Do Honey Bees have a Language?

[1973 Wells, P.H. and A.M. Wenner. Do honey bees have a language? Nature 241:171-175.]

Department of Biology, Occidental College, Los Angeles, California 90041, and Department of Biological Sciences, University of California, Santa Barbara, California 93106

Von Frisch and later adherents of the theory that honey bees communicate by means of an elaborate dance are challenged by controlled experiments which show that their data can be explained in terms of olfactory cues.

Whenever new data and interpretations are presented which cannot be reconciled with established beliefs, it is proper and desirable that supporters of the traditional view should examine these data and interpretations with great care, and offer such objections as can be generated. In 1969, on the basis of studies of the recruitment of honey bee foragers to food sources, we compared the predictive powers of the classical “language” hypothesis of von Frisch(1) with those of a simple olfaction model for forager recruitment and concluded that olfaction provides the best interpretation of our data(2).

Investigators who believe that honey bees have a language have since challenged our conclusion(3-5), and two other recent articles(6-7), while not including a discussion of our results, cannot be reconciled with our interpretations. Our purpose here is to examine the principal objections to our interpretations and present additional data.

Procedure and Results

From earlier experiments, we knew that scent fed into a hive on one day influences forager recruitment on the next. We established a hive and a line of three experimental sites, each 200 m from it, in a dry field. The sites were approximately 150 m apart and the prevailing light breeze never blew towards or from the hive. Instead, the wind moved from site 1 to site 3, with site 2 in the middle.

During experimental periods of three hours, we fed scented sucrose to ten marked regular foragers at each of sites 1 and 3. Recruited bees were captured and killed. On a following day we fed unscented sucrose at sites 1 and 3 and planted scented sucrose at site 2, which had not been visited by bees from our hive. In this situation, the language hypothesis predicts that recruited foragers would be captured at sites 1 and 3, both visited by successful foragers from the hive. If, however, recruited foragers locate food sources by olfaction, they should arrive at site 2 which was not visited by any bees from the hive.

In four repetitions of this experiment (on days 2, 7, 12 and 16) we captured a total of twenty-five new recruits at site 1, 224 at site 2, and eight at site 3. Thus, newly recruited foragers had located a site in the field not previously visited by them or their hivemates, and had failed to arrive at locations about which they should have been well informed (according to the language hypothesis).

When foragers are experienced at a food site, they are again recruited to it by odour cues rather than by languag(8, 9) and, when odour cues are rigorously excluded, recruitment of new foragers fails (2, 10). For example, on control days of our experiments reported in 1969 when no scented sucrose was made available (days 4, 9 and 14), arrival of recruits at our feeding stations was drastically curtailed(2). As a further control, we allowed ten bees to fly to each of four unscented feeders for 3 h (July 25, 1968) for a total of 1,374 round trips. During this period only five recruits from a hive of 60,000 bees found the feeders(2).

Thus, both experienced and inexperienced foragers seem to require olfactory cues to locate a food source in the field. In the absence of compelling evidence to the contrary, it is conservative to suggest that olfaction alone is sufficient to account for recruitment of honey bees to a food source.

Objections to Our Model

In spite of his polemic, Dawkins(3) does offer one substantive objection by arguing that bees are easily distracted from their “intended” (linguistically communicated) goals. In this view, the bees which arrive at site 2 have been distracted from sites 1 and 3, a possibility which deserves serious consideration. But in our experiment, site 2 was downwind from 1 and upwind from 3. Is it likely that recruited foragers could be simultaneously distracted upwind and downwind from their intended goals, or that any animal can be attracted to a downwind odour source? In any case, when all three sites were provided with unscented sucrose (days 9 and 14), recruits failed to arrive at sites about which they were supposedly linguistically informed(2). We conclude that Dawkins’s “distraction” model is not supported by the data.

Esch and Bastian(6) reported interesting new data but did not relate their findings to the literature on forager recruitment. In each of their experiments, a group of bees was trained to feed at a scent-marked site near the hive, and ten of the foragers were marked and confined temporarily in a cage. The feeder was next moved to a new location 200 m from the hive and the number of regular foragers reduced to one marked bee. The ten marked former foragers were then released and observed, to discover whether they would be re-recruited to the scent-marked feeder at the new location, as they should be on the hypothesis of linguistic communication by dance attendance.

