Behavioral Ecology doi:10.1093/beheco/arh166 Advance Access publication 17 November 2004
Countershading enhances crypsis with some bird species but not others Michael P. Speed,a,b David J. Kelly,b,c Andrew M. Davidson,b and Graeme D. Ruxtond School of Biological Sciences, Liverpool University, Crown Street, Liverpool L69 7ZB, U.K., b Biology, Liverpool Hope University College, Childwall, Liverpool L16 9JD, U.K., cDepartment of Zoology, Trinity College Dublin, Dublin 2, Ireland, and dInstitute of Biomedical and Life Sciences, Graham Kerr Building, University of Glasgow, Glasgow G12 8QQ, U.K.
a
ould (1991: 231) notably described the light bellies of countershaded animals as ‘‘perhaps the most universal feature of animal coloration,’’ yet a definitive explanation for dorsal darkening in countershaded prey animals remains elusive (see review in Ruxton et al., 2004). Because countershading is seen most frequently in inconspicuous animals, it has been widely assumed that it may add to the effectiveness of an animal’s crypsis, somehow reducing detection by predators. In the most famous and frequently applied theory of countershaded crypsis (Thayer, 1896), the combination of a dark dorsa and a light ventra is hypothesized to generate a property now called ‘‘self-shadow concealment’’ (SSC, see Edmunds and Dewhirst, 1994; Kiltie, 1988). Because illumination on prey is generally from above, shadows will be cast on ventral surfaces, and the resultant variation in shading may be used by predator’s visual systems to recognize and distinguish solid objects from their backgrounds: there is evidence that this is true for human vision (Ramachandran, 1988). Being darkest on top may reduce variation in shading caused by shadow and thus confound the ability of predators to recognize that a prey item is indeed a three-dimensional solid object when viewed from the side. This idea is generally attributed to Thayer (1896), who wrote that ‘‘mimicry makes an animal appear to be some other thing, whereas this newly discovered law [of gradation in the coloring of animals] makes him cease to appear at all.’’ Reporting Thayer’s work in Nature, Poulton (1902) commented that ‘‘no discovery in the wide field of animal coloration has been received with greater interest’’ (although we note that Poulton himself invented the idea some years earlier, see Poulton, 1888). This theory of SSC has been widely accepted as a possible explanation for countershading ever since (e.g., Braude et al., 2001; Bretagnolle, 1993; Edmunds and Dewhirst, 1994;
G
Address correspondence to M.P. Speed. E-mail:
[email protected]. Received 18 May 2004; revised 11 August 2004; accepted 3 September 2004.
Herring, 1994; Kiltie, 1988; Nagaishi et al., 1989; Phillips, 1962; Ruiter, 1956; Stauffer et al., 1999; Turner, 1961; Young and Roper, 1976). However, there is little experimental evidence that SSC really works in prey animals (see Ruxton et al., 2004) and, furthermore, there is very little experimental evidence that countershading enhances crypsis by this or any other means. Notably, however, a recent and large-scale comparative analysis concluded that countershading may aid concealment in even-toed ungulates (Stoner et al., 2003b) but that it is unlikely to do so in lagomorphs (Stoner et al., 2003a). Alternative explanations for countershading of course exist (see reviews in Kiltie, 1988; Ruxton et al., 2004). One alternative general explanation for some instances of countershading is that it is an epiphenomenon resulting from different demands applied to the dorsum and ventrum of an animal rather than a primarily adaptive antipredator trait. For example, in some terrestrial quadrupedal species with short limbs, the ventrum may be barely visible when the animal is viewed from above or from the side (see Braude et al., 2001). We might then expect such animals only to manifest camouflage pigmentation on the dorsum rather than on both the dorsum and the ventrum especially, if pigmentation incurs costs (see Kiltie, 1988; Ruxton et al., 2004). Of the relatively few experiments conducted to test the general prediction that countershading enhances crypsis (e.g., Edmunds and Dewhirst, 1994; Ruiter, 1956; Turner, 1961), Edmunds and Dewhirst’s (1994) is in our view the most rigorous. Edmunds and Dewhirst presented artificial prey on lawns to freely foraging birds (house sparrows, Passer domesticus; chaffinches, Fringilla coelebs; starlings, Sturnus vulgaris; blackbirds, Turdus merula; song thrushes, Turdus philomelos; robins, Erithacus rubecula; dunnocks, Prunella modularis; blue tits, Parus caeruleus; and great tits, Parus major). Prey were small green pastry cylinders that were uniformly light, uniformly dark, countershaded, or reverse shaded. Counter- and reverse-shaded baits were two-toned, being made from a thin strip of dark pastry laid on light pastry and
Behavioral Ecology vol. 16 no. 2 International Society for Behavioral Ecology 2004; all rights reserved.
