We separated excitation from emission wavelengths in the field under natural, day-time solar illumination by SCUBA-diving below the penetration depth of the red component of sunlight (15–30 m) using masks and cameras equipped with a red filter blocking wavelengths below 600 nm (Fig. 1). This revealed widespread, natural red fluorescence produced by many microorganisms, plants and invertebrates (Fig. 2, see Additional file 1). The latter included mostly corals 10111213, but also species for which red fluorescence has never been described before, such as a polychaete species, several sponges (not shown) and feather stars (details in Fig. 2).

Of central importance here is our discovery of red fluorescence in reef fishes. Using the principle described above to distinguish "regular" red colouration from red fluorescence (Fig. 3) we identified at least 32 fish species belonging to 16 genera in 5 families that fluoresced visibly in red (Fig. 4, 5, Table 1, see also Additional file 2). Fluorescent patterns usually included the eye ring and parts of the head or thorax and varied substantially between congeners (e.g. in the genera Eviota or Enneapterygius). Fins rarely fluoresced, except for the anal fin (some Gobiidae), the first dorsal fin (Tripterygiidae) or the tailfin (Syngnathidae). A 'whole body glow', including all fins, was present in the small wrasses Pseudocheilinus evanidus and Paracheilinus octotaenia. Visual and photographic evidence from the field was confirmed by extensive fluorescence microscopy and spectrometry of representative cases (Table 1). Fluorescence showed peak emission around 600 nm in most species (Fig. 6a, Table 1). Enneapterygius pusillus showed a second small peak at around 680 nm. P. evanidus differed from all others by having a double peak at 650 and 700 nm. We tested various light sources, including UV, but were not able to detect fluorescent emission at other (shorter) wavelengths in the fish described here. Anecdotal field observations suggest that yellow fluorescence may be present in other fishes (pers. obs.).

Dissection revealed that red fluorescence was associated with guanine crystals in pipefish, triplefins, blennies and gobies (Fig. 7, Table 1). Guanine crystals are produced by iridophores and are well known as the source of silvery reflection and iridescence in bony fish14. However, they have never been described to show strong red fluorescence. In E. pellucida and Ctenogobiops tangaroai, only about half of the crystals isolated from the eye rings showed bright red fluorescence (Fig. 7b), suggesting that the fluorescing substance is produced or sequestered independently from the crystals. Preparations of crystals maintained strong fluorescence after prolonged storage in a dried or liquid form (70% EtOH or 4% formaldehyde), allowing us to confirm fluorescence in preserved gobies collected up to 5 years before (Collection JH, Vienna, Table 1). This is in striking contrast to reflective red pigmentation, which bleaches out within hours after fixation. We did not find fluorescent guanine crystals in P. evanidus. Here, microscopic investigation suggests the presence of a fluorescent pigment associated with the bony tissue of scales and fin rays (Fig. 7f). This pigment has a chemical stability in preserved specimens that matches that of guanine-linked red fluorescence.

As a control, we isolated guanine crystals from non-fluorescing fish. In T. cana (Fig. 3), Psetta maxima (turbot) and Engraulis encrasicolus (anchovy), we found less than 1% of the crystals to fluoresce in red (Fig. 7e), suggesting that a very low level of red fluorescence may be widespread and is not limited to reef fish.

With few exceptions 151617 most marine fishes lack a sensitivity reaching into the red part of the spectrum5. But as far as we know, none of the fish genera listed in Table 1 has ever been tested for its spectral sensitivity. Here, we measured the retinal spectral sensitivity of wild-caught E. pellucida. Their retina contains rods and both single and twin cones as previously found in one other goby18. Mean wavelengths of maximum absorbance (λ_max) of the visual pigments in the rods and short-wavelength-sensitive (SWS) single cones were 497 and 458 nm, respectively. λ_max values in the twin cones were highly variable, ranging from 518–546 nm, although spectra were clustered into two groups with mean λmax values at 528 (medium-wavelength-sensitive, MWS) and 540 nm (long-wavelength-sensitive, LWS; Fig. 6b). The majority of twin cones had the 528 nm pigment in both members (528/528 twins) although both 528/540 and 540/540 twins were also observed. These results suggest that there may be more than two distinct visual pigments in the twin cones, and/or co-expression of opsin genes within the same outer segment, as in other fish19. Regardless, Fig. 6b shows that there is considerable overlap between the red fluorescence emission spectrum and the absorbance spectra of the visual pigments in the twin cones. It is highly likely, therefore, that this species can see its own fluorescence. A similar result is expected for Corythoichthys pipefish, given that long wavelength sensitivity is known from other Syngnathids 17.

