Robert L Vadas, Jr., Ecosystem Research, Canadian Wildlife Service, Pacific & Yukon Region, P.O. Box 340, Delta, BC, Canada V4K 3Y3

"Variety is charming and not at all alarming." (Old Essex Song)

"Relax! Were only between paradigm shifts anyhow." (Crutcher 1991, p. 213)

"Lack of progress in science is never so much due to any scarcity of factual information as it is to the fixed mindsets of scientists themselves." (Schram 1992, p. 357)

"Let them test their findings doubly and trebly before they regard any interpretation as certain. For often nature reaches her goal by another path, where man cannot see the way." (von Frisch 1971, p. 90)

"Without well-conceived, critical experimentation, many evolutionary theories are little more than "Just-So" stories... The worry, then, is... that we will fail to imagine the full range of possible answers-hypotheses to test, or be unable to dream up clever ways to perform critical experiments (Gould and Gould 1988, pp. 123 and 225)

"But are there not more than enough bee books?... the unscientific reader will find it hard to tell where the observation ends, and the poetic fancy begins... I have been careful not to embroider imaginatively on the facts, which are poetic enough in themselves... Anyone who wants facts about the life of bees and not the picture of them painted by the creative imagination can look at the textbooks." (von Frisch 1953, iii-iv)

Overview of the honey-bee controversy

The history of behavioral research on honey bees (Apis spp.) has been most interesting from sociological and philosophical perspectives (Wenner 1971, 1989. Wenner and Wells 1987, 1990. Veldink 1989). Of particular controversy was whether recruited honey bees found food (nectar and pollen), water, propolis (tree-bud resin), and new nests by sensory processes alone (odor-search hypothesis), or by using dance-language information provided by scout bees that had already visited these resources (Lindauer 1971, Seeley 1985, Gould and Gould 1988).

Although honey-bee foraging was examined by Aristotle and other ancient scientists, intensive study did not begin until the 20th century (von Frisch 1953, 1971, Lindauer 1971, Seeley 1985, Gould and Gould 1988. Wenner and Wells 1990). Early researchers, and later K. von Frisch, favored the odor-search hypothesis, as honey bees were shown to use their antennae to obtain olfactory cues during nectar and pollen seaching. Soon, however, some researchers hypothesized that honey bees also communicated in the hive with their fellow workers, particularly because recruited bees often reached food sources without having to follow flying scout bees. Communication apparently occurred via two dance patterns; whereas a simple 'round dance' was apparently adequate to recruit other worker bees to food sources near the hive, a 'wagging dance' was apparently necessary to give distance and direction information for farther foods (von Frisch 1953, 1971, Lindauer 1971). By midcentury, von Frisch and colleagues gathered what seemed to be convincing evidence in favor of the dance-language hypothesis for bee foraging, de-emphasizing earlier results on the importance of odor cues (von Frisch 1953, 1971, von Frisch and Lindauer 1956, Lindauer 1971, Wenner and Wells 1990). Von Frisch hypothesized that honey bees had more complex communication systems than did most birds and mammals, a counterintuitive idea that was at first shunned by the scientitic community, but later accepted by other bee researchers and biologists in general (Seeley 1985, Gould and Gould 1988, Wenner and Wells 1990). The honey bees were hypothesized to be intelligent creatures with photographic memories, space-age sensory and navigation systems, and possibly even insight skills, and von Frisch won a Nobel Prize for his findings (Lindauer 1971, Seeley 1985, Gould and Gould 1988, Wenner and Wells 1990). Contemporary bee-dance researchers such as J. L. Gould (Gould and Gould 1988, Gould 1992) and Seeley (1983, 1985, 1989, 1991) express little doubt that the dance language is an evolutionary adaptation for facilitating the foraging of bees in superorganismal fashion. Theoretical studies of honey-bee foraging behavior have thus been initiated by Seeley and colleagues (Camazine and Sneyd 1991, Bartholdi et al, 1993) to demonstrate that dances, competition, and optimal-foraging behavior among individual bees promote optimal foraging efficiency for the whole colony, via attainment of ideal-free distributions.

When A. M. Wenner and colleagues first began their honey-bee research in the early 1960's, they accepted von Frisch's dance-language hypothesis (Wenner et al. 1967, Wenner 1971, Wenner and Wells 1990). But as they performed experiments, they discovered several discrepancies with the prevailing paradigm (see below) and revived the odor-search hypothesis. Their criticism of the various experimental designs that had been used by von Frisch and other dance-language proponents highlighted the shortcomings of these tests and served to reinvigorate honey-bee research without dismantling the dance-language hypothesis (Lindauer 1971, Gould and Gould 1988, Wenner and Wells 1990).

To their surprise, Wenner and colleagues' evidence supporting the odor-search hypothesis and criticisms of the dance-language hypothesis were at best ignored, at worst chastized and censored (Veldink 1989, Wenner and Wells 1990). To some extent, it became a 'biopolitical' battle of the personalities, with von Frisch and Wenner being viewed as the gracious hero and paranoid villain, respectively (Veldink 1989). Many scientists assumed that Wenner's experiments were flawed, merely because they disagreed with von Frisch's, even though von Frisch did not present his experimental methods until several years after he did his research (making it difficult for Wenner and other researchers to replicate these experiments). Hence, later researchers often committed the 'appeal to authority' and assumed that von Frisch's dance-language hypothesis was fact, rather than viewing the data objectively and logically or performing their own experiments. The dance-language hypothesis became a paradigm through repetition in textbooks on animal behavior and sociobiology (e.g., Evans 1968) and honey-bee biology (Seeley 1985, Gould and Gould 1988).

The antagonism against Wenner impelled him to write a short book on the philosophical implications of the honey-bee controversy (Wenner 1971), and he switched his research field to marine invertebrates (Veldink 1989). But he did not give up the fight, expanding his philosophical perspective on honey-bee research to other biological topics and to scientific methodology in general (Wenner 1989, Wenner and Wells 1990). Wenner emphasized that scientists often were not objective, despite popular, stereotypic perception, and that several decades may often be required for biologists to shift away from nonviable hypotheses. That is, honey-bee researchers and other scientists were often susceptible to problems of 'paradigm hold' and 'logical gap' that Kuhn and Polanyi had earlier recognized; anyone who did not accept the prevailing paradigm was ostracized. Interestingly, Wenner and Wells' (1990) book has been vehemently criticized by bee-dance researchers (Seeley 1991, Gould 1992), but other scientists have heralded the book as an important advancement in science that should be required reading for biologists in general (Slobodkin 1991, Massaro 1992, Rojot 1992, Schram 1992).


