Aphids occur in a range of colour morphs that can differ in growth rates, host range, defensive behaviour, and susceptibility to parasitism3,4,5,6. Pea aphids, Acyrthosiphon pisum (Harris) (Homoptera: Aphididae), in south-central Wisconsin occur as two colour morphs, green and red. The colour morphs remain distinct through the summer months because the aphids reproduce parthenogenetically. Pea aphids experience high levels of parasitism by the wasp Aphidius ervi Haliday (Hymenoptera: Aphidiidae) and heavy predation by several predators, including ladybird beetles, especially Coccinella septempunctata L. (Coleoptera: Coccinellidae). A. ervi is a ‘parasitoid’ that attacks aphids by inserting an egg through the aphid's cuticle; the developing wasp larva feeds on and eventually kills the aphid7. Both the parasitoid and the predator can have a major impact on pea aphid populations8,9 and hence may be important selective agents. Although all three species are introductions to the Neartic, they have a long evolutionary history in their common Paleartic home range, where both aphid colour morphs also coexist10,11,12.

In field populations, we found that the relative level of parasitism and predation had a significant effect on aphid colour morph composition. Specifically, the proportion of red morphs increased following relatively high parasitism and decreased following relatively high predation (Fig. 1), implying balancing selection by parasitism and predation. This balanced parasitism/predation hypothesis was supported by our further studies demonstrating directly that parasitism by A. ervi is heavier on green morphs, whereas predation by C. septempunctata is heavier on red morphs. The parasitism rate on green morphs in the field (53%) was significantly higher than on the red morph (42%) (P < 0.001). Significantly higher parasitism rates on a green morph over a red morph have also been reported for the alfalfa aphid, Macrosiphon creelii Davis13. Another study found higher parasitism on red than on green morphs of A. pisum4: the difference between this result and ours may be due to differences in experimental protocol (that is, laboratory rather than field) or differences in the aphid or parasitoid strains used.

Figure 1: Change in proportion of red morphs in the aphid population between successive field samples versus the relative predation pressure.
figure 1

The sizes of the circles are proportional to weightings based on sample sizes. The regression is statistically significant at the P < 0.003 level (n = 20).

The high mobility of ladybird beetles made it impossible for us to measure predation in the field. However, we measured predation rates by C. septempunctata on caged plants in a greenhouse, and the predation rate on red morphs (0.91 ± 0.08 aphids eaten per hour) significantly exceeded that on green morphs (0.73 ± 0.06 aphids eaten per hour) (P < 0.04). Thus the observed pattern of parasitism and predation supports the hypothesis that the polymorphism is maintained by a balance of these two sources of mortality.

In all previous studies in which predation is a selective force maintaining a polymorphism, susceptibility to predation of some morphs is balanced against some other trait which makes susceptible morphs more competitive in the absence of predation. We found no evidence for this type of balance. Reproductive rates of the two morphs were not significantly different (P = 0.437). Thus, a hypothesis that greater parasitism or predation rates on one morph were balanced by greater reproductive rates is not supported. Furthermore, the propensity to drop from a plant when confronted with a foraging predator, an important aphid defensive behaviour, did not differ significantly between the two morphs (P = 0.226; see ref. 14 for methods). Although there may be other more subtle differences in the biology of the two colour morphs12, the evidence implicates balanced parasitism and predation as the factors that maintain the colour polymorphism.

The differential susceptibility of colour morphs suggested that the parasitoid and predator may be using prey colour as a foraging cue, and both A. ervi and C. septempunctata have been shown to use visual cues to locate prey15,16. To investigate the use of colour cues by C. septempunctata, we measured predation rates on both morphs in red, green or white containers. In green containers, predation by C. septempunctata was higher on the red morph, whereas in red containers predation was higher on the green morph. There was no difference in predation between the morphs in white containers (Table 1). These results indicate that red morphs are more susceptible on green plants because they are more visible.

Table 1 Numbers of morphs eaten on different backgrounds

The explanation for higher A. ervi parasitism on the green morph is not as obvious, as the green morph should be relatively more cryptic visually than the red morph on green plants. In an experiment similar to that on C. septempunctata, A. ervi showed no effect of container colour on relative parasitism rates on the two morphs. However, neither visual nor olfactory discrimination of the aphid morphs by the parasitoids can be completely ruled out by these experiments. In addition, differential encapsulation of parasitoid eggs is a possible mechanism for different parasitism rates on the morphs, which does not involve discrimination by the parasitoid4. However, we found no evidence of encapsulation in dissections of over 2,000 aphids containing eggs. Despite the clear evidence for higher parasitism rates on green morphs in the field, the mechanism for this pattern is unknown.

To examine the consequences of morph-biased parasitism and predation, we constructed a simple model of aphid morph–parasitoid–predator population dynamics (Box 1). A central assumption of the model is that one or both of the predatory species will exert density-dependent mortality on the aphids. As pea aphids make up the vast majority of A. ervi hosts in agricultural systems, A. ervi population dynamics are coupled to those of its hosts, so density dependence can be assumed. Although C. septempunctata is a generalist predator with a wide host range, it exhibits aphid-density-dependent migration9,17,18,19. We confirmed the importance of aphid-density-dependence in our system by showing that there were statistically significantly more C. septempunctata found in fields with more aphids (Table 2).