Downloaded from http://beheco.oxfordjournals.org at Liverpool University Library on March 11, 2010
Although the theory of self-shadow concealing countershading is over a century old, there are very few direct empirical tests to substantiate the prediction that prey that are dorsally darkened and ventrally lightened (generally termed countershaded) suffer lower rates of attack than other prey. In this paper, we report experiments designed to determine whether artificial, countershaded prey are chosen by predators less often than those that are all light, all dark, or reverse shaded (i.e., dorsally lightened and ventrally darkened). Artificial prey were presented in gardens and parks to free-living birds, either on white backgrounds or on backgrounds with some degrees of color matching. In one experiment, birds were unmarked, and in the other, they were individually identifiable. We found that in three experimental trials, countershaded baits were attacked at a rate not significantly different from that of uniformly dark baits. In one experimental trial, countershaded baits were at some advantage. When we examined the data set for this trial more closely, it was apparent that blackbirds were taking countershaded baits least often, but blue tits and robins conferred no special advantage to countershaded baits. Hence, the efficacy of countershading may vary with species of predator. Key words: countershading, crypsis, predator, prey. [Behav Ecol 16:327–334 (2005)]
Behavioral Ecology
328
Table 1 Summary of sites and experimental conditions for the two experiments (days with rain were avoided in both experiments)
Experiment, trial Experiment 1, garden 1 Experiment 1, garden 2
Experiment 2, trial 2
Site 1: Whitehaven, Cumbria, rural garden Site 2: Whitehaven, Cumbria, rural garden (2 miles from garden 1) Archbishop Ryan Park, Dublin
Archbishop Ryan Park, Dublin
Dates
Condition in second period of experiment
Visiting birds
Color of pastry
B, ST, TH, SP, CD, GT, CH, BT B, ST, TH, SP, GT
Green
26 June 2002 to 25 July 2002
Card present (14 days)
Card absent (14 days)
Green
26 June 2002 to 25 July 2002
Card absent (14 days)
Card present (14 days)
B, BT, R
Brown
11 March 2002 to 30 May 2002
B, BT, R
Brown
5 November 2003 to 1 December 2003
Brown pastry on brown background (11 March 2002 to 1 May 2002) Brown pastry on white background (5 November 2003 to 19 November 2003)
Brown pastry on white background (15 May 2002 to 30 May 2002) Brown pastry on brown background (23 November 2003 to 1 December 2003)
Key to abbreviations: B ¼ blackbird (Turdus merula), ST ¼ starling (Sturnus vulgaris), TH ¼ thrush (Turdus philomelos), SP ¼ sparrow (Passer domesticus), BT ¼ blue tit (Parus caeruleus), CD ¼ collared dove (Streptopelia decaocto), GT ¼ great tit (Parus major), R ¼ robin (Erithacus rubecula).
