Pollen. The pollens tested are listed in Table 1. Except for Typha latifolia L., which was handcollected from male inflorescences, all the pollen consisted of corbicular pellets obtained from pollen traps maintained on honey bee colonies. All collections occurred during the bloom periods listed in Table I and most were derived from the Tucson, Ariz., area. Exceptions included Cynara obtained from California south of San Francisco; Prunus dulcis Batsch from Winters, Calif.; Rubus from Kirkland, Wash.; and Taraxacurn from Laramie, Wyo.

Pollen for the tests was obtained by sorting beecollected pellets. A pollen was used only when it could be separated on the basis of color from a particular blend of known species. Any pellets (<3%) that contained a blend from two floral sources were discarded. All test pollens and the other pollens in the blends from which the test pollen was removed were determined by microscopic examination. After collection, all pollen was stored at -20*C until used in the tests. Several pollen mixtures were tested in addition to the pure pollens listed in Table 1. These included a standard identified mixture of 15 spring-collected species (Schmidt & Johnson 1984); a 2:2:1 (by weight) mixture of Cereus, Prosopis, and Larrea; and an equal mixture of Cereus, Prosopis, Larrea, Prunus, and Populus. These mixtures were selected because the standard mixture served as a reference for all of our experiments, the three- pollen mixture represented the three major pollen sources in the Tucson area, and the five-pollen mixture consisted of a blend of pollen with good as well as poor texture that was preferred as well as less preferred by the bees (Schmidt 1984). Properties of pollen, including grain size and protein content, were determined as in Schmidt & Johnson (1984).

Bees and Survival Tests. Honey bees used in the tests were obtained from colonies of mostly Italian honey bees located in Tucson, Ariz. Several frames of pupal brood were taken randomly from one to three colonies and maintained until emergence in boxes kept in an environmental room at 32-35'C in constant darkness and 70% RH.

Tests were conducted using 60 bees 1 d old removed from the emergence boxes and placed in acrylic plastic and screen cages (9 by 6 by 15 cm) in the environmental room. Each cage was provided water, a piece of beeswax foundation for the bees to rest upon, 50% sucrose solution, and a test pollen. The test pollen was formed by mixing corbicular pellets with distilled water sufficient to achieve a moist, kneadable texture. For Typha pollen, which was the only hand-collected pollen or spore, 30% sucrose by weight was added before mixing with water to provide a sugar level approximately equal to that normally added by bees in the process of making their corbicular loads (Schmidt & Buchmarm 1986). Each test pollen mixture was then placed in the hollows of polyethylene 24/40 tapered flask stoppers (taper from 24 to 21 mm diameter with a hollow 19 turn diameter by 33 turn long) (Nalgene Plastics), and the stoppers were inserted into holes slanted at 200 in the sides of the cages. Controls were identical to treatments except they received only sucrose solution and no pollen.

Three to six different tests plus a control were run concurrently. Each test and control consisted of three replicates. Each Monday, Wednesday, and Friday dead bees in each replicate were counted and removed, and pollen was replaced with fresh pollen. Weight of the consumed pollen was obtained by determining weight loss of the pollen remaining and subtracting from that loss the evaporative water loss of a matched stopper with pollen mix in a beeless cage in the environmental room. During the 2-d periods between changes, the diets did not appear to harden. Pollen-feeding continued for 20 d, by which time all measureable pollen-feeding had ceased.

Analysis. Mortality curves were generated with probit analyses using a SAS Proc Probit version 5.03 (SAS Institute 1982). From this program, 25, 50, and 75% mortality values plus the upper and lower 95% CL were generated. To simplify the data we report only the values plus or minus the interval equal to the greatest difference between either 95% CL and the central value. Kruskal-Wallis H test analyses were performed according to Wolff (1968), and linear regressions were calculated with SYSTAT version 2.0 (SYSTAT, Evanston, Ill.).

Typical mortality curves for bees fed various pollen diets are shown in Fig. 1. Control bees lived ca. 20 d, after which time mortality greatly increased, and all were dead by the 35th d (Fig. 1 A). When fed Kallstroemia, a species generally not collected or readily consumed by honey bees, survival curves were similar to those of controls (Fig. 1B). Larrea, one of the main honey bee nectar and pollen sources in the Sonoran Desert, promoted greater mean survival and increased the maximum life span by ca. 20 d (Fig. 1C). Finally, one of the best pollen sources for bees, the mixture of five different pollens, greatly enhanced survival, with some individuals living to >= d (Fig. 1D). Bees consuming pollen that promoted increased survival also tended to die more uniformly over time; controls had precipitous die-offs (Fig. 1D versus 1A).

Less typical mortality curves are shown in Fig. 2. Bees fed pollen from Washingtonia gradually died off over 75 d (Fig. 2A); those fed Simmondsia did not start dying until the 25th d, then died uniformly over the next 35-40 d (Fig. 2B); bees fed Cynara exhibited irregular periods of high dieoff (Fig. 2C); and those replicates fed Leucophyllum showed inconsistent patterns (Fig. 2D).

Each test series involved bees of similar genetics, nutritional background, and environmental situation. All of these conditions were different among bees of different test series. Thus, to avoid major influence of these variables, the best measure for determining the value of a diet is comparison of survival of the bees on that diet with survival of the bees in the matched control. When this is done, a value of days of additional survival time over that of the control is obtained. This added survival time value also allows more valid comparisons among all the different pollens.

The results of the mean increase in survival are given in Table 2. In this table, the best-performing diet measured at the 50% mortality time was the five-pollen blend. The bees on this diet survived 41 d longer than the controls. The poorest diet, Ambrosia, actually caused a decrease of 4 d in the average life span of the bees compared with the controls. The 25 and 75% mortality values followed trends similar to the 50% mortality curve.

