Excursus EXC: Exchange with the Journal Science

"[The] development becomes known to the public. Popular science books . . . spread the basic postulates of the theory; applications are made in distant fields, money is given to the orthodox and is withheld from the rebels. More than ever the theory seems to possess tremendous empirical support."

- Paul Feyerabend 1975:43

Scientists reputedly welcome challenge and readily adopt new interpretation whenever evidence no longer conforms to observed phenomena (Kneller 1978). Both Mahoney (1976) and Veldink (1989) found otherwise. Some comments can clarify why this discrepancy in attitudes about scientific objectivity exists.

Collectively, and over varying time spans, Kneller's assessment appears to be correct; scientific interpretation changes constantly and relatively rapidly. That change in commitment to interpretation is one feature that distinguishes science from religion.

When new evidence arises that conflicts with "conventional wisdom" on relatively important issues in science, however,individual scientists may momentarily rush to the defense of an existing paradigm with religious fervor. Feyerabend deplored this defensive behavior of scientists, and wrote: "To sum up: Unanimity of opinion may be fitting for a church. . . . Variety of opinion is necessary for objective knowledge" (1975:46; emphasis Feyerabend's).

Something akin to religious fervor was what we encountered when we challenged the dance language hypothesis. That was because that hypothesis had already evolved into a full-fledged paradigm spanning several fields. As might be expected, some felt that our impression - that the defense of the dance language hypothesis was extreme - was a gross exaggeration, since "everyone knows that" scientists shun "censorship" and suppression of new ideas. We were even considered by some scientists to be "paranoid" (Veldink 1989).

Prior to our challenge, and while we were still respected members of the animal language research community, our manuscripts received favorable reviews and were published without undue delay. Subsequent to that challenge, we encountered increasing hostility (as noted in excursus SI). Within a very few years, our chance of getting material into print or obtaining grant support went from difficult to nearly impossible. The negative evidence we had gathered relative to the dance language hypothesis seemed to be accorded little significance in this development (e.g., Veldink 1989).

Due to the extreme adverse reaction we encountered, we feel that this volume would not be complete - dealing as it does with the human side of scientists - unless we included some documentation for the comments made above. Accordingly, we provide here a chronological sequence of some pertinent material spanning those first few critical years.


Richard Dawkins, in a letter to Science, exemplified the confusion existing among some ethologists on the significance of our crucial experiment (Wenner, Wells, and Johnson 1969; see also our chapter 10). He objected to the interpretation we provided for the results of that experiment, in part, as follows:

Suppose a man tells me there is a bar three blocks down the street on the right. I set off thirsty, but on the way a strong smell of beer distracts me to another bar hidden up a side alley. Does this prove that human language does not communicate information? . . . it is entirely reasonable to suppose that bees have alternative ways of finding food - among them, the dance, smell, and the presence of other bees - and that each of these cues may predominate under different circumstances. For example, the artificial use of strong scent might cause olfactory cues to prevail. (1969:751)

Dawkins concluded his letter:

"In brief, bees are easily distracted. This modest and uncontroversial conclusion is all that can be drawn from the experiments purporting to disprove von Frisch's classic work"

Dawkins thereby adhered close to the "party line" of those committed to the dance language paradigm. From our perspective, on the other hand, Dawkins' ad hoc rationalization merely confused the issue, because he ignored the potential influence of wind direction, an essential element when one considers odor movement and resultant insect flight (see chapter 5 and excursus OS). Furthermore, Dawkins contradicted von Frisch's earlier assertions that recruit bees would not be distracted during their outgoing flight. For example, von Frisch had written:
There is no doubt that the bees understand the message of the dance. When they fly out, they search only in the neighborhood of the indicated range, ignoring dishes set closer in or farther away. Not only that, they search only in the direction in which the original feeding dish is located. (1962:78; see also our chapter 9)

Scent (notice in the above passage Dawkins' use of the word strong), strong or otherwise,ifof the right kind, can "distract" searching bees (see chapter 8), but only ifsearching bees are already downwind from such a source. In that sense odors are quite a different stimulus than either sound or light stimuli.

Dawkins, for example, could have been distracted by the smell of beer during his tavern hunt only ifthe second tavern had been upwind of his original path. But he could have been distracted by the sound or sight of the tavern, regardless of its direction from his original path. It is unlikely that Dawkins would have been distracted by the odor of a bakery under the circumstances of his example.