Interpretation of the honey-bee controversy

What went wrong in the progress of honey-bee behavioral research? First, there were problems with philosophical approach (Wenner 1971, Wenner and Wells 1990). Namely, (a) dance-language proponents did not formulate or adequately test alternative working hypotheses, particularly the odor-search hypothesis; and (b) their experimental tests were designed to verify the dance-language hypothesis, rather than to falsify it in Popperian fashion via 'crucial', unambiguous experiments (sensu Platt 1964).

Dance-language proponents did not give the odor-search hypothesis fair testing because they did not formulate realistic experimental expectations for this hypothesis (Veldink 1989, Wenner and Wells 1990, Rosin 1992). Namely, von Frisch concluded that honey bees did not use olfaction during foraging, because the bees did not need to open their abdominal Nasanov (scent) glands to stimulate other bees to forage nearby. However, this test showed only that the Nasanov gland did not recruit other foragers, and does not falsify the food-odor-search hypothesis. Rather, the function of the Nasanov gland is apparently to settle and orient swarming bees (during migrations to new hives) and dislocated bees of the colony, and not to strengthen food-site odors (Wenner 1992, Wells et al. 1993).

Typically, dance-language proponents used the verificatory mode in designing their experimental tests to demonstrate dance communication (Wenner 1971, Wenner and Wells 1990). Only Wenner and colleagues performed 'crucial' falsification tests, although some other researchers came close to performing decisive experiments. Namely, Wenner had noticed that extraneous odors, such as cut grass or researchers' body lotions, affected the foraging movements of bees. These observations led him to develop a 'crucial' experiment to determine if worker bees could be recruited away from the experimental station if it was no longer scented, unlike the control stations. The answer was "yes"; recruited bees used odor cues, and not dance-language cues, to locate the food at the control stations.

In response to these two sets of experimental problems, Wenner and Wells (Wenner 1989, Wenner and Wells 1990, Wells 1991) advocated that a composite philosophical approach was needed, as advocated by such researchers as Platt, Chamberlin, and Popper later in his life. Neither falsificatory tests (early Popper approach) nor verificatory tests were adequate by themselves to advance honey-bee and other scientific research. Whereas the verificatory approach biased findings in favor of current hypotheses because scientists merely searched for confirming evidence, the falsificatory approach sometimes led to the opposite problem, because scientists merely searched for negative evidence and were thus throwing away potentially useful hypotheses. In statistical terminology (sensu Peters 1991), verificationists and falsicationists commit too many type-I and type-II errors, respectively, if the 'alternate' hypothesis is considered to be the 'pet' theory. Instead, a healthy mix of verification and falsification studies, as well as multiple working hypotheses, is needed; honey-bee hypotheses should have been efficiently tested with 'crucial' experiments and refined so that further tests and progress could have been made (cf. Platt 1964). This 'strong-inference' approach is more flexible, multifaceted (pluralistic), and powerful for advancing scientific knowledge, and is more than a simple falsificatory approach (Wenner and Wells 1990).

The second problem was that null (random) mathematical models were not used by dance-language proponents in their experimental tests (Wenner and Wells 1990, Wenner et al. 1991), even though Wenner (1971) began using them in the late 1960s to criticize von Frisch's 'single-control fan' experiments. Namely, von Frisch and other researchers placed several food stations in a semicircle on the upwind side of the hive, the central ('experimental') station being the one that scout bees were exposed to first (the other food cues were the 'controls'). Hence, one might expect most bees to go to the experimental station merely because it was the center of the odor source, rather than being directed there by dancing bees, as schematized by Wenner's mathematical model. These experiments also highlight the problem of nonrandom, biased placement of treatments and controls, similar to von Frisch's early, mistaken hypothesis that sugar (nectar) and pollen foraging respectively elicit round versus waggle dances, based upon experiments with sugar dishes placed only near the hive (Gould and Gould 1988).

Wenner expanded his use of null models to test honey-bee hypotheses when he reviewed several studies showing that most bees search for food, water, and new hives close to their home hive (Wenner et al, 1991). That is, plots of the number of bees versus distance from the hive for a given colony usually gave a lognormal distribution (skewed right) regardless of where experimenters placed food cues. This is consistent with the odor-search but not the dance-language hypothesis, because the latter would predict different distributions depending upon where food was located. Wenner et al. (1991) suggested that the lognormal distribution may result from variation among individual bees and environmental variation during the expanding spiral (circle-out) searches of individual bees.

Third, von Frisch and other verificationists often ignored anomalous behaviors by honey bees, i.e., those that did not support the dance-language hypothesis and Nasanov-gland function in foraging (Veldink 1989, Wenner and Wells 1990, Rosin 1991, Wells et al. 1993). Most importantly, von Frisch disregarded his early evidence in favor of the odor-search hypothesis when he championed the dance-language hypothesis (Wenner and Wells 1990). That is, he had previously noted that worker bees follow the dancer around with their antennae touching the dancer's body, and that flower scents adhere well to the honey bee's body and may linger for several hours (Lindauer 1971, Seeley 1985, Gould and Gould 1988).

Moreover, when Wenner repeated von Frisch's experiment to tilt honey-bee nests in order to distort the directional information of dances, the recruited bees still found food and apparently ignored the still-accurate distance information provided by the dancers: the recruited bees did not scatter in all directions as von Frisch had found (Gould and Gould 1988). Gould and Gould (1988) interpreted Wenner's disparate results as an artifact of overly concentrated food solutions, which supposedly prevented bees from dancing (cf. Gould 1992). But such an interpretation was based on von Frisch's disparaging comments of Wenner's experiments, and not on empirical evidence per se (Gould and Gould 1988). Gould and Gould (1988) used a similar argument when Wenner redid J. L. Gould's experiments with stronger food odors and found that recruited bees flew to food, rather than to the bare spot that scout bees were dancing for (in contrast to Gould's results).