In general, the pollen mixtures increased survival time relative to pure pollen sources. The mean increase in survival for an average bee fed one of the four blends of pollen (two tests of the standard mixture and one each of the three-pollen and fivepollen mixtures) was 29 d. In comparison, an average increase of only 19.5 d was observed for all of the diets tested and 22.7 d for the separate pollen components (Larrea, Prunus, Prosopis, Cereus, Populus). Several of the diets were tested twice, often with significantly different lengths of survival. These differences are attributable to differing conditions before and during the trials for the two analyses, differences that are not easy to quantify (e.g., effect of season and juvenile hormone levels on worker life spans [Fluri et al. 19771).

The average total life span of the bees fed various diets ranged from 22 to 62 d, with a mean of 42 d. The time for 25% of the bees on a diet to die ranged from 17 to 42 d, with an average of 30 d; for 75% of the bees to die the time ranged from 25 to 84 d, with 54 d as the average. The average life span of the controls for the seven different experimental diet series ranged from 15 to 34 d, with a mean of 24 d; the time for 25% of the controls to die ranged from 9 to 29 d, with a mean of 19 d; the time for 75% to die ranged from 19 to 40 d, with a mean of 29 d. The values for added days of survival for bees at the 25, 50, or 75% mortality levels in Table 2 can be converted into values for total life spans from eclosion by adding 19, 24, and 29 d to the respective table values.

The bees generally consumed pollen at the greatest rate during the first 10 d. Thereafter, consumption decreased sharply and ceased by the 20th d. The rate of pollen consumption for six of the pollens is shown in Fig. 3. Two of the most preferred pollens, the standard mix (Fig. 3A) and Populus (Fig. 3B), illustrate typical rapid food intake; by ca. the 8th-10th d >75% of the total pollen consumption had occurred. A similar situation was observed for Baccharis, a less-preferred type (Fig. 3C); the main difference was that amounts consumed were lower and termination of feeding sharper at the 8th d. Ambrosia (Fig. 3D) and Typha (Fig. 3E) both were consumed in large-to-moderate amounts; nevertheless, both caused a net decrease in survival time relative to controls (Table 2). In the case of Ambrosia, pollen consumption ceased abruptly at the 5th d. A last example, Kallstroemia (Fig. 3F), illustrates a pollen that essentially is rejected by the bees. That they also died earlier than the controls suggests that Kallstroemia is at least deterrent to bees and probably toxic as well.

Table 3 lists the amount of pollen consumed per bee during the initial 10-d feeding period and for the entire 20 d of feeding. Also listed is the protein content in percentage of dry weight of each species and the total amount of pollen protein consumed per bee during the main 10-d feeding period. Pollen consumption ranged from 1.9 to 29 mg per bee, with a mean of 16.5 mg. At one extreme, Kallstroemia was essentially refused by the bees, whereas the five-pollen blend was avidly consumed. In terms of total pollen protein consumed, the values ranged from 0.3 to 7.1 mg, with a mean. value of 3.2 mg. The extreme values were again represented by Kallstroemia and the five-pollen blend.

The average life span of bees on test pollen increased significantly with increasing protein concentration in the pollen (r2 = o.33; df = 31; P = 0.001), with an increase in the 10- and 20-d amounts of pollen consumed r^ 2 = 0.44; df = 31; P < 0.001. r^2 = 0.49; df = 31; P < 0.001), and with an increase in the 10- and 20-d total amounts of protein consumed r^2 2 = 0.65; df = 31; P < 0.001. r^2 = 0.62; df = 31; P < 0.001).

Factors other than pollen consumption and protein levels were evaluated to determine if they affected the survival of bees. Of the 25 species of pollen tested, six were in the Compositae, two each were in the Rosaceae and Zygophyllaceae, and 15 belonged to unique families (Table 1). The only family with enough species to evaluate was the Compositae. The pollen from the six species in the Compositae was significantly poorer X^2 = 5.84; df = 1; P < 0.05; Kruskal-Wallis H test) in increasing life span than average pollen. Springblooming (January-June) species on average produce pollen that increases life span more than fall-blooming (July-December) species (x' = 3.93; df = 1; P < 0.05; Kruskal-Wallis H test). An additional point to note is that the last two pollen flows of the year, Haplopappus and Baccharis, both of which are major pollen flows, produced very poor pollen from the standpoint of protein content, honey bee consumption, and increasing life span of the bees.

Of the 25 tested pollen species, 17 are zoophilous and seven are anemophilous, or wind-pollinated (Table 1). Three of the anemophilous species, Ambrosia, Uromyces, and Typha, were the poorest species overall in increasing honey bee life span (they actually decreased it), whereas Populus, Ephedra, and Simmondsia were among the highest in increasing honey bee life span. Thus, the mechanism of pollen transfer (animal or wind) per se does not correlate with the ability of bees to survive on that pollen.

Finally, it is possible that the size or surface texture of the individual pollen grains play some role in the survival of bees fed different pollens. However, pollen size within the range of 20 to ca. 100 X 10^-6 m (greatest length or diameter) does not appear to play a role. For example, the three best and worst pollens in Table 3 all have average-sized pollen grains (<45 Am diameter), whereas the largest pollen grains, Agave (100 X 10^-6 m), Malva, Cynara, and Kallstroemia (60 X 10^-6 m), were, with the exception of Kallstroernia, mostly average in their ability to increase life span (size data in Schmidt & Johnson [1984]). Most of the pollen species have smooth or reticulate outer surface sculpture, but six (Haplopappus, Baccharis, Taraxacum, spring composites, Cynara, and Malva) have spiny pollen. Spininess, however, does not appear either to reduce consumption of the pollen or to depress bee life span (P > 0.1; Kruskal- Wallis H test).