Copeia, 2003(3), pp. 455–466

Visual Biology of Hawaiian Coral Reef . II. Colors of Hawaiian Coral Reef

N. J. MARSHALL,K.JENNINGS, W. N. MCFARLAND, E. R. LOEW, AND G. S. LOSEY

The colors of 51 species of Hawaiian reef fish have been measured using a spec- trometer and therefore can be described in objective terms that are not influenced by the human visual experience. In common with other known reef fish populations, the colors of Hawaiian reef fish occupy spectral positions from 300–800nm; yellow or orange with blue, yellow with black, and black with white are the most frequently combined colors; and there is no link between possession of ultraviolet (UV) re- flectance and UV visual sensitivity or the potential for UV visual sensitivity. In con- trast to other reef systems, blue, yellow, and orange appear more frequently in Hawaiian reef fish. Based on spectral quality of reflections from fish skin, trends in fish colors can be seen that are indicative of both visually driven selective pressures and chemical or physical constraints on the design of colors. UV-reflecting colors can function as semiprivate communication signals. White or yellow with black form highly contrasting patterns that transmit well through clear water. Labroid fishes display uniquely complex colors but lack the ability to see the UV component that is common in their . Step-shaped spectral curves are usually long-wave- length colors such as yellow or red, and colors with a peak-shaped spectral curves are green, blue, violet, and UV.

HE reasons why reef fish are so colorful ful discussion, see Endler, 1981, 1984), or dis- T have been the subject of speculation since ruption (Cott, 1940), aposomatism (Ehrlich et Darwin’s time (Darwin, 1859; Longley, 1917a; al., 1977), temperature control (Fox and Vevers, Lorenz, 1962), with the Hawaiian Islands and 1960). Some coloration could result from phy- their reef fish having been particularly closely logenetic constraint or a nonfunctional meta- scrutinized (e.g., Longley, 1918; Lorenz, 1962; bolic byproduct (Longley, 1917a,b; Fox and Vev- Barry and Hawryshyn, 1999). When faced with ers 1960). All of these possibilities are rendered the many colors of reef fish, it is tempting to more complex by the ability of almost all reef describe them using our human eyes. This can fish to change color on one or more of a num- lead to erroneous conclusions if we then try to ber of time scales from subsecond to ontoge- draw behavioral or physiological conclusions netic (Townsend, 1929; Crook, 1997a,b). Here based on this experience (Cuthill et al., 1999). we describe only the spectral distribution and Reef fish visual systems (Losey et al., 2003) and frequency of the most commonly seen colors of their light environment (Marshall et al., 2003) 51 species of Hawaiian reef fish and attempt to are quite different from ours, and the way they place this body of data in the context of previ- see their own colors must necessarily be quite ous work. different from the way we see them. Here we Use of the human visual system to describe attempt a general description of a variety of Ha- color and its uses can be very misleading (Ben- waiian reef fish colors that is independent of nett et al., 1994; Cuthill et al., 1999; Endler, human perceptual experience. Marshall et al. 1990). This is well illustrated by a consideration (2003) use this and other data to begin to mod- of the ultraviolet (UV) part of the spectrum to el the design of reef fish colors and vision. which most mammals, including humans, are These are early steps in the study of the visual relatively insensitive ( Jacobs, 1993) but which ecology of reef fish, and therefore we try to may play an important role in processes such as draw some general conclusions that we hope sexual selection (Andersson and Amundsen, will be of use in future work. 1997; Andersson et al., 1998; Bennett et al., Coral reef dwellers are certainly one of the 1996), camouflage (Lavigne, 1976; Shashar, most colorful and varied assemblages of ani- 1994), and food choice (Church et al., 1998). mals. Reasons suggested for the ‘‘bright’’ col- UV is of primary interest in this paper. It is im- oration of reef fish include territorial marking portant to realize, however, that just because we (Lorenz, 1962), sexual display (Thresher, 1984), do not see it, UV is not necessarily more im- camouflage through object or background portant to animals that do see it than is any oth- matching (Randall and Randall, 1960; for a use- er spectral region.

