The authors have declared that no competing interests exist.
Conceived and designed the experiments: MS AD AEL. Performed the experiments: AEL AD MS. Analyzed the data: MS. Contributed to the writing of the manuscript: MS AEL. Calibrated the images and measured them: AEL AD. Undertook visual modelling of the data: MS.
Camouflage is found in a wide range of species living in numerous habitat types, offering protection from visually guided predators. This includes many species from the intertidal zone, which must cope with background types diverse in appearance and with multiple predator groups foraging at high and low tide. Many animals are capable of either relatively slow (hours, days, weeks) or rapid (seconds and minutes) colour change in order to better resemble the background against which they are found, but most work has been restricted to a few species or taxa. It is often suggested that many small intertidal fish are capable of colour change for camouflage, yet little experimental work has addressed this. Here, we test rock gobies (
Predator-prey interactions have played a substantial role in shaping the diversity of life, leading to many adaptations and counter-adaptations for attack and defence
Almost certainly, the most common form of camouflage in nature, and the basis for many other types of concealment, is background matching, where an animal resembles the general colour and pattern of the background
While a great deal of recent research has investigated the different types of camouflage that may exist and how they work, mostly in artificial systems (e.g.
Fish make an ideal group to study colour change for camouflage because it is widely reported that many species have this ability
In this paper we study the colour change abilities of the abundant and widely distributed rock goby (
Gobies were collected by dip net in the intertidal zone from Gyllyngvase beach, Falmouth, Cornwall, UK (50° 8′33.4690″N, −005° 04′07.9716″W) between July 2013 and September 2013 for experiment 1 (40 individuals), and between October 2013 and July 2014 (40 individuals) for experiment 2. Once caught, fish were kept in fresh seawater in a grey bucket in order to minimise colour change prior to use. Both experiments were conducted
Our aim in experiment 1 was to test whether gobies are capable of changes in their luminance when placed on a black or white background. In experiment 2, we aimed to test for changes in colour. Our design here was intended to minimise perceived differences in brightness of the backgrounds by the fish, and to test whether gobies change their actual colours as opposed to just changes in luminance. We used red and blue as two colours at different ends of the visual spectrum that gobies are likely to be able to discriminate (see
Fish were placed in a 24 cm wide × 34 cm long × 5 cm deep (internal measurements) white tray that had been covered with a background of midpoint grey paper, calibrated as described above. The tray was filled 2 cm deep with fresh seawater and a spirit level was used to ensure trays were flat and the water level accurate to prevent variation in colour measurements due to water depth, and to ensure all areas of the tray had sufficient water to minimise stress to the fish by ensuring that even the largest individuals were fully submerged, while at the same time keeping water depth low so as to not affect the colour analyses. Fish were given 15 minutes to acclimatize on the grey background then photographed (see below) in the control tray before being transferred individually into a secondary experimental tray of 28.5 cm wide × 39 cm long × 7 cm deep (internal measurements) divided into eight compartments by thin plastic barriers attached with silicone sealant glue. These compartments were either four white and four black for experiment 1, or four red and four blue for experiment 2. Transfer of the fish between trays was done as quickly as possible and with a net in order to minimise any stress associated with capture and handling. Fish were unable to see each other, although barriers were not completely sealed and water was able to flow around the tray. Although this also meant that chemical cues could potentially transfer among individuals, any such effects should not produce directional colour changes in line with responses to background colours and brightness. The experimental trays also ensured that pairs of fish were tested under the same water conditions (e.g. temperature), and the relatively small size of the compartments prevented fish from swimming around too much, which would have made photography difficult.