Other dance language proponents began citing Dawkins' letter as refutation of our work, even though it was obviously nothing more than an expression of opinion, faulted at that by Dawkins' failure to consider wind direction in his polemical statement. This curious turn of events led one of us (Wells) to draft a reply to that Dawkins letter. We present that short draft here in its entirety so that the referee's comments, which follow and which formed the basis for rejection byScience, will fall into perspective.

Reply to Dawkins

Subsequent to our comparison of the predictive values of the "olfaction" and "language" hypotheses of honey bee forager recruitment [Wenner, Wells, and Johnson 1969], Richard Dawkins generated a challenge [Dawkins 1969] which confuses key issues. I am surprised to find that some ethologists have [now] accepted Dawkins' discussion as an excuse for disregarding our experimental work. It is necessary, therefore, to reply to his three objections to our paper.

First, Dawkins asserted that we "presume to challenge the findings of a great biologist" (Karl von Frisch). To this we plead guilty; but, the charge isn't serious enough. In so doing we also presume to challenge a host of subsequent investigators who have uncritically accepted any and all of von Frisch's interpretationsof available data; and who have used these interpretations as a foundation on which their own work is based. This may account for the emotion occasionally engendered by our papers.

Next, he objected to our citing (in our paper's introductory paragraphs) earlier studies which yielded data inconsistent with the hypothesis that bees use linguistic communication [Wenner, Wells, and Rohlf 1967; Wenner and Johnson 1966; Johnson 1967a; Lopatina 1964; Wells and Giacchino 1968; Wenner 1967] on grounds that the failure of bees to use language does not prove they have none. With this we heartily agree! Experimentation does not "prove" or "disprove" hypotheses, for formally this can never be done. Rather, it tests their usefulness in the a priori prediction of events. Usually one retains as his working hypotheses those with high predictive success under the most rigorously controlled conditions.

Rather than "purporting to disprove von Frisch's classic work" (Dawkins' misinterpretation of the intent of our experiments) our experimental results show that, under well controlled conditions, the olfactory hypothesis is successful while the language hypothesis fails to predict the distribution of recruit foragers in the field.

As a substantive objection to our paper, Dawkins suggested in an eloquent but anthropomorphic analogy that crowds of bees or strong scent might distract recruits just as "the smell of beer distracts me" from a prior destination. We purposely inserted controls against these possibilities in our experiments. These controls preclude his interpretation. As the legend to our Fig. 2 [in Wenner, Wells, and Johnson 1969] indicates, one experimental feeding station always was upwind and the other one downwind from a control station visited by no bees.

According to Dawkins' line of reasoning, bees would have had to be simultaneously distracted to the scent at that control site both upwind and downwind from their "intended" destinations. Either that, or recruited bees would have failed to verify von Frisch's contention that "if no clues are provided by scent the bees use information conveyed by the dance" [von Frisch 1968:532].

Furthermore, on days nine and fourteen of our experimental series [table 1 of Wenner, Wells, and Johnson 1969] no scent was placed at the control station. In the absence of "distracting" odor cues, recruit bees still failed to find the unscented experimental stations about which they might have been linguistically informed, even though "crowds of bees" were there.

But back to Richard Dawkins' delightful analogy. I feel confident that he and I would agree that scientific questions are more likely to be resolved by experimentation than by polemics such as our exchange of letters in Science. If I am right, and we can avoid being distracted from that one key point of agreement, next time he is in Los Angeles, I'll buy!

The letter by Wells attempting to reply to Dawkins was rejected by the editor of Science. The rejection was accompanied by a set of comments by an anonymous referee. We include that text, also in its entirety.

Referee's Comments

This paper is uncomplimentary to all scientists who do not accept the author's theory. Specifically the author accuses ethologists of accepting Dawkins' discussion as "an excuse for disregarding our experimental work." Can't he accept the thought that they reject his work because after critical study they do not consider it irrefutable?

And if that were not infuriating enough he accuses a "host of investigators for uncritically accepting any and all of von Frisch's interpretations." How can he presume to know that? Is any new theory accepted without critical study? Perhaps if his own work were as thorough, complete and convincing as that of von Frisch he would not need to berate other scientists for not accepting it. The use of polemics is a poor substitute for irrefutable facts.