q 2003 by the American Society of Ichthyologists and Herpetologists 456 COPEIA, 2003, NO. 3

Approximately half of the species of reef fish- spectrometer, or ‘‘Sub-Spec,’’ an Andor Tech- es examined to this date are largely insensitive nology submersible spectrometer that is fully to UV based on ocular media transmission mea- described in Marshall (1996, 2000a). These in- surements (Thorpe et al., 1993; Siebeck and struments and their use are detailed elsewhere Marshall, 2001; Losey, et al., 2003). Although (Endler, 1990; Marshall, 1996, 2000b). the remainder may allow UV to reach the retina The sampling area of the Sub-Spec spectrom- and be detected, the value of UV-sensitive vision eter, which could be as small as 0.3 mm2 (de- to such fishes is unclear. Therefore, it is impor- pending on the range to the object being mea- tant to examine functional and ecological dif- sured) was visualized through a sighting optic ferences between reef fishes that do and those that is much like the viewfinder for a single lens that do not detect UV. This is particularly in- reflex (SLR) camera but with a sampling fiber- triguing for body colors and color patterns be- optic embedded centrally in the mirror-plane of cause it raises the possibility of a communica- the optical system. The light reflected from a tion channel open only to species with UV visual colored area on a fish was sampled through the sensitivity and closed to all others. fiber optic and stored by the spectrometer. With S2000 measurements, the bare end of a fiber MATERIALS AND METHODS optic probe attached to the spectrometer was placed close to the fish at a 458 angle so that it All fish were caught using hand nets and bar- sampled from that colored region alone. Each rier nets or, occasionally, rod and line either by measurement was an average of three to 10 sam- researchers or local tropical fish suppliers. Fish ples of the same colored zone on the fish. De- were housed in marine aquaria at the Hawai’i tailed descriptions of spectrometers in general Institute of Marine Biology, University of Ha- and their design are available at the Ocean Op- wai’i, on Coconut Island, Kaneohe Bay. Fish col- tics website: www.Oceanoptics.com. ors were occasionally measured in situ or in Sampling color patches on a fish is a subjec- small aquaria but normally out of water. Efforts tive process in the choice of color patches to were made to reduce stress (and in some cases sample. We strove to include all qualitatively dif- the resulting color change) by anaesthetizing ferent hues. Toward this goal, we also used a fish with MS222 or clove oil or by removing UV-sensitive camera to image each fish to reveal them from the water for only a few seconds for possibly inconspicuous UV reflecting areas. De- measurement. They were placed in cloth soaked tails of UV camera design are found in Marshall in seawater and the skin was kept wet during (1996) and Losey et al. (1999a). measurements. For most species, results of this technique are equivalent to those when fish are Definition of color categories and spectral shapes.— kept in water (Marshall 1996, 2000a), and mea- The same 21 human-subjective color categories surements made out of water allow more exact- identified in Marshall (2000a) are used here to ing optical alignment with small spots and provide labels and ease description and com- stripes. Clove oil occasionally results in fish tak- parison of reef fish colors, not for quantitative ing on a dark coloration after several minutes, purposes. Colors discussed in relation to these so we made measurements before this com- specific categories are written in italics: color menced. Some species (the rapid color chang- without UV-Blue, Green, Yellow, Orange, Brown/Ol- ers such as some balistids and pomacentrids, for ive, Red, Blue/Red, Blue/Far-Red, Magenta, Black, example) changed color too rapidly and fre- Labriform-Green Complex; UV containing colors— quently for us to be confident that we were mea- UV, White, Violet, Blue-UV, Blue-UV-Hump, Green- suring their ‘‘normal’’ coloration. Data from UV, Yellow-UV, Orange-UV, Red-UV, Blue/UV/Red such fish were discarded. Whether this was the Labriform-Purple Complex. These categories are case for other fish was judged subjectively (by based on the appearance of the color to us and us closely examining fish in and out of water) the shape of the spectrum of the color (Fig. 1). and it remains possible that we overlooked For example, both Red and Red/UV may appear some colors that ‘‘disappear’’ or change without identical to our visual system; however Red/UV our knowledge or are outside of the range of contains a second region of reflectance in the human vision. Attention should be paid to such UV. Good examples of Yellow/UV and Orange/UV problems in more detailed descriptions of in- can be seen in Figure 1B. dividual species in future work. General descriptions of colors or collections of a number of color categories that may all ap- Spectrophotometric measurements.—The reflectance pear similar to human perception are written in spectra from different areas of fish skin were normal font to contrast with the italicized colors measured using either an Ocean Optics S2000 defined above. For example, ‘‘yellow’’ is used to MARSHALL ET AL.—COLORS OF HAWAIIAN FISHES 457

