1 Supplemental Figures and Tables:
2
3 Figure S1. Data on the visual system of Neolamprologus brichardi used to build visual
4 models
5 A) Relative opsin expression of N. brichardi determined by RNAseq. B) Normalized
6 transmittance of N. brichardi crystalline lens. C) Normalized absorbance of SWS1, RH2B and
7 RH2A. D) Estimated visual sensitivities incorporating the crystalline lens filtering medium. E)
8 Normalized downwelling and F) sidewelling irradiance at two depths. Increasing water depth
9 differentially reduces available longer light wavelengths. Refer to Figure 2. 10
11 Figure S1 (continued).
12 G) 95% confidence intervals of spectral reflectance in dominant and subordinate individuals.
13 Only horizontal facial stripe spectra do not overlap between dark and pale fish, other colors
14 overlap to various degrees. Colored curves refer to color patches in individuals with dark
15 horizontal stripe, while light grey curves refer to equivalent patches in individuals with pale
16 horizontal stripe. H) Chromatic contrast (mean ± SEM) is achieved by adjacency of color
17 elements and is not influenced by darkness of horizontal stripe. I) Achromatic contrast (mean ±
18 SEM) is achieved by paling of the horizontal stripe and has a differential influence on adjacent
19 and non-adjacent color elements. Refer to Figure 2. 20 21 Figure S2. Horizontal facial stripe provides reliable information on aggressive intent.
22 A) and B) Winners are larger and fight more aggressively than losers. Lines connect fish dyads.
23 C) and D) Principal components analysis of spectral data and signal experimental manipulation.
24 Principal Components 1 and 2 explain 96.5% of the variance (90.8% and 5.7%, respectively) and
25 clearly separate different colors. D) Zoom in at the melanistic area of the plot shows that
26 ‘Eyeliner’ is similar to dark melanic stripes of dominant fish, while ‘Wound Snow’ is similar to
27 pale melanistic stripes of subordinates. See also Table S1 for clustering of Principal
28 Components. E), F) and G) Reaction norms of amount of aggression incurred by individuals with
29 out-of-equilibrium signals and controls. Retaliation costs are highest for individuals with
30 artificially darkened stripe (‘bluffers’, E and F), followed by individuals with paled stripe
31 (‘Trojans’, E and G), and then controls (i.e. reliable signalers, F and G), which were the ones that
32 received the lowest amount of aggression. Lines connect the same individual, tested twice with
33 two different treatments. Refer to Figure 4.
34 35
36 Table S1. Cluster analysis of principal components of spectral data
Clusters
Color patch 1 2 3 4 5
Eyeliner 0 1 0 0 0
Wound Snow 1 0 0 0 0
Vertical facial stripe (D) 0 10 0 0 0
Vertical facial stripe (P) 0 10 0 0 0
Horizontal facial stripe (D) 1 9 0 0 0
Horizontal facial stripe (P) 10 0 0 0 0
Lachrymal stripe (D) 8 2 0 0 0
Lachrymal stripe (P) 10 0 0 0 0
Head (D) 9 1 0 0 0
Head (P) 10 0 0 0 0
Blue (D) 0 0 10 0 0
Blue (P) 0 0 10 0 0
Branchiostegal (D) 0 0 0 0 10
Branchiostegal (P) 0 0 1 0 9
Yellow (D) 0 0 0 9 1
Yellow (P) 0 0 0 9 1
37 Five clusters were identified. ‘Eyeliner’ clusters together with black melanistic stripes (cluster 2),
38 while ‘Wound Snow’ clusters with pale horizontal stripe, lachrymal stripe and head (cluster 1).
39 D: dark. P: pale. Refer to Figures 4 and S2C. 40 Supplemental Experimental Procedures
41 Choice of species
42 The Princess of Burundi, Neolamprologus brichardi (Teleostei: Cichlidae), is a small (up to 8 cm
43 in standard length) fish native to Lake Tanganyika, eastern Africa. Together with N. pulcher, it
44 has emerged as a model system in studies on the evolution of cooperative breeding behavior [1].
45 Recently, substantial genomic and transcriptomic resources have become available for N.
46 brichardi [2]. Such resources make this species an excellent system for the study of speciation,
47 evolution of cooperative behavior and communication also from a genetic perspective.
48 Neolamprologus brichardi performs the complete range of behaviors observed in the wild under
49 lab conditions, which makes it an optimal species for behavioral studies [3]. Phylogenetic
50 relationships have been recently studied [4], which led some authors to synonymize it with N.