presented with either the dark or the light area upperside. Edmunds and Dewhirst reported a significant advantage to countershaded baits over the uniformly dark forms. However, on its own, Edmunds and Dewhirst’s experiment does not constitute a definitive test of the prediction that countershading enhances crypsis for at least two reasons. First, Edmunds and Dewhirst’s data set is small (they report 9 sampling days), and even though they showed statistical significance, there is a good case for repeating this work. Second, Edmunds and Dewhirst were unable to consider whether different bird species responded to countershading in prey in different ways. We therefore report two experiments designed to test whether countershading enhances protection by diminishing the probability of detection by predators. We modified Edmunds and Dewhirst’s basic experimental design to
incorporate a ‘‘cryptic condition’’ (in which prey more or less match their background color) and a ‘‘conspicuous condition’’ (in which all prey are presented on a white background). In our initial experiment we used gardens as arenas and unmarked birds as predators; in our second experiment we used individually identifiable blackbirds (T. merula) and robins (E. rubecula) in a single park, whose behavior was monitored throughout. If countershading does reduce the likelihood of detection in the artificial prey, then we would expect that in the cryptic condition countershaded baits are taken less often than the other forms. Furthermore, if the primary cause of prey choice is ease of visual detection, then in the conspicuous condition we expect prey to be taken at similar rates. We show that although countershading cannot always be demonstrated to enhance crypsis, there is some evidence in our data set to support Edmunds
Table 2 Results of GLM analysis of bird-feeding data in experiment 1 Source
Type III sum of squares
df
Mean square
F
Significance
a. Whitehaven, garden 1 Day Background Prey Background 3 prey Error Total Corrected total
30.958 36.262 1.529 1.576 68.205 937.000 108.051
1 1 3 3 103 112 111
30.958 36.262 0.510 0.525 0.662
46.752 54.761 0.770 0.794
,.001 ,.001 .514 .500
b. Whitehaven, garden 2 Day Background Prey Background 3 prey Error Total Corrected total
23.630 3.103 4.398 3.224 34.467 1032.000 89.844
1 1 3 3 103 112 111
23.630 3.103 1.466 1.075 0.335
70.616 9.272 4.381 3.211
,.001 .003 .006 .026
Downloaded from http://beheco.oxfordjournals.org at Liverpool University Library on March 11, 2010
Experiment 2, trial 1
Site
Condition in first period of experiment
Countershading enhancement of crypsis
and Dewhirst’s results. Crucially though, countershading may be effective with some species of predators but not others. EXPERIMENT 1: UNMARKED INDIVIDUALS Methods
Statistical analysis Data for each garden were analyzed separately using the GLM ANOVA procedure in SPSS v.11.0. Prey and background were fixed factors, and day of experiment was included as a covariate. The square root transform of the number of each prey type attacked per day had a good match to a normal distribution and was the dependent variable in each case. In all GLM analyses described in this paper, we used planned Bonferroni post hoc tests in the estimated marginal means procedure of SPSS. In most cases, we compared mean numbers of different bait types attacked within but not between an experimental condition (i.e., within the background-matching or white background conditions).
a
16 14 12 10 8 6 4 2 0 L
D
C
R
C
R
Prey type
b
16 14 12 10 8 6 4 2 0 L
D
Prey type
Figure 1 Mean numbers of each prey type attacked in experiment 1 in the presence/absence of card. Shaded bars represent presentations without white card; unshaded bars represent presentations with card. L ¼ light, D ¼ dark, C ¼ countershaded, R ¼ reverse shaded. Error bars represent 2 SEM (solid lines connect means significantly different, p , .01; dashed line, p , .05). (a) Whitehaven site 1: card used in first presentation. (b) Whitehaven site 2: card absent in first presentation.
Results There was a significant effect of day as a covariate in data sets from both gardens, in which there was a positive correlation of total number of prey attacked and day of experiment (see Table 2 and Figure 1). In both gardens background was recorded as a significant factor; a higher mean number of prey were attacked in the presence of card. In garden 1, in which baits were presented first on card, there were no significant effects of prey or prey 3 background interaction. In contrast, in garden 2 with card absent as the first condition, all factors and the prey 3 background interaction were significant. With card present, the Bonferroniadjusted post hoc test showed no significant differences between mean numbers of each prey attacked (p . .98 in all comparisons); with card absent, comparisons of light and dark (p ¼ .03) and light and countershaded were significant (p , .01; in all other comparisons p . .05). Hence, countershaded baits were taken at a rate not significantly different from that of dark or reverse-shaded baits.
