Thresholds of polarization vision in octopuses

Polarization vision is widespread in nature (Roberts et al., 2011), mainly among the invertebrates, where it contributes to a variety of behavioural tasks including navigation (Wehner, 1976), habitat localization (Schwind, 1991) and communication (Chiou et al., 2008). In terrestrial environments, this sensory capacity is best understood in the dorsal rim area of the eye, which is directed towards the sky to enable the detection of the pattern of celestial polarization (Wehner, 1976). Whole-field (whole-eye) polarization sensitivity also exists in many insects, for example butterflies (Kelber et al., 2001), moths (Belušič et al., 2017), dragonflies (Laughlin, 1976) and biting flies (Meglič et al., 2019), but the functional significance is less well studied (Heinloth et al., 2018; Kelber et al., 2001). Underwater, polarization sensitivity can also be found across the whole eye of many animal species, such as all cephalopods and most crustaceans (Fineran and Nicol, 1978; Moody and Parriss, 1961; Talbot and Marshall, 2011; Waterman et al., 1969). With polarization sensitivity across the whole visual field, polarization information can be used for diverse functions such as prey or predator detection (Temple et al., 2012) and communication (Cronin et al., 2003; Marshall et al., 2019; Marshall et al., 2014). This is partly because many aquatic environments are surrounded by stable polarization backgrounds (Cronin and Shashar, 2001; but see Johnsen et al., 2011; Shashar et al., 2000; Shashar et al., 1998). Additionally, several species of crustacean and cephalopod incorporate optical structures in their cuticle or skin, which produce strongly polarized signals for communication (Chiou et al., 2008; Marshall et al., 2019; Shashar et al., 1996).

It has been proposed that, in some underwater situations, polarization vision may prove more reliable than colour vision (Marshall and Cronin, 2011). Light transmission through seawater is strongly wavelength dependent, with short and long wavelengths being attenuated more rapidly than medium wavelengths (∼475 nm) (Smith and Baker, 1981). As a result, colour vision becomes less useful with depth (Bowmaker, 1995; Lythgoe, 1979; McFarland and Munz, 1975) and animals tend to tune their colour vision systems to take advantage of the wavelengths of light available to them (Cheroske and Cronin, 2003; Cheroske and Cronin, 2005; Cummings and Partridge, 2001; Lythgoe, 1968). The full range of polarization contrasts, however, can be present at any depth, so perhaps for this reason many marine animals may rely more heavily on polarization vision than colour. It must be noted though, that polarization contrasts are quickly eroded by veiling light and so are only effective over relatively short distances underwater (<12 m; Johnsen et al., 2016).

Cephalopods are a prime example of polarization specialists. Their advanced camera-type eyes use only a single type of visual pigment, rendering their visual world entirely monochromatic (Chung and Marshall, 2016; Marshall and Messenger, 1996; Mathger et al., 2006; Messenger, 1977). Instead, their rhabdomeric photoreceptors are precisely ordered into two channels of polarization sensitivity, resulting in dipolatic vision across the whole visual field (Labhart, 2016; Moody and Parriss, 1961; Talbot and Marshall, 2011). This provides a highly sensitive polarization vision system ideally suited to detecting object-based contrasts in their marine environment (Shashar and Cronin, 1996; Temple et al., 2012), roughly analogous to how we use our own colour vision system. Previous studies with cephalopods have demonstrated both neural and behavioural responses to very small contrasts in polarization (Saidel et al., 2005; Temple et al., 2012). However, these studies used stimuli with degrees of polarization much higher (∼1.0) than those found in nature (<0.7) (Horváth et al., 2015; Novales Flamarique and Hawryshyn, 1997) and so do not represent the range of contrasts typically encountered by these animals in the wild. To investigate this further, we examined the threshold of detection of polarization contrasts in two species of octopus (Abdopus aculeatus and Octopus cyanea) by varying both angle of polarization (AoP) and degree of linear polarization (DoLP) using a modified liquid crystal display (LCD) to deliver dynamic polarization stimuli (Basnak et al., 2018; Glantz and Schroeter, 2007; How et al., 2012; Pignatelli et al., 2011; Temple et al., 2015). We compared their polarization sensitivity with measurements of polarization taken from their natural environment.