Experimental trials were undertaken in blocks, with a single block consisting of a pair of fish, with one fish placed on each background colour, and with those individuals approximately matched by size to remove bias that may occur due to variation in colour change with individual size. Twenty fish were tested on each background colour for each experiment (40 fish in total per experiment). Fish were subsequently photographed again at 1–2, 10 and 60 minutes while remaining in the tray to establish the extent of colour change over time. Photos were taken using a Nikon D90 SLR camera, which had undergone a quartz conversion to enable ultraviolet sensitivity (Advanced Camera Services, Norfolk, UK) and fitted with a Nikon 105 mm Nikkor lens. In both experiments photographs were taken in human visible (400–700 nm) and ultraviolet (300–400 nm). For the human visible photos a UV/IR blocking filter was used (Baader UV/IR 2″ Cut Filter) and a UV pass filter was used during the ultraviolet photographs (Baader U 2″ Cut Filter). All photographs included a Spectralon 40% grey reflectance standard (Labsphere, Congleton, UK) next to the tray and a ruler. Due to changing light conditions and reflectance from the water surface, a black and silver photographic umbrella (Neewer, Guangdong, China) was used to shade the trays from direct sunlight.
Images were taken in RAW format with manual white balance and fixed aperture settings. Images were then linearized with regards to light intensity based on camera responses to a set of eight Spectralon grey standards with reflectance values ranging from 2 to 99% (in custom programs written in Image J) in order to correct for the non-linear responses in image values many cameras produce in response to changes in light levels
We wanted to analyse colour change with regards to one of the likely main predator groups of rockpool fish: shore birds. To obtain data corresponding to avian vision, we transformed the reflectance based image based on spectral sensitivity data from the peafowl (
Once calibrated, the outline of each goby was drawn around by hand using Image J and the region of interest (ROI) saved. Each image layer was measured to acquire values for photon catch. We then calculated a series of metrics to analyse the appearance of each goby. Saturation (the amount of a given colour compared to white light) was defined as the distance an object is in a tetrahedral colour space from the achromatic grey point
Finally, we calculated how changes in the appearance of fish equated to differences in their level of match to the experimental backgrounds. To do so we used a log form of a model of visual discrimination, the Vorobyev-Osorio model
We did not specifically expect an overall difference in appearance between fish on each background at all time points. Instead, our key prediction was that there should be no difference at the start of the experiment (time zero) when fish have been on the same intermediate grey background, whereas there should be differences as the experiment progresses. The exact time where differences arise should also depend on the speed of colour change. As such, we conduced a series of planned comparisons
For luminance, there was no significant difference between fish on black or white backgrounds at time 0 (W = 414.0, n = 20, p = 0.925), but there were significant differences at one minute (W = 587.0, n = 20, p<0.001), 10 minutes (W = 600.0, n = 20, p<0.001), and at 60 minutes (W = 610.0, n = 20, p<0.001), with fish on white backgrounds having higher luminance values (
Panel D shows the level of similarity for fish against the white and black backgrounds for JNDs in luminance (see main text) at 0, 1, 10, and 60 minutes. Graphs A, C, and D show medians plus inter-quartile range (IQR), whiskers are lowest and highest values that are within 1.5*IQR from the upper and lower quartiles, asterisks represent outliers. B shows means plus standard error.
In terms of colour change, for saturation, there was also no significant difference between fish on black or white backgrounds at time 0 (W = 406.0, n = 20, p = 0.925), but significant differences occurred at one minute (W = 305.0, n = 20, p = 0.005), 10 minutes (W = 268.0, n = 20, p<0.001), and 60 minutes (W = 308.0, n = 20, p = 0.006), with saturation values being higher on the black backgrounds (
On the white background there was no significant reduction in JNDs over time (better match to the substrate) for colour (H = 0.46, df = 3, p<0.927), but there was for luminance JNDs (H = 31.77, df = 3, p<0.001;
There was no significant difference between fish on red or blue backgrounds for luminance at time 0 (W = 410.0, n = 20, p = 1.000), at one minute (W = 373.0, n = 20, p = 0.324), or at 60 minutes (W = 346.0, n = 20, p = 0.086), nor was there a significant difference at 10 minutes when controlling for multiple testing (W = 327.0, n = 20, p = 0.025);
Panel D shows the level of similarity for fish against the red and blue backgrounds for JNDs in colour at 0, 1, 10, and 60 minutes. Graphs A, C, and D show medians plus inter-quartile range (IQR), whiskers are lowest and highest values that are within 1.5*IQR from the upper and lower quartiles, asterisks represent outliers. B shows means plus standard error.
