Recruitment, Search Behavior, and Flight Ranges of Honey Bees*

[1991 Wenner, A.M., D. Meade, and L. J. Friesen. Recruitment, search behavior, and flight ranges of honey bees. Amer. Zool. 31(6):768-782.]

* An Invited Review Paper for the Division of Invertebrate Zoology.

Department of Biological Sciences,
University of California, Santa Barbara,
Santa Barbara, California 93106


Department of Biological Sciences,
Santa Barbara City College,
Santa Barbara, California 93109

SYNOPSIS. During the past three decades, considerable evidence has been gathered in attempts to understand more fully honey bee recruitment to food sources. Those efforts also apply directly to two long-standing and competing recruitment hypotheses: odor search vs. “dance language” communication. However, whereas most researchers have focused on individual interactions and behavior, the colony can also be viewed as a unit. A review of evidence from a colony perspective reveals that colony members range an average distance from their home base, whether while foraging on food sources, while collecting water, or while relocating as swarms. Those averages, based on the logarithm of the distance from the colony, vary with the type of resource exploited and size of the odor field. Such a mathematical correspondence between distances travelled from parent colonies may well agree with an odor-search recruitment model, but is hardly reconcilable with the “dance language” hypothesis.


Any assessment of what might be true changes constantly, since science is a process rather than a “thing” or an accomplishment (Hull, 1988). An example of that process is the ongoing question of how honey bees (Apis mellifera) might forage and recruit hivemates to resources. During the past 45 years one opinion (“language” use) has prevailed; more recently, a second interpretation (odor use) has been regaining credibility. Few realize that, throughout recorded history, those who have studied honey bee exploitation of food sources have embraced either one or the other of those interpretations (Wenner and Wells, 1990:chap. 4).

Aristotle (~330 B.C.), Charles Butler (1609), Ernst Spitzner (1788), John Burroughs (1875), Maurice Maeterlinck (1901), Bruce Lineburg (1924), R. W. Gowland (1927), Karl von Frisch (1937), C. N. Buzzard (1946), Hans Kalmus (1960), Adrian Wenner and co-workers (1960s to the present), and R. Rosin (late 1970s to the present) felt that newly recruited honey bees behaved like many other flying insects and relied upon odor cues as they searched for food sources exploited by successful hivemates.

The work of Ring Carde and others (e.g., Kennedy, 1983; Carde, 1984; Carde and Charleton, 1984) these past 30 years have provided much input on the theory of odor search modalities in other insects, theory that can well apply to the flight behavior of newly recruited honey bees (Wenner and Wells, 1990:chap. 5).

By contrast, Thomas Wildman (1768), F. Dujardin (1852), J. Emery (1875), G. Bonnier (1906), E. R. Root (1908), Julien Francon ([1938] 1939), Karl von Frisch ([1946] 1947), Adrian Wenner (initially: e.g., 1962, 1964), James Gould (e.g., Gould et al., 1970; Gould, 1975, 1976), and Axel Michelsen et al. (1989) all worked under the assumption that a “language” among honey bees was possible and that recruitment to food sources was a “language” use phenomenon.

Recurring questions persist: Do recruited honey bees use only odor in their search for food sources exploited by successful hive-mates? Do they obtain “symbolic language” information about distance and direction of a food source from a successful forager before they leave the hive? Can they use that descriptive information about location and “fly directly out” to that same food source, as von Frisch (1947) claimed (using odor only when in the immediate vicinity of the “target” source)?

During the last 65 years, experimental results from field studies of colony foraging patterns (as against behavior of individual bees) have been accumulating, results that apply directly to the applicability of the two competing hypotheses: “language” use and odor-search. Roubik (1989) noticed an emerging pattern and provided a theoretical “search area” model to results he had at hand. We present here a different model after providing background on the two extant hypotheses.

The odor-search hypothesis

Although Aristotle was the first person known to propose the use of odor by searching recruits, Lineburg (1924) was apparently the first to invoke the use of odor search behavior to describe the flight patterns of searching recruit bees. Von Frisch ([1937] 1939) later outlined a portion of the odor-search hypothesis we eventually embraced. He wrote:

I fed some of the numbered bees of the observation hive at a feeding-place 40 feet to the west of the hive. In the meadow round the hive to the north, south, west, and east I put glass dishes with sugar water and a little honey on the ground. If the dancer bee dancing in the hive [had] reported where the feeding-place was, [then] the new bees would all fly to the west feeding-place. As a matter of fact, a few minutes after the commencement of the dance new bees appeared at the same time at all the little dishes to the north and south, to the west and east. They did not know where the food was. They flew out in all directions and looked for it. . . .

But not only in the neighborhood! In further experiments I left the feeding-dish, visited by some numbered bees, at a short distance from the hive. And I put some other dishes farther and farther away in the meadow, observing whether they would be found or not. The farther they were the longer time it took till they were found by the bees sent out by the dancer. In the last experiment [searching bees] were found after 4 hours in a meadow a full kilometer from the hive. . . . It is clear from a long series of experiments that after the commencement of the dances the [recruited] bees first seek in the neighborhood, and then go farther away, and finally search the whole flying district. . . .