Fig. 1. Examples of Hawaiian fish colors plotted as percent reflectance relative to a white ‘‘Spectralon’’ standard (Endler, 1990; Marshall, 1996). The colors of the curves are approximate matches to the color in life as seen by humans. Note that color reflectance takes two forms: step-shaped and peak-shaped. Colors with one of either of these shapes are classed as simple, colors with two or more of either shape (i.e., two peaks or a peak and a step or two peaks and a step as in Chlororus sordidus) are classed as complex (Marshall, 2000a). For each example, the numbers on the body areas in the photographs correspond to the numbered reflection curves for the fish. (A–B) Zanclus cornutus. (C–D) Chromis ovalis. (E–F) Chlororus sordidus. (G–H) Coris gaimard. The photographs show close-ups of body regions in this brilliantly colored fish. They are (clockwise) the dorsal fin, the base of the pectoral fin, and the tail close to the caudal peduncle. describe both color categories ‘‘Yellow’’ and ‘‘Yel- can be expressed as the wavelength correspond- low/UV.’’ ing to the 50% reflection point (R50) in the The spectrum of a color, or regions of spec- step (the half-way point from top to bottom of tra, can be described as steplike shapes (e.g., the step). This is similar to the 50% transmis- yellow or red, color 1, Fig. 1B) or peaklike (e.g., sion (T50) used to quantify ocular media blue, color 4, Fig. 1H, Chittka and Menzel, (Douglas and McGuigan, 1989; Siebeck and 1992). The wavelength of a step-shaped color Marshall, 2001; Losey, et al., 2003). Peaks are 458 COPEIA, 2003, NO. 3 5 4 5 5 4 4 4 5 6 5 5 5 4 4 4 5 lumn Media Ocular cant UV 2 2 2 2 1 1 1 1 1 1 pig- t-wavlength- V V V V V V ment Visual UV UV UV fish are those nsmission data 1 als and Methods. 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 UV nent compo- UV Red- UV Orange- X UV Yellow- UV Green- X UV- Blue/ Hump UV- Blue let Vio- ., 2003). An ‘‘X’’ indicates that the color listed in the top row approximates plex White Lab. Com- Purple ’’ unless white is the only UV-reflecting color, in which case they are scored with a OSEY ET AL plex Lab. 1 Com- Green (L XX X Ma- genta Black Far- Red Blu/ Blu/Red XX W2 XX Olive Red Brown, are those fish with UV- and violet-sensitive cone cells and UV-fish are known to lack both. Parenthetic UV XX X X X XX X X XX 1 for which we suspect UV sensitivity that has been overlooked (Losey et al., 2003). OCULAR MEDIA TYPE AND VISUAL , XXXX X XXX XXX X X XXX XXX XXXXXX XXXXXW XXX XXXX XXX XXX X X X X W5 X W 4 W UV and V 1 XX XXXXXX XXX X X X XX XX AWAIIAN REEF FISH H OLOR OF Species Blue Green Yellow Orange 1. C Taeniura lymma Acanthurus achilles Acanthurus nigrofuscus Ctenochaetus strigosus Aulostomus chinensis Rhinecanthus aculeatus Bothus pantherinus Chaetodon auriga Cirrhitops fasciatus Acanthurus dussumieri Acanthurus leucosternon Acanthurus lineatus Acanthurus olivaceus Naso lituratus Naso unicornis Zebrasoma flavescens Sufflamen bursa Chaetodon ephippium Chaetodon ocellatus Chaetodon ornatissimus Chaetodon trifascialis Chaetodon unimaculatus Forcipiger flavissimus Heniochus diphreutes ABLE Dasyatididae the color seen. AllCells fish, to unless marked the ‘‘j’’,component left were to of mature their or the reflection. terminalsummarizes dark That phases. the vertical Color is, UV line categories they content are code possess for those colors a all of peak in Marshall colors. or (2000a) which UV-reflecting and portion fish very are of are detailed low their scored in or reflectance Materi with no of a UV at ‘‘ least reflection 20% exists. between Cells 200 to and the 400 right nm. of The this ‘‘UV components’’ line co mark colors with a signifi T Acanthuridae and ecological considerations (Losey et al., 2003). Ocular media scored with a three or less pass some UV (Losey et al., 2003) and are likely to have shor ‘‘W’’. The last two columns to the right summarize cone cell visual pigment sensitivity data and vision liklihood estimates made from ocular media tra sensitive vision (UV or Violet). UV Aulostmidae Balistidae Bothidae Chaetodontidae Cirrhitidae MARSHALL ET AL.—COLORS OF HAWAIIAN FISHES 459 5 6* 6 6 5 5 6 6* 5 5 3 5 4 5 4 5 2 6 6* 5 5 5 3 Media Ocular )1 ) 2? 1 1 2 1 1 1 1 pig- V V V V ment Visual (UV (UV 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 UV nent compo- UV Red- UV Orange- UV Yellow- UV Green- UV- Blue/ Hump UV- Blue let Vio- plex White Lab. Com- Purple XX X plex Lab. Com- Green 1. Continued. X X W UV ABLE T Ma- genta Black XXX Far- Red Blu/ Blu/Red XXXX XX XX XXX XX Olive Red Brown, XX X X X XXXXXXX X X XXX XX X X XXXXXXXX XXXXX XXXX XXXXX XX XXXXX XXXXX X X W 5 XX X X X X W 5 XXXXX XX X X X X XXX X XXXX XX XX Species Blue Green Yellow Orange Ostracion meleagris Chlororus sordidus Fistularia commersonii Sargocentron spiniferum Coris gaimard Pervagor aspricaudus Centropyge flavissimus Chromis ovalies (j) Pseudanthias bicolor Arothron hispidus Zanclus cornutus Coris venustra Gomphosus varius Halichoeres ornatissimus Labroides phthirophagus Thalassoma lutescens Pervagor spilosoma Parupeneus cyclostomus Centropyge loricula Centropyge potteri Chromis vanderbilti Scarus psittacus Arothron melaegris Thalassoma duperry Pseudochelinus tetrataenia Mulloidichthys vanicolensis Parupeneus multifasciatus Ostraciidae Scaridae Fistulaeriidae Holocentridae Labridae Monacanthidae Mullidae Pomacanthidae Pomacentridae Serranidae Tetradntidae Zanclidae 460 COPEIA, 2003, NO. 3

10,000 times to create a random expected dis- tribution of spacing and clump size for compar- ison with the observations to determine wheth- er the data were, overall, nonrandomly clumped. As a guide toward a post hoc decision as to which of the visually apparent clumps in the observations were likely to be nonrandom, iterations were repeated until 6000 randomly generated clumps of at least two adjacent cells were generated. Results indicated that the prob- ability of obtaining a clump size of four or great- er was 0.051, so four was taken as the threshold clump size to qualify as a clump.