Downloaded from http://beheco.oxfordjournals.org at Liverpool University Library on March 11, 2010
The design of experiments with unmarked birds closely followed the method of Edmunds and Dewhirst (1994), in which pastry prey were made from a 3:1 mixture of plain white flour and lard. To make pastry of a light hue, 25 ml of green ‘‘Supercook’’ dye and 50 ml of water were added to 600 g of pastry; for pastry of a dark hue, 75 ml of dye was added to a second 600-g batch of pastry. Edmunds and Dewhirst’s recipe generates a large difference (to human eyes) in contrast between the light and dark pastries. The prey were created by molding the pastry into approximately 3 mm-diam cylinders and then cutting them into 10 mm lengths. Four prey types were made using this method; light and dark prey were cut directly from rolled pastry and rolled into cylinders. For the countershaded and reverseshaded prey the pastry was first rolled until it was approximately 1.5 mm thick and then the light and dark pastries were laid on top of one another to produce a two-toned prey. Countershaded prey were placed with the dark pastry on the uppermost side and the light pastry on the underside; reverse shaded were placed with light pastry uppermost and the dark pastry on the underside. This produces a contrast boundary between light and dark sections of the bait. Following Edmunds and Dewhirst (1994), we used 25 of each prey type randomly distributed throughout a 10 by 10 matrix so that each bait was 0.5 m from the nearest other baits (note this was 1 m in Edmunds and Dewhirst’s original experiment). Prey were distributed at 0830 h and collected when approximately 50–60% of the baits had been taken or at around 1600 h on each day, whichever came first. We split the experimental trials into two phases, one in which prey were placed directly onto grass (‘‘card absent’’) and the other in which they were placed on 1.5 3 1.5-cm squares of white card (‘‘card present’’). Because all prey types would be clearly visible against the white card, the card-present condition was included to control for the possible existence of preferences that operate after detection and are unrelated to crypsis. Any differences in predation rates in the card-present treatment are therefore likely to reflect postdetection preferences rather than difference in crypsis. We ran this experiment simultaneously between 26 June 2002 and 28 July 2002 in two similar gardens in Whitehaven, Cumbria, U.K., that were 2 miles apart. In garden 1, card present was the first condition and card absent the second condition, both conditions lasting for 14 days. In garden 2, this order was reversed (see summary in Table 1). Because water degrades the coloration of the artificial prey, we did not collect data on rainy days.
329
Mean number attacked
•
Mean number attacked
Speed et al.
Behavioral Ecology
330
Table 3 Output of GLM for first experiment with individually marked Dublin birds Source
Type III sum of squares
df
Mean square
b. Second experiment in which white board was presented first Trial 0.038 1 0.038 Prey 0.136 3 0.045 Background 1.544 1 1.544 Species 212.380 2 106.190 Prey 3 background 6.716 3 2.239 Prey 3 species 2.265 6 0.377 Background 3 species 4.087 2 2.043 Prey 3 background 3 species 8.599 6 1.433 Error 436.663 1783 0.245 Total 2766 1808 Corrected total 691.300408 1807
One argument that can be made against the quality of data generated with unmarked free-living birds is that the individuals participating may change during the course of an experiment. We therefore decided to follow-up this experiment with a second experiment in which a group of individually ringed birds was presented with baits in a similar experimental setup. Here we were able to check that all the identified individuals were present throughout the duration of the experimental trials. EXPERIMENT 2: INDIVIDUALLY RINGED BIRDS Methods The trials were conducted in Archbishop Ryan Park in Merrion Square, Dublin, Ireland (Table 1). We performed two experiments, one between 11 March 2002 and 30 May 2002 and the other between 5 November 2003 and 1 December 2003. All presentations were made on boards. Some years earlier (1998), green baits had been presented to birds as part of an early-feeding experiment; hence, we used brown baits and brown-painted boards here. To color the baits, a brown dye was manufactured by combining 20 parts of Green (90) with 1 part of Christmas Red (4R) (both from J E O’Brien & Sons Ltd., Dublin, Ireland). To produce the dark brown color, we mixed 37.5 ml of dye with 37.5 ml of water and added this to 600 g of pastry. To produce the light brown coloration, we mixed 7.5 ml of dye with 68.5 ml of water and again added this to 600 g of pastry. This recipe generates a large difference (to human eyes) in contrast between the light and dark pastries. Countershaded and reverse-shaded baits were then produced as in Edmunds and Dewhirst (1994), with the dark and light forms providing a conspicuous contrast boundary. The board used was a plain white Corriboard square (5-mm thick 3 50 cm 3 50 cm). For color-matching conditions, we painted the board with a dark brown paint (Dulux Weathershield—Bitter
Significance
11.005 19.585 0.225 611.014 17.327 7.848 15.665 8.728
.001 ,.001 .635 ,.001 ,.001 ,.001 ,.001 ,.001
0.154 0.185 6.305 433.599 9.141 1.541 8.344 5.852
.695 .906 .012 ,.001 ,.001 .161 ,.001 ,.001