I succeeded with all kinds of flowers with the exception of flowers without any scent. And so it is not difficult to find out the manner of communication. When the collecting bee alights on the scented flowers to suck up the food, the scent of the flower is taken up by its body-surface and hairs, and when it dances after homing, [then] the interested bees, following the movements of the dancer bee and holding their antennae against its body, perceive the specific scent on its body and know what kind of scent must be sought to find the good feeding-place announced by the dancing bee. That this view is correct can be proved easily. . . .” (von Frisch, [1937] 1939; emphases ours)

We found that this earlier model outlined by von Frisch meshed better with the observed behavior of searching recruits (e.g., Friesen, 1973; Rosin, 1990; Wenner and Wells, 1990) than did the dance language hypothesis proposed by von Frisch (e.g., 1947, 1967). Since the first formal presentation of our odor-search model (Wenner, 1974), based largely on research by Friesen (1973), we have improved our understanding of the behavior of recruited honey bees during our search for feral honey bee colonies on Santa Cruz Island off the coast of Santa Barbara, California (Wenner et al., 1990).

A characteristic flight pattern reported repeatedly in the literature, seen on the island during “bee hunting,” and during years of experience in commercial beekeeping has application here. Buzzard (1946, p. 166) described that orientation flight earlier “those that found no honey would circle in ever-increasing circles in an attempt to find it.” As indicated earlier, von Frisch (1939, p. 430) wrote: “Flying out in all directions, they find out in the shortest time the plant which has commenced to bloom, wherever it is in the entire flying district.” More recently, Southwick (1991, p. 227) reported: “We have observed recruits leaving the hive, and they seem to circle out in an ever increasing spiral before disappearing.”

A portion of our odor-search model can be illustrated by a simple diagram (Fig. 1), details of which can be found in Wenner and Wells (1990:chap. 5, excursus OS). In essence:

1) Newly recruited bees leave the colony and fly in an ever-expanding spiral. When they perceive odor molecules similar to those present on the dancing bee they had encountered before leaving the colony, they begin a zigzag flight upwind and continue that flight pattern as long as they encounter such odor molecules. If they miss the target and end upwind of the odor cues drifting downwind, they loop up and fly (or are carried) downwind until they again perceive target molecules, etc. The pertinent odors can be either of the food, of the locality, or those emanating from the bodies of regular foragers that fly the resultant aerial pathway between colony and food source.

2) In the case of multiple odor sources (such as a line or an arc of feeding stations), searching recruits apparently cannot recognize the fact that more than one station might exist. Instead, the population of searching bees ends up at an array of stations in a mathematically predictable pattern, provided that those stations are equally attractive.

The odor-search hypothesis gained support from some of the results gathered by Esch and Bastian (1970), by Gould et al. (1970), and by Friesen (1973). When Esch and Bastian relocated a familiar station to a new site, a few of the foragers found the new location and danced upon their return to the colony; however, only 14 of the 70 bees attending the dances of foragers found the new location. Twenty others attended dances and left the hive but failed to arrive at the transposed station. The remaining 36 former foragers did not contact the dancers.

Ten of the 14 “successful” recruits in the Each and Bastian experiments required between two and nine exploratory flights (after repeated contacts with the dances of regular foragers between flights) before they finally located the station. Only four of the 14 successful recruits located the food on the first flight, with search times of 56, 58, 90, and 360 seconds. By contrast, experienced foragers would have expended less than 30 seconds flying on their “beeline” flight paths between hive and food (Wenner, 1963).

In the Gould et al. (1970) experiments, 277 different bees left the hive after having attended 155 observed dances. Of those 277 recruits, only 37 found either of the two stations located 120 m from the hive (the fate of the other 240 recruits was undetermined). Twenty-five of the searching bees ended up at a station in the “correct” direction, but 12 of them ended up at a station in the opposite direction that had not been “indicated” in the dance maneuver.

Search times for recruited bees in those experiments were even more revealing (Fig. 2). Whereas the outgoing flight time between colony and food source for experienced foragers was less than 20 seconds (see Wenner, 1963), recruit bees searched for up to 75 minutes (median time of 8 min) for the stations, even if they had later arrived at the station in the “correct” direction, as supposedly indicated by information contained in the dance maneuver.

Friesen (1973) measured the elapsed time between hive departure and arrival at a feeding station. Maximum search times for recruits that found a downwind station were more than 24 min but slightly less than 9 min for an upwind station.

The dance language hypothesis

H. Eltringham, Fellow of the Royal Society in London, translated Julian Francon’s book, The Mind of the Bees, and wrote a preface to that English edition. Francon (1939) had proposed a sophisticated “language” hypothesis for recruitment of honey bees, and Eltringham felt that Francon had gained convincing evidence that “the original bee can, in some way unknown to us, give to the assistant bees the most precise and accurate instructions.”

Experiments by von Frisch followed shortly thereafter; he concluded on the basis of those experiments (von Frisch, 1947, p. 5): “Today, after two years of experimenting, I have come to realise that these wonderful beings can, in a manner hitherto undreamt of, give each other exact data about the source of food.” He later wrote (at various times):

“We see that the majority of searching bees fanning out, moved within an angle deviating not more than 15 degrees each to the left and to the right from the direction leading towards the feeding place” (1948, p. 10). “For almost two decades my colleagues and I have been studying one of the most remarkable systems of communication that nature has evolved. This is the ‘language’ of the bees; the dancing movements by which forager bees direct their hivemates, with great precision, to a source of food” (1962, p. 78). “This description of the location enables the newcomers to fly rapidly and with certainty to the indicated flowers, even when these are kilometers away – an accomplishment on the part of the bees that is without parallel elsewhere in the entire animal kingdom” (1967, p. 57).

In contrast to these statements by von Frisch, proponents of the dance language hypothesis today no longer seem to have a clear notion of what one should expect from that hypothesis.


