We investigated how regularity of prey color pattern affects crypsis and how visual complexity of the background affects prey detection. We performed 2 predation experiments with artificial prey and backgrounds, using blue tits (Cyanistes caeruleus) as predators. In experiment 1, we found that contrary to a previous hypothesis, a pattern with repeated background-matching pattern element shapes was not easier to detect than a pattern with variable background-matching shapes. Increased background complexity with respect to shape diversity and complexity made prey detection more difficult. In experiment 2, we tested how spatial regularity of background-matching pattern elements affects crypsis. We found that spatially irregular prey with randomly placed pattern elements were harder to detect on both simple and complex backgrounds compared with spatially regular prey that had the elements aligned. Increased background element shape complexity made both prey categories harder to detect. In conclusion, our study shows that spatial regularity of prey pattern but not regularity due to invariable pattern element shapes deteriorates crypsis. Visually complex backgrounds and specifically those consisting of elements with complex shapes make detection of cryptic prey difficult.
The most obvious way for a prey to use its body coloration to decrease its risk of being detected by predators is through visual resemblance to its background (Wallace 1889; Thayer 1918; Cott 1940; Edmunds 1974; Endler 1978; Ruxton et al. 2004). This principle of concealment is generally known as background matching. Background matching was recently described as coloration that “generally matches the color, lightness and pattern of one (specialist) or several (compromise) background types” (Stevens and Merilaita 2009a). In other words, we know that high visual similarity between the appearance of the prey body pattern and its background increases the concealment of the prey and makes it more difficult to be detected by visual predators. Although there exist a number of conceptual and empirical studies on background matching (e.g., Endler 1978, 1984; Merilaita et al. 1999, 2001; Merilaita and Lind 2005; Houston et al. 2007; Sherratt et al. 2007), surprisingly little is still known about how background matching is maximized and how natural selection should be expected to shape the appearances of prey relying on background matching (Merilaita and Stevens 2011).
A prey color pattern can resemble or deviate from its background with respect to several different aspects of coloration and patterning, such as color, lightness, size, shape, and spatial distribution of pattern elements. However, not all such aspects need to be equally important in background matching. It is possible that while a close match to background with respect to 1 aspect of color pattern crucially improves the degree of crypsis (i.e., decreases the risk of detection), another aspect may only have a slight effect. Hence, to understand how natural selection for background-matching coloration has influenced the appearance of animals, it is important to study the role of different aspects of coloration in background matching.
Background matching is not the only factor that can influence the probability of prey detection, and there exist both theoretical and empirical evidence indicating that a given level of background matching does not necessarily result in equal probability of detection in different backgrounds. Hence, in addition to prey color pattern and its visual interaction with the background, also the characteristics of the background itself can influence the risk of detection of the prey (Merilaita 2003; Dimitrova et al. 2009; Dimitrova and Merilaita 2010). It has been suggested that high level of visual complexity of a habitat background (e.g., high variability or complexity in shapes of the elements constituting the background) can make prey detection difficult due to the high amount of visual information that a predator has to process to find useful information (cf. a low signal-to-noise ratio). Also studies on human visual search have found that target detectability is related to the variability of background elements and indicate the importance of the properties of the background on the difficulty of a search task (Gordon 1968; Duncan and Humphreys 1989; Troscianco et al. 2009). A recent study showed that blue tits (Cyanistes caeruleus) took longer to find prey in a geometrically complex than in a geometrically simple background (Dimitrova and Merilaita 2010). Thus, when studying prey detectability, it is also important to consider how the intricate effects of different background appearances affect prey detection.
In the present study, we have 2 objectives. Our main objective is to explore prey pattern regularity and how it influences background matching. We define a regular pattern as a pattern with a geometric transformation applied to one of its elements, thus resulting in a pattern that repeats an element according to a simple rule. In animals, one reason for such regular repeated patterns are various developmental constraints, such as bilateral symmetry and segmental body structure. Pattern regularity has been suggested to amplify visual signals (Kenward et al. 2004), which implies that it may be disadvantageous for cryptic color patterns. Indeed, Cott (1940) suggested that crypsis through disruptive effect is improved if the elements constituting the pattern are dissimilar in shape. Also, bilateral symmetry in which the patterning of one side of an animal is the mirror image of the patterning on the other side, contributes to pattern regularity (Stevens and Merilaita 2009b). It has been shown that such symmetry is in general terms detrimental for crypsis (Cuthill et al. 2006; Merilaita and Lind 2006).