51 pulcher. The two species differ in their facial pigmentation patterns and are thought to behave
52 similarly. We adopt the pre-synonymy taxonomy because it highlights the differences in facial
53 pigmentation pattern, which are the focus of our study. This should not create taxonomic
54 confusion and favors the accumulation of clear information for each of the pigmentation
55 phenotypes. Like most other species of the Tanganyikan cichlid tribe Lamprologini, N. brichardi
56 is sexually monochromatic, i.e. coloration patterns are identical between males and females [5].
57 The facial pigmentation of N. brichardi consists of two black melanic stripes, arranged in a
58 horizontal T-shape, surrounded by structural blue coloration, yellow pigmentation elements and
59 a white branchiostegal membrane. Another less conspicuous stripe is present in the pre-orbital
60 (lachrymal) area. The species has a beige body with fine orange elements in the posterior half
61 and white-fringed fins (see Figure 2 in main text). 62 Substantial data on life history and behavioral traits have been documented for these
63 species. The dominant, breeding couple has the peculiarity of being aided by up to 25
64 subordinate helpers in their tasks, and the social group is organized in a strict linear hierarchy [6–
65 8]. As a consequence of cooperative breeding and colony life, individuals repeatedly and
66 regularly interact [9] and it is known that vision plays a role in individual recognition of mates,
67 kin and neighbors [3, 10, 11]. Social groups of N. brichardi can be found on coastal rocky
68 substrates of Lake Tanganyika, usually between 3–50 m deep. The rocky substrate provides a
69 territory with shelters and breeding grounds where adhesive eggs are spawned [12, 13]. The
70 breeding male is always the largest individual of the group, usually followed by the breeding
71 female and subordinate helpers are the smallest [1]. Groups aggressively defend their territory
72 and dominant females have dominance behavior similar to dominant males, show high
73 testosterone levels and brain arginine vasotocin expression (a neuropeptide involved in
74 vertebrate territorial, reproductive and social behaviors) [14]. Group hierarchy is based on size, is
75 relatively stable over time and fish adopt one of two queuing strategies for breeding. Fish either
76 queue and breed in their natal group, or they disperse to a new group where they queue and breed
77 [1]. The mean distance between adjacent territories is 1.6 m and territories are clustered into
78 colonies [15]. These life history and behavioral traits create conditions for repeated interactions
79 among individuals. Most of them involve submissive behaviors, followed by aggressive
80 behaviors and only then territory maintenance (such as digging) and broodcare [16].
81
82 Study animals
83 Neolamprologus brichardi were raised and kept under controlled captive conditions at the
84 Zoological Institute, University of Basel, Switzerland, on a 12:12 h light:dark regime, in tanks 85 with about 1.5 cm of sand, a foam filter, a heater and terracotta flowerpots as shelters. Fish were
86 fed a combination of newly hatched Artemia nauplii, commercial flakes and frozen cichlid food
87 once or twice a day. All experiments were authorized by the Cantonal Veterinary Office, Basel,
88 Switzerland (permit numbers 2317 & 2356).
89
90 Color reflectance spectra
91 Spectral reflectance measurements of N. brichardi facial patterns were taken using a USB4000
92 spectrophotometer (Ocean Optics Inc.) and DH-2000-DUV Mikropack deuterium-halogen light
93 source, connected to a laptop computer running Ocean Optics SpectraSuite software. Twenty
94 individuals were tranquilized using a solution of KOI MED® Sleep (KOI&BONSAI, 0.5% v/v
95 2-Phenoxyethanol) before being transferred to a shallow tray filled with sufficient water to fully
96 cover the fish. Because tranquilizing the fish before measuring their spectral reflectance may
97 induce a short term darkening of their skin pigmentation we took care to measure reflectance
98 after original conditions were re-established (~15 seconds). Spectral reflectance of various facial
99 color patches (Figures 2A, 2B, S1G) was measured with a 200 µm bifurcated optic UV⁄visible
100 fiber. The bare end of the fiber was held at a 45º angle to prevent specular reflectance. A
101 Spectralon 99% white reflectance standard was used to calibrate the percentage of light reflected
102 at each wavelength from 350–750 nm. At least ten measurements per facial pattern per
103 individual were taken and subsequently averaged. Spectra were assessed based on the
104 wavelength at which light was reflected and the shape of the reflectance curves, and classified
105 into previously established categories of reef fish colors [17].