Chocolate) to match the dark brown bait color used, otherwise the board was presented unpainted. Presentations were made at 25 different locations around the park that were chosen to be well within any given bird’s territory. No individual was allowed more than three presentations at any given site during a day’s data collection, and presentations at a location ceased when approximately 50% of baits had been taken or when 15 min had passed, whichever came sooner. The experimenter (D.K.) had greater knowledge of the numbers of baits taken in this experiment and was better able to stop a presentation when the 50% criterion had been met than was possible in experiment 1. As several birds could visit and remove prey items quickly, it was not feasible to randomly distribute the prey items in a manner that ensured correct identification of every prey taken. Therefore, the presentation was divided into four equal sectors (25 3 25 cm), and each of the four treatments (light brown, dark brown, countershaded, and reverse shaded) was placed in a different sector. The treatments were always placed in the same order on the board, but the board was rotated through 90 with respect to the observer between consecutive presentations. This system corrected for any distance from cover considerations, as well as favored perches. At the start of each presentation, there were 10 of each bait type spaced evenly across the relevant sector. This was the minimum number of baits that ensured a period of uninterrupted feeding by the birds. The observer stood close to the board to record both the number of baits eaten and the order in which they were eaten. This ensured that all baits eaten were attributable to individual birds. In the first phase of the first experimental trial (from 11 March 2002 to 1 May 2002), the board initially was brown; subsequently we changed the board color to white (unpainted) (fully conspicuous condition from 15 May 2002 to 30 May 2002). In the second experimental trial (from 5 November 2003 to 1 December 2003), we reversed this order of presentation (Table 1).
Downloaded from http://beheco.oxfordjournals.org at Liverpool University Library on March 11, 2010
a. First experiment in which brown board was presented first Trial 3.662 1 3.662 Prey 19.552 3 6.517 Background 0.075 1 0.075 Species 203.326 1 203.326 Prey 3 background 17.298 3 5.766 Prey 3 species 7.835 3 2.612 Background 3 species 5.213 1 5.213 Prey 3 background 3 species 8.713 3 2.904 Error 573.359 1723 0.333 Total 4564.000 1740 Corrected total 866.177 1739
F
•
Countershading enhancement of crypsis
a
3.0
2.5
2.0
1.5
1.0
0.5
0.0 L
Statistical analysis
C
R
C
R
Prey type
b
1.8 1.6 1.4
Mean numbers attacked
We used the GLM analysis in SPSS v. 11.0 on square root transformed numbers of baits taken per presentation with bird species, prey, and background as fixed factors and presentation number as covariate. Planned Bonferroni post hoc tests were again used to compare mean attack rates within experimental conditions. In the white-board presentation of the first experimental trial, blue tits made only 7 visits compared to 119 in the preceding part of this experimental trial. Blue tits were therefore excluded from the overall GLM analysis and were examined separately. In the second experimental trial, all three species made similar numbers of visits between experimental conditions and hence, all three were included in the GLM.
D
1.2 1.0 0.8 0.6 0.4
Results
0.2
In the first experimental trial in which baits were initially presented against a brown background, all main factors (except background) and all interactions were significant (Table 3, Figure 2). We consider two levels of analysis from this point: first, the prey 3 background interaction, looking at the total effect of avian predation on the baits, and second, the species 3 prey interactions for both backgrounds. Blue tits were excluded from the main GLM analysis for reasons stated (however, we confirmed that their exclusion did not affect the conclusions for the brown-board presentation by running a similar GLM with the blue tits present and the ‘‘white-board’’ condition excluded; see Appendix). When the background was brown, only mean numbers of dark and countershaded prey were not significantly different (p ¼ .697, otherwise all other mean comparisons were significant, Figure 2a). When the background was subsequently changed to white, no means were significantly different (p ranged from .980 to 1). Thus, even though there is no main effect of background in the number of baits taken (in a large part because the experimenter determined when to stop a presentation), there is strong evidence of differential choice when the baits matched the background but no evidence of differential choice when they contrast with the background. This is consistent with our expectation that the white background would diminish any differences in detection rates between baits. If we look at each species (Figure 3a,b), we see marked differences in bait choices when the background is brown. While the blue tits showed no differences in bait choices at all (one-way ANOVA, F3,476 ¼ 0.873, p ¼ .455), the robins took light baits over all others and the blackbirds showed more complex order of choice: L(ight) . R(everse shaded) . C(ounter shaded) ¼ D(ark). In contrast, when the back-
0.0 L
D
Prey type
Figure 2 Mean numbers of prey attacked by all predators at Archbishop Ryan Park, Dublin. Shaded bars represent data for brown boards; unshaded bars represent data for white boards (solid lines connect means significantly different, p , .01; dashed line, p , .05). (a) First experimental trial in which brown backgrounds are used first. (b) Second experimental trial in which white backgrounds are used first.