Research has focused primarily on individual foragers and searching recruits, but the colony can also be studied as a unit (e.g., Wheeler’s 1923 “super-organism” concept). However, while individual bees may range far and wide, simple energetics would impose a limit on the distances at which colonies could effectively exploit resources.

An early study by Eckert (1933) reviewed previous work on colony foraging. Eckert found, as Gowland (1927) had found earlier, that colonies concentrated on food sources in certain regions near the hive and that colonies too far removed from resources steadily lost weight. Other reviews have covered work since that time (Waddington, 1983; Seeley, 1985; Winston, 1987).

Visscher and Seeley (1982) extracted distance and direction information from dance maneuvers of foragers in an observation hive and also found that their colonies concentrated on food sources in particular areas. Foraging patterns changed through time; they attributed those changes to the discovery of richer food sources and recruitment thereafter by means of “dance language” use. They did not consider other possible explanations for their results, such as changes in fragrance concentration and dispersal or changes in wind direction.

Roubik (1989) recognized that the Visscher and Seeley (1982) results obtained in New York for the distances bees foraged from their colonies bore a striking resemblance to results Vergara (1983) obtained in Panama. Roubik proposed an area exploitation model for colonies; however, his model 1) implied a mechanistic foraging pattern rather than the use of dance maneuver information and 2) did not fit either the Visscher and Seeley (1982) or the Vergara (1983) results especially well.

Wenner (1967) earlier had proposed that an expanded binomial pattern could explain the eventual arrival of recruits at equaliy attractive stations located at various distances from a colony.

Experiments with a scented computer-driven mechanical bee (Michelsen et al., 1989), provided a pattern of results similar to that obtained by Visscher and Seeley (1982) and to that obtained by Vergara (1983). Moreover, the results Michelsen et al. (1989) obtained admittedly did not mesh closely with expectations of the von Frisch dance language hypothesis.

The similarity of results obtained from studies in Wyoming, New York, California, Panama, and Europe to one another all suggest that a unifying principle exists. These various studies of colony foraging patterns, as covered below, also provide new insight into the applicability of the two competing honey bee recruitment hypotheses.


Researchers usually use one of three methods to estimate the distances honey bees travel to resources or to new home sites: 1) they interpret the information contained in forager or scout dances executed within the colony or on the surface of swarms, 2) they sample bee visitation at different distances from a colony to determine the density of foragers known to be from that colony, and/or 3) they mark bees in the field and recover them later at the parent colonies.

Accuracy of distance information in waggle dances

The straight run portion of the dance maneuver is apparently the most consistent of available measures if one wishes to determine distances travelled by foragers (Wenner, 1962; von Frisch, 1967, p. 104) and is best delineated by the sounds emitted by dancing bees during the straight run portion of their dance (Wenner, 1962:fig. 1).

However, researchers at best obtain only imprecise distance information from waggle dances. Appreciable variability occurs both for repeated runs of the dance maneuver performed by individual bees (i.e., within bee variation) and for dances executed by different bees (between bee variation) travelling to any given site. Table II and figure 5 in Wenner, 1962 provide an analysis of variance, based on the results obtained from 86 different bees and 629 straight runs performed by foragers. In those studies, between bees and within bees variance combined increased steadily with increasing distance travelled by foragers from their colony (Wenner, 1962:fig. 5b).

Distance experiments run by von Frisch

Von Frisch (1967, pp. 88-97) summarized the results from several experiments designed to test the ability of recruits to find a station located at a particular distance from the colony after they had contacted a dancing bee visiting that same station. The results in his summary permit us to address the question: How accurately did recruits in the von Frisch experiments perform compared to any distance information they could have obtained from the dance maneuver? We can also ask: 2) How well did searching recruits in the Michelsen et al. mechanical bee experiments perform compared to the performance of recruited bees in the von Frisch experiments?

By extracting information from figures 1 and 5 in Wenner (1962), one also can derive histograms illustrating between-bee variation in distance information, as contained in the dance maneuver, for any particular distance from the colony. That is, one can estimate the expected distribution of recruited bees in the field if they would be able to use distance information without error that they had extracted from dancing bees.

Figure 3 provides a triple comparison: 1) the best results that could be expected for distance recruitment on the basis of dance maneuver information (Fig. 3A), 2) results obtained by von Frisch in his experiments (Fig. 3B), and 3) results obtained by Michelsen and co-workers in their studies of recruitment with a mechanical bee (Fig. 3C). The examples chosen are for feeding stations placed at about 1 km from the colony.

The searching recruits in von Frisch’s 1,050 m experiment (Fig. 3B) performed much better than expected on the basis of dance maneuver information (Fig. 3A). Furthermore, recruited bees should have exhibited, on average, some error on their outward flight and performed even less well than indicated by the distance information present in dance maneuvers.

When von Frisch was informed about the inaccuracy of dance maneuver information, he resorted to an ad hoc modification of his hypothesis and proposed that recruit bees “averaged” the information from several dances of the attended bee before they left the hive. However, averaging the information, as he proposed, would only reduce within-bee variance, not between-bee variance (Wenner, 1962).

Von Frisch (1967) had many more foragers travelling between colony and feeding station when his feeding stations were at greater distances than when they were closer to the colony. The resultant larger aerial pathway for greater distances would provide a stronger odor trail for stations more remote from the hive. He also did not refer to his earlier mention (von Frisch, 1937) of the long time required for searching: “The farther [the stations] were the longer time it took. . . . In the last experiment they were found after 4 hours in a meadow a full kilometer from the hive. . . .”