Fourteen of seventy experimental bees did attend dances and subsequently found the new feeder location. Nineteen additional marked bees attended dances and flew from the hive without arriving at the feeder. The remaining thirty-seven marked experimental bees had no contact with the dancer and did not fly to the new location of the feeder.

Ten of the fourteen successfully re-recruited bees required between one and nine exploratory flights with intermediate contacts with the dancer. Only four succeeded on the first flight, two of them within 1 min of flight time. Many of the nineteen unsuccessful dance attenders also made several flights from the hive; they attended between five and thirty-one dances (mean, 17.5) and made between one and nine exploratory flights (mean, 3.4), without finding the feeder.

For the fourteen successfully re-recruited bees the mean time in flight between first attending the dancer and locating the feeder was 8.5 min, compared with less than 0.5 min for experienced foragers(11). Esch and Bastian did not report the durations of flights by unsuccessful dance attenders.

In spite of these negative results, Esch and Bastian consider that their data support the language hypothesis. Because four bees found the feeder on their first flights, it is inferred that, “since they could not have searched the entire experimental area in such a short time . . .”, successful re-recruits must have had prior knowledge (obtained from the dance) of the feeder location.

By focusing attention on the performance of successful bees Esch and Bastian have failed to recognize that most of their data actually contradict the predictions of the language hypothesis. The population of experimental animals must be viewed as a whole. In reality, there was a total of 100 exploratory flights by marked bees after attending the dancer. If recruited bees simply flew out equally in all directions from the hive, like spokes in a wheel, 2% would be headed within 7 degrees of the feeder location, a result which is consistent with the observations of Esch and Bastian.

More generally, the data obtained by Esch and Bastian do not bear out the claim by von Frisch(12): “. . . the tail-wagging dance makes known the distance and the compass direction to the goal. . . . This description of the location enables the newcomers to fly rapidly and with certainty to the indicated flowers, even when these are kilometres away . . . an accomplishment on the part of the bees that is without parallel elsewhere in the entire animal kingdom”. Nor do they eliminate the alternative possibility, that re-recruited foragers were flying about, using wind and odour, while seeking the scent to which they had been trained.

The article by Gould, Henerey and MacLeod(7) contains many data not available earlier, but some of their data do not fall in line with what one might expect from the dance language hypothesis. As in the study by Esch and Bastian, their experiments were “. . . designed to examine the behaviour of individual recruits as each attended a dance and subsequently arrived at a feeding station”. In their series 1 experiment, marked bees were trained to two feeding stations 120 m distant from and in opposite directions from the hive, and many workers in the hive were given individual marks. During experiments, of 277 potential recruits observed to attend dances, 240 failed to arrive at any station, and only thirty-seven were subsequently captured at the feeding stations. Of these, a third arrived at a station in a direction opposite to and 240 m distant from that “indicated” in the dance maneuver. This result contradicts the prediction of the language hypothesis that all thirty-seven of the successful recruits should have arrived at the “correct” station.

These investigators also found that successful recruits spend a considerable amount of time in flight before reaching the food source. The direct line flight time between hive and a feeding station located at 120 m is less than 25 s(11), and recruited bees generally fly from the hive within a minute and a half after leaving a dancing bee (50% leave within 30 s)(13). The data in Table 4 of Gould et al.(7) reveal that the twenty-five bees which did arrive at the “correct” station flew an average of more than thirty times longer than would be necessary to “fly rapidly and with certainty” to the food source. The twelve bees which ended up at the station in the opposite direction averaged only thirty-six times as long as necessary for a direct flight.

A greater number of marked bees arrived at the statjon regularly visited by the foragers which recruited them than to the one 240 m from it, and some successful bees found the food source quickly (two in less than a minute), so Gould et al. concluded that quantitative information was communicated by the dance. The authors thereby focus attention on successful performers as did Esch and Bastian.

In another experiment, these authors fed sucrose of high molarity at one station and very low molarity at the other. Most of the dancing in the hive was by foragers regularly visiting the high molarity food source and during the experiments most of the recruits captured were at that location. This correlation is accepted by the authors as evidence of linguistic communication.