106
107 108 Visual system
109 To characterize the visual system of N. brichardi, we used published quantitative opsin data [2,
110 18] and amino acid sequences from eye RNAseq data [2] done on our stock of N. brichardi, and
111 collected new ocular transmission measurements from wild specimens. Our N. brichardi express
112 the UV-sensitive SWS1, and the two green-sensitive RH2A and RH2B opsin genes, which is a
113 common opsin expression palette in cichlid species, including the ones from Lake Malawi
114 (Figure S1) [19]. On comparisons of amino acid sequences of these three genes to the sequences
115 of their Lake Malawi relatives we found that there are only minor differences between species. In
116 particular, N. brichardi and Metriaclima (Maylandia) zebra show amino acid similarity of 95.4%
117 (± 1.1%, SEM) at SWS1, 98% (± 0.7%, SEM) at RH2Aα and 97.7% (± 0.8%, SEM) at RH2B
118 (similarities calculated in MEGA6) [20]. We measured ocular media transmission of the whole
119 eye (by cutting a window in the back), cornea, and the lens from wild caught N. brichardi (Cape
120 Kachese, Zambia, Lake Tanganyika; n = 3) to gain an understanding of the physical light
121 filtering properties of the eye, following previously established protocols [19, 21]. Light from a
122 pulsed xenon light source (Jaz-PX, Ocean Optics Inc.) was directed through the pinhole using a
123 400 µm optical fiber and the ocular media and collected by a 100 µm optical fiber attached to a
124 Jaz spectrometer (Ocean Optics Inc.). A Spectralon 99% white standard was used as a reflection
125 standard. At least three measurements per media were taken and subsequently averaged. Spectra
126 were thereafter normalized using their maximum transmission and the wavelength at which 50%
127 transmission (T50) was reached was determined within the 300–750 nm interval [19, 21]. We
128 found the lens to be the limiting light transmission medium of the N. brichardi eye with a T50
129 cut-off value of 359 nm (Figure S1). Based on our molecular assessment we followed Dalton et
130 al. [22], and used opsin absorbance spectra (λmax = 368 nm for short wavelength (SWS), λmax = 131 488 nm for mid wavelength (MWS), λmax = 533 nm for long wavelength (LWS) [23, 24] from M.
132 (M.) zebra, a rock-dwelling cichlid species from lake Malawi, to reconstruct the visual
133 sensitivities of N. brichardi (Figure S1). We then incorporated our lens transmission
134 measurements to create a template of the visual system of N. brichardi to model how N.
135 brichardi perceives color (Figure S1).
136
137 Underwater light environment in Lake Tanganyika
138 We took measurements of the natural ambient light under which the fish color patterns evolved.
139 Measurements were taken at Isanga Bay (Zambia, Lake Tanganyika) in September 2011 at
140 depths of 3 m and 7 m (Figure S1). Illumination was measured using a USB2000 spectrometer
141 attached to a PALM-SPEC computer running native software (Ocean Optics), enclosed in an
142 underwater housing (Wills Camera Housings, Victoria, Australia). We used a shortened (60 cm)
143 1000 µm UV/visible optical-fiber with a cosine corrector to provide an 180º hemisphere to
144 measure both, downwelling (by pointing the fiber upwards) and sidewelling light (pointing the
145 fiber horizontally into the middle or towards the shore of the lake). However, there was no
146 significant difference in our overall conclusion when using either of the measurements.
147
148 Visual modeling
149 We used a theoretical visual model [25, 26] to quantify the chromatic (‘color’ or ‘hue’) and
150 achromatic (‘intensity’ or ‘luminance’) color contrasts between the facial pattern elements using
151 the N. brichardi visual system and assuming ambient light conditions as measured from their
152 natural habitat (see above). The chromatic model calculates the color distance (ΔS) within the
153 visual space of the fish, where low values of ΔS denote similar colors and high values of ΔS 154 indicate chromatically different colors. When calculating ‘hue’ distances between color patches,
155 luminosity is disregarded within the model, the colors are assumed to be encoded by an
156 opponency mechanism based on the sensitivities of the fish visual system, and chromatic
157 discrimination is thought to be limited by photoreceptor noise determined by the relative
158 proportion of each photoreceptor [25, 26]. The receptor quantum catch (qi) in the photoreceptor
159 cell of type i is calculated as
160 qi = ∫ Ri(λ)S(λ)I(λ)dλ
161 where λ denotes the wavelength, Ri(λ) the spectral sensitivity of the photoreceptor cell, S(λ) the
162 spectral reflectance of the color patch, I(λ) the illumination spectrum entering the eye and
163 integration is over the range of 350–750 nm (equation 1 of Vorobyev & Osorio [25]).
164 Illumination was set as measured at a depth of 7 m and coming from above (no difference was
165 found when using spot tests and illumination at 3 m). In the absence of retinal morphological and
166 physiological data for our study species, we based the relative proportion of cone receptors on
167 studies from other cichlid fishes, which frequently have a square pattern [27–29] with a ratio of
168 1:2:2 (SWS:MWS:LWS), which is consistent with RNAseq data for N. brichardi (Figure S1).