RESULTS Colors and color combinations.—The colors of 51 Fig. 2. The 50% reflection points (R50) for step- species of Hawaiian reef fish were quantified shaped colors (A) and positions of the peak in peak- (Fig. 1, Table 1). These data can usefully be shaped colors (B) in Hawaiian reef fish. The curve compared to the larger sample (264 species) for plotted through the data is a five-point moving aver- age with 1-nm wide points. the Great Barrier Reef and Caribbean (GBR&C; Marshall, 2000a). Two of the color categories identified for GBR&C fish, UV and blue/UV/red, quantified simply as the wavelength of highest were not found in Hawaiian fish but were found reflection. To examine the frequency distribu- only in one and three fish, respectively, in the tion of peak and step wavelengths, a five-point GBR&C sample (Table 1). It is unclear whether moving average with 1 nm in each step was plot- this has any phylogenetic and/or adaptive ex- ted through the frequency distribution data to planation and may simply reflect undersam- aid visual estimation of the distribution. pling of Hawaiian reef fish. All but six species possessed at least one color Monte Carlo test.—Visual inspection of the fre- with a significant UV component. No clear quency distribution for the wavelengths peaks trend was noted with regards to body region and R50 steps (Fig. 2) suggested nonrandom containing UV or human visible color with clumping of colors around certain wavelengths. which UV was associated although there was a A Monte Carlo test was performed to test the general trend in GBR&C fish for UV to be significance of this impression. Each cell of this found in facial markings and fin margins. This distribution (each bar of the histogram) repre- is true for a few Hawaiian species, such as Naso sents a 5 nm wide sample. Two features that de- lituratus that possesses Blue/UV fin margins and fine the degree to which clumps appear in a Parupeneus multifasciatus that has Red/UV eyes frequency distribution were tabulated. First we and Violet facial markings. found the average count for all nonzero cells of Of the 46 species for which we have vision the distribution. We then tabulated ‘‘spacing’’ likelihood estimates and coloration measure- as the mean distance between nearest-neighbor ments, only 17 are likely to have short-wave- cells of the frequency distribution that were length-sensitive vision (Table 1), and only three greater than the mean count. As a second fea- of these lack UV-reflective colors. This is not, ture, we defined a clump to be a series of ad- however, unexpected since only five of the 51 jacent cells that were all greater than the mean species sampled lacked UV-reflective colors. If count. We then measured mean ‘‘clump size’’ we ignore the uneven sampling of families in as the number of adjacent cells in a clump our data and perform a contingency test for where all cells were greater than the mean likely possession of short-wavelength-sensitive vi- count. Nonrandom clumping will result in a sion versus UV-reflective colors, we find a com- small spacing statistic and a large clump size. plete lack of relationship (chi-square 5 1.3, df During each iteration of the Monte Carlo test, 5 1, P . 0.1). A similar lack of relationship be- a random frequency distribution was created tween UV colors and UV sensitivity was found over the same range of wavelengths and total in the larger sample of GBR&C fish examined count as that of the observations. Spacing and by Siebeck and Marshall (2001). clump size were measured for the random dis- The most frequent colors (Table 2) were yel- tribution. This iterative step was repeated lows (35/51 species), blues (28/51 species), MARSHALL ET AL.—COLORS OF HAWAIIAN FISHES 461 black (24/51 species), and white (25/51 spe- and bright (saturation, chroma, and brightness cies). The most frequent combinations were yel- are often impossible to completely disentangle). low with blue (24/51 species), black with white A saturated color, at some point in the spec- (17/51 species) and yellow with black (16/51 trum, contains a considerable and often sudden species). Combinations found in Australian fish- change in reflectance, for example, the Yellow es (Marshall, 2000a), but not in Hawai’i, were of Zanclus cornutus (Fig. 1). This color is also greens with yellows, greens with reds, and brighter than Z. cornutus orange (from 350—700 greens with white. Our relatively small sample nm at least, Fig. 1) as it reflects more light but size, the depauperate Hawaiian fish fauna, and, is only slightly different in hue content. That is, in particular, uneven sampling of species be- the overall shapes of the curves are similar. In tween families, preclude strong conclusions. any detailed study of fish coloration, it is critical One obvious trend is that butterflyfish are fre- to describe the colors in these terms and not by quently colored black with white, yellow with matching the color against some human per- white and yellow with black, usually with all ceptual standard. three colors present. What is the significance of color clusters?