Why do some reef fish fluoresce strongly in red, whereas most do not? Although non-fluorescent (reflective) red pigmentation is widespread in reef fish20, it appears grey or black at depth (Fig. 3), allowing fish to blend in with their background21. Red fluorescence, however, does the opposite: by emitting a colour that is lacking from the environment, a fish contrasts more against its background. As a result, red fluorescence may function as a communication or attraction signal, as proposed for red-bioluminescent deep sea fishes16 and siphonophores22, ultraviolet-reflecting reef fishes23 and green fluorescent mantis shrimp24 and parrots25.

We see four reasons why red fluorescence may be part of a private communication system in fish. Firstly, peaking mostly around 600 nm, red fluorescence is at the borderline of what is visible to many marine fishes, and due to rapid attenuation of red light by water, even those that can see red will be able to see it over short distances only. This is suggestive of an adaptive shift away from the "public area" of the visual spectrum into a bordering "private area". Secondly, most species found to fluoresce are small, benthic, pair- or group-living fishes, often with conspicuous intra-specific behaviours, but quite cryptic colouration in other parts of the visible spectrum. Thirdly, there is strong inter-specific variation within and between closely related genera suggestive of species-recognition (Fig. 4, 5). Finally, fluorescence is present in structures that are used in intra-specific signalling. This is true for the first dorsal fin in triplefins (Fig. 8a, Additional file 3, sequence 2b), the tailfin in pipefish (Fig. 8b, Additional file 3, sequence 1) and possibly the anal fin in E. pellucida (Fig. 3) and E. guttata. Fluorescent eye rings may function as an indicator of presence (Fig. 8c) or reveal the direction of gaze (Fig. 4e–g, Fig. 8a, d, Additional file 3, sequence 2b). Since iridophores are involved in rapid colour change in other fishes 2627, fluorescent fish may also be able to change the strength of fluorescence.

One reason why red fluorescence in fishes may have escaped attention is because fluorescence is usually observed during night dives using strong UV light sources. Although this reveals the spectacular bluish-green fluorescence typical of many corals, it is not a good strategy to visualize red fluorescence because excitation and emission wavelengths are far apart, making red fluorescence weak relative to the shorter wavelength emissions. Moreover, most fish are hiding at night, explaining why red fluorescence had not yet been described for fishes. Up till now we have not yet found a nocturnal fish that fluoresces in red, which fits well with widespread reflectant red colouration indicating that crypsis is more important in these species.

For a correct interpretation of our results, it is important to stress that it is not required to look through a red filter to "see" red fluorescence during daytime, particularly not for the stronger cases. Using a filter merely facilitates its detection in the field. Without a filter, red fluorescence at depth is usually only one element in complex, multi-spectral, red-containing colours, such as orange-brown, red-brown, pink, lilac, violet or even bright white in some encrusting corals. Cases of pure fluorescence-based red are thus far limited to a few corals and sponges. Whether pure or mixed, red-containing colours cannot be generated by reflectance and therefore sources of red fluorescence can be identified without a filter at depth. This merely requires special attention by the diver: Although e.g. red-brown is prevalent on reefs, it is not perceived as unusual or unexpected by most. Consequently, while red masks are invaluable for initial detection, fluorescence photography is easiest using a regular mask and a camera with a red filter. An alternative is to refrain from a red filter altogether and to adjust the white balance of the camera manually to local light conditions using a white slate. This will emphasize reds, while suppressing, but not removing, shorter wavelengths. At depth, this will highlight any red fluorescence. Applying this to fish colour vision, we suspect that, if red fluorescent fish can indeed see their fluorescence, they see it as enhanced contrast involving pink, red-brown or other red-containing, mixed colours in an otherwise blue-green environment. Consequently, behavioural experiments should not test the ability of fish to see weak red light, but their ability to distinguish between multi-spectral signals with and without a weak red component in a blue-green flooded environment.