In the present study, we investigate 2 different types of pattern regularity. With 2 separate experiments, we wanted to determine if 1) repeated pattern elements or 2) spatial regularity incur a cost for background-matching prey patterns in terms of increased risk of detection. Moreover, it is possible that the difference in detectability between different prey color patterns is influenced by the difficulty of the search task (Merilaita 2003). Therefore, we also manipulated the difficulty of the search task by using backgrounds of 2 different levels of complexity but without changing the resemblance level between the prey and the background (Dimitrova and Merilaita 2010). This allowed us to study whether high crypsis is amplified or low crypsis is compensated for by a background that makes prey detection difficult.
Our second objective is to further investigate the effect of background complexity on prey detection. In our previous study, showing that increased visual complexity of the background increases prey search time (Dimitrova and Merilaita 2010), we manipulated background complexity by increasing the number of different shapes constituting the backgrounds, but without controlling for the possible effect of the complexity of the shape itself. In the present study, we therefore test if the complexity of the shape of the elements constituting the visual background influences prey detection.
MATERIAL AND METHODS
The predators and the prey
This study was conducted at Tovetorp Zoological Research Station (Stockholm University) in south-eastern Sweden (58°56′N, 17°08′E). Experiment 1 was conducted between December 2007–March 2008 and January–March 2010, whereas experiment 2 was conducted between January–March 2009 and January–March 2010. We used wild-caught blue tits, Cyanistes caeruleus, as predators in the experiments. The experiments were performed with permission from the Swedish ethical board in Linköping (D. nr.: 56-05 and 62-08). A similar experimental protocol has also been used in Dimitrova et al. (2009) and Dimitrova and Merilaita (2010). Before the experiments, the birds were trained to search for artificial prey items, covering a piece of peanut as a reward, on an artificial background. During the experiments, we measured the effective search time and used it as a measure of camouflage.
The prey items and the backgrounds were made of paper, allowing an easy manipulation of their patterning. The use of artificial prey and backgrounds also minimize the risk that our results would be biased by the predators' previous experience of prey and backgrounds (cf. Alatalo and Mappes 1996). We created the prey patterning, using a purpose-written program in Matlab R2008b (Mathworks). The exact placement of the pattern elements was chosen randomly for each prey item. The patterning of the backgrounds was created with the software Corel Draw 11 (Corel Corporation). All prey items and backgrounds were reproduced with a laser printer (HP LaserJet 4000 Series PS with 1200 dpi resolution) on white copying paper (Canon Office). Each printed paper background (A4-sized) was glued on an equally sized corrugated cardboard using solvent free glue stick (Scotch, 3 M) to form the experimental boards.
We captured blue tits with mist nets and kept them indoors individually in cages (80 × 60 × 40 cm3). The room temperature was about 18 °C, and the light:dark rhythm (with 30 min dusk and dawn) was adjusted according to the prevailing day length. In their cages, the birds were provided with suet, sunflower seeds, peanuts, and water ad libitum. The birds were kept indoors for 6 days (median), before they were released in the area of capture.
Prior to the experiment, we individually trained the blue tits to search for the prey items by associating the prey items with pieces of peanut that they covered. The training of a bird began the day after its capture. Each bird was assigned to 1 prey series and the prey series was presented on both simple and complex background. Each training session took place in an experimental cage with water ad libitum.