—Cluster- Color distribution throughout the spectrum.—The ing could result from constraints in the chemi- peaks and steps of the colors of Hawaiian reef cal or physical nature of pigments and structur- fish are not randomly distributed over the spec- al colors (for an excellent resume of the nature trum (P , 0.025 by Monte Carlo test, Fig. 2). of animal colors, see Fox and Vevers, 1960). Of There are regions where steps or peaks occur far more interest to us would be clustering that more often. For steplike colors, R50 clumps are results from adaptive interaction among color- evident at 347, 385, 515, 575, and 730 nm, but ation, environment, and visual pigments. the Monte Carlo does not support the signifi- Our data agree with two, seemingly universal, cance of the visually apparent 730 nm peak. aspects of natural biological colors found in Peak-shaped colors have maximal reflection flowers (Chittka and Menzel, 1992; Kevan et al., with significant clumps at 400, 450 and 505 nm, 1996), fish (Marshall, 2000a,b), and birds (Fin- but the Monte Carlo did not support the visually ger and Burkhardt, 1994; Vorobyev et al., 1998). suggestive clump at 355 nm. On a larger scale, First, colors constructed from steps mostly lie the region between 400 and 500 nm contains beyond 500 nm, that is, yellow, orange, and red. only one step-shaped color, but contains 55 of It is rare to find a body or flower color, that is, the 104 peak-shaped colors that were plotted. for instance, red and has a peak-shaped spec- Because we made every attempt to include any trum (an exception to this is the red feathers perceptually different hue, it is unlikely that the of birds of paradise and hummingbirds, Voro- gaps in this distribution are caused by sampling byev et al. 1998). Second, the converse is true error. In addition, Marshall (2000a) found an at the other end of the spectrum. Greens, blues, almost identical distribution in a much larger violets, and UV colors tend to possess peak- sample of Australian reef fishes. shaped spectra. Almost no steplike colors have R50 points be- tween 400 and 500 nm. A similar but less clear- DISCUSSION cut trend is seen in the R50 points for Pacific Cautionary remarks on color descriptions.—Describ- reef fish (Marshall, 2000a; NJM, unpubl.). The ing colors in a meaningful way is not trivial, and exception to the above trends, step-shaped there is some confusion of nomenclature, es- clumps at 347 and 385 nm, all belong to colors pecially where the terms ‘‘bright’’ or ‘‘colorful’’ of white or almost white appearance to us and, are used. Endler (1990) provides a useful look therefore, reflect across much of the spectrum. at the quantifiable attributes of colors and, as As is the case for ‘‘white’’ flowers and bees (Kev- others have done, identifies hue, chroma (or an et al., 1996), these fish whites (to humans) saturation) and brightness as words that de- will not appear white (or at least chromatically scribe quantifiable aspects of colors. Hue refers flat) to fish with short-wavelength-biased visual to the spectral location of the step(s) or peak(s) systems, especially those with UV receptors. of the color. Chroma defines the restriction of The reason for the gap in steplike R50 points the reflection to only certain portions of the between 400 and 500 nm is elusive. Colors that spectrum such as a very steep step or narrow appear to be UV, blue or green to visual systems peak. Brightness refers to the overall number of must necessarily not reflect strongly beyond 500 photons reflected. When reef fish are described nm or they would acquire a reddish tint. The as colorful or bright, this often means they con- sensation of color in this short wavelength part tain many different hues that are both saturated of the spectrum is the result of peak-shaped 462 COPEIA, 2003, NO. 3 OS— 1 ); Bls UV / This coding for color yellow and white. green W—blues with 1 ) with greens ( ; Bls black blue far red and red / B—blues with 1 ); Bls UV / red UV, blueUV hump, blue / and red blue, blue Gs—blues ( . Fish that have the color combinations listed in the top row are scored with an X in the table. Color combinations 1 ISH Rs–blues with reds ( F 1 EEF R ); Bls UV / AWAIIAN H for instance are Bls orange blue and combinations is repeated for greens, yellows, oranges, and reds. Labriform color combinations are shown in the last column. orange OMBINATIONS OF C OLOR 2. C ABLE T blues with oranges ( for colors found adjacent to MARSHALL ET AL.