As Gould et al. carefully pointed out, the validity of their interpretations for both of the above experiments depends upon the assumption that perfect odour symmetry existed between the two stations. In experiments of this type even small asymmetries of location odour do influence recruitment(13). Gould et al. located their stations in aromatic vegetation. They got uniformly high recruitment whether or not they added scent to the food, yet there is considerable evidence that recruitment is minimal to an unscented location(2, 10). The high recruitment they obtained indicates that the feeding locations had odours detectable by bees. The unanswerable question is: Did asymmetries of these odours exist between two feeding locations 240 m apart?

Mautz(5) also was concerned with the behaviour of individually marked workers which came in contact with a forager dancing in the hive. In his experiments, 32% of marked workers that attended dances found the feeder, and the amount of time the potential recruits attended the dancer was positively correlated with success at finding the food. In addition, the average flight time for successful recruits was more than ten times that expected of experienced foragers and the mean flight time of bees which flew out but failed to find the feeder was twice that of the successful recruits. Mautz noted that one bee found a goal at 400 m after following only five waggle cycles of the dance.

As with Gould et al., and with Esch and Bastian, Mautz assumed that his feeder locations were not recognizable by bees as olfactorily distinctive and that only linguistic information could be communicated more effectively by increased duration of contact with the dancer. He does not discuss the possibility that a recruit which spends more time attending a dancing bee might, by so doing, gain a more accurate impression of the odour characteristics of a specific location. Again we see a supporter of the language hypothesis focusing attention on the successful recruits.

Of the several papers we review, Lindauer’s(4) was the most deliberate and direct effort to obtain data contradictory to our findings. The first type of experiment undertaken by Lindauer resembled those of Gould et al.(7) in design, and yielded similar data. Forager bees were trained from a hive to two feeding sites, and were then fed high molarity sugar at one and low molarity sugar at the other. Virtually all dancing was by bees visiting the high molarity sugar site, and most successful recruits landed there. Lindauer assumed, as did Gould et al. and Mautz, a perfect symmetry (to bees) of environmental odours at his locations. From this he inferred that, “if the recruits followed only the odour signals given by the dancers . . . they should have appeared in equal numbers at (both) sites”. In accordance with his assumption Lindauer attributed the asymmetry in recruit arrivals to linguistic communication. Thus, the untested assumption of station symmetry (to bees) is central to the reasoning of Lindauer as well as to the above authors.

In his second experiment, Lindauer put out three stations in a geometry similar to that used in our experiments(2) and trained bees to stations 1 and 3 but not to 2. Then, with scented sucrose at all three stations (a variation from our design), he collected recruits. Although the stations were deliberately asymmetrical (no foragers at station 2), approximately one-fourth of all captured recruits arrived at station 2. The language hypothesis predicts that the recruits should have travelled only to stations 1 and 3.

In order to reconcile these data with the language hypothesis Lindauer generated ad hoc the auxiliary hypothesis that potential recruits alternate their attentions between dancers visiting sites 1 and 3, and subsequently “integrate the directions communicated by both groups of foragers”. Although the “integration” model satisfies Lindauer as an explanation of his data, it could not possibly predict the distribution of recruits in our earlier experiments(14, 15) or, indeed, in the recent experiments of Gould et al.(7).

Lindauer got good recruitment at his stations 1 and 3 when presenting unscented food at those locations. Interpreted in the light of our data(2, 10) these results suggest that Lindauer’s feeding sites were in some measure distinctive to bees. This may explain the asymmetry of recruitment he observed when foragers to station 1 were killed and bees visited only a single station. Unfortunately, the precise location and ecology of the sites for these experiments are not given in his paper (except that it was near Frankfurt), nor does he provide dates, wind speed and directions, number and sequence of observations contributing to the total data and other relevant information. These omissions make any interpretation difficult and speculative. Another portion of Lindauer’s paper offers some philosophical and rhetorical objections to our paper, but we will consider these later.