The training of the birds consisted of 3 steps. In the first step, the prey items were glued directly on a brown, 10 × 15 cm2 corrugated cardboard and were thus easy to detect: 1) 1 nonpatterned, white prey item that had a piece of fully visible peanut glued on it; 2) 1 (experiment 1) or 2 (experiment 2) patterned prey items that were glued on the cardboard from 1 point, with a piece of peanut glued to its underside, but not sunk in a hole so that it was partly visible; and 3) 2 patterned prey items that were glued from 1 point to cover a hole with a piece of peanut in it. One prey item (experiment 1) or 2 prey items (experiment 2) from each prey category were used as the patterned prey items (2 and 3). In the second training step, the 10 × 15 cm2 corrugated cardboard was covered with both experimental backgrounds. Half of the cardboard plate was covered with a simple and the other half with a complex background, making the prey items cryptic. One prey item from each category was lightly glued from 3 points on randomly placed holes with pieces of peanut. No prey items were placed on the border between the 2 backgrounds. A tip of one of the prey items was slightly bent up to initiate searching. The third training step was performed in the same way as the second step, but this time we used an A4-sized corrugated cardboard instead, and all prey items were glued from 3 points.
Every training session and the experiment were preceded by 45–60 min without food in the experimental cage. A bird advanced to the next step if it had found all prey items on the preceding training step within 1 h. Otherwise, it had to redo that step in the next training session.
The experimental cages were made of plywood (width × height × depth: 55 × 90 × 70 cm3) and were lit from the ceiling with 2 high-frequency fluorescent lamps (15 W, BIOlight, Narva). The observation window (10 × 12 cm2) was covered with one-way see-through plastic sheet because the experimental room was kept dark when a bird was in an experimental cage, and this hindered the birds to see the observer. There was always water ad libitum in the cage and the temperature was about 17 °C. A perch was located 20 cm below the ceiling and on the opposite side; near to the floor, there was an opening through which the experimental boards could be inserted.
For presentation of prey items in the experiment, 1 prey item was lightly glued at 3 points on each experimental board. The prey items covered a randomly placed hole, in which there was a piece (about 2 × 2 × 2 mm3) of an organically grown peanut as a reward for a bird that found the prey item and tore it off the background. During the experiment, each bird was used once and was presented 3 (experiment 1) or 4 times (experiment 2) 1 prey series on both backgrounds (simple and complex). This resulted in 18 (experiment 1) or 16 (experiment 2) successive presentations for each bird. For each bird, we randomized the order of the prey categories (repeating the order equally many times for both backgrounds) but made sure that equal numbers of birds started the experiment with each prey category.
For each presentation the observer (M.D.) recorded the effective search time, that is, the time during which the bird actively searched for a prey item on an experimental board. The timing was stopped when the bird had found the prey item and pecked it to tear it off. The bird was then allowed to eat the reward before the board was replaced.
In the first experiment, we investigated how pattern regularity due to repeated element shapes affects detection times of background-matching prey. We studied this under more and under less challenging search conditions, these conditions being varied by using backgrounds of 2 different levels of visual complexity (Dimitrova and Merilaita 2010).
We used 3 prey pattern categories with patterns consisting of 4 black elements (Figure 1A). 1) One prey category had a regular background-matching pattern consisting of 4 identical element shapes that matched 1 of the element shapes in the background. We compared it with 2) a prey with variable background-matching pattern, to find out if a pattern consisting of more repeated identical elements is easier to detect than a pattern with fewer repeated elements (i.e., higher element diversity). The variable background-matching pattern consisted of 4 background-matching elements of 2 different shapes (2 of each shape). To find out if element shape diversity could compensate for poor matching and if the possible cost of regularity is comparable with conspicuousness caused by mismatching elements, we also presented the predators with 3) a third prey category that had a variable mismatching prey pattern. Its pattern consisted of 2 different shapes, 1 background-matching and 1 background-mismatching element (2 of each shape). To be better able to generalize our results from the prey category comparisons, we created 2 different series of the 3 prey categories (hereafter called Square and Triangle; Figure 1A). The prey category with regular background-matching pattern 1) consisted in prey series Square of a square element shape, whereas in series Triangle the element shape was triangular (Figure 1A). In both prey series, the variable background-matching prey pattern 2) had 2 squares and 2 triangles (Figure 1A). In both prey series, the variable mismatching prey category 3) had 1 cross-like element that deviated from all the elements in the background, but in prey series Square, the background-matching shape was a square, whereas in prey series Triangle it was a triangle (Figure 1A). Each prey item was unique so that the location and the rotation angle of the pattern elements were chosen randomly and with the condition that they would not overlap.