In 1970 Gould, Henerey, and MacLeod, three undergraduate students at the California Institute of Technology, published their ten-page lead article in Science (see chapter 12). By then we were becoming aware of the sociological implications of the dance language controversy. Nevertheless, we were still ill prepared for what transpired next.

A study of the paper by Gould and co-workers revealed that these undergraduates had gathered data for only a few hours at the end of the summer in 1969, just after the publication of our crucial experiment paper (Wenner, Wells, and Johnson 1969). It was immediately apparent to us that the results they had obtained were completely at variance with the expectations of the dance language hypothesis (see chapters 1 and 13). At the same time their results supported our interpetation (see excursus NEG).

One of us (Wenner) submitted a letter to Science that stressed the fact that their results, in fact, supported our position. In the cover letter to that manuscript Wenner included the following comments:

After carefully studying that article, I have concluded that it is not really a challenge of our work but a substantiation of our earlier findings. Unfortunately, this aspect of their study is clouded by the investigators efforts to discredit our earlier work and by their conclusion which does not necessarily follow from their data [see our chapter 12].

I think that your readers should have an opportunity to read this divergent opinion. Toward this end I have written the enclosed comment, "A Divergence from the Expected," and hope you will publish it. I have kept my statement short and in the nature of opinion so that it can appear as a Letter in your journal.

Wenner also included a list of names of reputable people who could serve as referees, as follows: Bernard Abbott, Kenneth Armitage, Vincent Dethier, Jerry Downhower, W. George Evans, and Edward Glassman. We do not know if any of them were used.

The manuscript was rejected, and comments from only a single referee were appended, with the last portion of those comments removed. Furthermore, the number 2 circled at the top of the page (in the manner of Science editors in those days) indicated that the comments of referee number 1 were not furnished to us. We reproduce that one referee's comments verbatim:

Dr. Wenner should read the paper he is criticizing! The average time spent by successful recruit bees was 3 to 4 minutes as clearly stated in the second paragraph on p. 551 of Gould et al and in their Table 2. It can only be perversity which led Wenner to utilize, instead, the data in Table 7, which contains (clearly stated!) data on delay between dance attendance and arrival at a feeding site. To repeat, Gould et al.clearly show the successful recruit takes 3 to 4 minutes to fly to the feeding site not "40 times the usual flight time."

Contrary to Dr. Wenner's incredible "interpretation" of Gould et al. all 225 recruits arrived at a station. Only 37 were marked, the 188 not listed in Table 4 were unmarked recruits, not unsuccessful recruits! Thus, we have 25 marked bees that arrived at the "correct" station and 12 that arrived at a station +/- 180 degrees from the correct direction. No marked recruits arrived at stations +/- 90 degrees from the correct direction. This may be an interesting comment on bee language. I have performed bee-language experiments with my animal behavior classes and we have occasionally seen: (a) foragers which indicated both the training direction and the opposite direction in a single bout of dancing, (b) foragers which indicated only the opposite to the training direction [sic]. In both of these cases, it has been my impression that these errors were the result of interference in the dances by too many attendants to the dance. However, occasional bees may be misled by such errors, or as some of our data also suggest, there may be some ambiguity or confusion in the dance which can lead recruits to seek food in a direction 180 degrees from the dance direction. If the editors of Science would like such a comment on the excellent paper by Gould et al. I should be happy to write one. (unpublished comments by an anonymous referee; emphasis his)


When Wenner showed the above comments to Larry Friesen, Friesen replied, "But virtually everything the referee wrote is wrong in point of fact!" We then sat down and listed some of the errors in the referee's comments as follows:

1. The "3 to 4 minutes as clearly stated" applied to forager bees, not recruits.

2. There are no data on flight times in table 2.

3. There is no table 7 in the Gould, Henerey, and MacLeod paper.

4. From page 550 one can conclude that 225 recruits arrived, but on page 552 Gould and co-workers clearly stated: "Only 37 (that is, 13 percent) of these 277 attendants were successfully recruited and later caught at a feeding station. . . ."

5. There were no stations at locations 90 degrees from the correct direction.

6. The original dance language hypothesis does not permit recruits to arrive at a station located 180 degrees from the "correct" direction.