Another factor von Frisch did not consider fully was the importance of odors during recruitment, both those provided by regular foragers and by odors peculiar to the station locality. Others (e.g., Gould, 1975, p. 686) eventually realized that von Frisch did not have adequate controls against odor artifacts.

The Michelsen and co-workers 1,000 m results

In the distance experiments (100, 250, 500, 1,000 m) run by Michelsen et al. (1989), both the within – and between – bee variances should have been negligible. That is, they should have been able to provide a rather precise signal for distance of a food source (mean of means), since they would not have to contend with the natural variation present in dances by real bees.

In three of the Michelsen et al. experiments, test stations were located at distances similar to those employed by von Frisch (1967, p. 88-97), including one experiment with a target station at about 1,000 m. In none of their mechanical bee experiments did results correspond with those obtained by von Frisch for the same distance (e.g., Fig. 3C). Almost all recruits in mechanical bee experiments ended up at stations closer to the colony than at the one presumably indicated by the programmed dance maneuver.

Michelsen and co-workers also obtained results completely at variance with the dance language hypothesis (and with von Frisch’s results for a 450 m station) when they had two real foragers visiting a station 500 m from the colony (Fig. 4A). Once again more recruits arrived at stations closer to the colony than at the station presumably indicated by forager dances.

A recurring lognormal pattern

Instead of supporting the notion that recruits use “dance language” information as they search for food sources exploited by regular foragers, the pattern of results obtained by Michelsen et al. (1989) matched the patterns Roubik (1989) relied upon (Visscher and Seeley, 1982; Vergara, 1983) to develop his area exploitation model. Actually, all three of those patterns appear to be lognormal (i.e., random) and thereby hardly consistent with the “language” hypothesis.

Figure 4B depicts as a histogram all the Michelsen et al. data for both mechanical and real bees, plotted against a theoretical lognormal distribution for the same distance. A simple test of lognormality for such results can be run. The results can be cumulated, converted to percentiles, and plotted as a function of distance on lognormal probability graph paper. One can then fit a straight line as closely as possible, percentage-wise, to all of the points so plotted (Siegel, 1956). A Kolmogorov-Smirnov test can then be applied to indicate whether the straight line fit is significantly different, statistically, from the results obtained (Siegel, 1956; Zar, 1984).

Michelsen et al. (1989) had in fact obtained a distribution for their searching recruits that did not differ statistically from a random and mathematically predictable lognormal distribution (D = 0.40; P> 0.20).

A re-interpretation of the 1967 Wenner results

Johnson (1967), Wenner (1967), and Johnson and Wenner (1970) recognized that the von Frisch distance experiments lacked an important control and that he did not demonstrate searching bees use dance maneuver information. Wenner then repeated a key von Frisch distance experiment with the original experimental protocol. With a feeding station at 400 m from the colony and with three scented control stations at 200 m, 300 m, and 500 m, recruits arrived, as in the von Frisch experiments, predominantly at the 400 m distance (Wenner and Wells, 1990:fig. 9.2).

After Wenner had provided bee visitation from another colony at two of the other scented stations as well (200 and 300 m), hereby having bee odor at three stations, the recruit arrival pattern from the experimental colony altered dramatically (Wenner, 1967:table 1, experiment 2). Most recruits from the experimental colony now arrived at the 200 and 300 m stations (as in the experiments by Michelsen et al., 1989). That result occurred despite the fact that only 400 m distance information was present in the dance maneuvers of foragers in the hive.

Wenner (1967) concluded that the distribution of recruits at the three now more nearly equivalent stations matched a binomial distribution (1:2:1). However, a lognormal fit matches the same set of results (D = 0.25, P > 0.20) better than does a binomial distribution.

We re-examined the experimental results that Wenner obtained with a feeding station at a site 500 m distant from the colony, on the possibility that the correlation observed above was spurious. When bees from the experimental colony visited only a station located 500 m from the colony, as in the von Frisch protocol, once again nearly all recruits ended up at that station and ignored a station at 350 m (Wenner and Wells, 1990:fig. 9.3).

However, when bee visitation was provided at three control stations located closer to the colony as well as at the 500 m experimental station, a very different distribution emerged for recruitment from the experimental colony (Wenner, 1967:table 1, experiment 3). Then, very few of the recruits arrived at the 500 m station presumably indicated by dance maneuvers in the experimental colony; nearly all recruits arrived instead at stations closer to the experimental colony.

At the time, Wenner had concluded that the distribution of searching recruits was similar to what one would expect from an expanded binomial distribution (1:3:3:1). However, statistical tests revealed that the results were not as close to a 1:3:3:1 ratio as they were to a lognormal fit (D = 0.20; P > 0.20).

The Cornell and Panama patterns

Consider now the two sets of patterns upon which Roubik based his area model for distance foraging.

Whereas Roubik (1989:fig. 2.24) combined the Visscher and Seeley (1982) data gathered at different times during the season, we recognized that the mid-July segment of their original data (see Fig. 5) did not differ from a lognormal pattern (D = 0.25; 0.20 > P > 0.10). The close fit to a single lognormal distribution might imply that one or at most two food sources dominated input of resources to the colony at that time.

The other results gathered by Visscher and Seeley did not form unimodal lognormal patterns but were polymodal lognormal. The data gathered in mid-June, for example, fell into four quite distinct lognormal modes (by application of techniques outlined by Cassie

The Vergara results for the foraging pattern of African bees in Panama (Roubik, 1989:fig. 2.24) were not statistically different from a lognormal distribution (D = 0.25, P > 0.20).