Furthermore, behavioral observations showed that recruited bees did not fly in 'beelines' to the food source as von Frisch had claimed; their routes were more circuitous (Seeley 1985, Gould and Gould 1988) and suggestive of search by odor, as in many other insects (Wenner and Wells, Rosin 1991). That is, bees typically flew down-wind from the hive, then zigzagged upwind towards food resources. Moreover, recruited bees that flew the correct distance (or direction) provided by dance information usually did not fly in the correct direction (or distance). Rather, most of the supporting research for the dance-language hypothesis was for the minority of bees that were successful in finding the correct feeding locations. Although Seeley (1985) and Gould and Gould (1988) still favored the dance-language hypothesis because even partial success might be worth something in a large bee colony, Wenner and Wells (1990) correctly criticized such data analyses for being biased in (statistically) unacceptable ways.

Fourth, dance-language proponents' experiments were often confounded by lack of adequate controls (Wenner 1971, Wenner and Wells 1990). Recruited bees were often experienced foragers, i.e., aware of food and odor locations before being exposed to dances by scout bees. This would enhance the ability of recruited bees to find food sources even if they did not get any distance or direction information from the dances. As collectively emphasized by various bee researchers (von Frisch 1953, Lindauer 1971, Seeley 1985, Gould and Gould 1988), honey bees can memorize the odors, shapes, colors, opening times, and access points of flowers for their entire foraging careers.

Moreover, the "control" food sources (dishes) set out in single-control experiments were not adequate controls because they had not been visited (conditioned) by any bees, in contrast to the experimental station, which was visited by scout bees before their dances at the hive (Wenner 1971, Wenner and Wells 1990). Hence, possible odor cues left by scout bees needed to be at all feeding stations, which Wenner accomplished via 'double-control' (sham) experiments; the control dishes were allowed to be colonized by bees from other colonies. This increased the proportion of recruited bees going to the control stations after dances, evidence for the odor-search hypothesis.

Fifth, teleological reasoning and adaptive stories were inappropriately used to verify dance communication (Veldink 1989, Wenner and Wells 1990). Forager bees' waggle dances did contain distance and/or direction information to food sources, albeit imprecise information, but this by itself does not prove that dancing or recruit bees are aware of this information. Rather, Wenner and colleagues (Wenner et al. 1967, Wenner and Wells 1990) found that sound production by dancing bees was better correlated with food distance than were dance movements, and Gould and Gould (1988) emphasized that sound production was necessary in dark hives to allow worker bees to follow dancing scouts. Nevertheless, Wenner et al. (1967) found that sound production was better correlated to temperature fluctuations than food distance. Similarly, the frequency of firefly flashes and cricket chirping increase with ambient temperature, and correlations between behavior and environmental condition are common in other insect species not considered to communicate with each other (Gould and Gould 1988, Massaro 1992).

Although von Frisch and other dance-language proponents assumed that bees would not waste energy dancing unless they were communicating, these researchers failed to consider alternative hypotheses (Jamieson 1986, Wenner and Wells 1990, Rosin 1992). For example, (a) the forager bees may dance as a reaction to harassment from other worker bees crowding around them (Rosin 1992); (b) they may dance to allow other bees to smell their bodies and learn the 'hot' odors (Wenner and Wells 1990); (c) the individual components of the dance may be adaptive, rather than the dance as an aggregate (Rosin 1992); and/or (d) the dance may reflect excitement, as honey bees are known to dance on newly found hives (Wenner 1992) and scouts of the more primitive stingless bees successfully recruit workers despite their undirected hive dances (von Frisch 1953, Lindauer 1971).

Sixth, many of these adaptive arguments were not parsimonious explanations of bee behavior. (1) Von Frisch (1953) himself noted that although honey bees showed retentive memory for colors and times that provided best foraging conditions, bee behavior was mostly stereotypic as a result of the bee's small, primitive brain (cf. Wenner and Wells 1990). That is, the ability to communicate via dances probably would require extensive sensory integration (Lindauer 1971, Seeley 1985, Gould and Gould 1988). (2) Waggle dances took place on both vertical and horizonal surfaces, and in the dark and sunlight (von Frisch 1953, Lindauer 1971), such that worker bees would have to use their antennae (for dancer vibrations), proprioreceptors, and/or eyes, depending upon the circumstances, to translate the information in the dance (cf. Seeley 1985, Gould and Gould 1988). As worker bees are already known to touch the dancer with their antennae (von Frisch 1953, Lindauer 1971), it is more parsimonious to assume that tactile and olfactory cues alone are the information communicated. (3) It is unlikely that bee dances contain distance and directional information to new hives, because the scout bees fly out in circles around the hive (Wenner 1992). That is, the irregular flight paths of scouts may require sensory innovations much too complicated for translating this information to distance and directional information as the crow flies. (4) The bias of the recruited bees for the experimental station in single-control experiments was more than would be expected by the dance-language hypothesis, based upon the inaccuracy of bee dances for providing distance and direction information (Wenner 1971, Wenner and Wells 1990). Lindauer (1971) hypothesized (ad hoc) that worker bees averaged the information from several dances to achieve this superior accuracy, but this is neither parsimonious nor consistent with Wenner's double-control experiments (see above).

Acceptance of the unparsimonious assumption that honey bees had extraordinary sensory and communication skills relative to other terrestrial insects and vertebrates may reflect bias by dance-language proponents, in their admiration for honey bees (Veldink 1989, Wenner and Wells 1990, Rosin 1992). Indeed, Gould (Gould and Gould 1988, Gould 1992) admitted that the odor-search hypothesis may explain much of the foraging behavior of honey bees, but pleaded that dance communication not be totally ignored as a possibility. However, use of parsimony (Occam's razor) would eliminate the dance-language hypothesis for explaining bee behavior, particularly because (a) bee brains probably lack the infrastructure for higher-level sensation and communication, and (b) the odor-search hypothesis provides reasonable predictive power by itself.