Finally, we want to make a cautionary remark on diving safety. Diving with a red mask is similar to night diving, with dramatically reduced light intensities and viewing distances. Disorientation becomes a serious problem. Moreover, it takes several minutes to adapt to the darkness. Staying in a small, familiar area and moving slowly and carefully is crucial. Furthermore, it is essential to take a torch to read equipment. Dials, indicators and computer backlights are either reflectant or luminesce blue or green, making them effectively illegible at depth in the absence of a local white (red-containing) light source. To circumvent this problem, we also used the Oceanic DataMask which has a built-in dive computer that can be read irrespective of any filter attached to the front. Because of these unfamiliar restrictions, we recommend that only experienced divers use this procedure and that only one partner in a buddy team uses a red mask at any given time. We also recommend attaching filters in such a way that they can be instantly removed without having to change to a spare mask, which one should carry nevertheless.

Fish were collected in Dahab (Dahab Marine Research Centre, Egypt, with permit from NCS/EEAA) and at Lizard Island Research Station (Australia) (GBRMPA permit G05/13668.1 to UES) by anaesthetising individuals with clove oil (10 ml in 50 ml EtOH added to 200 ml sea water). Fish for laboratory work in Tübingen were obtained from the sustainable aquarium trade and kept in accordance with German animal care legislation. Species were identified using standard20 and specialised 282930 literature.

Underwater pictures of fluorescence under natural illumination were taken with a Canon PowerShot G7 or G9, Canon Ixus 750 and Sony HD HDR-CX6EK using Lee Medium Red filters or Edmund Optics Y59-642 glass filters. Macro photography in the laboratory was done using Canon PowerShot G7 or G9 with blue LED light source (LUXEON K2 LXK2-PB14-P00, λmax = 470 nm) and Edmund Optics Optical filter G43-943. All of these filters block most light below 600 nm, but allow some blue or green light to leak through (e.g. Fig. 4a–b). This blue-green hue was removed in most pictures by only showing the red channel (590–700 nm) of the RGB digital image. Other pictures were taken using a Leica MZ16F fluorescent stereo-microscope on anaesthetised, freshly killed or preserved animals. A Leica DM5000 fluorescence microscope was used to document guanine crystals.

Fluorescent emission spectra were recorded using an AvaSpec-2048-USB2 spectrometer with a green laser (λ_exc = 532 nm) as excitation light source, yielding reproducible emission spectra of from live and anaesthetised fish. We supplemented these measurements with high-sensitivity spectrometry of fluorescent guanine crystals and scales of C. tangaroai, E. pellucida, E. sebreei, E. zebrina, F. longispinus, I. decoratus, P. micheli and P. evanidus. Here, emission spectra were induced using λ_exc = 473 nm laserlight (LDH – P400B Picoquant) on a custom-built Zeiss Axiovert 135 TV confocal laser scanning microscope equipped with an avalanche photodiode (APD, SPCM-AQR-14, Perkin Elmer) and a spectrometer with cooled CCD camera (Spec-10:100B, Princeton Instruments). Since these measurements coincided well with those obtained with the AvaSpec-2048-USB2, data from both methods were pooled (Fig. 6). Spectra were normalized by setting maximum emission equal to 1. Curves for genera (Fig. 6) are averages of the average spectra of each available species.

For electron scanning micrographs, guanine crystals were isolated14 from E. pellucida, coated with 20-nm Au/Pd and photographed using a Cambridge Stereoscan 250 Mk2 electron microscope (Fig. 7d).

For microspectrometry of E. pellucida visual pigments, five gobies (13–18 mm standard length) were dark adapted for at least 1 hour and their retinas removed under infrared illumination with the aid of image converters. Each retina was dispersed mechanically and mounted in 340 mOsm kg-1 phosphate-buffered saline containing 10% dextran. Absorbance spectra (325–800 nm) of visual pigments housed in individual photoreceptor outer segments were measured using a wavelength-scanning microspectrophotometer, as described elsewhere31 (Fig. 6b).