Gould 315 m “misdirection” experimental results

In his book on honey bee ecology, Seeley selected the results of one of Gould’s “misdirection” experiments (Seeley, 1985:fig. 7.2) to illustrate the apparent success of those experiments in directing bees to a site at a given distance from the colony. However, an analysis of Gould’s data once again yielded a distribution that did not differ significantly from a random lognormal distribution (D = 0.15; P > 0.20).

Magnetic retrieval of metal tagged bees

Gary et al. (1978) collected thousands of foraging bees in the field, some of them from common Italian colonies and some of them from hybrid bee colonies. They tagged them with metal discs that were later recovered with the aid of bar magnets attached to the parent colony entrances. From their paper (Gary et al., 1978:fig. 2, table 2), we recovered the original data for distances routinely travelled by foragers.

The combined data for the two strains of bees did not differ from a lognormal distribution (D = 0.37; P > 0.20). Hybrid bees ranged a lognormal average distance of 75 m from their colonies (D = 0.22; P > 0.20). Forager distribution for Italian bees was bimodal lognormal, with 70 percent of them foraging only slightly further than the average distance travelled by the hybrid bees. The other 30% of the Italian foragers travelled a somewhat greater average distance, a distance that can be determined readily by separating modes (Cassie, 1954).

Swarm relocation behavior

Schmidt and Thoenes (1990) tallied the distances that swarms moved from their parent colonies by placing a set of concentric rings of empty swarm hives (scented with pheromones) around the experimental colonies. They already knew that swarms are more likely to occupy these specifically designed cavities than natural cavities (Schmidt, 1990).

The distances that swarms relocated from their parent colonies did not differ from a lognormal pattern (Fig. 6; D = 0.40, P > 0.20). Coincidentally, the distances Schmidt and Thoenes selected for their placement of scented empty swarm hives matched those distances chosen by Michelsen and co-workers for placement of test stations in their mechanical bee experiments. Both patterns were approximately logarithmic and matched one another.

Both Lindauer (1955) and Seeley and Morse (1977) tallied distance information provided by foragers on the surfaces of swarm clusters to determine likely sites those swarms would move after having exited from their parent colonies. The distance distribution patterns gathered in both studies were quite similar to one another. The combined set of results (Seeley and Morse, 1977:fig. 1) did not differ from a lognormal pattern (D = 0.33; 0.10 > P > 0.05).

Water gathering

Colonies usually exist close to a water source (e.g., Columella, ~50 A.D.). During our honey bee removal project on Santa Cruz Island, California (Wenner, 1989; Wenner et al., 1990), we gathered data on the distances from colonies where we found bees at water. The pattern we obtained from data gathered for 53 colonies was binomial lognormal (Fig. 7A), with ~80% of the colonies exploiting water sources closer than 400 m (lognormal mean of ~160 m). The other 20% of the colonies gathered water from a greater distance (Fig. 7B), apparently forced to do so because of the prolonged drought in Southern California these past few years.


Experiments testing the presumed “use” of direction information, as interpreted from dance maneuvers, pose even greater problems than those testing “use” of distance information. First, we seem to have no good measures of the degree of variation present in direction information contained in dance maneuvers, compared to what is known for variation in distance information in dances (Wenner, 1962 and above).

Second, stations set out to test the “use” of direction information have almost always been placed in an arc at one distance from the colony, without an equivalent arc of stations placed in the opposite direction from the colony. Earlier Wenner (1962) warned against use of that station arrangement, because that procedure generates an odor field in only one direction from the colony and a consequent “odor center” problem.

Third, very little attention has been paid to wind speed and direction in those experiments designed to measure effectiveness of “direction communication.” Yet, odors (as physical particles) can only travel downwind, and stations must be scented or one gets no recruits (e.g., von Frisch, 1937; Wenner et al., 1969; Wells and Wenner, 1971). In areas with prevailing wind directions, as on Santa Cruz Island, one can also readily perceive that bees forage primarily upwind from their colonies (Wenner et al., 1990).

Recruit distribution: A center of odor field phenomenon

Whereas recruitment and foraging at different distances from a colony fit a lognormal pattern, a different mathematics applies when stations are all at the same distance from a colony. This topic is already treated fully in Wenner and Wells (1990:chap. 5, excursus PN) and will be covered only briefly here.

An expanding spiral flight pattern (e.g., Fig. 1) would result in many searching bees ending up downwind from scented food. Von Frisch did not recognize the potential consequences of the geometry of unevenly placed stations in his experiments (e.g., Wenner, 1962). Instead, when he conducted his “fan” experiments, he always placed his experimental station behind the central portion of an arc of stations. The geometry of station placement could then dictate the results (e.g., Johnson, 1967).

Goncalves (1969) later placed test stations in all directions from the colony and obtained recruitment in all directions, rather than only in the direction indicated in the dance maneuver. A downwind station bias in his results was explained by Friesen’s (1973) results.

Despite the fact that Johnson (1967) pointed out the unacceptability of arc station placement, others continued to employ that flawed design (e.g., Stephen and Schricker, 1970; Gould, 1975; Michelsen et al., 1989).

In one case, Michelsen and co-workers placed an arc of three stations in one direction from the colony and a single station in the opposite direction. They then attempted to direct recruits to the central station of the arc; their results seemed compatible with the notion of use of direction information provided by their mechanical bee (Fig. 8A). However, their results also closely match what one might expect if recruits ended up at the various stations in inverse proportion to the distance of all stations from the center of them all (Fig. 8B; see also Wenner and Wells, 1990:excursus PN).