And seventh, dance-language proponents used too many ad hoc explanations to explain away discrepant results, further intensifying the overuse of adaptationist, "just-so" arguments. For example, the dances of temperate honey bees (Apis mellifera) were known to give less accurate distance and direction information than dances of tropical (African) species (Gould and Gould 1988). But instead of downplaying the importance of A. mellifera dances in foraging, Gould and Gould (1988) suggested that such inaccuracy might even be adaptive. Namely, sloppy dances might spread recruited bees out across more flowers within a patch, given that patches should be larger and persist longer in the temperate than tropical zone. But swarm dances were just as inaccurate as food dances in A. mellifera, even though new nest sites were small in size unlike patches of flowers (Gould and Gould 1988). Again, instead of rejecting their adaptive story, Gould and Gould (1988) suggested that bees attended more swarm dances to better estimate where new nest sites were (via the above-discussed averaging hypothesis of Lindauer).

Another example of ad hoc explanations concerned the inability of honey bees to recruit to food sources on boats in a lake, despite extensive dances for the food by scout bees (Gould and Gould 1988). Only when the boat was moved closer to shore did bees find the food, such that Gould and Gould (1988) considered honey bees to be intelligent enough to "know" that scout bees dancing for sites in the middle of the lake were mistaken and to be ignored. Clearly, dance-language proponents indiscreetly used ad hoc explanations to stack assumption upon assumption and thus preserve von Frisch's communication hypothesis.

In sum, Wenner (Wenner 1971, Wenner and Wells 1990) spearheaded a useful critique of a paradigm that had not been adequately tested, as such dance-language proponents as Gould and Gould (1988) were willing to admit. But the latter researchers still asserted that Wenner was mistaken about the irrelevancy of honey-bee dances for recruiting bees to food and other resources. As emphasized by Gould (Gould and Gould 1988, Gould 1992), Massaro (1992), and even Wenner and Wells (Wenner and Wells 1990, Wells 1991), different sensory processes may be used in different situations to find resources. Even if the odor-search hypothesis does apply most of the time, particularly when food odors are strong, it cannot be concluded that the dance is never important for recruiting bees. The point is, scientists cannot definitively prove or disprove an hypothesis with limited data, as I will further discuss below.

What bee researchers have not done, however, is to determine under what circumstances dances might be important for recruiting bees to food and other resources. While this approach is verificatory (ad hoc) rather than strict Popperian falsification, it would prevent an unceremonious discarding of an hypothesis that still may be of some use. But this means that dance-language proponents should stop attacking Wenner (e.g., Lindauer 1971, Gould and Gould 1988, Seeley 1991, Gould 1992) and begin designing tests to pinpoint the spatiotemporal scale(s) that dancing is useful as a communication tool. Until then, commercial bee keepers will continue to ignore the "ivory-tower" findings of dance-language proponents, because the research programme has not provided a predictive framework (sensu Peters 1991) that is useful for managing bees (Wenner and Wells 1990, Schram 1992).


Scientific problems in the ecological and behavioral sciences

Clearly, the honey-bee controversy brings to light several important guidelines for biologists and other scientists to make the research environment more habitable and productive. These include (1) the need for both verificatory and falsificatory approaches to balance research programmes; (2) the need for null (random) mathematical models to test alternative hypotheses; (3) examination of all relevant data, even if they contradict prevailing hypotheses; (4) the need for adequate controls and 'crucial' experiments; (5) avoidance of teleological reasoning to justify adaptive stories; (6) the need for parsimony and realism in assessing organisms' mental and physical abilities; (7) circumspect use of ad hoc explanations; and (8) avoidance of scientific censorship.

To further understand the antagonism between proponents of the dance-language and odor-search paradigms, readers will have to delve into the honey-bee literature him- or herself. But a review of other scientific literature is appropriate here, to demonstrate that the honey-bee controversy is not atypical in biology.

First, the relative importance of verificatory and falsificatory approaches has been heavily debated by population (Cousens 1985, Calow 1987, Grime 1987, Mentis 1988), community (Poore 1962, MacFayden 1975, Saarinen 1980, Salt 1983, Elner and Vadas Sr. 1990), and theoretical ecologists (Haila 1988, Taylor 1989, Scheiner et al. 1993), evolutionary biologists (Van Valen 1976, Hull 1988), environmental biologists (Romesburg 1981, Murphy 1989, Eberhardt and Thomas 1991, Nudds and Morrison 1991), social scientists and statisticians (Tukey 1969, 1980, Tweney et al. 1981, Loehle 1987, Nesselroade and Cattell 1988), and science teachers (Zielinski and Sarachine 1990, Stinner 1992). Although several of these investigators argued for one approach over the other, many of them (particularly the social scientists and science teachers) accept both verification and falsification as complementary approaches that should be used hand-in-hand for generating and testing hypotheses. That is, exploratory (empirical or modelling) analyses to search for patterns and generate hypotheses, as well as synthetic, inductive reviews that establish which patterns are general and applicable in given situations, are valid verificatory approaches. Likewise, Popperian ('hypothetico-deductive' or 'confirmatory') analyses that test whether a given pattern is present in a new data set, or that test predictions to be expected if a given ecological mechanism is operating, are valid falsificatory approaches. Descriptive and experimental scientists need not operate only in verificatory and falsificatory modes, respectively (Nudds and Morrison 1991). For example, descriptive data can be used to falsify ecological theories (Shipley and Keddy 1987, Vadas Jr. 1990a, Douglas and Matthews 1992), whereas dance-language experimentalists often worked in verificatory modes to examine honey-bee behavior (see above).

Moreover, the complexity of interactions between species and the hierarchy of factors affecting species suggests that strictly falsificatory approaches cannot be used because of the uncertainty of what biological (rather than statistical) hypotheses are being tested (Saarinen 1980, Bradley 1983, MacNally 1983, Underwood 1986, Matter and Mannan 1989). That is, correlation rather than causation is established by most descriptive and experimental studies (Wenner et al. 1967, Peters 1991). For example, in experiments in which one species depresses the abundance of another, natural-history observations would be necessary to determine if predation, competition, and/or other mechanisms was responsible for the changes in species' abundances. Similarly, the consistency of empirical data with theoretical predictions does not prove that the ecological mechanisms built into the model are actually operating (Slatkin 1983).