Thus all the authors we have discussed(4-7) explicitly or implicitly assumed that the feeding stations they established in the field were in no way distinctive to bees and that the environment can be made symmetrical with respect to odours and other factors. They also assumed that, in order to find the food, a recruit must have prior quantitative information about its location. These assumptions lead the authors to interpret arrival of new workers at sites visited by dancing foragers as definitive evidence of linguistic communication. As pointed out earlier they have focused their attention on successful recruits. Experiments based on these assumptions invariably lead to affirmation of the consequence of the hypothesis (if bees have a language, recruits will reach the food. Some recruits find the food. Therefore, bees have a language). This reasoning is deductively invalid(19).

The experimental designs of these investigators suffer yet another serious disadvantage. It is not logically possible experimentally to establish the validity of the station symmetry assumption (failure to display asymmetry does not prove that none exists). It is quite practical, however, to seek positive evidence of station asymmetry in experiments of this design. One way would be to remove all directional information from dances performed by foragers regularly visiting food sites. If, in spite of this, recruits preferentially arrive at the feeder regularly visited by the foragers which recruited them, it could be concluded that they had done so without use of prior directional information, and that honey bees can exploit subtle environment asymmetries in the recruitment of new foragers to food sources.

We have incorporated the use of “directionless” dances in the design of the experiments reported below.

Conditions and Observations

In the summer of 1970 we did experiments using a single frame observation hive and two feeding stations located 150 m from it in approximately opposite directions (west and east of the hive, but slightly north of it). The hive and stations were located at the University of California, Santa Barbara. Both stations were located in dry fields, devoid of green vegetation. The West Station was nearer the sea and upwind the East Station somewhat closer to the University buildings. Otherwise, the sites appeared to be symmetrical.

Ten marked foragers were trained to feed at each station; all successful recruits were either killed or marked as replacements for regular foragers which were then killed. Observations were made at the hive and stations during 3-h periods (0830-1130), August 21 to September 4, 1970, and food was provided only during those periods. A westerly breeze prevailed during this period (up to 5 mile h-1 on most days) and the usual temperature was 65 – 75 degrees F.

A principal objective of the experiments was to examine relationships among food molarity, forager behaviour, and recruitment. Therefore, scent added to the food was held constant. All solutions used were scented with lavender oil (32 l. l.-1). Sucrose solutions of 0.8 M, 1.3 M, 1.8 M and 2.3 M were used. On any given day a solution of lower molarity was offered at one station and one of higher molarity at the other. Rotation of molarities between the stations controlled for possible inherent differences in location attractiveness, and on each day the stations acted as controls against each other. The experiments were run blind: the observers were not informed of molarities of sucrose provided at the feeding stations on any given day.

The hive had glass sides and was equipped with a clear plastic atrial chamber inside a diffusely lighted observation room. Hive conditions (well populated, and with little natural forage) ensured that our marked foragers danced in the atrial chamber, rather than on the comb. These dances were on a horizontal surface, and contained no discernible directional information. The straight (waggle) portions of the dance were apparently randomly oriented and highly variable within each dance episode.

Altogether, a total of 1,793 dances by our marked bees and 136 dances by unmarked bees foraging on natural sources were observed during eleven days (33 h) of experimentation. Thus, during our study periods, 93% of all dancing in the hives was by our marked foragers.

Although we were largely successful in eliminating oriented dances from our hive, occasionally a marked forager did enter and perform a dance episode on the vertical surface of the comb. These occurrences will be of great interest to proponents of the language hypothesis, so we will present data on oriented dances before proceeding to other aspects of the study.

There was a total of only 50 apparently oriented dances recorded during 33 h of experimentation; 1.5 such dances h-1, on the average, or approximately 2.8% of all dance episodes recorded for our marked foragers. These occurred equally (24 to 26) for the two stations and did not correlate with recruitment (Table 1).

These oriented dances were not distributed evenly throughout the experimental period, but occurred sporadically. For example, 40% of them occurred on one day, August 30, 1970. Occurrence of a block of oriented dances on a given day did not appear to influence recruitment. A comparison of data for August 30, 1970 (which had the greatest number of on-comb dances during the eleven days of experimentation) with the following (more typical) day is illustrative, and is presented as Table 2.