We successfully apply our knowledge of odor-search behavior in our research project on Santa Cruz Island (Santa Barbara County, California). We treat colonies as foraging units while we locate and remove all feral colonies from that 25,000 hectare mountainous terrain (Wenner, 1989; Wenner et al., 1990). Our perception, as shown in part in Figure 1, permits us to find colonies in only a few hours instead of the days formerly required (e.g., Visscher and Seeley, 1989). We have now located more than 120 colonies in only four seasons.

The canyon topography and uniform Mediterranean climate on Santa Cruz Island most of the year results in wind moving past most colonies from only one direction all summer. That circumstance provides opportunity to study foraging patterns.

In the seasonal drought circumstances prevailing on that island, most colonies are very small, numbering only a few thousand individuals. Average distances foragers travel upwind apparently vary with quality of crop, but those distances in any case have not been very great. The maximum distances noted so far have been only upwind to introduced European weed patches, such as 1,500 m to horehound (Marrubium vulgare) and yellow mustard (Brassica species) and 2,500 m to sweet fennel (Foeniculum vulgare).

By contrast, as Friesen (1973) found earlier in studies of recruitment, we have found that island colonies forage only a few hundred meters downwind, regardless of the type or quality of crops that might be in that direction.

Studies of pollen gathered by colonies, in conjunction with Steve Buchmann at the USDA bee laboratory in Tucson, have just begun. These studies will reveal the maximum distances travelled for various plant species. Fortunately, the island vegetation has already been mapped and, aside from natural changes (e.g., fire), will change only in a prescribed and planned manner (e.g., removal of foreign exotics), because Santa Cruz island is now a portion of the Channel Islands National Park.


A recurring pattern is evident in the average distances that bees range from their colonies, a pattern found so far in four different circumstances: 1) distribution of searching honey bees at test stations in field experiments, 2) distances foragers travel to food sources, 3) distances bees travel to water, and 4) distances swarms relocate from their parent colonies. The distributions, based on the logarithm of the distance (i.e., lognormal), indicate that honey bee colonies function as units.

The lognormal pattern applies to the two extant hypotheses concerning honey bee recruitment to food sources. These two persistent and competing hypotheses, with us for centuries, may both be considered to be supported by any given set of experimental results; however, neither should have become a “ruling theory” (Chamberlin, 1890). The question is not which hypothesis is the correct one, but which hypothesis is most applicable (i.e., which fits the largest body of facts now) and/or which is most useful for explaining foraging patterns.

The new lognormal concept opens research possibilities, and one should note that many different researchers contributed to its formulation. Roubik (1989) took an important step toward breaking the long-standing impasse, not because the model he proposed fit the examples of distance foraging results he chose especially well, but because he introduced a mathematical model to explain the regular pattern of forager distributions.

The mechanistic approach that Roubik used, in contrast to the vitalism attitude (functionalism) that has prevailed these past 45 years (see Rosin, 1980), is the same type of approach that led to rapid advances in other areas of biology (e.g., molecular biology and genetics). As Visscher and Seeley stressed (and as others have noted), the distribution of resources is patchy in nature. The lognormal recruitment patterns observed are certainly not what one would expect if bees could use a “dance language.”

This new perspective has really been a joint effort. Esch and Bastian (1970), Gould et al. (1970), and Friesen (1973) provided accurate data on time taken for recruits to reach a station and on the percentage of success for those searching recruits. Wenner, Wells, and Johnson (reviewed in Wenner and Wells, 1990) made the test stations more nearly equal to one another, demonstrated that the dance language model no longer fit the results, and quantified the importance of odor to searching recruits. Friesen (1973) thoroughly documented the importance of wind direction for recruit success.

Michelsen and co-workers, by having either none or only two bees visit a test station, also made their test stations more nearly equal to one another. Their results did not agree with expectations of the dance language hypothesis; however, by having their stations placed at unequal distances from one another (with no concentration of stations near the test station as von Frisch had done), they obtained a recognizable lognormal pattern.

Visscher and Seeley gathered ample data on how far foragers ranged from large colonies in a natural setting and thereby provided an opportunity for a comparison of the results that Michelsen et al. obtained with few stations and with only two bees, to that which might happen for a colony foraging as a unit. Vergara studied African bee foraging; his find of a similar pattern to that obtained by Visscher and Seeley opened the way for Roubik to recognize the similarity between the two patterns and possible implications of that similarity.

What factors contribute to the average lognormal distance that bees range from their colony or relocate as swarms? The expanding spiral flight of naive bees as they leave their colony has been reported by students of bee behavior from the time of Columella (~50 A.D.) to Southwick (1991). Nor is the “Golden Section” pattern a rare phenomenon in nature (Cook, [1914] 1979; Ghyka, [1946] 1977).

The radius of an expanding spiral (perhaps logarithmic) would differ somewhat for each departing recruit, and searching bees would end up at varying distances from their colony on their first flight out. Furthermore, several factors would influence the length of that radius at any given time; richness of food, strength of odor cues, size of odor field, wind speed, relative abundance of stores in the colony, and colony size are a few examples.

Aside from the above general statement about direction orientation (and stressing the importance of wind direction), we do not cover here problems that arise when colonies forage in different directions. That is because a different mathematics applies; our mechanistic odor-search model (Fig. 1) is but a start. We note that nearly all direction experiments have been run with stations at a single distance. Once all stations are made equal in attractiveness, the geometry of station placement apparently dictates the results one can obtain (e.g., Johnson, 1967; Wenner and Wells, 1990:excursus PN).