—COLORS OF HAWAIIAN FISHES 463 . ONTINUED 2. C ABLE T 464 COPEIA, 2003, NO. 3 spectra (Fig. 1). Yellow, orange, and red colors Is there functional significance to the combinations of could also be the result of peak-shaped spectra; colors and are there phylogenetic trends?—The color however, for unknown reasons (perhaps chem- categories Blue, Blue/UV, Yellow, and Orange are ical constraints to pigments), fish skin colors at approximately twice as abundant in Hawaiian this end of the spectrum all result from steplike reef fish as elsewhere (compare Table 1 with spectra whose R50 points range from 500—700 appendix 1 in Marshall, 2000a). Approximately nm. There is no apparent reason why steplike the same proportion (20%) of relatively abun- spectra should not exist between 400 and 500 dant Hawaiian fish as Great Barrier Reef and nm except that they would appear spectrally Caribbean fish has been examined and both broad, whiteish or grey/greenish, to any visual samples of fish were caught by similar methods. system and may not convey useful information Whether such a dataset can be linked to phy- for contrast with longer or shorter wavelength logenetic trends, especially in the relatively spe- colors. cies-depauperate Hawaiian waters (Randall, 1998) or to more functional reasons such as dif- Is there any relationship between possession of UV-re- ferences in water quality remains obscure. How- flective colors and UV-sensitive vision?—As noted ever, if we confine the comparison to the three previously for Pacific and Caribbean species families for which we have the most data in both (Losey et al., 1999; Marshall 2000a; Siebeck and Hawaii and Australia (Labridae, Chaetodonti- Marshall, 2001), not all species in possession of dae, and Acanthuridae) the same observation UV colors transmit UV to the retina (Table 1; of increased relative abundance of Blue, Blue/ Losey et al., 2003). As a result, these species can- UV, Yellow, and Orange in Hawaiian waters is not see their own ultraviolet colors, leaving two maintained. possibilities: the UV part of the color is non- As noted previously (Loew and McFarland, adaptive (at least visually), or it is adaptive but 1990; Longley, 1917a; Lythgoe, 1979), the com- only with respect to the visual system of another binations of colors used by reef fish are not ran- species such as a predator. In fact, possession of dom. The color pair combination of yellows or UV-reflective colors appears to have no value as oranges with blues, Black with White, and yellows a predictor of short-wavelength-sensitive vision. with Black are frequently found (lower case, nor- mal font, denotes a generalized, human percep- tual, description of more than one color cate- Do colors containing a UV component have a special gory). Possible functional reasons for these function?—The colors of most Hawaiian reef fish combinations are examined in detail in Mar- contain a UV reflective component (Table 1) shall et al. (2003), but it is worth noting here and, as UV-blind humans, it is tempting to spec- that all are highly contrasting. Also these colors ulate that UV colors combined with an ability transmit particularly well and form strong dis- to see UV may open the possibility for a ‘‘se- ruptive patterns in marine waters making them cret’’ or limited-audience communication chan- ideal for communication and/or disruptive nel. UV colors have the potential to be useful camouflage in the marine environment (Lyth- for close-range communication as these wave- goe, 1979; Marshall, 2000a). lengths of light are rapidly attenuated in turbid Phylogenetic trends are difficult to identify media such as reef water (Lythgoe 1979; Sie- without more extensive sampling, but two beck and Marshall 2000, 2001). Although obvi- trends are obvious. First, labroid fishes, the par- ous to nearby conspecifics or competitors, sig- rotfishes and wrasses, possess