Regarding colour, for saturation, there was no significant difference between fish on red or blue backgrounds at time 0 (W = 378.0, n = 20, p = 0.394), but there were significant differences at one minute (W = 311.0, n = 20, p = 0.008), 10 minutes (W = 307.0, n = 20, p = 0.006), and 60 minutes (W = 289.0, n = 20, p = 0.001), with fish being more saturated on the red background (
Three individuals are shown on the left having been placed on a black background, and then the same individuals are shown on the right after being on a white background.
On a red background, there was a significant reduction in JNDs for colour (H = 25.31, df = 3, p<0.001), but not for luminance JNDs (H = 3.10, df = 3, p = 0.376);
Here, we tested whether rock gobies can change either their luminance (lightness) or colour depending on the background on which they are placed. As predicted, in experiment 1, fish changed in their overall luminance when put onto either a white or a black background, with individuals getting lighter or darker respectively. This led to changes in the level of similarity of fish to each background in terms of luminance, improving camouflage matching over time. In contrast, although there were some statistically significant changes in hue and saturation in this experiment too, these generally did not affect the overall match to the background, indicating that these changes were perceptually small and unlikely to be of significance in terms of camouflage, similar to other work
In contrast, in experiment 2 where fish were placed onto either red or blue backgrounds, individuals underwent marked changes in colour with regards to both hue and saturation. At least some goby species have been shown to have three cone types, sensitive to relatively shorter and medium/longer parts of the spectrum
The result that fish changed to become more red in coloration on the red background, yet that changes towards the blue colour were much smaller (they mostly become more grey in colour) is interesting. It suggests that some types of colour are easier for the fish to adopt than others. This fits with the background environment in the habitat where the fish were collected, whereby blue colours are rare, yet red encrusting algae and brown stones and seaweed are common. Past work has shown that different types of chromatophore control different colours, with black melanin being controlled by melanophores, yellow pteridine controlled by xantophores, red carotenoids controlled by erythrophores and more rarely blue cyanophores controlling a yet unknown cyan biochrome
The levels of change in luminance were relatively small in this study even for the fish that changed the most. Thus the decrease in difference to the background, although significant, was not very large. However, this did equate to a decrease in discrimination thresholds of almost 10 JNDs on average for fish on the white backgrounds. In general across the experiment fish were quite dark, and in nature they are likely to be a better match to the general colour and brightness of the substrate in the rockpools (especially the dark rocks). Therefore, in such cases in the wild when fish are already well matched in appearance to the background even relatively small differences may equate to a valuable benefit in improved camouflage. Furthermore, to our eyes, changes in the brightness of fish are clearly perceptible (
It is interesting that the ability of fish to change colour seems to be better than their ability to change brightness. Until we test fish on more natural coloured backgrounds we can only speculate as to why this may be. In the rockpool environment, the background is highly heterogeneous in terms of brightness, with stones and gravel of a range of shades occurring on a small scale (smaller than the size of the fish). Thus, overall changes in brightness may have a relatively small benefit. In contrast, some rockpools and larger backgrounds seem to have broadly different colours, meaning that colour change may be more valuable. This is likely to be especially the case with changes in shore height too, whereby there are changes in the amount of substrate types, especially greater brown and green algae cover lower down the shore.
Here, we have focussed on changes in colour and brightness using relatively artificial background appearances. Next, it will be important to test for colour change and camouflage ability on backgrounds that more closely resemble those in the environment where the individuals live. Moreover, given that gobies often have strongly contrasting patterns that appear disruptive, it would be important to test whether individuals have the capacity to change their markings on backgrounds of different marking sizes and contrasts. Previous studies have shown that flatfish are capable of impressive changes in pattern depending on the substrate appearance
While in the present study we have focussed on colour change for camouflage a number of recent studies on colour change in gobies report change colour in response to breeding, with individuals becoming less camouflaged and more attractive to mates
(XLSX)
We thank Zeehan Jaafar and another anonymous referee for a range of helpful comments on the work and manuscript. We also thank Jolyon Troscianko for help by writing the Image J calibration programmes and assistance with camera calibration.