ground was white, all significant effects of color difference on choice rates disappear. In the second experimental trial, in which baits were initially presented against a white background, trial number, prey, and the prey 3 species interaction were nonsignificant; all other components of the ANOVA were significant (Figure 2b, Table 3, b). Overall, more baits were taken in the second part of the experiment when brown boards were presented. This is explained in part by a larger number of total visits made by the birds (215 in the first half of the trial and 241 in the second), which may reflect growing demand for food as the winter progressed during the experiment. When the background was white, there were overall no significant differences between bait types; however, when the background was changed to brown, countershaded baits were taken at a significantly lower rate than light baits, and there were no other significant differences noted. Hence, there is some advantage to countershaded prey in this part of the experiment. This level of analysis, however, hides important differences between species (Figure 3c,d). When the background was white, the blackbirds took countershaded baits more often
Downloaded from http://beheco.oxfordjournals.org at Liverpool University Library on March 11, 2010
Five identified blackbirds and eight identified robins took part in the first experimental trial for its entire duration. These birds had been previously marked (by mist netting and color ringing) and trained to associate the experimenter with food. In addition a number of unidentified individuals of both species and unidentified blue tits, P. caeruleus, visited the feeding boards. Between experimental trials, a new set of birds was marked to replace individuals that had moved or died in the interim. Fifteen robins took part in all the second experimental trial, and only four were identified as individuals present in the first experimental trial. Four blackbirds took part in the second experimental trial; two of which were identified as having previously encountered the baits. As in the first experimental trial, a number of unringed blue tits also fed from the boards.
331
Mean numbers attacked
Speed et al.
Behavioral Ecology
332
b
10
8
8
Mean numbers attacked
10
6
4
L D C R
Blue tits n=119
L D C R
Robins n=227
0
L D C R
Blackbirds n=88
L D C R
Blue tits n=7
L D C R
Robins n=41
L D C R
Blackbirds n=79
d 10
10
8
8
Mean numbers attacked
Mean numbers attacked
c
4
2
2
0
6
6
4
4
2
2
0
6
L D C R
Blue tits n=45
L D C R
Robins n=156
L D C R
Blackbirds n=12
0
L D C R
Blue tits n=54
L D C R
Robins n=167
L D C R
Blackbirds n=18
Figure 3 Results for experiment 2 by bird species; experimental trials 1 (brown board first) and 2 (white board first). (a) Trial 1, brown board. (b) Trial 1, white board. (c) Trial 2, white board. (d) Trial 2, brown board. Post hoc tests: ---- ¼ 0.1%; — ¼ 1%; ...... ¼ 5%. ‘‘n’’ refers to the number of presentations in which birds of each species visited the boards during each trial.
than light and dark baits, but when the background was changed to brown, they took the countershaded baits significantly less often than the others (we note that the number of visits by blackbirds in this second experimental trial is lower than in the first). In contrast, the blue tits never took baits at significantly different rates, whereas the robins took the light baits over the others, when the background was brown (Figure 3) but not when it was white.
DISCUSSION Though SSC is often viewed as the correct explanation for countershading (e.g., Gould, 1991, and see review in Ruxton et al., 2004), Edmunds and Dewhirst’s (1994) study stands out as the only good direct experimental demonstration that countershading can enhance crypsis. In this paper, we have attempted to replicate and extend important features of Edmunds and Dewhirst’s experiment.
We draw a number of conclusions from the first experiment performed with unringed birds on two neighboring lawns. First, when presented on a lawn initially without cards (in garden 2), the birds chose some types of bait at different rates, but there was no special advantage to countershading, in that the countershaded baits were taken at a rate not significantly different from that of dark or reverse-shaded baits. When subsequently placed on white cards, these preferences disappeared altogether, indicating that the observed patterns of predation were based on detection rather than on some postdetection preference (such as visual familiarity or bait taste). We therefore were unable to replicate Edmunds and Dewhirst’s demonstration that countershading enhances crypsis (compared to uniform dark baits) but were able to show that observed choices were not likely to be due to postdetection preferences. Second, order of presentation has an important though not unexpected effect on behavior; if birds find the baits on a white background first (garden 1), they do not significantly discriminate between baits even when
Downloaded from http://beheco.oxfordjournals.org at Liverpool University Library on March 11, 2010
Mean numbers attacked
a
Speed et al.