At that point Friesen insisted to Wenner, "if you point out the errors made by the referee, they will have to publish your letter." Wenner replied, "Larry, you don't know what's going on, do you?"

Nevertheless, Wenner wrote a detailed comparison of the comments made by the referee to the actual facts in the case, revised the manuscript slightly, and resubmitted it with a cover letter. Some of the points made in the cover letter are as follows:

Normally I would not appeal such a decision on your part (This is the third consecutive rejection of such a manuscript by Science [on this issue]); but in this latest case I feel the referee has done both Science and me a disservice. The referee has made some serious errors in fact, a point which will undoubtedly make a difference to you as editor. (Attached herewith please find documentation for this claim.)

Even if the referee had been correct, my opinion should be printed in Science. In this regard I have always admired Science's stated policy:

"Science serves its readers as a forum for the presentation and discussion of important issues related to the advancement of science, including the presentation of minority or conflicting points of view, rather than by publishing only material on which a consensus has been reached."

I would hope that this particular controversy does not fall outside that general policy. That is, if Science can print a 10-page challenge of our work, surely space can be provided for a few paragraphs of divergent opinion. Even if my facts had been wrong (which they are not), I do not see why the referee should object to an airing of my opinion. If I had been as incorrect as he implied, supporters of the language hypothesis should be pleased to see such gross errors on my part in print.

On the positive side, the referee's comments did help reveal two minor errors in my manuscript (neither of which he pinpointed and neither of which makes any difference in my argument). For purposes of clarification, however, I have modified the manuscript slightly. The revised manuscript is enclosed.

I trust you will pass favorably on my manuscript this time around. Thank you for considering the topic once again.

We provide here the entire manuscript, as submitted the second time to Science.

A Divergence from the Expected

A recent article in Science, "Communication of direction by the honey bee," is of more than incidental interest, as it contains an extensive amount of data not available earlier. However, some important aspects of their data do not fall into line with what one might expect from the classic honey bee "dance language" hypothesis. The series 1 experiments were especially pertinent in this regard, since they were: ". . . designed to examine the behavior of individual recruits as each attended a dance and subsequently arrived at a feeding station."

The first item of interest is that approximately a third of those marked recruits which succeeded in finding a station arrived at one in a direction opposite to that "indicated" in the dance maneuver. This result clearly contradicts the expectations of the classic hypothesis [von Frisch 1947, 1950; Wenner, Wells, and Johnson 1969]. All 37 of the successful bees should have arrived at the "correct" station.

In their study these investigators also found that successful recruit bees spend a considerable amount of time in flight, on the average, before reaching a food source. The direct line flight time between hive and feeding stations in their experiments would normally be only about 24 sec [Wenner 1963]; and recruited bees generally fly from the hive within a minute and a half after leaving a dancing bee (50% leave within 30 sec) [Wenner 1963]. With these facts in mind, an examination of the data in Table 4 of the Gould, et al. paper reveals that the 25 bees which arrived at the "correct" station flew an average of about 30 times longer than necessary if they were to "fly directly out" to the food source. (Interestingly enough, the 12 bees which ended up at a station in the opposite direction averaged only 36 times as long as necessary.)

Clearly the above results do not match the expectation that ". . . recruits alarmed by these dancers . . . find the feeding place with surprising speed and precision" [von Frisch 1967].

Another interesting point is that Gould and co-workers have confirmed earlier findings of my co-workers and myself [Johnson and Wenner 1970; Wenner, Wells, and Rohlf 1967]; that is, that bees which leave the hive are not likely to find a food source, even under the best of circumstances [Note: Some confusion results here because Gould and co-workers apparently define "recruits" as those bees which succeed in reaching a station. I prefer to consider recruits as those bees which leave the hive after contact with a successful forager]. [For example, they stated]: "Only 37 of 277 attendants were successfully recruited and later caught at a feeding station, even though the high molarity of the sucrose used reportedly produces maximum dancing and recruitment.

If one works in unscented localities, rather than in the highly pungent area such as that chosen by the authors, the recruitment efficiency becomes even worse. As we reported earlier in Science, while working with unscented sucrose in a relatively odor-free locality, ". . . in the absence of a major nectar source for the colony, we received only five recruits from a hive of approximately 60,000 bees after ten bees had foraged at each of four stations for a total of 1374 round trips during a 3-hour period" [Wenner, Wells, and Johnson 1969]. This result is especially interesting since we have subsequently found that ever smaller amounts of odor in the food and in the locality result in an ever higher frequency of dancing, all other factors being equal [unpublished data, available upon request].