Wildlife biologists (Romesburg 1981, Nudds and Morrison 1991) and community ecologists (Strong 1980, Connor and Simberloff 1986) in particular have criticized biologists who infer ecological mechanisms from patterns; this improper use of induction has been distinguished by the term 'retroduction' (Romesburg 1981, Nudds and Morrison 1991). Clearly, a single verificatory or falsificatory test done by ecologists and other organismal biologists cannot definitely make or break an hypothesis as easily as in the physical sciences and molecular biology (cf. Platt 1964), because of the greater spatiotemporal variability of patterns and quantitative ambiguity of theoretical predictions for the former softer, less mature sciences (Hammond 1978, Loehle 1987, Peters 1991). That is, scientists often make inductions (inferences) without enough replication of experimental or descriptive data to properly verify or falsify hypotheses (Hacking 1981, Hurlbert 1984, Loehle 1987), particularly because a pattern falsified at one spatial, temporal, or biological scale may be a valid pattern at other scales (O'Neill et al. 1986, Murphy 1989, Romesburg 1989, Vadas Jr. 1989, Peters 1991). Clearly, not all science is or should be based upon preconceived (biased) notions and hypothetico-deductive reasoning (Brush 1974, Mentis 1988), nor should new hypotheses be disregarded simply because they have not been properly tested yet (Loehle
1987).

Second, the use of null models has become important for testing whether nonrandom ecological and evolutionary phenomena (e.g., competitive exclusion and catastrophy-induced extinctions) are plausible (Raup 1977, Strong 1980, Gould 1981, Connor and Simberloff 1986). Although many of these "null" models involve assumptions that detract from what one might expect under random conditions, they are nevertheless laudable attempts to quantitatively test whether alternative, less parsimonious models are necessary to explain patterns in nature (Peters 1991).

Third, scientific revolutions cannot occur unless researchers stop ignoring and suppressing anomalous results and alternative hypotheses (Hacking 1981, Loehle 1987, Crutcher 1991, Peters 1991). Unfortunately, biologists and physical scientists often lack proper objectivity when undertaking and writing up research, because of their preconceived notions of how the world works (Brush 1974, Jackson and Prados 1983, Cole 1985, Erman and Pister 1989, Keddy 1990, 1991, Underwood 1990), although fraudulence itself is rather rare in science (Hull 1988). Admittedly, it is self-defeating to throw away theories because of occasional anomalies, because such anomalies could be the result of experimental error or be restricted to specific conditions (Ballard and Sparberg 1962, Poore 1962, Loehle 1987), and successful biological theories are often built from older, outdated ones (Hacking 1981, Stamp 1992). Nevertheless, models should not be mindlessly bandaged with ad hoc explanations when they clearly have little predictive power (Hacking 1981, Keddy 1991, Peters 1991). Scientific judgment and statistical analyses are clearly crucial for deciding when to discard old theories for new ones (Peters 1991).

Fourth, adaptationist stories have fallen into disfavor in evolutionary biology and sociobiology because teleological reasoning and theory are not good substitutes for observational and experimental data (Gould 1977, May and Robertson 1980, Stenseth 1983, Peters 1991). That is, one cannot validly assume that behavior is adaptive to corroborate theories, although teleological reasoning can have heuristic (and verificatory) value in generating hypotheses about functional adaptations (Ruse 1989). In particular, optimal-foraging behavior at the individual and population levels is energy-expensive because of the brain capacity and psychic power required, and thus unlikely unless there are strong selection pressures for optimal behavior (Westman 1977, Schmid-Hempel 1985).

Fifth, parsimony (Occam's razor) is well-accepted as a tool in ecology and evolutionary biology, to keep hypotheses simple when greater complexity is unnecessary to explain ecological patterns and mechanisms (Holsmger 1981, Hull 1988, Peters 1991, but see Dunbar 1980). Moreover, concepts that are too complex and ambiguous to test properly should be replaced with operational, simpler theories that allow falsificatory tests to be done (Fryer 1987, Elner and Vadas Sr. 1990, Peters 1991).

And sixth, censorship of manuscripts and proposals is common in the biological and physical sciences (Barber 1961, Jaksic 1985, Peters 1991). Pielou (1981) and Wenner and Wells (1990), for example, noted that censorship of papers such as theirs took several forms, which I can corroborate based upon attempts to publish an anti-paradigm paper on competition theory (Vadas Jr. 1990b) and a "backburner" manuscript on fish ecomorphology. Namely, reviewers (or editors alone) may criticize and reject a single manuscript because it is "all" of the following: wrong and unscientific (critic #1), too controversial (critic #2), hackneyed because it merely provides common knowledge (critic #3), and boring because it addresses a dead topic that nobody cares about anymore (critic #4). Such bias is particularly apparent when a critic's comments are caustic, contradictory, or do not otherwise make sense. Unfortunately, the same scientist may be a censor in one situation and the victim of censorship in another, such that it is hard to discern the heroes and villains (Barber 1961).

Censorship takes several forms in the biological and physical sciences (Barber 1961), most of which are apparent in the honey-bee controversy. (1) 'Methodological bias' is discrimination against theoretical, mathematical, or abstract approaches to science (Barber 1961). We could add here antagonism between descriptive and experimental biologists (Diamond 1983, Hart 1983, Wilson 1989, Underwood 1990), analytical and simulation modellers (Hall and DeAngelis 1985), reductionists and holists (Dunbar 1980, Saarinen 1980, Ulanowicz 1986), pheneticists versus cladists in systematics (Hull 1988), and honey-bee researchers that use strong food odors versus those who use weak odors (see above). (2) 'Religious bias' is discrimination against such ideas as Darwinian evolution and the 'big bang' theory of cosmology (Barber 1961). We could add here societal and cultural biases, which are particularly prevalent among ecologists (Risch and Boucher 1976, Smith 1978, Keddy 1990). For example, it is quite possible that dance-language proponents and opponents (such as myself) have socialistic and agnostic biases, respectively. (3) 'Authority bias' causes scientists to accept the views of older, respected scientists, i.e., the appeal-to-authority fallacy (Barber 1961, Dayton 1979). Ancient ecological data (e.g., animal population size from pelt counts) and authoritative reviews (textbooks, review papers, and popular-science articles) are often accepted uncritically, particularly by researchers outside the immediate research field and laypeople, worsening the paradigm hold (Gilpin 1973, Forman and Russell 1983, Elner and Vadas, Sr. 1990). Veldink (1989) and Wenner and Wells (1990) clearly documented these problems in honey-bee research (see above). (4) 'Xenophobic bias' causes researchers in one research field, society, or 'school' (clique) to reject outside views (Barber 1961). This bias hurts multidisciplinary efforts and is typical of 'normal science', during which the status quo is stable against scientific revolutions (Hacking 1981, Tweney et al. 1981, Crutcher 1991, Romesburg 1991). In the honey-bee controversy, Wenner became an outsider when he started publishing negative evidence for the dance-language, causing other bee researchers to exclude him from scientific meetings and to reject his manuscripts (Veldink 1989, Wenner and Wells 1990).