169 The weber fraction (ω) was set to assume a 0.05 LWS noise threshold, which is a conservative
170 approach representing approximately half the sensitivity of the human LWS cone system [30]. In
171 addition to chromatic contrast we also calculated achromatic ‘luminance’ contrast as a second
172 property of the visual signal. Long wavelength receptors are thought to be responsible when
173 perceiving differences in luminance (for discussion see Marshall et al. [31]) and we therefore
174 used the differences in the natural logarithm quantum catch (Q) of the long wavelength receptor
175 (L) to calculate luminance differences between color patches using
176 ΔL = ln(QLpatchX) – ln(QLpatchY). 177 We predict that the more ΔS increases above the threshold of 1 JND (just noticeable
178 difference), the more distinguishable colors become from one another, which might be especially
179 important for long-range signals where intervening water and particles start to blur colors [31].
180
181 Resource contest experiment
182 A total of 40 N. brichardi (20 males and 20 females) originating from several stock tanks were
183 sexed (by examination of the genital papilla), measured (standard length [SL], taken as the
184 distance between the tip of the snout and the insertion of caudal fin rays) and weighed (body
185 mass [BM], taken after one day fasting). Female SL was 5.41 ± 0.55 cm (mean ± standard
186 deviation) and BM was 4.30 ± 1.34 g. Male SL was 5.62 ± 0.56 cm and BM 4.67 ± 1.48 g. To
187 control for daily variation in behaviors, all territorial dyadic combats were conducted between
188 11AM and 1PM [32]. Combats were performed in an aquarium (60 × 30 × 30 cm) divided in the
189 middle of the longer side into two equal compartments by a removable opaque grey plastic
190 barrier. The conditions in both compartments were the same: each had a filter, a heater, ca. 2 cm
191 of sand and a quarter of a terracotta flowerpot (12 cm in diameter, 10 cm long) adjacent to the
192 barrier. Due to the social nature of these fish, small opposite-sexed conspecifics (one per
193 compartment) were introduced in transparent plastic bottles to encourage territory establishment
194 of the focal fish. Dyads of fish were matched by sex, SL (Mann-Whitney U test, V = 233, P =
195 0.27) and BM (V = 316, P = 0.21). Test fish were caught from their tanks of origin and released
196 in one of the two compartments to establish their territory in the quarter flowerpot for three days.
197 The compartment was randomly selected. Fish were fed commercial flakes or frozen cichlid food
198 twice a day and at least one hour before starting the trial to control for effects of feeding regime
199 [33]. Following this acclimation period the visual barrier was removed, merging the territories of 200 the two fish, and merging the flowerpot shelter into one. This procedure guarantees that both fish
201 have simultaneous ownership over a territory and that they cannot divide this resource after the
202 barrier has been removed. Neolamprologus brichardi is a highly territorial species and
203 immediately starts to combat for ownership of the shelter [9, 34]. To avoid disturbances by the
204 experimenter, the interactions of the fish were videotaped with a Sony HDR XR 550VE
205 camcorder. After each trial ended, fish were moved to separate holding nets in their original
206 tanks. (see end of file)
207 Intensity of the horizontal facial stripe (pale or dark) was recorded at the beginning and
208 end of experiments. Outcome and behaviors of the 20 min combats were recorded via video
209 analysis (for a detailed ethogram of the species, modified after Sopinka et al. [35] and Taves et
210 al. [34], see below). The winner was declared as the fish from which the loser fled three times
211 without counter-strike or constantly held a submissive posture [9, 36]. Alternatively, a fish was
212 declared winner if it owned the flowerpot in the end of the combat (i.e. the most valuable
213 resource of the territory). Behaviors of both fish were counted and fit into four categories of
214 recorded behaviors due to their very diverse behavioral repertoire (see below), which is
215 consistent with other studies of Neolamprologus [7, 34, 35]. A fighting ability index for each fish
216 was calculated by subtracting the total of submissive behaviors from the sum of territorial,
217 display and contact aggressive behaviors (called dominance index by Aubin-Horth et al. [14]).
218 Factors determining the outcome of resource contests were identified with linear mixed
219 models with ‘body mass’ and ‘fighting ability’ (the difference between aggressive and
220 submissive behaviors) as response and ‘outcome’, ‘sex’ and their interaction as explanatory
221 variables. We fit a generalized linear mixed model (GLMM) with binomial error distribution,
222 logit link function and ‘pair’ as random effect, to test whether outcome is significantly associated 223 to the intensity of the facial stripe at the beginning or at the end of the contest. For this analysis
224 we used the R package lme4 [37].