•
Countershading enhancement of crypsis
species and not others, which could go some way to explaining why we failed to replicate all components of Edmunds and Dewhirst’s original results in the first experiment with unmarked birds.
Conclusions In one of our four experimental trials, countershading did enhance protection. Most notably perhaps, when we examined the data for this experimental trial at the species level, it was clear that only the blackbirds were taking the other baits at higher rates than countershaded baits. Furthermore, the three focal bird species took the baits in very different ways in this experiment, ranging from no selection by color to a high level of choosiness. If such high levels of predator heterogeneity have some generality, then cryptic appearances of prey animals may be optimized to minimize overall predation levels from a suite of different predators. We have supporting evidence for the hypothesis that countershading enhances crypsis, but we also have a number of data sets in which countershading provided no enhancement to crypsis compared to plain dark baits. It may be that countershading varies in its effectiveness at enhancing crypsis in visual conditions other than those used in this experiment. In future experiments, we intend therefore to vary lighting conditions and the degree of contrast and sharpness of the contrast boundary between dorsum and ventrum. Our results do show, however, that in most instances (i.e., three of four experiments in color-matching conditions), our countershaded baits were on average not more or less vulnerable to predation than the simple, dark ‘‘phenotype,’’ despite having a relatively large area of light pigment. An important conclusion is that SSC could still apply here even if the net effect of countershading is to maintain but not to actually enhance crypsis. If defensive pigmentation incurs costs, then countershading may be a good way of maintaining cost-effective crypsis.
APPENDIX Demonstration that the inclusion/exclusion of the blue tits does not affect overall conclusions in experiment 2, experimental trial 1. Here we included the blue tits but excluded the white-board section of the experiment in which the blue tits were barely represented. Table A1 shows the results from GLM analysis; Figure A1 shows data for all bird species for this experimental trial. Overall conclusions about the pattern of predation are unchanged by removing the blue tits from the analysis.
Table A1 Results from GLM analysis in which all three bird species are represented, but white-board data are excluded (experiment 2, trial 1)
Source
Type III sum of squares
df
Trial Prey Species Prey 3 species Error Total Corrected total
2.664 40.885 150.729 34.680 388.268 3697.000 610.186
1 3 2 6 1723 1736 1735
Mean square 2.664 13.628 75.365 5.780 0.225
F
Significance
11.822 60.477 334.443 25.650
.001 ,.001 ,.001 ,.001
Downloaded from http://beheco.oxfordjournals.org at Liverpool University Library on March 11, 2010
they are subsequently placed against the green backgrounds of the lawn. One explanation for this is that in the initial presentation with the cards, the birds had time to learn that all squares in the matrix were likely to contain a bait. Hence, when baits were presented without cards, it may be that the birds simply return to places where they previously found food rewards and choose them in a manner unrelated to their underlying relative conspicuousness. An implication is that in factorial, ordered experiments performed using a matrix such as this, we cannot rely on bait choices from the ‘‘cards-first’’ garden to tell us much about crypsis and the efficacy of countershading when the background is changed to the colormatching condition. It could be argued that in this sort of experiment the birds might approach and view the baits from above, in which case countershading and reverse shading could not be expected to have effects. However, it is clear that baits were not exclusively viewed from above because, for example, the light and reverse-shaded baits were generally treated differently in relation to the other baits, even though their upper surfaces were identical (Figures 1b and 2). Hence, we must conclude that when countershaded and dark baits are treated as equivalent, as they were in most experimental trials, it is not primarily because the birds viewed the baits exclusively from above. Having marked individuals in the second experiment enabled us to learn about the different contributions of bird species to the overall levels of predation seen on each bait. The first experimental trial of the second experiment, with individually marked birds, produces similar results to those seen in garden 1. Overall, the birds chose the baits at different rates when placed on a color-matching background and showed no differences on a white background. Again, there