Because the disadvantage of regularity or mismatching may vary with the difficulty of the search conditions, we used a more and a less difficult background. The 2 backgrounds (Figure 1B,C), referred to from now on as simple and complex, were used in a previous experiment by Dimitrova and Merilaita (2010), where high-background complexity was found to increase prey search time when the degree of background matching was kept constant. Also in the present experiment, we kept the degree of background matching of the prey categories constant between the backgrounds. Hence, the backgrounds had the same number of black and gray elements of every shape, 5 in the simple and 8 in the complex background, the 5 shapes being common for both backgrounds. Moreover, 2 shapes unique to the complex background had a highly complex shape (i.e., long perimeter in relation to area). To ensure that the degree of background matching did not vary between the 2 backgrounds, the density of the 2 shapes that were included both in the background and in the prey patterning (the square and the triangle; Figure 1) were kept constant. The densities of the pattern elements, both in terms of number and area they covered were equal in both the backgrounds and the prey.
Each bird was assigned to either prey series Square or Triangle. We presented each prey category–background combination 3 times for a bird. Thus, each bird was presented with the same prey item on the simple and the complex background. This resulted in 18 consecutive experimental board presentations per bird. In total, we used 50 blue tits, 25 birds for prey series Triangle, and 25 birds for prey series Square.
In the second experiment, we investigated if a spatially regular prey pattern is easier to detect than a spatially irregular prey pattern. We also investigated if increased element shape complexity in the background would make the search task more difficult.
In this experiment, we used 2 prey pattern categories consisting of 4 black elements on white ground color (Figure 2A,B). The prey pattern categories differed with respect to the spatial distribution of the elements. In the spatially regular category, 1) there were 4 elements of 1 shape arranged on a line with constant intervals. The line of elements was placed randomly on the triangular prey, and the rotation angles of the 4 elements were identical and randomly chosen. We compared the regular prey with an irregular prey 2), which was produced by placing the 4 elements randomly (but so that they would not overlap with each other), each with a randomly chosen rotation angle. Also in this experiment we created 2 prey series to improve the generality of our results (Figure 2A,B) and for all prey items, the exact distribution of the pattern elements differed. The element shapes used for the 2 prey patterns were a pentagon (prey series Pentagon) and a shape created from 2 circular sectors (prey series Flame).
In this experiment, we also wanted to investigate if backgrounds with 2 different levels of complexity of element shape would induce a difference in search time for prey items that match both backgrounds to an equal degree. Therefore, we created for both prey series 2 different backgrounds referred to as simple and complex (i.e., 4 backgrounds in total; Figure 2C,D). All backgrounds consisted of 8 element shapes, but because we kept the numbers and areas of elements equal between the simple and the complex backgrounds, only the complexity of element shape differed between them. The only shapes in common for all backgrounds were the pentagon and the flame used in the prey patterns (Figure 2). We used shape complexity index to measure shape complexity of elements. We defined the index as the perimeter of the element divided by the square root of its area. Thus, a larger value indicates a higher degree of shape complexity and larger deviation from circular shape. In the simple backgrounds, the shape complexity index values of the elements were 4 or lower and in the complex backgrounds they were 6 or higher. Again, half of the background elements were black and half were gray to add shade diversity in the backgrounds.
The blue tits were assigned to 1 prey series, either Flame or Pentagon. Prey series Flame was displayed only on the first pair of simple and complex background (Figure 2C), whereas prey series Pentagon was displayed only on the second set of backgrounds (Figure 2D). For each bird, we used 2 times the 4 different combinations of background categories (simple and complex) and prey categories (spatially regular and irregular), which resulted in 16 consecutive presentations per bird. In total, we used 40 blue tits, 21 birds for series Flame, and 19 birds for series Pentagon.