225
226 Ethogram of behavior repertoire of Neolamprologus brichardi
Category Behaviors Description Contact aggression Bite Focal fish bites another fish Chase Focal fish follows another fish Displace Focal fish swims towards another fish, forcing it to move Mouth-lock* Two fish lock jaws and push against each other Ram Focal fish hits another fish with its head, but jaws remain closed Display aggression Aggressive posture Focal fish lowers its head towards another fish and shows the side of its body with spread fins Head shake Fish tosses its head from left to right Puffed throat Fish opens operculum and lower jaw cavity Submission Bitten Focal fish gets bitten by another fish Flee Focal fish swims away from another fish Submissive posture Focal fish has a (nearly) vertical position, with the head directing upwards Submissive display Focal fish is positioned with a submissive posture accompanied by a quivering caudal fin Territoriality Body digging Focal fish quivers its body on the substrate and moves sand Digging Focal fish takes sand in its mouth, sometimes swims to a different area and spits it out Lookout Focal fish observes another fish from its shelter Hover Focal fish defends brood chamber, inhibit other fish from entering Cleaning Focal fish removes algae from shelter by nibbling on them 227 * both fish get the score 228 To determine which color elements changed in chromatic or achromatic contrast between
229 dominant and subordinate fish, pairwise Mann-Whitney U tests were calculated for each
230 comparison. False discovery rate (FDR) was applied to correct for multiple testing. To detect
231 overall differences between adjacent and non-adjacent color patches in chromatic and achromatic
232 contrasts in territorial and non-territorial fish, we ran linear mixed-effects models (LMM) with
233 the R package nlme [38]. As we measured several color patches per fish and then used them in
234 different comparisons, all adjacent and all non-adjacent chromatic or achromatic contrasts were
235 averaged per individual. ‘Individual’ was then used as random effect. Shapiro tests confirmed
236 normality of the residuals. As achromatic contrast deviated from normality, it was square-root
237 transformed. First, to test whether the facial color pattern is conspicuous to the fish eye, we
238 compared chromatic and achromatic contrasts between adjacent and non-adjacent color patches
239 of dominant fish (i.e. fish with dark horizontal stripes, which is the state in which the phenotype
240 is normally expressed). In a second step we further analyzed subordinate fish (i.e. fish with pale
241 horizontal stripes) to investigate how changes in facial stripe intensity affected the phenotype.
242 We ran mixed models with ‘adjacency’, ‘stripe intensity’ and their interaction as fixed effects
243 and ‘individual’ as random effect.
244
245 Facial stripe out-of-equilibrium manipulation and standard mirror image stimulation
246 experiment
247 Standard mirror image stimulation (MIS) experiments were used to determine if N. brichardi are
248 able to recognize and punish unreliable signaling by measuring the response of each individual to
249 its image. MIS provides instantaneous feedback without some of the confounding factors
250 resulting from using other individuals as stimuli [39]. Cichlids, including Neolamprologus, are 251 known to react aggressively towards their mirror images [3, 36, 40]. Additionally, it has been
252 shown that N. pulcher behaves similarly towards mirror images as towards conspecifics [41].
253 The test setup consisted of an aquarium (40 × 25 × 25 cm) with a 2.84 mm-thick plane glass
254 mirror (25 × 25 cm), placed inside the tank, behind a terracotta flowerpot arch (10 cm in
255 diameter; 3 cm wide) at a sidewall. Using a flowerpot arch instead of a closed flowerpot
256 guaranteed that the fish could see their reflection at all times, including inside the shelter,
257 avoiding the generation of impossible reflection angles that could confuse them. At the
258 beginning the arch and the mirror were hidden behind an opaque grey plastic barrier. After
259 removal of the opaque barrier, the mirror image should reflect a conspecific territory owner to
260 the test fish. This setup further addresses common limitations faced when presenting
261 manipulated individuals to focal dominant, territorial individuals [42]. In our setup, the focal fish
262 act as intruders and test the repellent effect of the manipulated signals in individuals of the same
263 size they perceive as territory owners. (see end of file)
264 A total of 49 N. brichardi (25 males and 24 females) originating from several stock tanks
265 were sexed, measured (SL), and weighed (BM) as in the resource contest experiment (above).