Friesen’s 1973 paper could have provided an important lead for investigating that direction orientation further, but wind speed and direction were ignored by others at that time and have been largely ignored in publications since that time. So, also, have others ignored the importance of odor drifting downwind from the aerial pathway of foragers flying their beelines (e.g., Friesen, 1973; Wenner, 1974). That omission is striking, since everyone knows odor is important for successful recruitment and that odors travel only downwind.

In any event, the way is now open for some giant strides in the study of honey bee foraging ecology. We now perceive that searching behavior in honey bees is much like that exhibited by other flying insects. We will continue to pursue research with this notion in mind and hope that others join us in this new adventure.


We thank J. Alcock, S. Bambara, S. Buchmann, J. Dugan, M. Page, J. Schmidt, E. Sugden, S. Thoenes, H. Wells, P. H. Wells, and H. E. Wenner for helpful advice on the manuscript. We also thank various authors for the excellent data they provided to make this review possible, as well as the many volunteers who have assisted us in the Santa Cruz Island feral bee removal project. The Nature Conservancy and the University of California Faculty Research Committee provided partial funding.


Aristotle. ~330 B.C. (1931 translation). Historia animalium. Book IX. 40 (Vol. III; Vol. IV). Oxford Univ. Press, London.

Bonnier, G. 1906. Sur la division du travail chez les abeilles. C. r. hebd. Seanc. Acad. Sci., Paris 143:941-946. (Cited in Ribbands, 1953.)

Burroughs, J. 1875. Birds and bees and other studies in nature. Houghton Mifflin Co., New York.

Butler, C. 1609 (1969 reprint). The feminine monarchie. Da Capo Press, New York.

Buzzard, C. N. 1946. Shining hours. Collins, St. James’s Place, London.

Carde, R.T. 1984. Chemo-orientation in flying insects. In W. J. Bell and R. T. Carde (eds.), Chemical ecology of insects. pp. 111-124. Chapman and Hall, New York.

Carde, R. T. and R. E. Charlton. 1984. Olfactory sexual communication in Lepidoptera: Strategy, sensitivity and selectivity. In T. Lewis (ed.), Insect communication, pp. 241-265. Academic Press, New York.

Cassie, R. M. 1954. Some uses of probability paper in the analysis of size frequency distributions. Aust. J. Mar. Freshwat. Res. 5:513-522.

Chamberlin, T. C. 1890 (1965 reprint). The method of multiple working hypotheses. Science 148:754-759.

Columella, L. J. M. ~50 A.D. (1954). Lucius Junius Moderatus Columella on agriculture. Translation by E. S. Forster and E. H. Heffner. Harvard Univ. Press, Cambridge, Massachusetts.

Cook, T. A. 1914 (1979 reprint). The curves of life. Dover, New York.

Dujardin, F. 1852. Quelques observations sur les abeilles, et particulierement sur les actes qui, chez les insectes peuvent etre rapportes a l’intelligence. Annls. Sci. nat. (B) Zool. 3 Ser. 18:231-240. (Cited in Ribbands, 1953.)

Eckert, J. E. 1933. The flight range of the honeybee. J. Agric. Res. 47:257-285.

Emery, J. 1875. Ants and bees. Nature, London 12:25-26.

Esch, H. and J. A. Bastian. 1970. How do newly recruited honey bees approach a food site? Z. vergl. Physiol. 68:175-181.

Francon, J. 1938 (1939 translation). The mind of the bees. Methuen, London.

Friesen, L. J. 1973. The search dynamics of recruited honeybees, Apis mellifera ligustica Spinola. Biol. Bull. 144:107-131.

Frisch, K. von. 1937 (1939 reprint). The language of bees. In Annual report of the Smithsonian Institution for the year ended June 30, 1938, pp. 423-431. Publication 3491. U.S. Gov. Printing Office, Washington: 1939.

Frisch, K. von. 1947. The dances of the honey bee. Bull. Anim. Behav. 5:1-32. (Translated from: 1946. Die Tanze der Bienen. Oesterr. Zoolog. Zeitsch. I:1-48.)

Frisch, K. von. 1948. Solved and unsolved problems of bee language. Bull. Anim. Behav. 9:2-25.

Frisch, K. von. 1962. Dialects in the language of the bees. Sci. Am. 207:78-87.

Frisch, K. von. 1967. The dance language and orientation of bees. (Translated from the 1965 German edition by Leigh E. Chadwick.) Harvard Univ. Press, Cambridge, Massachusetts.

Gary, N. E., P. C. Witherell, and K. Lorenzen. 1978. The distribution and foraging activities of common Italian and “Hy-Queen” honey bees during alfalfa pollination. Env. Ent. 7:233-240.

Ghyta, M. [1946] 1977. The geometry of art and life. Dover, New York.

Goncalves, L. 1969. A study of orientation information given by one trained bee by dancing. J. Apicult. Res. 8:113-132.

Gould, J. L. 1975. Honey bee recruitment: The dance-language controversy. Science 189:685-693.

Gould, J. L. 1976. The dance-language controversy. Q. Rev. Biol. 51:211-244.

Gould, J. L., M. Henerey, and M. C. MacLeod. 1970. Communication of direction by the honey bee. Science 169:544-554.

Gowland, R. W. 1927. Effect of honey bees on sweet clover pasture. In J. H. Sheppard. Sweet clover experiments in pasturing, pp. 52-56. N. Dak. Agr. Expt. Sta. Bull. 211 (57 pp., illus.)

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

Johnson, D. L. 1967. Honeybees: Do they use the direction information contained in their dance maneuver? Science 155:847-849.