We analyzed the prey search times with a mixed-model ANOVA in SPSS 19.0 for Windows. To achieve normal distribution and homoscedasticity, we applied ln-transformation suggested by the Box–Cox analysis on the search time data of both experiments. If the sphericity assumption was not met, we used the Greenhouse–Geisser correction. We used prey presentation order, prey category, and background as within-subject factors and prey series as a between-subject factor. As post hoc tests, we used paired t-tests on the mean values of the search times of each prey and background category with α-values corrected with the sequential Bonferroni method (Sokal and Rohlf 1995).
In the first experiment, blue tits took longer to detect the prey items on the complex background than on the simple background (F1,48 = 38.20, P < 0.0001; Figure 3). Presentation order had a significant effect on search times (F2,96 = 8.40, P < 0.0001). Figure 3 suggests that, as expected, the birds became faster between the first and last presentation. Prey series did not significantly affect search times (F1,48 = 3.48, P = 0.068).
We found a significant difference in search time between the 3 prey categories (F2,96 = 4.71, P = 0.011; Figure 3). Because prey series did not significantly affect search times (F1,48 = 3.48, P = 0.068) and none of the three-way and two-way interactions were significant (P > 0.18 for all interactions), we pooled together the search times for the 2 different prey series and backgrounds to further analyze the effect of prey category with 3 post hoc tests. The variable background-matching prey was more difficult to detect than the variable mismatching prey (t99 = −2.94; P = 0.004). This difference remained significant after the sequential Bonferroni correction (corrected α = 0.017). However, there was no difference in search time between the variable background-matching prey and the regular background-matching prey (t99 = −0.99; P = 0.33; corrected α = 0.05), nor between the regular background-matching prey and the variable mismatching prey (t99 = 1.95; P = 0.054; corrected α = 0.025).
In the second experiment, we found that there was a significant difference between the spatially regular and spatially irregular patterned prey (F1,38 = 10.88, P = 0.002). There was also a significant difference in search time between the 2 backgrounds (F1,38 = 35.53, P < 0.0001; Figure 4). The blue tits searched for a longer time for the prey items on the backgrounds consisting of complex elements than the backgrounds consisting of simple elements. Furthermore, the interaction between prey category and background was significant (F1,38 = 4.34, P = 0.044; Figure 4). Again, presentation order had a significant effect on search time (F3,114 = 4.66, P = 0.004; data not shown), such that search times generally decreased between the first and last presentation. Prey series did not affect search times (F1,38 = 1.62, P = 0.21). None of the other two-way or three-way interactions were significant (P > 0.37).
To further analyze the interaction between the prey and the background categories, we pooled the data from the 2 prey series and performed post hoc tests for the mean values of all prey categories and background combinations. Of the 6 comparisons, 4 were found to be significant (Figure 4). These differences remained significant after the sequential Bonferroni correction. Both the spatially irregular prey (t39 = −4.736, P < 0.0001; corrected α = 0.0083) and the spatially regular prey (t39 = −2.713, P = 0.01; corrected α = 0.017) were more difficult to detect on the complex than on the simple background. Furthermore, the spatially irregular prey on the complex background were more difficult to detect than the spatially regular prey both on the simple background (t39 = −6.102, P < 0.0001; corrected α = 0.010) and on the complex background (t39 = −3.741, P = 0.001; corrected α = 0.013). There was no difference in the search times between the spatially irregular prey on the simple background and spatially regular prey on the complex background (t39 = −1.34, P = 0.19; corrected α = 0.025), nor between the spatially irregular and the spatially regular prey on the simple background (t39 = −0.84, P = 0.40; corrected α = 0.05).
Our 2 experiments revealed interesting results about how regularity and background matching of prey pattern elements influence camouflage and about how background pattern complexity influences prey detection by birds. Below, we will first discuss the effect of prey pattern regularity and then the effect of background complexity on the detection of cryptic prey.