Experiments were performed during four visits to the Lizard Island Research Station, Australia (location: 14°40′03″S; 145°26′49″E) between 2012 and 2015. Ten octopuses [8 Abdopus aculeatus (d'Orbigny 1834) and 2 Octopus cyanea Gray 1849] were collected opportunistically at low tide on coral reefs around Lizard Island. Only individuals with a mantle length of approximately 7 cm or less were retained for testing; most were closer to 3–4 cm mantle length. Direct length and mass measurements were not made, to reduce stress to the animals. Octopuses were kept individually in glass aquaria (15×15×15 cm) continually supplied with flow-through filtered and oxygenated ocean water and maintained under a natural day–night light cycle. Glass lids were placed over the aquaria, weighted down with lead diver's weights to prevent escapes. Fresh crabs and stomatopods were fed to the animals daily, with the size/number of food items matched to the animal's size; if the animal fed eagerly, more was provided if it was available that day. Octopuses were released back to the immediate vicinity of their capture location after testing (<14 days).

Visual stimuli were presented to each individual octopus by placing the octopus, in its home aquarium, in front of a modified LCD computer monitor (15 inch LCD, Type: VPC15AS1, Viglen, St Albans, Hertfordshire, UK). By moving the animal in its home tank, we avoided the undue stress of repeatedly capturing the animals each time they were tested. The animal was given time to acclimatize to the new visual surroundings after being moved. Stimulus presentations commenced when the animal was stationary yet awake and at least one eye was looking at the stimulus screen; the time taken for this to occur varied greatly between animals. The aquarium containing the octopus could be rotated so that the octopus had a clear view of the screen and was in the back third of the tank, 10–16 cm from the screen. During testing, the bottom and the two sides of the tank perpendicular to the viewing surface of the LCD were lined with white felt to reduce internal reflections that create intensity artefacts (Foster et al., 2018). The LCD and aquarium were covered with a black-out cloth to avoid movements in the testing room disturbing the subjects and therefore the testing tank was illuminated by the light emitted by the LCD screen and a small amount of room light that leaked through and around the black-out cloth; the precise intensity of this background light was neither controlled nor measured and therefore may have added some variance to the threshold values recorded.

The LCD was modified to display video in polarization contrast only, as described in Temple et al. (2012), and further modified so that the DoLP could be varied. Briefly, the front polarizer was removed such that the images varied in AoP rather than intensity. To control the DoLP, the LCD was further modified by removing the light supply and rear polarizer and replacing them with an LED light source (6 W, 4000 K, Master LED spot; MV GU10, Philips, Eindhoven, The Netherlands) that projected (off-axis at an angle of 30 deg) onto one of a series of custom designed DoLP filters positioned against the back side of the LCD. The DoLP filters included: (1) a thin sheet of acrylic; (2) a neutral density filter (varied ND filters, Lee Filters, Andover, Hampshire, UK); (3) a green gelatin filter (fern green #122, Lee Filters); (4) a thin (0.28 mm) sheet of Teflon to completely depolarize and spatially homogenize the light; (5) a sheet of polarizer (#7300, Rosco, London, UK); and (6) a diffusing/scattering tank to reduce the DoLP before it entered the liquid crystal matrix (Fig. 1A). The DoLP was varied by using different densities of hollow glass spheres (10 µm; Dantec Dynamics, Skovlunde, Denmark) suspended in water in the diffusing tanks (20×20×1 cm tanks constructed of 6 mm acrylic). A homogeneous distribution of the hollow glass spheres was maintained with vigorous flow provided by a 12 V water pump (automobile windshield washer fluid pump). Varying the DoLP resulted in changes in the overall intensity transmitted through the filters: the average change was 7%, with the total change in intensity from highest to lowest DoLP filter being 51% (Fig. 1B). To compensate for this, the intensity among the different DoLP filters was roughly matched by the addition of neutral density gelatin filters (item 2 in the list above) on the back (LED light) side of the Teflon sheet. The absolute spectral radiance of the LCD monitor was measured (Fig. S1) using a calibrated spectrophotometer (USB65000, Ocean Optics, Largo, FL, USA). The AoP of light emitted by the modified LCD varied from 45 to 130 deg in relation to the Uint8 value (8 bit LCD scale) that can be varied from 0 to 255. The background upon which the stimulus was shown was oriented horizontally (0 deg) by rotating the LCD by 45 deg. The depolarizing filters permitted the entire image to vary in DoLP (from 0.0 to 1.0) without altering the AoP contrast of the images displayed (Fig. 1B). The precise polarization characteristics of the monitor were measured using a Glan–Thompson Fresnel Rhomb assembly, coupled to a spectrophotometer (USB2000, Ocean Optics). Full methods for LCD measurement are published in Foster et al. (2018).