Table 1 also shows that dancing by unmarked bees foraging on natural food sources declined as the experiment progressed. This is consistent with our findings that insertion of a new food source in the hive-environment system may change the behaviour of bees regularly foraging on an established one(10). Presumably this alteration is mediated through a change in the available recruit pool.

From our data on oriented dances (Tables 1 and 2) it is not possible to infer that these dances are responsible for the observed recruitment of new foragers to those feeders we had established in the field. There is, however, a positive correlation between the number of disoriented dances (or total dances) and recruitment of new foragers (Table 3).


The experimental design is not rigorous enough to define the asymmetry in environmental cues used by recruits while searching for and locating a particular feeder in the field, but quantitative directional information does not appear to be a part of that system.

While doing this set of experiments, we also gathered information on Nasanov gland exposures at the food source by bees foraging on various molarities of sucrose. Our previous studies indicate that Nasanov gland exposure by regular foragers fails to attract undisturbed bees but apparently does provide a point of reference for disoriented members of the colony. Gland exposure is apparently a function of the interest potential recruits may have in a food source, rather than food quality per se(2, 10). Nasanov gland exposure by foragers visiting established food sources in the field was depressed at the lowest molarity we used (Table 4), a fact which indicates that few recruits were in the field searching for that source.


Further Considerations

We have done several experiments which were designed to determine whether honey bees use linguistic communication under defined conditions(2, 8, 9, 14-16). Results indicated that the language hypothesis has little heuristic value and fails as a predictive tool. The results further suggest that if one wished bees to pollinate a particular crop, it should be useful to regulate the odour carried into the hive by foragers, rather than the angle of dance on the comb.

More generally, we feel that honey bee foraging ecology is regulated by a complex system of hive, environmental and behavioural variables. With sufficient knowledge, it should be possible to manipulate the system and predict the consequences for the foraging ecology of a honey bee colony. We then might say that we understand colony function in terms which are neither teleological nor anthropomorphic; and we would be in a position to attack pragmatically problems of a comparative or practical nature. Accordingly, we have examined interrelationships among Nasanov gland exposure, visits of marked foragers to a food site, dancing in the hive, and recruitment of new foragers, and correlated these with the amount of odour in sucrose solutions provided in the field. In these experiments we have had some success in manipulating the system(10), as have Waller(17) and Friesen(18).

Waller showed that association of odours with food in the field, but not odours alone, could increase bee populations in experimental plots(17). Friesen examined interrelationships among wind speed and direction, forager flight paths, odour levels and locations, forager visitation and recruitment of new foragers to sites in the field. He concluded that, after recruits leave the hive, their success and distribution depend upon an interaction of field variables affecting the distribution of odours(18). Thus, evidence from several sources indicates that odour is of great importance in the system of variables which regulate honey bee foraging ecology.

Our comparison of the predictive values of the olfaction and language hypotheses of honey bee forager recruitment was an attempt to determine just how important odour is in this system(2). To our surprise, our results and interpretations generated a controversy which deeply polarizes interested biologists. Along this line we must stress that the controversy does not emerge from a difference of opinion about the acceptability of various parts of a body of evidence.

The ingenious step and fan experiments of von Frisch have been repeated many times, and our own published data clearly show that when his procedures are closely followed, without insertion of additional controls, distributions of successful recruits may be expected to resemble those obtained by him(14, 15). Similarly, when Lindauer used an experimental geometry modelled after ours, he obtained data consistent with ours(4); and the data presented in Table 3 of this paper resemble those obtained by Gould et al. and by Lindauer when they used similar experimental designs(4, 7). Neither the repeatability of experiments on honey bee behaviour nor the care and accuracy with which workers in this field gather and report data appears to be in question.

Proponents of the language hypothesis argue that the many repetitions of the von Frisch experiments make it highly probable that his interpretation is correct. One must realize, however, that the amount of additional confirmation affected by each new favourable repetition of an experiment becomes smaller as the number of previous repetitions grows(19). Thus, even under the most favourable conditions, sheer number of repetitions is never sufficient to render a hypothesis immune to challenge. Furthermore, in the case of the bee language hypothesis, the existence of a considerable body of unfavourable evidence greatly diminishes the weight of even a large body of confirming data(20).