highly complex nals using UV will not transmit well to more dis- colors that are almost unique to this group: La- tant eyes of eavesdropping predators (Losey et briform purple, Labriform green, and Blue/red or al., 1999a; Marshall 2000a). Chromis ovalis and Blue/far red (Tables 1—2 and Fig. 1B, and fig. 1 Parupeneus multifasciatus, for example, both pos- and appendix 1 in Marshall, 2000a). These fish sess colors that are particularly strongly reflec- lack short-wavelength-sensitive vision but have tive in the UV, with peak reflectance over 40% complex ocular filters (Siebeck and Marshall, at around 360 nm in both species, and both are 2000, 2001) and striking ontogenetic color UV sensitive (either with peak sensitivity in the changes that are important in their social or- UV or violet peaking sensitivity, which also con- ganization and recognition of sex-changed in- tains UV sensitivity, Fig. 1, Table 1). Losey dividuals (Warner, 1984). Unfortunately, their (2003) has described a similar system for young retinas have proven to be extremely difficult to Dascyllus spp., damselfish that possess a very analyze (Losey et al., 2003), but these complex short-wavelength color patch that would be vis- colors suggest that renewed efforts should be ible to nearby conspecifics but largely escape made to carefully describe their visual systems the attention of predators. and relate them to body coloration. Second, MARSHALL ET AL.—COLORS OF HAWAIIAN FISHES 465 both chaetodontids and pomacanthids are of- jective assessment of avian sexual dichromatism. ten colored in bold black, white, and yellows and Am. Nat. 160:183–200. blues. Such bold markings may be important in DARWIN, C. 1859. On the origin of species by means maintaining contact between conspecifics and of natural selection. Murray, London. DOUGLAS, R. H., AND C. M. MCGUIGAN. 1989. The have even been implicated in aposomatism (Er- spectral transmission of freshwater teleost ocular lich et al., 1977). media—an interspecific comparison and a guide to potential ultraviolet sensitivity. Vision Res. 29:871– 879. ACKNOWLEDGMENTS ENDLER, J. A. 1981. An overview of the relationships between and . Biol. J. Linn. Soc. 16: Many thanks to U. Siebeck for useful discus- 25–31. sions and data, T. Yoshikawa, K. Asoh, J. Zam- ———. 1984. Progressive background matching in zow, and P. Nelson for ideas and fish. J. Randall moths, and a quantitive measure of crypsis. Ibid. 22: provided the photographs in Figure 1C and E. 187–231. N. Hart and two anonymous reviewers supplied ———. 1990. On the measurement and classification helpful suggestions on an earlier draft. This of color in studies of animal color patterns. Ibid. 41: work was funded by the ARC in Australia, Na- 315–352. tional Science Foundation (OCE9810387) in FINGER, E., AND D. BURKHARDT. 1994. Biological as- pects of bird coloration and avian color vision in- the United States and NERC and the BBSRC in cluding ultraviolet range. Vision Res. 34:1509–1514. the United Kingdom. Work was conducted in FOX, H. M., AND G. VEVERS. 1960. The nature of ani- accordance with University of Hawaii IACUC mal colors. Sidgwick and Jackson, London. protocol 95–012. Fishes were captured in accor- JACOBS, G. H. 1993. The distribution and nature of dance with the Hawaii Institute of Marine Biol- color vision among the mammals. Biol. Rev. 68: ogy collection permit. This is contribution 1160 413–471. of the Hawai’i Institute of Marine Biology. KEVAN, P., M. GIURFA, AND L. CHITTKA. 1996. Why are there so many and so few white flowers? Trends Plant Sci. 1:280–284. LITERATURE CITED LAVIGNE, D. M. 1976. Counting harp seals with ultra- violet photography. Polar Rec. 18:269–277. ANDERSSON, M., AND T. AMUNDSEN. 1997. Ultraviolet LOEW, E. R., AND W. N. MCFARLAND. 1990. The un- color vision and ornamentation in bluethroats. derwater visual environment. p. 1–43. In: The visual Proc. R. Soc. Lond. B. Biol. Sci. 264:1587–1591. system of fishes. R. H. Douglas and M. B. A. Djam- ———, J. ORNBORG, AND M. ANDERSSON. 1998. Ultra- goz (eds.). Chapman and Hall, London. violet sexual dimorphism and assortative mating in LONGLEY, W. H. 1917a. Studies upon the biological blue tits. Ibid. 265:445–450. significance of . I. The colors and BARRY, K. L., AND C. W. HAWRYSHYN. 1999a. Effects of color changes of West Indian reef-fishes. J. Exp. incident light and background conditions on po- Biol. 1:533–601. tential conspicuousness of Hawaiian coral reef fish. ———. 