Table 2 Distribution of C. septempunctata relative to aphid density

The model shows that the coexistence of the two colour morphs is facilitated by both density-dependent parasitism resulting from the coupling of host and parasitoid population dynamics, and density-dependent migration that leads to greater predator densities in fields with high aphid abundance. Although the density dependence of A. ervi that results from its population dynamics being coupled to those of the aphids promotes morph coexistence, even in the absence of C. septempunctata density dependence, the potential for coexistence is greatly enhanced if the predator exhibits density dependence as well (Fig. 2). Thus the maintenance of the colour polymorphism is not simply a matter of two predatory species preferentially attacking alternate morphs: it also requires density-dependent mortality from the parasitoid and/or predator.

Figure 2: Theoretical conditions leading to coexistence of green and red aphid morphs.
figure 2

Combinations of z0 and α are shown that satisfy the condition for coexistence of colour morphs. Although details depend on the particular equations and values of the parameters used (Box 1), the qualitative structure of the figure—particularly the existence of a region of morph coexistence—is the same for all biologically plausible models.

The discovery that balancing parasitism/predation can maintain a polymorphism indicates that other unexplained polymorphisms may be maintained by balanced density-dependent mortality from predators, parasitoids and/or pathogens, as almost all organisms are killed by a complex of predatory species. Determining an underlying selective mechanism of polymorphism maintenance bears directly on the long-standing debate on whether the bulk of polymorphisms should be classified ‘adaptive’ or ‘neutral’20,21. Our results also have implications for the conservation of biodiversity, given the importance of preserving genetic diversity within insect species in addition to conserving the species themselves22,23,24. An understanding of the mechanisms that maintain genetic diversity will enable conservation programs to be more effective.


Assessment of parasitoid, predator, and aphid densities in the field. We estimated the density of the parasitoid, the predator, and the two aphid morphs by sampling 12 alfalfa fields roughly every 6 days throughout the summer of 1996. In each field, aphid density and colour were recorded for 100 stems in eight locations. Twelve three-minute walking scan samples of parasitized aphids and C. septempunctata adults were also made throughout the field. Parasitoid counts were limited to ‘mummified’ pea aphids which are immobile, skeletonized aphids containing one parasitoid larva or pupa. Sampling dates followed by an alfalfa harvest and fields without aphids were excluded from the analysis.

Impact of parasitoids and predators on relative proportions of colour morphs. We measured the impact of parasitoids and predators on the relative proportions of colour morphs by regressing the change in proportion of red morphs in the aphid population between successive field samples versus the relative predation pressure. The relative predation pressure was measured as log[(C7t−1 + 1)/(C7t−1 + mt)], where C7t−1 is the number of C. septempunctata observed during 12 three-minute scan samples in a field in week t − 1, and mt is the number of mummies observed in the field during scan samples in week t. We used mt as a measure of parasitism in week t − 1 because mummies form roughly one week after parasitism. The regression of Δr = rtrt−1 = constant + slope* log[(C7t−1 + 1)/(C7t−1 + m1)] was done while weighting by the square-root of the total number of aphids counted in either week t − 1 or week t, whichever had the fewest aphids. Weighted regression was necessary to correct for higher variance in Δr created when the number of aphids counted per field was small.

C. septempunctata density dependence. We tested for aphid-density-dependent migration by C. septempunctata by regressing aphid density (at each sample date) versus C. septempunctata density. The regression model was = z0* date + α*A/Ā, where C7 is the number of C. septempunctata observed per field in 12 three-minute scan samples, A is the number of aphids per 800 alfalfa stems, Ā is the mean number of aphids per 800 stems for all fields, and ‘date’ is a categorical variable for the date of the sample. Only those dates on which C. septempunctata were found in two or more fields are included.

Assessing parasitism rate. To estimate parasitism rates, we sampled five alfalfa fields on five dates in May, June and August 1996, and dissected a total of 643 aphids for larval A. ervi. Data were analysed with χ2 test, blocked by field and date.

Susceptibility of colour morphs to C. septempunctata on plants. We measured the susceptibility of the two aphid colour morphs to predation by allowing a single adult C. septempunctata to forage on caged alfalfa plants with 15 adults of each morph for 4h; the experiment was replicated 27 times. After removing the predator from each plant, the number of remaining aphids of each morph was recorded and subtracted from 15 to determine the number consumed. The data were analysed with a 2-tailed, paired t-test.

Use of colour cues by C. septempunctata. We investigated the use of colour cues by C. septempunctata by allowing single adult beetles to forage for aphids of both colour morphs in red, green and white containers. To maximize our ability to detect differences in predation rates between morphs, we designed the experiment as a split-plot, with background as the whole plot and aphid colour morph as the split plots. Each arena had a single background colour and 5 similar-sized (third or fourth instar) aphids of each colour morph. Predators foraged for 30min, and at the end of the experiment the number of remaining aphids of each colour morph was recorded and subtracted from 5 to determine the number consumed. All predators foraged actively and consumed at least one aphid. Each treatment was replicated 20 times. Data were analysed as a split-plot analysis of variance.

Comparison of reproductive rate of the colour morphs. The reproductive rates of the aphid colour morphs were compared by pairing 10 of each morph on individual caged alfalfa plants and allowing them to reproduce for 14 days. At the end of the experiment, all aphids of each morph was recorded. The experiment was started with 14 plants, but 2 plants were severely wilted after two weeks so these were excluded from the analysis. Data were analysed with a 2-tailed paired t-test.