Any hypothesis which is generated as an explanation of certain observed events will, of course, imply their occurrence. The events to be explained will then be taken as supporting evidence for it. If the hypothesis is valid it also may lead to a priori predictions of facts and events in conditions different from those leading to its formulation.

It is exactly here that we have difficulty with the language hypothesis of honey bee orientation. Each time we have inserted previously omitted controls or altered the experimental design to create new test implications, the distribution of successful recruits to feeders located in the field is no longer that predicted by the language hypothesis. Similarly, much of the evidence recently obtained by Esch and Bastian, by Mautz, by Gould et al. and by Lindauer is contrary to the predictions of that hypothesis(4-7).

The difficulties we had in reconciling our data(14-16) with the language hypothesis led us to propose the alternative olfaction model(21, 22). The olfaction hypothesis allowed a satisfactory explanation of the distribution of successful recruits in those experiments which incorporated new controls(14-16) and could encompass the large body of existing data on honey bee forager recruitment. Furthermore, it had the advantage of simplicity.

Thus, in 1967 we had two hypotheses, olfaction and language, each generated a posteriori to explain an existing body of evidence. The language hypothesis had been well articulated by von Frisch twenty years earlier(1), but it had been expanded and modified through the years in an attempt to explain newer results. Olfaction, while new in the sense that it challenged the then currently accepted hypothesis and two decades of thought habits, was essentially a return to the position held by von Frisch and others in the earlier years of this century.

Lindauer not only presented new data but challenged our interpretation on philosophical grounds when he invoked Aristotelian (Darwinian) teleology as an argument in favour of language. He argued that in nature “each morphological structure and behavioural act is associated with a special function. On this basis alone, it would seem unlikely that information contained in the waggle dance of a honey bee is not transmitted to her nest mates”(4). We have answered his challenge in part by discussing the relevant data in terms of philosophy of science which we consider to be more powerful(19, 20).

An inherent weakness of Lindauer’s teleological argument can be illustrated by giving one interesting example from another field of biology. Methyl eugenol is extremely attractive to male oriental fruit flies. In one test in Hawaii 1,300 male Dacus dorsalis were attracted a half-mile upwind to a muslin screen that had been treated with it. Methyl eugenol is not produced by the female fruit flies nor does it attract them. It is not a component of the natural food of this fly and probably has no nutritional value. Yet male oriental fruit flies are irresistibly attracted to it and “apparently cannot stop feeding when they have free access to it, and they kill themselves with over indulgence”(23). Their behaviour under these circumstances certainly cannot be construed as adaptive.

Admittedly, this is an extreme example of non-adaptive behaviour, but it does reveal a weakness in the teleology argument. The mere presence of a characteristic behavioural pattern in an animal cannot be construed as purposeful, adaptive or “associated with a special function”.

Lindauer(4) also challenged us on grounds that we have not individually discussed three specific situations investigated earlier by von Frisch and his colleagues. First, at distances quite close to the hive, recruited foragers may be captured at scented dishes other than the one visited by the foragers which recruit them, while at distances of 100 m and beyond “the recruited workers all fly in the direction of the feeding plate”. That the latter part of this quoted statement does not hold is well documented by the data of Gould et al(7) as well as by ours(2, 14, 15), and even by Lindauers own data(4) unless one is willing to accept his “integration” model. As for the first part of the statement, when the feeding dishes are close to the hive, we agree with von Frisch and Lindauer; the bees seem to be using odour.

A second situation involves disoriented dances performed by foragers on a horizontal surface. In an experiment using a tilted hive, von Frisch got approximately equal recruitment in four directions while regular foragers were fed in only one direction. Later that day, with the hive put upright again, he got preferential recruitment in the direction of the feeder visited by foragers(12). We prefer a design in which two or more stations visited by bees simultaneously serve as controls against each other, and we set up our experiment accordingly. Our data on the effectiveness of disoriented dances (Table 3) seem to be in disagreement with those of von Frisch. During our 11 days of observation, new recruits preferentially arrived at the sites visited by dancing foragers even though there was no directional component in the dances.