266 Female SL was 5.90 ± 0.74 cm (mean ± standard deviation) and BM was 5.19 ± 2.21 g. Male SL
267 was 5.92 ± 0.84 cm and BM 5.20 ± 2.62 g. Before fish were tested, they were separated from
268 their social group for two days and kept in a pre-test tank (40 × 25 × 25 cm), covered on all four
269 sides to minimize disturbance. This tank contained a flowerpot arch placed adjacent to one wall
270 so fish learned to use this as a shelter instead of a closed flowerpot. After this isolation period of
271 two days, during which they became territorial, fish were gently netted out of the pre-test tank,
272 partially anesthetized with KOI MED® Sleep (KOI&BONSAI, 0.5% v/v 2-Phenoxyethanol) and
273 the horizontal facial stripe was manipulated randomly in one of three different ways: 274 1. Darkened facial stripe: The facial stripe was brushed with black waterproof eyeliner
275 (Collection 2000, USA). To control for the paling treatment (see below), wound snow and
276 wound spray (KOI MED®) were applied in the head region above the facial stripes.
277 2. Paled facial stripe: To control for the darkening treatment, the facial stripe was first
278 painted with the black waterproof eyeliner (Collection 2000, USA) and then covered up with
279 ‘Wound Snow’ and ‘Wound Spray’ (KOI MED®).
280 3. Control sham-manipulation: Same treatment as 2. applied in the head region above the
281 facial stripe as in 1., so the facial stripe was left un-manipulated.
282 Spectral reflectance measurements show that treatments result in the desired effect of
283 extreme darkening and paling that was similar to non-manipulated horizontal stripes. A principal
284 components analysis (PCA) of spectral reflectance data clearly groups black ‘eyeliner’ with dark
285 melanistic stripes (typical of dominant fish) and groups ‘Wound Snow’ with pale horizontal
286 stripe, lachrymal stripe and head (typical of subordinate fish) (Figure S2). A cluster analysis
287 conducted with R package mclust [43] confirms this visual assessment (Table S1).
288 After facial stripe manipulation fish were released into the test tank compartment without
289 the flowerpot arch and mirror, and allowed to recover for 5 min from anesthesia and treatment.
290 For motivational purposes and to ease acclimation to new surroundings, fish were fed a little
291 amount of newly hatched Artemia nauplii. After the recovery period, the opaque barrier was
292 removed and the fish could interact with its mirror image. To control for daily variation in
293 behaviors, all experiments were conducted between 9AM and 11:30AM [32]. To control for
294 individual effects of aggression levels between behavioral types [44], fish were tested twice with
295 two different treatments. Order of treatment was randomized. All aggressive (display and
296 contact) and submissive behaviors towards the mirror image were counted during a period of 2.5 297 min from a video recording (Sony camcorder, see above), starting after the removal of the
298 opaque barrier.
299 We fit a linear mixed model with ‘aggressive bouts’ as response and ‘treatment’, ‘sex’
300 and their interaction as explanatory variables. To normalize the residuals, ‘aggressive bouts’ was
301 square-root transformed. As fish were tested twice, ‘individual’ was added as a random effect.
302 Tukey’s HSD post-hoc analysis was performed to test for differences among treatment levels.
303
304 References
305 1. Wong, M., and Balshine, S. (2011). The evolution of cooperative breeding in the African 306 cichlid fish, Neolamprologus pulcher. Biol. Rev. Camb. Philos. Soc. 86, 511–530.
307 2. Brawand, D., Wagner, C. E., Li, Y. I., Malinsky, M., Keller, I., Fan, S., Simakov, O., Ng, 308 A. Y., Lim, Z. W., Bezault, E., et al. (2014). The genomic substrate for adaptive radiation 309 in African cichlid fish. Nature 513, 375–381.
310 3. Balshine-Earn, S., and Lotem, A. (1998). Individual recognition in a cooperatively 311 breeding cichlid: evidence from video playback experiments. Behaviour 135, 369–386.
312 4. Duftner, N., Sefc, K. M., Koblmüller, S., Salzburger, W., Taborsky, M., and Sturmbauer, 313 C. (2007). Parallel evolution of facial stripe patterns in the Neolamprologus 314 brichardi/pulcher species complex endemic to Lake Tanganyika. Mol. Phylogenet. Evol. 315 45, 706–715.
316 5. Gante, H. F., and Salzburger, W. (2012). Evolution: cichlid models on the runaway to 317 speciation. Curr. Biol. 22, R956–958.
318 6. Taborsky, M., and Limberger, D. (1981). Helpers in Fish. Behav. Ecol. Sociobiol. 8, 143– 319 145.
320 7. Balshine, S., Leach, B., Neat, F., Reid, H., Taborsky, M., and Werner, N. (2001). 321 Correlates of group size in a cooperatively breeding cichlid fish (Neolamprologus 322 pulcher). Behav. Ecol. Sociobiol. 50, 134–140.