Coda

To synthesize, much of the problems faced by Wenner and colleagues (Wenner 1971, 1989, Wenner and Wells 1987, 1990) in their attempts to test the dance-language hypotheses of honey bees are common in biology and science in general. Fortunately, bias and censorship are less common among scientists than among laypeople (Pease and Bull 1992), but it is ironic that scientists, particularly biologists, are often their own worst enemies when it comes to attacks upon science (Barber 1961, Rosenzweig et al 1988, Peters 1991). Nevertheless, examination of other research fields such as community ecology (but see Risch and Boucher 1976, Smith 1978, Keddy 1990), systematics, and evolution suggest that 'religious' and 'authority' biases are not always prevalent among biologists (Hull 1988). And although researchers in opposing camps may be more likely to tests opponents' hypotheses in falsificatory fashion and give unfairly negative critiques of opponents' grants and manuscripts, systematists have often been more evenhanded (objective) in their actions (Hull 1988). On the other hand, important ecological papers have sometimes been unfairly rejected by early reviewers (Cook 1977), and experimental manuscript submissions have corroborated the importance of reviewer bias against evidence contradictory to prevailing paradigms (Loehle 1987). Hence, the problems emphasized by Wenner and Wells (1990) in honey-bee research are applicable to at least some other scientific fields. Nevertheless, to avoid making inferences on limited case studies, further philosophical studies by biologists and sociologists are needed, because most science philosophers have focused their efforts on physical scientists (Warren et al. 1979, Tweney et al. 1981, Loehle 1987, Veldink 1989, Stinner 1992). Certainly, Wenner and Wells (1990) have made a good start in this direction, as have Saarinen (1980), Salt (1983), Hull (1988), and Peters (1991).

Acknowledgements - Helpful critiques of manuscript drafts were provided by P. L. Angermeier, C. A. Dolloff, R. W. Elner, W. E. Ensign, J. M. Morton, J. J. Ney, D. J. Orth, M. J. Sabo, R. L. Vadas, Sr., C. Veldink, P. H. Wells, A. M. Wenner, and Yrjo Haila, R. Vadas, Sr., P. Wells. A. Wenner also gave me several relevant reference materials. I thank the Dept of Fisheries and Wildlife Sciences (Virginia Tech) for its financial support during writing and revising of the manuscript. I also thank D. E. Gill, G. Morrison, and L. A. Nielsen for their stimulating graduate seminars on science philosophy, and the Univ. of Michigan biologist who unintentionally introduced me to the problems of censorship and dogmatism in science during 1986.


References

Ballard, R. M. and Sparberg, E. B. 1962. Resistance to discovery. - Science 135: 981-982.

Barber, B. 1961. Resistance by scientists to scientific discovery. - Science 134: 596-602.

Bartholdi, J. J. III, Seeley, T. D., Tovey, C. A. and Vande Tate, J. H. 1993. The pattern of effectiveness of forager allocation among flower patches by honey bee colonies. - J. Theor. Biol. 160: 23-40.

Bradley, R. A. 1983. Complex food webs and manipulative experiments in ecology. - Oikos 41: 150-152.

Brush, S. G. 1974. Should the history of science be rated X? - Science 183: 1164-1172.

Calow, P. 1987. Towards a definition of functional ecology. - Funct. Ecol. 1: 57-61.

Camazine, S. and Sneyd, J. 1991. A model of collective nectar source selection by honey bees: self-organization through simple rules. - J. Theor. Biol. 149: 547-571.

Cole, K. C. 1985. Is there such a thing as scientific objectivity? - Discover 6(9): 98-99.

Connor, E. F. and Simberloff, D. 1986. Competition, scientific method, and null models in ecology. - Am. Sci. 74: 155-162.

Cook, R. E. 1977. Raymond Lindeman and the trophic-dynamic concept in ecology. - Science 198: 22-26.

Cousens, R. 1985. Theory, hypothesis and experimental design in ecology. - Brit. Ecol. Soc. Bull. 16: 76-77.

Crutcher, K. A. 1991. How to succeed in science. - Perspect. Biol. Med. 34: 213-218.

Dayton, P. K. 1979. Ecology: a science and a religion. - In: Livingston, R. J. (ed.), Ecological processes in coastal and marine systems. Plenum Press, New York, pp. 3-18.

Diamond, J. M. 1983. Laboratory, field and natural experiments. - Nature 304: 586-587.

Douglas, M. E. and Matthews, W. J. 1992. Does morphology predict ecology? Hypothesis testing within a freshwater fish assemblage. - Oikos 65: 213-224.

Dunbar, M. J. 1980. The blunting of Occam's Razor, or to hell with parsimony. - Can. J. Zool. 58: 123-128.

Eberhardt, L. L. and Thomas, J. M. 1991. Designing environmental field studies. - Ecol. Monogr. 61: 53-73.

Elner, R. W. and Vadas, R. L. Sr. 1990. Inference in ecology: the sea urchin phenomenon in the northwestern Atlantic. - Am. Nat. 136: 108-125.

Erman, D. C. and Pister, E. P. 1989. Ethics and the environmental biologist. - Fisheries 14(2): 4-7.

Evans, W. F. 1968. Communication in the animal world. - Thomas Y. Crowell, New York.

Forman, R. T. T. and Russell, E. W. B. 1983. Evaluation of historical data in ecology. - Bull. Ecol. Soc. Am. 64: 5-7.