In the first experiment, we found that a background-matching prey pattern that was highly regular with respect to pattern element shapes was not easier to detect than a pattern consisting of variable background-matching shapes. This result suggests that a highly repetitive pattern may not necessarily incur an extra cost compared with a less repetitive pattern and may not be selected against in background-matching animals. In his classic book on animal coloration, Cott (1940) hypothesized that a color pattern consisting of 1 repeated element shape would be easier to detect than a color pattern consisting of elements varying in shape, because such variable elements would more likely be perceived as a group of separate objects instead of a single object. Our experiment does not lend support to his idea that a pattern consisting of variable element shapes would improve the camouflage of a prey. However, Cott (1940) discussed this in the context of disruptive coloration, and it is possible that variability in element shapes is more important in disruption than in background matching. Moreover, although the geometric shapes of the elements constituting the regular background-matching prey type were identical, the variable rotation of the elements introduced some variation between them. It is possible that this may have increased the perceived difference between identical geometric shapes and decreased the perceived difference between the regular background-matching prey type and the variable background-matching prey type.
The variable prey pattern that included mismatching elements was easier to detect than the variable background-matching prey pattern. This result suggests a higher cost for mismatching than for lack of variation in pattern element shapes for background-matching animals. On the other hand, we did not find any difference in search times between the highly regular prey pattern and the variable mismatching prey pattern. One interpretation of this result is that the cost of having a variable mismatching pattern is not much higher than if the pattern consisted of regular background-matching elements. If this result is also more generally true, then we could expect it to have important implications when considering prey living in heterogeneous habitats consisting of several patch types. It has been suggested that one way to maximize crypsis in such habitats is through a coloration that is a compromise between the requirements for crypsis in several of the patch types (Merilaita et al. 1999, 2001; Houston et al. 2007; Sherratt et al. 2007). We could consider the background used in experiment 1 as one patch type of a heterogeneous habitat and the mismatching prey as a compromise coloration with elements that match different patches. In that case, our result would imply that such compromise may do equally well within 1 patch type in spite of the partial mismatching, when compared with coloration consisting of regular background-matching elements (and probably even better within all patch types of the habitat). This clearly warrants further investigation.
In the second experiment, we found that the effect of repetitive element shapes in prey patterning was dramatically changed when they were arranged in a spatially regular fashion, with constant intervals and similar orientations. Here, the blue tits detected prey items with spatially random distributed pattern elements at a lower rate than prey items with pattern elements arranged in a spatially regular fashion. This clearly indicates that the specific spatial arrangement of pattern elements can influence the camouflage of a prey. Another factor that contributes to pattern regularity and has previously been shown to be important is bilateral symmetry. Bilaterally symmetric patterns have been shown to be easier to detect, increase the potency of a warning signal, and make avoidance of such warning signal more easily learned (Forsman and Merilaita 1999; Forsman and Herrström 2004). For camouflaged prey, on the other hand, it is not beneficial as such patterns tend to make prey easier to detect (Cuthill et al. 2005; Merilaita and Lind 2006). Importantly, the regular patterns in our experiment were not bilaterally symmetric (the patterns were aligned randomly with respect to the symmetry axis of the prey “body” and also the quadruplets of pattern elements deviated from bilateral symmetry). Hence, our study demonstrates that other types of spatial regularities other than bilateral symmetry also affects prey concealment, and then probably also the function of antipredator signals, such as warning coloration.
In addition to effects of prey patterning on prey detection, we also investigated the effect of complexity of the shapes constituting the visual background. Our first experiment confirmed our previous finding that visual complexity of the background, on which a cryptic prey occurs, does as such make a prey more difficult to be detected (Dimitrova and Merilaita 2010). In the present study, we found a similar effect of background, even though we used different prey item shapes and prey patterns. In experiment 1, the visually complex background provided a harder search task and made all prey items harder to detect, even when some element shapes mismatched the background. Thus, these experiments provide more robust support for the importance of background complexity for the probability of prey detection.