In summary, these workers have obtained results which generally agree with what we have published earlier and which differ markedly from those obtained earlier by von Frisch and co-workers or with those expected on the basis of the classic language hypothesis [von Frisch 1967]. And, in contradistinction to the authors' conclusion, the results are generally in excellent agreement with what we would expect under the circumstances. The final distribution of successful recruits does differ significantly from what one might obtain if these animals had flown at random about the countryside, it is true, but there is no compelling reason for assuming that the dance language hypothesis is a necessary explanation for this divergence from randomness [Wenner, Wells, and Johnson 1969]. I feel that a slight difference in composite location odor near the experimental sites, perceptible to bees but not to investigators, could have been responsible [Johnson and Wenner 1970].

This second time around, the manuscript was again rejected. The comments of only one referee were enclosed, as follows:

As one of the referees who urged Science to publish earlier research reports by Dr. Wenner and his associates, I now recommend that you do not publish this particular critique of the article by Gould, Henery, and MacLeod. Enough is enough! His critique, the referee's criticism of the critique, and his countercritique of the referee's criticism, have become labyrinthine. It is conceivable that the Gould et al. paper should not have been published; however, you would require a jury discussing all this material for hours to decide. I no longer believe it is worth the effort, because Dr. Wenner's own criticisms of the von Frisch explanation are mostly beside the point, and they consistently fail to mention a growing body of positive evidence strongly favoring the von Frisch explanation. To publish an increasingly complex and ambiguous debate on but one aspect of negative evidence is an inefficient - and misleading - use of space in Science. If and when Dr. Wenner comes up with solid evidence to support his views, he will find an interested and sympathetic audience waiting.

(Note in the above referee comments the reliance on the verification approach.)


Five years after the Gould, Henerey, and MacLeod lead article, Science published an eight-page lead article by Gould, based upon work he did for his doctoral dissertation (see our chapter 12). Again, this article was a challenge of our work. Once again, Wenner "tested the water" to see if Science was by now receptive to the publication of a rebuttal of the paper by Gould (actually a clarification). The letter of inquiry included the following statements:

Anyone well-versed in the work of my colleagues and myself who carefully reads the Gould article will certainly recognize a convergence of viewpoints, with Gould now coming very close to our position in the bee language controversy. Unfortunately, he does not mention the existence of that convergence, and those not overly familiar with our work will likely miss that important point.

As evidence of the foregoing, Wenner provided some appendices to the letter, documentation that juxtaposed quotations from our earlier work and quotations from the Gould paper so that the editor could compare the statements. Wenner continued:

Dr. Wells and I would like to know whether Science might publish something similar to that which is enclosed. While it is true that there is a normal process which should be followed (submitting a manuscript), our recent attempts at getting material into print in Science have failed because of an intensely hostile peer review. . . . Dr. Wells and I eventually got our message into Nature ([1973] 241:171-175), but only after a considerable delay. Ironically, Gould has now gotten that same message into print in Science - See the two starred items in the enclosed Appendix A.

Gould's publication of the same train of thought as we have expressed earlier would indicate that the climate has now changed atScience. Is this true? Do our comments now have a chance of getting into print? We realize that space is at a premium and would be happy to have only the cover statement considered [two manuscript pages], provided some reference would be made to the availability of the two appendixes for those who might wish to obtain them directly from us.

The cover statement included two quotations from the Gould paper, as follows:

Excepting the experiments reported here, the locale-odor hypothesis can effectively account for all the results achieved to date [including those of von Frisch] without recourse to the dance-language theory. . . . recruitment to odors alone might be the usual system in honey bee colonies not under stress. (1975b:686, 691)

The same assistant editor again rebuffed our attempt, and wrote:

When you have an experimental paper we would be glad to consider it. However, we would not be interested in publishing your cover statement, with or without the appendices. . . . As for the circumstances that prevail here, bee classes [sic] are not among the controversies on which we have a position.

In the years that followed, however, Science continued to publish papers supportive of the dance language hypothesis and peripheral research areas, but one finds no research papers reporting results that do not fit within the dance language paradigm.