Fryer, G. 1987. Quantitative and qualitative numbers and reality in the study of living organisms. - Freshw. Biol. 17: 177-189.

Gilpin, M. E. 1973. Do hares eat lynx? - Am. Nat. 107: 727-730.

Gould, J. L. 1992. Interpreting the honeybee's dance (book review). - Am. Sci. 80: 278-279.

- and Gould, C. G. 1988. The honey bee. - Science American Library, New York.

Gould, S. J. 1977. Ever since Darwin: reflections in natural history. - W. W. Norton, New York.

- 1981. Palaeontology plus ecology as paleobiology. - In: May, R. M. (ed.), Theoretical ecology: principles and applications. 2nd ed. Blackwell, Boston, MA. pp. 295-317.

Grime, J. P. 1987. Research philosophies in functional ecology. - Funct. Ecol. 1: 71-72.

Hacking, I. (ed.) 1981. Scientific revolutions. - Oxford Univ. Press, New York.

Haila, Y. 1988. The multiple faces of ecological theory and data. - Oikos 53: 408-411.

Hall, C. A. S. and De Angelis, D. L. 1985. Models in ecology: paradigms found or paradigms lost? - Bull. Ecol. Soc. Am. 66: 339-346.

Hammond, K. R. 1978. Towards increasing competence of thought in public policy formation. - In: Hammond, K. R. (ed.), Judgment and decision in public policy formation. Westview Press, Boulder, CO. pp. 11-32.

Hart, D. D. 1983. The importance of competitive interactions within stream populations and communities. - In: Barnes, J. R. and Minshall, G. W. (eds), Stream ecology: application and testing of general ecological theory. Plenum Press, New York, pp. 99-136.

Holsinger, K. E. 1981. Comment: the blunting of Occam's Razor, or to hell with parsimony. - Can. J. Zool. 59: 144-146.

Hull, D. L. 1988. Science as a process: an evolutionary account of the social and conceptual development of science. - Univ. of Chicago Press, Chicago.

Hurlbert, S. H. 1984. Pseudoreplication and the design of ecological field experiments. - Ecol. Monogr. 54: 187-211.

Jackson, C. I. and Prados, J. W. 1983. Honor in science. - Am. Sci. 71: 462-464.

Jaksic, F. M. 1985. Controversies in community ecology and their consequences for ecological practice and funding allocation: a plea for common sense. - Revista Chilena de Historia Natural 58: 5-7.

Jamieson, I. G. 1986. The functional approach to behavior: is it useful? - Am. Nat. 127: 195-208.

Keddy, P. 1990. Is mutualism really irrelevant to ecology? - Bull. Ecol. Soc. Am. 71: 101-102.

- 1991. Forward towards the machine guns, lads! - A reply to E.I. Newman. - Brit. Ecol. Soc. Bull. 22: 15-17.

Lindauer, M. 1971. Communication among social bees. - Harvard Univ. Press, Cambridge, MA.

Loehle, C. 1987. Hypothesis testing in ecology: psychological aspects and the importance of theory maturation. - Q. Rev. Biol. 62: 397-409.

MacFayden, A. 1975. Some thoughts on the behavior of ecologists. - J. Anim. Ecol. 44: 351-363.

MacNally, R. C. 1983. On assessing the significance of interspecific competition to guild structure. - Ecology 64: 1646-1652.

Massaro, D. W. 1992. Anatomy of a controversy: the question of a "language" among bees (book review). - Am. J. Psychol. 105: 653-659.

Matter, W. J. and Mannan, R. W. 1989. More on gaining reliable knowledge: a comment. - J. Wildl. Manage. 53: 1172-1176.

May, R. M. and Robertson, M. 1980. Just so stories and cautionary tales. - Nature 286: 327-329.

Mentis, M. T. 1988. Hypothetico-deductive and inductive approaches in ecology. - Funct. Ecol. 2: 5-14.

Murphy, D. D. 1989. Conservation and confusion: wrong species, wrong scale, wrong conclusions. - Cons. Biol. 3: 82-84.

Nesselroade, J. R. and Cattell, R. B. (eds) 1988. Handbook of multivariate experimental psychology. 2nd ed. - Plenum Press, New York.

Nudds, T.D. and Morrison, M. L. (eds) 1991. Ten years after "reliable knowledge": are we gaining? - J. Wildl. Manage. 55: 757-799.

O'Neill, R. V., De Angelis, D. L., Waide, J. B. and Allen, T. F. H. 1986. A hierarchical concept of ecosystems. - Monogr. Popul. Biol. 23: 1-253.

Pease, C. M. and Bull, J. J. 1992. Is science logical? - Bio-Science 42: 293-298.

Peters, R. H. 1991. A critique for ecology. - Cambridge Univ. Press, New York.

Pielou, E. C. 1981. The broken stick model: a common misunderstanding. - Am. Nat. 117: 609-610.

Platt, J. R. 1964. Strong inference. - Science 146: 347-353.

Poore, M. E. D. 1962. The method of successive approximation in descriptive ecology. - Adv. Ecol. Res. 1: 35-68.

Raup, D. M. 1977. Probabilistic models in evolutionary paleobiology. - Am. Sci. 65: 50-57.

Risch, S. and Boucher, D. 1976. What ecologists look for. - Bull. Ecol. Soc. Am. 57(3): 8-9.

Rojot, J. 1992. Anatomy of a controversy: the question of a language among bees (book review, in French). - Revue de Gestion des Ressources Humaines 5 (Oct.): 53-54.

Romesburg, C. 1981. Wildlife science: gaining reliable knowledge. - J. Wildl. Manage. 45: 293-313.

- 1989. More on gaining reliable knowledge: a reply. - J. Wildl. Manage. 53: 1177-1180.

- 1991. On improving the natural resources and environmental sciences. - J. Wildl. Manage. 55: 744-756.

Rosenzweig, M, L., Davis, J. I. and Brown, J. H. 1988. How to write an influential review. - Bull. Ecol. Soc. Am. 69: 152-155.

Rosin, R. 1991. Much abuzz about nothing! (the honeybee "dance language"). - Am. Bee J. 131: 525-526.

- 1992. Do honeybees dance? - BioScience 42: 83-85.