As reported previously (Foster et al., 2018; How et al., 2012; Temple et al., 2012), small changes in radiant energy are produced by our modified LCDs when viewed at oblique angles (away from normal to the plane of the screen). To minimize these effects, the looming image was centred on the animal's eyes for each presentation by adjusting the vertical and horizontal position of the centre of the looming object in the PowerPoint presentation in relation to the animal's position in the aquarium. This ensured that the largest angle subtended by the edge of the image was less than 10 deg from normal to the LCD surface. Additionally, small changes in DoLP accompanied the different AoP settings because of the presence of a variable component of ellipticity (Stokes vector S3) in the transmitted light from the display (see Foster et al., 2018, for details). These were measured and incorporated in the final analysis.

Octopuses were shown videos of a looming object (rapidly expanding circle) in polarization-only contrast. Looming stimuli were created using the ‘zoom in’ animation in PowerPoint (Microsoft Corporation, Albuquerque, NM, USA) (duration of expansion: 200 ms) and the stimulus disappeared 5.0 s later using the ‘shrink’ animation. The appearance and disappearance of the stimulus was marked by a click sound that was only audible on the video camera audio track and to the tester via a set of headphones. The stimulus was 4.0 cm onscreen and subtended an angle of less than 20 deg when viewed by the octopus. To determine an animal’s threshold, the AoP of the stimulus relative to the background was decreased until the animal stopped responding; the AoP was then decreased a further 1–2 steps to ensure that the animal could not detect the stimulus. Step sizes were 15 on the Uint8 scale (range of 0–255). This step size was decreased to steps of 5 Uint8 values when threshold values approached the limit of the monitor, e.g. background value of Uint8=0. When the threshold AoP was determined at one DoLP level, the same descending process through AoP values was repeated at the next lowest DoLP. At any DoLP value being tested, the lowest AoP at which a response was detected concomitant with the stimulus was determined as the animal's threshold (see below for a description of positive responses). Multiple passes of the descent through AoP values (repeats) were not completed because of the high number of different stimuli that needed to be presented (several AoP values at each of 9 DoLP settings), which would have put undue stress on the subjects. To maintain the subject's attention on the screen, periodically, higher contrast AoP stimuli were interjected into the descending pass. This enabled the observer to verify that the octopus was attentive to the stimuli (i.e. that the animal's eyes were open and that it was awake and looking in the direction of the stimulus, usually with one eye).

At the time of testing, the experimenter was able to observe the octopus’ responses on closed-circuit video capture and then replay the video recordings to determine whether a response had occurred. Responses took the form of changes in body colour pattern or movement of arms or papillae that followed the stimulus by precisely 200 ms and observably differed from any background variations in body pattern or movement. The presence of a response was evident to the trained observer; however, detection of very weak responses often required watching the video footage several times to confirm the response. If at the time of testing there was any doubt about whether the individual had responded or not, or if a background body pattern or movement change may have coincidently occurred simultaneously with the stimulus presentation, then that stimulus step was repeated, as the next stimulus or after a repeat of the previous few higher contrast stimuli. For scoring, all videos were analysed ‘blind’ by a trained observer at the end of the study without knowledge of stimulus setting presented. Subject responses to looming stimuli were categorized into one of five subjective response strength categories (0=no response, 1=very weak/just perceptible response, 2=weak response, 3=medium response, 4=strong response). Examples of responses for each category have been included in Movies 1 and 2 (also available from https://doi.org/10.5523/bris.1r4kwj2eu0tnq1yj9b9pdeb5bg). Stimuli were only presented to the octopus if it was awake, and had at least one eye open facing the screen, which was ascertained from the closed-circuit video feed. Data analysis was performed graphically with calculation of median, using Microsoft Excel and IBM SPSS V24.

Looming stimuli were presented every 2–5 min, a frequency that was found to allow octopuses to maintain strong responses for several hours over multiple days with no ill effects (i.e. the animals continued to be interactive during regular husbandry interactions and fed well throughout their stay in the laboratory). The threshold for angular contrast was determined at nine settings of DoLP (0.005, 0.04, 0.08, 0.15, 0.22, 0.31, 0.55, 0.75, 0.98). Note that not all individuals contributed data for each DoLP setting, because of experimental limitations.