1917b. Studies upon the biological signifi- J. Mar. Biol. Ass. U.K. 79:1–14. cance of animal coloration. II. A revised working BENNETT, A. T. D., I. C. CUTHILL, AND K. J. NORRIS. hypothesis of mimicry. Am. Nat. 11:257–285. 1994. Sexual selection and the mismeasure of color. ———. 1918. Habits and coloration of Hawaiian Am. Nat. 144:848–860. Brachyura and fishes, with a note on the possibility ———, ———, AND E. J. MAIER. 1996. Ultraviolet vi- of submarine color-photography. Year Book Car- sion and mate choice in zebra finches. Nature 380: negie Inst. 17:158–163. 433–435. LORENZ, K. 1962. The function of color in coral reef CHITTKA, L., AND R. MENZEL. 1992. The evolutionary fishes. Proc. R. Inst. Gt. Brit. 39:282–296. adaptation of flower colors and the insect pollina- LOSEY, G. S. 2003. Crypsis and communication func- tors’ color vision. J. Comp. Physiol. A 171:171–181. tions of UV-visible coloration in two coral reef dam- CHURCH, S. C., T. D. BENNETT, I. C. CUTHILL, AND J. selfish, Dascyllus aruanus and D. reticulatus. Anim. C. PARTRIDGE. 1998. Ultraviolet cues affect the for- Behav. In press. aging behavior of blue tits. Proc. R. Soc. Lond. B. ———, W. N. MCFARLAND, E. R. LOEW, J. P. ZAMZOW, Biol. Sci. 265:1509–1514. P. A. NELSON, AND N. J. MARSHALL. 2003. Visual bi- COTT, H. B. 1940. Adaptive coloration in animals. Me- ology of Hawaiian coral reef fishes. I. Ocular trans- thuen and Co. Ltd., London. mission and retinal sensitivity. Copeia 2003:433– CROOK, A. C. 1997a. Color patterns in a coral reef 454. fish. Is background complexity important? J. Exp. LYTHGOE, J. N. 1979. The ecology of vision. Clarendon Mar. Biol. Ecol. 217:237–252. Press, Oxford. ———. 1997b. Determinants of the physiological col- MARSHALL, N. J. 1996. Measuring colors around a cor- or patterns of juvenile parrotfish, Chlorurus sordidus. al reef. Biophoton. Int. July/Aug.:52–56. Anim. Behav. 53:1251–1261. ———. 2000a. The visual ecology of reef fish colors, CUTHILL, I. C., A. T. D. BENNETT, J. C. PARTRIDGE, AND p. 83–120. In: Animal signals. Signalling and signal E. J. MAIER. 1999. Plumage reflectance and the ob- design in animal communication. Y. Espmark, Y. 466 COPEIA, 2003, NO. 3

Amundsen, and G. Rosenqvist (eds.). G. Tapir, ing pigments in the fish lens. I. Phylogenetic distri- Trondheim, Norway. bution and identity. Ibid. 33:289–300. ———. 2000b. Communication and camouflage with THRESHER, R. E. 1984. Reproduction in reef fishes. the same bright colors in reef fish. Philos. Trans. R. T.F.H. Inc. Ltd., Neptune City, NJ. Soc. Lond. 355:1243–1248. TOWNSEND, C. H. 1929. Records of changes in color among fishes. Zoologica 9:321–378. ———, K. JENNINGS, W. N. MCFARLAND, E. LOEW, AND VOROBYEV, M., D. OSORIO, A. T. D. BENNETT, N. J. G. S. LOSEY. 2003. Visual biology of Hawaiian coral MARSHALL, AND I. C. CUTHILL. 1998. Tetrachromacy, reef fishes. III. Environmental light, and an inte- oil droplets and bird plumage colors. J. Comp. Phy- grated approach to the ecology of reef fish vision. siol. A 183:621–633 Copeia 2003:467–480. WARNER, R. R. 1984. Mating behavior and hermaph- RANDALL, J. E. 1998. Zoogeography of shore fishes of roditism in coral reef fishes. Am. Sci. 72:128–136. the Indo-Pacific region. Zool. Stud. 37:227–268. ———, AND H. A. RANDALL. 1960. Examples of mim- (NJM) VISION,TOUCH AND HEARING RESEARCH icry and protective resemblance in tropical marine CENTRE,UNIVERSITY QUEENSLAND,BRISBANE, fishes. Bull. Mar. Sci. Gulf Carib. 10:444–480. QUEENSLAND,AUSTRALIA; (WNM) SCHOOL OF SHASHAR, N. 1994. UV vision by marine animals: main- FISHERIES AND FRIDAY HARBOR MARINE LABO- ly questions, p. 201–206. In: Ultraviolet radiation RATORY,UNIVERSITY OF WASHINGTON,FRIDAY and coral reefs. D. Gulko and P. L. Jokiel (eds.). HARBOR,WASHINGTON 98250; (ERL) DEPART- Hawai’i Institute of Marine Biology, Honolulu. MENT OF BIOMEDICAL SCIENCES,VETERINARY SIEBECK, U. E., AND N. J. MARSHALL. 2000. Transmis- COLLEGE,CORNELL UNIVERSITY,ITHACA,NEW sion of ocular media in labrid fishes. Philos. Trans. YORK 14853; AND (GSL) HAWAI’I INSTITUTE OF R. Soc. Lond. B Biol. Sci. 1257–1262. MARINE BIOLOGY,UNIVERSITY OF HAWAI’I, P.O. ———, AND ———. 2001. Ocular media transmission BOX 1346, KANEOHE,HAWAII 96744. E-mail: of coral reef fish—can coral reef fish see ultraviolet [email protected]. Send re- light? Vision Res. 41:133–149. print requests to NJM. Submitted: 13 Feb. THORPE, A., R. H. DOUGLAS, AND R. J. W. TRUSCOTT. 2001. Accepted: 1 Feb. 2003. Section editor: 1993. Spectral transmission and short-wave absorb- W. L. Montgomery.