323 8. Zöttl, M., Heg, D., Chervet, N., and Taborsky, M. (2013). Kinship reduces alloparental 324 care in cooperative cichlids where helpers pay-to-stay. Nat. Commun. 4, 1341. 325 9. Reddon, A. R., Voisin, M. R., Menon, N., Marsh-Rollo, S. E., Wong, M. Y. L., and 326 Balshine, S. (2011). Rules of engagement for resource contests in a social fish. Anim. 327 Behav. 82, 93–99.
328 10. Frostman, P., and Sherman, P. T. (2004). Behavioral response to familiar and unfamiliar 329 neighbors in a territorial cichlid, Neolamprologus pulcher. Ichthyol. Res. 51, 8–10.
330 11. Le Vin, A. L., Mable, B. K., and Arnold, K. E. (2010). Kin recognition via phenotype 331 matching in a cooperatively breeding cichlid, Neolamprologus pulcher. Anim. Behav. 79, 332 1109–1114.
333 12. Taborsky, M. (1984). Broodcare helpers in the cichlid fish Lamprologus brichardi: their 334 costs and benefits. Anim. Behav. 32, 1236–1252.
335 13. Heg, D., Bachar, Z., Brouwer, L., and Taborsky, M. (2004). Predation risk is an ecological 336 constraint for helper dispersal in a cooperatively breeding cichlid. Proc. Biol. Sci. 271, 337 2367–74.
338 14. Aubin-Horth, N., Desjardins, J. K., Martei, Y. M., Balshine, S., and Hofmann, H. A. 339 (2007). Masculinized dominant females in a cooperatively breeding species. Mol. Ecol. 340 16, 1349–1358.
341 15. Stiver, K. A., Desjardins, J. K., Fitzpatrick, J. L., Neff, B., Quinn, J. S., and Balshine, S. 342 (2007). Evidence for size and sex-specific dispersal in a cooperatively breeding cichlid 343 fish. Mol. Ecol. 16, 2974–2984.
344 16. Taborsky, M., and Grantner, A. (1998). Behavioural time – energy budgets of 345 cooperatively breeding Neolamprologus pulcher (Pisces : Cichlidae). Anim. Behav. 56, 346 1375–1382.
347 17. Marshall, N. J. (2000). The visual ecology of reef fish colours. In Animal signals: 348 signalling and signal design in animal communication, Y. Espmark, T. Amundsen, and G. 349 Rosenqvist, eds. (Trondheim, Norway: Tapir Academic Press), pp. 83–120.
350 18. Schulte, J. E., O’Brien, C. S., Conte, M. A., O’Quin, K. E., and Carleton, K. L. (2014). 351 Interspecific variation in Rx1 expression controls opsin expression and causes visual 352 system diversity in African cichlid fishes. Mol. Biol. Evol. 31, 2297–2308.
353 19. Hofmann, C. M., O’Quin, K. E., Justin Marshall, N., and Carleton, K. L. (2010). The 354 relationship between lens transmission and opsin gene expression in cichlids from Lake 355 Malawi. Vision Res. 50, 357–363.
356 20. Tamura, K., Stecher, G., Peterson, D., Filipski, A., and Kumar, S. (2013). MEGA6: 357 Molecular Evolutionary Genetics Analysis. Mol. Biol. Evol. 30, 2725–2729. 358 21. Siebeck, U. E., and Marshall, N. J. (2001). Ocular media transmission of coral reef fish — 359 can coral reef fish see ultraviolet light? Vision Res. 41, 133–149.
360 22. Dalton, B. E., Cronin, T. W., Marshall, N. J., and Carleton, K. L. (2010). The fish eye 361 view: are cichlids conspicuous? J. Exp. Biol. 213, 2243–2255.
362 23. Levine, J. S., and MacNichol, E. F. (1979). Visual pigments in teleost fishes: effects of 363 habitat, microhabitat, and behavior on visual system evolution. Sens. Processes 3, 95–131.
364 24. Carleton, K. L., Ha, F. I., and Kocher, T. D. (2000). Visual pigments of African cichlid 365 fishes: evidence for ultraviolet vision from microspectrophotometry and DNA sequences. 366 Vision Res. 40, 879–890.
367 25. Vorobyev, M., and Osorio, D. (1998). Receptor noise as a determinant of colour 368 thresholds. Proc. Biol. Sci. 265, 351–358.