Ruse, M. 1989. Teleology in biology: is it a cause for concern? - Trends Ecol. Evol, 4: 51-54.

Saarinen, E., ed. 1980. Conceptual issues in ecology. - D. Reidel, Boston, MA.

Salt, G. W. (ed.) 1983. A round table on research in ecological and evolutionary biology. - Am. Nat. 122: 583-705.

Scheiner, S. M., Hudson, A. J. and VanderMeulen, M. A. 1993. An epistemology for ecology. - Bull. Ecol. Soc. Am, 74: 17-21.

Schmid-Hempel, P. 1985. Foraging theory (book review). - J. Evol. Biol. 1: 86-88.

Schram, F. R. 1992. Anatomy of a controversy (book review). - Am. Zool. 32: 357.

Seeley, T. D. 1983. The ecology of temperate and tropical honeybee societies. - Am. Sci. 71: 264-272.

- 1985. Honey bee ecology: a study of adaptation in social life. - Princeton Univ. Press, Princeton, NJ.

- 1989. The honey bee colony as a superorganism. - Am. Sci. 77: 546-553.

- 1991. Bee warned (book review). - Nature 349: 114.

Shipley, B. and Keddy, P. A. 1987. The individualistic and community-unit concepts as falsifiable hypotheses. - Vegetatio 69: 47-55.

Slatkin, M. 1983. Models of coevolution: their use and abuse. - In: Nitecki, M. H. (ed.), Coevolution. Univ. of Chicago Press, Chicago, pp. 339-370.

Slobodkin, L. B. 1991. Skepticism in the apiary (book review). - BioScience 41: 262-263.

Smith, F. E. 1978. Episodes in ecology (book review). - Science 200: 526-527.

Stamp, N. E. 1992. Theory of plant-insect herbivore interactions on the inevitable brink of resynthesis. - Bull. Ecol. Soc. Am. 73: 28-34.

Stenseth, N. C. 1983. Grasses, grazers, mutualism and coevolution: a comment about armwaving in ecology. - Oikos 41: 152-153.

Stinner, A. 1992. Science textbooks and science teaching: from logic to evidence. - Sci. Educat. 76: 1-16.

Strong, D. R. Jr. 1980. Null hypotheses in ecology. - Synthese 43: 271-285.

Taylor, P. 1989. Revising models and generating theory. - Oikos 54: 121-126.

Tukey, J. W. 1969. Analyzing data: sanctification or detective work? - Am. Psychol. 24: 83-91.

- 1980. We need both exploratory and confirmatory. - Am. Stat. 34: 23-25.

Tweney, R. D., Doherty, M. E. and Mynatt, C. R. (eds) 1981. On scientific thinking. - Columbia Univ. Press, New York.

Underwood, A. J. 1986. The analysis of competition by field experiments. - In: Kikkawa, J. and Anderson, D. J. (eds), Community ecology: pattern and process. Blackwell, Boston, MA, pp. 240-268.

- 1990. Experiments in ecology and management: their logics, functions and interpretations. - Aust. J. Ecol. 15: 365-389.

Ulanowicz, R. E. 1986. Growth and development: ecosystem phenomenology. - Springer, New York.

Vadas, R. L. Jr. 1989. Food web patterns in ecosystems: a reply to Fretwell and Oksanen. - Oikos 56: 339-343.

- 1990a. The importance of omnivory and predator regulation of prey in freshwater fish assemblages of North Amenca. - Environ. Biol. Fish. 27: 285-302.

- 1990b. Competitive exclusion, character convergence, or optimal foraging: which should we expect? - Oikos 58: 123-128.

Van Valen, L. 1976. Domains, deduction, the predictive method, and Darwin. - Evol. Theor. 1: 231-245.

Veldink, C. 1989. The honey-bee language controversy. - Interdisc. Sci. Rev. 14: 166-175.

von Frisch, K. 1953. The dancing bees: an account of the life and senses of the honey bee. - Harcourt, Brace & World, New York (transl. from the German by D. Ilse).

- 1971. Bees: their vision, chemical senses, and language. 2nd ed. - Cornell Univ. Press, Ithaca, NY.

- and Lindauer, M. 1956. The "language" and orientation of the honey bee. - Annu. Rev. Entomol. 1: 45-58.

Warren, C. E., Allen, M. and Haefner, J. W. 1979. Conceptual frameworks and the philosophical foundations of general living systems theory. - Behav. Sci. 24: 296-310.

Wells, P. H. 1991. Forager recruitment: examining evidence. - Am. Bee J. 131: 747-749.

- , Wells, H., Vu, V., Vadehra, N., Lee, C., Han, R., Han, K. and Chang, L. 1993. Does honey bee Nasanov pheromone attract foragers? - S. Calif. Acad. Sci. Bull. 92, in press.

Wenner, A. M. 1971. The bee language controversy: an experience in science. - Biological Sciences Curriculum Study, Educational Programs Improvement Corp., Boulder, CO.

- 1989. Concept-centered versus organism-centered biology. - Am. Zool. 29: 1177-1197.

- 1992. Swarm movement: a mystery explained. - Am. Bee J. 32: 27-31.

- and Wells, P. H. 1987. The honey bee "dance language" controversy: the search for "truth" vs. the search for useful information. - Am. Bee J. 127: 130-131.

- and Wells, P. H. 1990. Anatomy of a controversy: the question of a "language" among bees. - Columbia Univ. Press, New York.

- , Wells, P. H. and Rohlf, F. J. 1967. An analysis of the waggle dance and recruitment in honey bees. - Physiol. Zool. 40: 317-344.

- , Meade, D. E. and Friesen, L. J. 1991. Recruitment, search behavior, and flight ranges of honeybees. - Am. Zool. 31: 768-782.

Westman, R. S. 1977. Environmental languages and the functional bases of animal behavior. - In: Hazlett, B. A. (ed.), Quantitative methods in the study of animal behavior. Academic Press, New York, pp. 145-201.

Wilson, E. O. 1989. The coming pluralization of biology and the stewardship of systematics. - BioScience 39: 242-245.

Zielinski, E. J. and Sarachine, D. M. 1990. Creativity and criticism. - Sci. Teacher 57(8): 18-22.