Johnson, D. L. and A. M. Wenner. 1970. Recruitment efficiency in honeybees: Studies on the role of olfaction. J. Apicult. Res. 9:13-18.

Kalmus, H. 1960. Training bees to smells and exciting bees in a hive. 101 simple experiments with insects. pp. 96-97. Doubleday, Garden City, New York.

Kennedy, J. S. 1983. Zigzagging and casting as a programmed response to wind-borne odour: A review. Physiol. Ent. 8:109-120.

Lindauer, M. 1955. Schwarmbienen auf Wohnungssuche. Z. vergl. Physiol. 37:263-324.

Lineburg, B. 1924. Communication by scent in the honeybee – a theory. Am. Nat. 58:530-537.

Maeterlinck, M. 1901. The life of the bee. (Translated by Alfred Sutro.) Dodd, Mead, and Company, New York.

Michelsen, A., B. B. Anderson, W. H. Kirchner, and M. Lindauer. 1989. Honeybees can be recruited by a mechanical model of a dancing bee. Naturwissenschaften 76:277-280.

Root, E. R. 1908. (No title.) Gleanings in Bee Culture. 36:830, 868. (Cited in Ribbands, 1953.)

Rosin, R. 1980. The honey-bee “dance language” hypothesis and the foundations of biology and behavior. J. Theor. Biol. 87:457-481.

Rosin, R. 1990. Honey bee dance language challenged again. Am. Bee J. 130:672.

Roubik, D. W. 1989. Ecology and natural history of tropical bees. Cambridge Univ. Press, New York.

Schmidt, J. 0. 1990. Swarm traps: An example of research and technology transfer. Am. Bee J. 130:333-334.

Schmidt, J. 0. and S. C. Thoenes. 1990. The efficiency of swarm traps: What percent swarms are captured and at what distance from the hive? Am. Bee J. 130:811-812.

Seeley, T. D. 1985. Honeybee ecology: A study of adaptation in social life. Princeton Univ. Press, Princeton, New Jersey.

Seeley, T. D. and R. A. Morse. 1977. Dispersal behavior of honey bee swarms. Psyche 84:199-209.

Siegel, S. 1956. Nonparametric statistics for the behavioral sciences, pp. 47-58. McGraw-Hill, New York.

Southwick, E. E. 1991. Bee research digest: Bee dance language? Am. Bee J. 131:226-228.

Spitzner, M. J. E. 1788. Ausfuehrliche Beschreibung der Korbbienenzucht im sachsischen Churkreise, ihrer Dauer und ihres Nutzens, ohne kunstliche Vermehrung nach den Grunden der Naturgeschichte und nach eigener langer Erfahrung. Leipzig. (Reference as in von Frisch, 1967.)

Stephen, W. P. and B. Schricker. 1970. The effect of sublethal doses of parathion, part 2: Site of parathion activity, and signal integration. J. Apicult. Res. 9:155-164.

Vergara, C. 1983. Rango de vuelo y cuantificacion de los recursos colectados por abejas Africanizadas en un bosque tropical Panama. Smithsonian Tropical Research Institute, Balboa, Panama, unpublished data report (data in Roubik, 1989, p. 88).

Visscher, P. K. and A. D. Seeley. 1989. Bee-living as a research technique in ecological studies of honey bees. Am. Bee J. 129:536-539.

Waddington, K. D. 1983. Foraging behavior of pollinators. In L. Real (ed.), Pollination biology, pp. 213-239. Academic Press, Orlando, Florida.

Wells, P. H. and A. M. Wenner. 1971. The influence of food scent on behavior of foraging honeybees. Physiol. Zool. 44:191-209.

Wenner, A. M. 1962. Sound production during the waggle dance of the honeybee. Anim. Behav. 10:79-95.

Wenner, A. M. 1963. The flight speed of honeybees: A quantitative approach. J. Apicult. Res. 2:23-32.

Wenner, A. M. 1964. Sound communication in honeybees. Sci. Am. 210:116-124.

Wenner, A. M. 1967. Honeybees: Do they use the distance information contained in their dance maneuver? Science 155:847-849.

Wenner, A. M. 1974. Information transfer in honeybees: A population approach. In L. Krames, T. Alloway, and P. Pliner (eds.), Nonverbal communication, Vol. 1, Advances In the study of communication and effect, pp. 133-169. Plenum Press, New York.

Wenner, A. M. 1989. “Bee-lining” and ecological research on Santa Cruz Island. Am. Bee J. 129:808-809.

Wenner, A. M., J. E. Alcock, and D. Meade. 1990. Removal of feral honey bee colonies from Santa Cruz Island. Am. Bee J. 130:818-819.

Wenner, A. M. and P. H. Wells. 1990. Anatomy of a controversy: The question of a “language” among bees. Columbia Univ. Press, New York.

Wenner, A. M., P. H. Wells, and D. L. Johnson. 1969. Honeybee recruitment to food sources: Olfaction or language? Science 164:84-86.

Wheeler, W. M. 1923. Social life among the insects. Harcourt, Brace, New York.

Wildman, T. 1768. A treatise on the management of bees; wherein is contained the natural history of those insects; with the various methods of cultivating them, both ancient and modern, and the improved treatment of them. To which are added, the natural history of wasps and hornets, and the means of destroying them. T. Cadell, London.

Winston, M. L. 1987. The Biology of the honey bee. Harvard Univ. Press, Cambridge, Massachusetts.

Zar, J. H. 1984. Biostatistical analysis. 2nd ed., pp. 53-60. Prentice-Hall, Englewood Cliffs, New Jersey.