There may, however, be various aspects of complexity that makes a background difficult for predators. To understand how various natural backgrounds influence prey search and natural selection on prey defenses, we must specifically know which aspects of background complexity can influence the difficulty of the search task. In Dimitrova and Merilaita (2010) and in experiment 1 of the present study, the main difference between the simple and the complex backgrounds was in the level of element shape diversity (5 different elements in the simple vs. 8 different elements in the complex background). However, in the complex background there are 2 element shapes that may appear more complex than the rest, and this too might contribute to the difference in search performance observed between these backgrounds. Therefore, in experiment 2 of the present study, we specifically investigated the effect of element shape complexity (here defined as perimeter-to-√area ratio). We found that increased complexity of background element shape made prey search markedly more difficult, when we controlled for the level of background matching by keeping the density of matching shapes in the backgrounds constant. Because we only manipulated the complexity of the element shapes and kept the density and area of the elements constant, we are confident that the result reflects the effect of shape complexity. There is, however, a possibility that in addition to its direct effect, the increase in shape complexity might also have had some effect on the difficulty of the search task through a change in the perceived heterogeneity or variability between the background elements. Nonetheless, our finding is important because it indicates that natural backgrounds, whether they are leaf works of herbs or trees, tree trunks or rocks covered by lichen, debris-covered stream beds or other habitat patches that consist of a large set of elements, make the search task more difficult if the shape of the elements is complex than if the shape of the elements is simple. In other words, it gives us a guideline for comparing natural backgrounds and what may influence their difficulty for a given predator.
Interestingly, in experiment 2, we found that the effect of visual complexity of the background on prey camouflage was also dependent on the appearance of the prey. Although both prey categories (spatially regular and irregular) were more difficult for the blue tits to find on the complex than on the simple background, the effect of background complexity was much stronger on the spatially irregular prey pattern. On the simple background, both prey categories were equally easy to detect, but on the complex background, the search time of the spatially irregular prey was on average about 35% longer than the search time of the spatially regular prey. This suggests that the probability of a prey being detected is affected by rather complex interrelationships between the prey color pattern and the visual background. Probability of detection is not only influenced by the interaction of prey patterning and the visual background (i.e., background matching) but also by the effect of background characteristics (in this case complexity) as such. Furthermore, background characteristics seem to shape the interaction between prey and background appearances (i.e., the effect of background matching, in this case, the match in the spatial distribution of the prey pattern elements).
This result also shows that the spatial arrangement of pattern elements is an important aspect of background matching. The spatially regular arranged elements, which deviated from the more irregular spatial arrangement of the elements in the background, resulted in decreased camouflage. In a recent study, Cuthill and Székely (2009) showed that phase matching impedes detection. Collectively, these studies indicate the importance of matching spatial aspects of coloration in camouflage.
Our study suggests that there is a cost for spatial pattern regularity for cryptic animals living in backgrounds that are visually less regular, particularly if the habitat is complex. Yet, in spite of this suggested selection pressure many animals do have spatially regular color patterns. Perhaps, there are other forces that act upon the appearance of the prey patterns and maintain prey pattern regularity, such as developmental constraints. In addition, prey may actively seek out complex habitats and thus decrease the cost of having a spatially regular prey pattern during times of high predation pressure.
In both experiments, the birds took longer to find the prey in the beginning than at the end of the experiment. This improvement in search performance is probably a result of learning or forming of search image (Dukas and Kamil 2001). The factor presentation order did not interact with any other factors, suggesting that this improvement in search performance did not differ between the different prey types or the backgrounds. Importantly, this also suggests that the effect of background complexity on prey search does not disappear with learning or experience. We think that this may be due to the fact that the neural capacity of visual information processing is limited (Dukas 2004). Therefore, searching prey in a complex background containing more information to process is consistently slower than searching prey in a simple background containing less information (Merilaita 2003).
In summary, in this study, we found that prey pattern regularity is an aspect that may affect the camouflage of background-matching prey. Especially, spatial pattern regularity of prey pattern clearly decreases prey camouflage. And finally, we showed that one specific aspect of background complexity namely complexity of the shape of the visual elements constituting the background is an important determinant of the difficulty of the search for cryptic prey.
Royal Swedish Academy of Sciences (to M.D. and S.M.); the Lars Hierta Memorial Foundation (to M.D.); the Swedish Research Council (d.nr. 2007-5683) and the Academy of Finland (nr. 125518; to S.M.).
We are grateful to A. Vallin for assistance. We would also like to thank the staff at Tovetorp Zoological Research Station and A. Forsman, G. Gamberale Stille, and 2 anonymous reviewers for helpful comments on earlier versions of the manuscript.