369 26. Vorobyev, M., Brandt, R., Peitsch, D., Laughlin, S. B., and Menzel, R. (2001). Colour 370 thresholds and receptor noise: behaviour and physiology compared. Vision Res. 41, 639– 371 653.
372 27. Fernald, R. D. (1981). Chromatic organization of a cichlid fish retina. Vision Res. 21, 373 1749–1753.
374 28. Van der Meer, H. J., and Bowmaker, J. K. (1995). lnterspecific variation of photoreceptors 375 in four co-existing haplochromine cichlid fishes. Brain. Behav. Evol. 45, 232–240.
376 29. Dalton, B. E., Loew, E. R., Cronin, T. W., and Carleton, K. L. (2014). Spectral tuning by 377 opsin coexpression in retinal regions that view different parts of the visual field. Proc. R. 378 Soc. B Biol. Sci. 281.
379 30. Wyszecki, G., and Stiles. W. S. (2000). Color science — concepts and methods, 380 quantitative data and formulae 2nd ed. (New York: Wiley).
381 31. Marshall, N. J., Jennings, K., McFarland, W. N., Loew, E. R., and Losey, G. S. (2003). 382 Visual biology of Hawaiian coral reef fishes. III. Environmental light and an integrated 383 approach to the ecology of reef fish vision. Copeia 2003, 467–480.
384 32. Oliveira, R. F., Lopes, M., Carneiro, L. A., and Canário, A. V. M. (2001). Watching fights 385 raises fish hormone levels. Nature 409, 2001.
386 33. Fisher, H. S., and Rosenthal, G. G. (2006). Hungry females show stronger mating 387 preferences. Behav. Ecol. 17, 979–981.
388 34. Taves, M. D., Desjardins, J. K., Mishra, S., and Balshine, S. (2009). Androgens and 389 dominance: sex-specific patterns in a highly social fish (Neolamprologus pulcher). Gen. 390 Comp. Endocrinol. 161, 202–207. 391 35. Sopinka, N. M., Fitzpatrick, J. L., Desjardins, J. K., Stiver, K. A., Marsh-Rollo, S. E., and 392 Balshine, S. (2009). Liver size reveals social status in the African cichlid Neolamprologus 393 pulcher. J. Fish Biol. 75, 1–16.
394 36. Dijkstra, P. D., Schaafsma, S. M., Hofmann, H. A., and Groothuis, T. G. G. (2012). 395 “Winner effect” without winning: unresolved social conflicts increase the probability of 396 winning a subsequent contest in a cichlid fish. Physiol. Behav. 105, 489–492.
397 37. Bates, D., Maechler, M., Bolker, B., and Walker, S. (2014). lme4: Linear mixed-effects 398 models using Eigen and S4.
399 38. J, P., D, B., S, D., and D, S. (2012). nlme: Linear and Nonlinear Mixed Effects Models.
400 39. Rowland, W. J. (1999). Studying visual cues in fish behavior: a review of ethological 401 techniques. Environ. Biol. Fishes 56, 285–305.
402 40. Reddon, A. R., and Balshine, S. (2010). Lateralization in response to social stimuli in a 403 cooperatively breeding cichlid fish. Behav. Processes 85, 68–71.
404 41. Balzarini, V., Taborsky, M., Wanner, S., Koch, F., and Frommen, J. G. (2014). Mirror, 405 mirror on the wall: the predictive value of mirror tests for measuring aggression in fish. 406 Behav. Ecol. Sociobiol. 68, 871–878.
407 42. Bradbury, J. W., and Vehrencamp, S. L. (2011). Principles of Animal Communication, 408 Second Edition (Sinauer Associates, Inc.; 2 edition).
409 43. Fraley, C., Raftery, A., and Scrucca, L. (2014). mclust: Normal Mixture Modeling for 410 Model-Based Clustering, Classification, and Density Estimation.
411 44. Bell, A. M. (2007). Future directions in behavioural syndromes research. Proc. R. Soc. B 412 Biol. Sci. 274, 755–761.
413
414 415
416
417 Schematic representation of behavior experimental setups. In the territory resource contest
418 setup (upper row), fish are visually separated and allowed to establish their territory in the
419 terracotta flowerpot for three days (a), after which the opaque divider is removed and
420 territories/shelters merged (b) and individuals are allowed to fight over non-divisible territorial
421 resource for 20 minutes. Observer could see the fish without being seen to control for dangerous
422 levels of aggression. In the standard mirror image stimulation setup (bottom row), fish with
423 manipulated signals are introduced to a bare side of the tank with no shelters (c). After opaque
424 divider is removed (d), individuals can see a shelter and a fish holding a territory next to it (i.e.
425 their mirror image) and are allowed to interact with it.