are as likely to be signallers' behaviour and identities as events that must be defined apart from the siguallers. In fact, central to Green's definition of a situation is 'demeanor and motor patterns of behavior' (Green 1975, pages 4, 38). Behaviour is also incorporated by using apparent function in the criteria. Any signal with behavioural and identity referents can be rephrased in terms of a 'situational' description. For example, a phoebe (Sayornis phoebe) utters 'twh-t' in diverse situations in which it may abort flights and may interact or select incompatible alternatives. However, since social, functional, or other relationships cannot be predicted from 'twh-t' without information from other sources, this rephrasing adds nothing to the behavioural interpretation. At present it is not obvious that 'situational' correlates of signals involve classes of referents that exist apart from signallers, although the possibility merits careful investigation. Confusion about referents hampers understanding of the semantic contributions of signals to interactional behaviour. We can lessen this by: (1) defining the behavioural, situational, or other correlates sought in studying each signal, and the extent to which correlates change as the signal's form does; (2) stipulating how we define 'external' referents; (3) recognizing that there is at present neither theoretical nor empirical justification for excluding behavioural referents; (4) recognizing that a signal may have several referents; and (5) distinguishing among different sources of information. To ask about referents is to ask a question about information made available by specified sources. It is not to pose the very different question: in what ways is an individual's uncertainty reduced in any event? Many sources contribute, and a signal should be credited with neither more nor less information than it makes available. I thank J. Sabini, H. Oster, P. Rozin, D. Leger, S. Zigmond, D. Owings, D. Hennessy, K. Derrickson, and T. Dickinson. W. JOHN SMITH

Department of Biology, University of Pennsylvania, Philadelphia, Pa. 19104 U.S.A.

References Altmann, S. A. 1967. The structure of primate social communication. In: Social Communication Among Primates (Ed. by S. A. Altmann), pp. 325-362. Chicago: University of Chicago Press. Green, S. 1975. Communication by a graded vocal system in Japanese monkeys. In: Primate Behavior, Vol. 4 (Ed. by L. A. Rosenblum), pp. 1-102. New York: Academic Press. Green, S. & Marler, P. 1979. The analysis of animal communication. In: Handbook of Behavioral Neurobiology 3, Social Behavior and Communleation (Ed. by P. Marler & J. Vandenburgh), pp. 73-158. New York: Plenum Press.

SHORT C O M M U N I C A T I O N S Marler, P. 1967. Animal communication signals. Science, N.Y., 157, 769-774. Owings, D. H. & Leger, D. W. 1980. Chatter vocalizations of California ground squirrels: predator- and social-role specificity. Z. Tierpsychol., 54, 163-184. Seyfarth, R. M., Cheney, D. L. & Marlcr, P. 1980. Vervet monkey alarm calls: semantic communication in a free-ranging primate. Anim. Behav., 28, 1070-1094. Smith, W. John. 1977. The Behavior of Communicating. Cambridge, Mass.: Harvard University Press. Wittgenstein, L. 1953. Philosophical Investigations. New York: MacMillan.

(Received 13 February 1981; revised 21 May 1981; MS. number: AS-143) Volatiles from Empty Comb Increase Hoarding by the Honey Bee Substantially increased amounts of empty comb in the nests of honey bees (Apis melliferu L.) during periods of abundant nectar secretion and availability, significantly increased the amount of honey the bees stored (Rinderer & Baxter 1978). Also, bees in laboratory hoarding cages (Kulin~evi6 et al. 1973) showed increased rates of sucrosesolution hoarding in the presence of increased amounts of empty comb (Rinderer & Baxter 1979). Laboratory hoarding cages house bees with a piece of empty comb and feeders containing sucrose solution and water. Bees remove the sucrose solution from the feeder and hoard it in the comb (Free & Williams 1972; Kulin~evi6 & Rothenbuhler 1973). Bees in hoarding cages can only occasionally be observed on comb not being used for immediate storage. Nonetheless, the presence of such comb greatly influences hoarding rates. This observation suggests the hypothesis that the stimuli provided by empty comb which lead to increased hoarding are chemical. Testing this hypothesis required the use of modified hoarding cages. The modification permitted air to be pumped first through a 2-1itre Plexiglass box filled with comb pieces and then into hoarding cages just above the

1275

comb. Connections between the air pump, the Plexiglass box, and the cages were made with tight-fitting inert plastic tubing. Each cage received 235 em z of air per rain from the air flow system throughout the experiments. Control cages received air flows from a similar system, except that the Plexiglass box contained only air. In the first of three experiments, air was passed over empty comb held at 35 C; in the second, air was passed over empty comb held at 5 C and was then warmed to 35 C; in the third, air was passed over comb filled with stored honey held at 35 C. The honey was completely ripened and sealed in the cells with wax by bees. Each experiment was composed of both experimental and control cages of bees. Combs of emerging adult worker bees were obtained for each experiment from the broodnests of seven colonies to obtain bees for seven different colony-source experimental replications. Different groups of seven colonies were used for each experiment. The combs of emerging bees were held in an incubator (35 C and 50 % relative humidity) until the bees were 0 to 24 h old. Bees from each colony were then placed in groups of 30 into five experimental and five control cages. These cages were attached to the appropriate air flow system and placed in an incubator (35 C and 50~0 relative humidity). Each cage was inspected daily for seven days; the amount of sucrose solution removed from the feeders was measured, and all feeders were refilled. Data from each experiment on the volume of sucrose solution removed during the seven days were submitted to a two-way analysis of variance. Bees in cages exposed to volatiles from 35 C empty comb hoarded significantly more sucrose solution (P < 0.001) than did bees in control cages (Table I). Differences were significant between the hoarding responses of bees from different colonies (P < 0.001). While bees from most colonies increased their hoarding by about 20% over controls, bees from one colony increased their hoarding by more than 100~o. This difference resulted in a significant (P < 0.001) treatment by colony interaction term in the analysis. Bees exposed to air containing volatiles from cold comb (5 C) or to volatiles from warm comb (35 C) containing honey did not h o a r d more

Table I. Analysis of Variance of Millilitres of Sucrose Solution Hoarded by Bees in Response to Airflows Carrying Volatiles from Three Types of Comb Analysis Comb type in experiment

Treatment

ml hoarded X 4- sE

Source of variation

df

F

P

Empty comb (35 C)

Experimental Control

17.0 4- 0.5 12.9 4- 0.3

Treatment Colonies Interaction Error

1 6 6 56

28.7 13.1 14.9

0.001 0.001 0.001

Empty comb (5 C)

Experimental Control

10.6 4- 0.3 10.0 -f- 0.3

Treatment Colonies Interaction Error

1 6 6 56

0.5 4.7 0.9

NS 0.005 NS

Comb with stored honey (35 C)

Experimental Control

10.0 4- 0.2 10.4 4- 0.3

Treatment Colonies Interaction Error

1 6 6 56

0.8 19.8 0.8

Ns 0.001 Ns