The script used for calculating PD is included in the supplementary information (Script 1). Written in Matlab (Natick, MA, USA), Script 1 implements Bernard and Wehner's (1977) receptor sensitivity equations and How and Marshall's (2014) polarization distance calculations. Script 1 is fully annotated and provides a step-by-step approach to the method.

Octopuses responded to looming stimuli typically within milliseconds, with changes in body colour pattern (Fig. 2A). These changes in body pattern were approximately proportional to the contrast of the stimulus relative to the background, as observed previously in cuttlefish (Temple et al., 2012), such that high contrast stimuli invoked a full body colour change often combined with movement, while low contrast stimuli invoked small changes in body colour often restricted to one part of the body, e.g. a few square millimetres of one arm (see Movies 1 and 2, also available at https://doi.org/10.5523/bris.1r4kwj2eu0tnq1yj9b9pdeb5bg). We saw no difference (quantitative or qualitative) in the responses of the two species, and the threshold values of the two O. cyanea individuals fell within the range measured for the eight A. aculeatus.

When the DoLP was high (>0.3) the animals were able to respond to differences in the AoP of the stimulus relative to the background (ΔAoP) at a median value of 1.3 deg. Below a DoLP of 0.3, the minimum angular contrast between stimulus and background required to elicit a response increased rapidly as DoLP approached 0 (Fig. 2B).

An alternative system for representing polarization contrasts is to use a measure of ‘polarization distance’ or PD. Roughly analogous to the better-known ‘colour distance’ measure, PD provides an estimate of the amount of contrast detectable to a given polarization vision system (in this case, a dipolat with horizontal and vertical polarization receptors). For a full explanation of PD, see How and Marshall (2014); the equations and Matlab script used to calculate PD in this study are included in Script 1. When plotted as a function of PD, the measured thresholds line up approximately along a median PD value of 0.010 (Fig. 2C). The relationship between this threshold value and the range of DoLP and ΔAoP stimuli is illustrated by projecting the threshold onto the initial stimulus axes (Fig. 2B, dotted black line). This demonstrates that much of the variance in the data can be explained by the orientation of the underlying polarization-sensitive photoreceptors on which the polarization distance model is based. Some of the remaining variance can be explained by performance differences between individuals, with mean PD threshold across all ΔAoP/DoLP combinations ranging from 0.0049 to 0.024.

Octopuses responded to looming stimuli varying in polarization contrast alone and were sensitive to very small ΔAoP, particularly when the DoLP was high. The large variation in ΔAoP required to elicit a response (1 deg at high DoLP, to 53 deg for low DoLP) may be accounted for using the ‘polarization distance’ neurophysical model first suggested by Bernard and Wehner (1977) and then later expanded upon by How and Marshall (2014). This model takes a neural-processing approach to understanding contrast from the perspective of the animal's polarization vision system. Because cephalopods use a dipolatic system based on two polarization channels oriented horizontally and vertically relative to the outside world (Labhart, 2016; Moody and Parriss, 1961; Talbot and Marshall, 2011), AoP contrasts at low DoLP need to be larger to elicit an equivalent contrast in the photoreceptor output compared with stimuli at high DoLP. Along with the work of Basnak et al. (2018), this study is among the first to show how the systematic probing of polarization contrast sensitivity across the range of DoLP can converge on a single value of PD, reinforcing the validity of this approach to the study of polarization-based contrast vision in animals.

The sensitivity of polarization-based contrast vision has been measured previously for several other animals (see below), but the median behavioural threshold of PD=0.010 recorded in this study is the most acute recorded so far. Furthermore, several individuals responded to stimuli with PD values well below 0.010 (Fig. 2C). For example, the 10 most acute thresholds of response measured in the study ranged between PD=0.0017 and 0.0039. That responses were not always detected below PD=0.010 may reflect variance in motivation state, habituation or stress of the animals at the time of testing, which is difficult to control, or the lack of sensitivity of our response detection system, which relied on observing small changes in colour pattern only. As such, these results could be considered as conservative, as we suspect that the actual behavioural threshold may be closer to PD=0.002–0.004, but confirmation of this will require further investigation.