The third situation involves the detour of foragers around an obstacle to a feeder on the other side of it. In one experiment performed by von Frisch, marked foragers were trained to fly around a rocky ridge to a scented feeder. Scent plates were placed on top of the ridge and 50 m laterally to the direct line between hive and feeder. During a 90 min period, three new recruits were captured at the lateral stations, eight on top of the ridge and twenty-three at the feeder. Because the new recruits apparently failed to follow the flight path used by the foragers which recruited them, von Frisch feels that they were linguistically informed of the feeder location(12).

We must argue that the flight paths and duration of searching by the twenty-three successful recruits are not known, and that a failure of bees to use the detour (if indeed they did not use it) does not differentially support either language or olfaction. Since neither hypothesis predicts that recruits will follow the same path as the marked foragers, the experiment is not crucial and the results do not support or refute either hypothesis.

We believe that experimentation does not “prove” or “disprove” hypotheses, but rather affects their credibility. If the credibility of a hypothesis at any given time is determined by the total body of relevant information available, then the language hypothesis of forager recruitment was very credible in 1946 when von Frisch proposed it. It is less so now, for the body of relevant information is quite different and includes much unfavourable evidence. In fact, it is so much less credible that we no longer can believe that honey bees communicate linguistically.

Do honey bees have a language? That is a question which may never be answered with certainty. It may be more useful to examine assumptions critically, state hypotheses and their consequences with precision, review the evidence objectively, and ask: can we now believe that honey bees have a language? Thus, it appears that the honey bee forager recruitment controversy is not about the nature of evidence but rather about the nature of hypotheses(25). It is not what investigators observe (the data) but what they believe (infer) that is at the heart of the controversy.

We thank Nelson Dee and Stephanie Niebuhr for technical assistance. The research was supported by the US National Science Foundation. An analysis of earlier events and attitudes leading up to the bee language controversy is available elsewhere(24).

1 Von Frisch, K., Osterr. Zool. Z., 1, 1 (1946); translation, Bull. Anim. Behav., 5, 1 (1947).
2 Wenner, A., Wells, P., and Johnson, D., Science, 164, 84 (1969).
3 Dawkins, R., Science, 165, 751 (1969).
4 Lindauer, M., Amer. Nat., 105, 89 (1971).
5 Mautz, D., Z. Vergl. Physiol., 72, 197 (1971).
6 Esch, H., and Bastian, J. A., Z. Vergl. Physiol., 68, 175 (1970).
7 Gould, J. L., Henerey, M., and MacLeod, M. C., Science, 169, 544 (1970).
8 Johnson, D. L., and Wenner, A. M., Anim. Behav., 14, 261 (1966).
9 Johnson, D. L., Anim. Behav., 15, 487 (1967).
10 Wells, P. H., and Wenner, A. M., Physiol. Zool., 44, 191 (1971).
11 Wenner, A. M., J. Apic. Res., 2, 25 (1963).
12 Von Frisch, K., Tansprache und Orientierung der Bienen (Springer-Verlag, Berlin, 1965); translation, The Dance Language and Orientation of Bees (Harvard University Press, Cambridge, 1967).
13 Johnson, D. L., and Wenner, A. M., J. Apic. Res., 9, 13 (1970).
14 Johnson, D. L, Science, 155, 844 (1967).
15 Wenner, A. M., Science, 155, 847 (1967).
16 Wenner, A. M., Wells, P. H., and Rohlf, F. J., Physiol. Zool., 40, 317 (1967).
17 Waller, G. D., J. Apic. Res., 9, 9 (1970).
18 Friesen, L., Biol. Bull. (in the press).
19 Hempel, C. G., Philosophy of Natural Science (Prentice-Hall, Englewood Cliffs. 1966).
20 Popper, K., in British Philosophy in Mid-Century (edit. by Mace, C. A.) (Macmillan, New York, 1957).
21 Wenner, A. M., and Johnson, D. L., Science, 158, 1076 (1967).
22 Wenner, A., and Wells, P., XXI Int. Apic. Congr. Summ., 88 (1967).
23 Steiner, L. F., J. Econ. Ent., 45, 241 (1952).
24 Wenner, A. M.. The Bee Language Controversy (Educational Programs Improvement Corporation, Boulder, 1971).
25 Altmann, S. A., Nature, 240, 361 (1972).