was no significant benefit to countershading in relation to uniformly dark baits that matched the brown board most closely. However, consideration of the data at the level of bird species revealed a hierarchy of choosiness; the blackbirds were most discriminating (taking the baits in an order of preference L . R . D ¼ C), robins next (L . D ¼ C ¼ R), and finally blue tits (no preferences at all). It follows that if we had rerun the experiment at a different site, with a different composition of individuals from these species, we might find quite different results. One reason that the blackbirds may show higher levels of prey discrimination is that during feeding they stand on the board as dominant predators taking multiple baits, whereas the robins and blue tits flew in from neighboring perches and generally took one or two baits at a time, with limited time for visual inspection and perhaps different viewing angles. It is possible that the blue tits simply followed a spatial rule, flying to the nearest part of the board rather than focusing on the most conspicuous baits, as the robins appear to have done. In addition, we cannot rule out the differences in perceptual system between the bird species as a contributory factor; nor can we rule out the possibility that the color of the substrate on which these species normally forage differs and affects their behavior at the feeding board in complex ways. Most notably perhaps, in the second trial of this experiment, the blackbirds took countershaded baits more often than the others when they were placed against a white background and least often when they were placed against the brown background. This strongly suggests that for these birds at this point in the experiment, countershaded baits really were least readily detected when placed against a colormatching substrate. Furthermore, we can discount the possibility that the birds are averse to countershaded prey because they favored them in the contrasting background condition. Hence, we appear to have evidence that countershading may work to diminish detection rates with some
333
Behavioral Ecology
334
3.0
Mean numbers attacked
2.5 2.0 1.5 1.0
0.0 L
D C Prey type
R
Figure A1 Data for experiment 2, experimental trial 2 with all bird species present.
We thank Tim Caro, Candy Rowe, Ian Harvey, Katherine Allen, Malcom Edmunds, and an anonymous referee for their help. Technical support was provided by D. Sennett, Liverpool Hope University College, and D.K. was funded by the Liverpool Hope research fund.
REFERENCES Braude S, Ciszek D, Berg NE, Shefferly N, 2001. The ontogeny and distribution of countershading in colonies of the naked mole-rat (Heterocephalus glaber). J Zool 253:351–357.
Downloaded from http://beheco.oxfordjournals.org at Liverpool University Library on March 11, 2010
0.5
Bretagnolle V, 1993. Adaptive significance of seabird coloration: the case of procellariiforms. Am Nat 142:141–173. Edmunds M, Dewhirst RA, 1994. The survival value of countershading with wild birds as predators. Biol J Linn Soc 51:447–452. Gould SJ, 1991. Bully for brontosaurus. London: Penguin. Herring PJ, 1994. Reflective systems in aquatic animals. Comp Biochem Physiol 109A:513–546. Kiltie RA, 1988. Countershading: universally deceptive or deceptively universal. Trends Ecol Evol 3:21–33. Nagaishi H, Nishi H, Fujii R, Oshima N, 1989. Correlation between body colour and behaviour in the upside-down catfish, Synodontis nigriventis. Comp Biochem Physiol 92A:323–326. Phillips GC, 1962. Survival value of the white coloration of gulls and other seabirds. University of Oxford. Poulton EB, 1888. Notes in 1887 upon lepidopterus larvae &c. Trans Entomol Soc Lond 1888:595–596. Poulton EB, 1902. The meaning of the white undersides of animals. Nature: 596. Ramachandran VS, 1988. Perception of shape from shading. Nature 331:163–166. Ruiter LD, 1956. Countershading in caterpillars: an analysis of its adaptive significance. Arch Neerl Zool 11:285–341. Ruxton GD, Speed M, Kelly DJ, 2004. What, if anything, is the adaptive function of countershading? Anim Behav 68:445–451. Stauffer JAJ, Hale EA, Seltzer R, 1999. Hunting strategies of a Lake Malawi Cichlid with reverse countershading. Copeia 1999:1108– 1111. Stoner CJ, Bininda-Emonds ORP, Caro T, 2003a. The adaptive significance of coloration in lagomorphs. Biol J Linn Soc 79:309– 328. Stoner CJ, Caro TM, Graham CM, 2003b. Ecological and behavioral correlates of coloration in artiodactyls: systematic analyses of conventional hypotheses. Behav Ecol 14:823–840. Thayer AH, 1896. The law which underlies protective coloration. Auk 13:124–129. Turner ERA, 1961. Survival value of different methods of camouflage as shown in a model population. Proc Zool Soc Lond 136:273–284. Young RE, Roper CFE, 1976. Intensity regulation of bioluminescence during countershading in living animals. Fish Bull 75:239–252.