In other animals, the performance of the dorsal rim area of insects has been investigated in crickets and bees, which show electrophysiological and behavioural responses to dorsally presented polarization patterns down to a DoLP of 0.05–0.10 (Henze and Labhart, 2007; Labhart, 1996; von Frisch, 1967). This kind of celestial polarization vision operates using different requirements to object-based vision, in that the aim is to encode information about the overall AoP of the wide-field sky pattern rather than detecting contrasts between objects and background in an image parsing approach. As such, it is not meaningful to compare these thresholds directly with those measured in the current study. In crustaceans, the dorsal light reflex of crayfish can be reliably elicited by polarization contrasts as low as a ΔAoP of 15.2 deg (at DoLP of 1.0) and a DoLP of 0.13 (at ΔAoP of 20 deg) (Glantz and Schroeter, 2006). The conversion of these values to PD is not possible because of a lack of information about absolute AoP within the stimuli, but as an approximate comparison, the octopuses responded to a median ΔAoP of 1.3 deg when the DoLP was >0.3, an order of magnitude more acute than the equivalent threshold in the crayfish. In other studies, the startle behaviour of fiddler crabs was measured to a threshold ΔAoP of 3.2 deg and DoLP of 0.08, equivalent to PD=0.040 and 0.075, respectively (How et al., 2014; How et al., 2012). Stomatopods performed less well, only responding to contrasts of DoLP that are greater than ∼0.2 (when AoP for stimulus and background is 90 deg) (How et al., 2014), equivalent to PD=0.16, although the behaviours tested may not have revealed the absolute limit of PD threshold in these species. The cuttlefish Sepia plangon has also been tested using a very similar approach to the current study and reached a threshold performance of ΔAoP of 1.05 deg at a DoLP of 1.0 (PD=0.013) (Temple et al., 2012). The median value of PD=0.010 reported here is similar to that reported in S. plangon, but the frequency of responses at thresholds as low as PD=0.002 suggests that technological and methodological improvements in testing these and other cephalopods may well be rewarded with a more accurate estimate of their true abilities.

Assessing the functional advantage of the sensitive polarization vision in colourblind octopuses requires investigation of the types of visual scenes these animals may experience. Photographic polarimetry from the natural environment shows a range of polarization contrasts that fall close to or within the range of detection of the octopus visual system (Fig. 3) (Johnsen et al., 2016; Marshall et al., 2019). Cues and signals from fish predators and prey, as well as the communication signals of cephalopods and other animals, all generate polarization contrasts well within the detection range measured in this study.

Polarization contrast may add another channel of visual information that can be used in conjunction with intensity for parsing objects from the polarized backdrop of the underwater light field, or for identifying hitherto undiscovered communication signals from conspecifics. Whether polarization contrast acts as an independent channel of information to the intensity variations that frequently co-occur, or whether it simply modulates intensity to enhance a single contrast channel (analogous to wearing polaroid sunglasses) remains to be demonstrated. Inspiration for how to approach this question could come from recent work on fiddler crabs Afruca tangeri (Smithers et al., 2019), which made use of spatiotemporally synchronized intensity and polarization LCD displays to show that they respond to intensity and polarization contrasts as separate channels of information.

While our results point to a remarkably sensitive polarization contrast detection system in octopuses, it is worth mentioning two things: firstly, it is unlikely to be unique to the specific species studied here; and secondly, we know very little about how octopuses might use polarization sensitivity in their everyday behaviours. Our preliminary results with other cephalopods (cuttlefish and squid) point to equal sensitivity in other species and these are known to specifically target silvery fish against open waters where polarization vision may help (Shashar et al., 2000). We hope that our experimental paradigm combined with other new approaches like 3D glasses mounted on cuttlefish (Feord et al., 2020) and more behaviourally relevant stimuli may lead to a comprehensive understanding of this exciting alternative to colour vision.

Thanks to the staff at Lizard Island Research Station for practical and logistical support. Colleagues from the Cronin, Marshall and Roberts labs played a vital part in helping to catch animals and contributed to valuable discussions throughout the project. Special thanks to Roy Caldwell for his boundless knowledge of octopuses and to Daniel Osorio for early discussions on the subject of polarization contrast. Fieldwork was conducted under Great Barrier Reef Marine Park Association (GBRMPA) research permits G12/35005.1 and G12/35042.1; GBRMPA Limited Impact Permit 145 UQ006/2014; and Queensland General Fisheries Permit 140763.