All animal care and experimental protocols complied with local and federal laws and guidelines and were approved by the appropriate governing body in Germany, the Landesamt für Gesundheit und Verbraucherschutz (LaGeSo 52/16).
Experimental procedures
Study animals and housing conditions
The rainbow krib is a small cichlid originating from rivers and streams in West Africa (Heiligenberg, 1965; Nwadiaro, 1985). Females are shorter and deeper bodied, and often more intensely colored, compared to the more streamlined, less colorful males (Heiligenberg, 1965; Martin & Taborsky, 1997; Reddon & Hurd, 2013). Males occur in two color morphs with red versus yellow opercula (Heiligenberg, 1965). Both sex and male color morph determination are influenced by the pH during early development (Martin & Taborsky, 1997; Reddon & Hurd, 2013; Rubin, 1985).
Rainbow kribs used in this study were obtained from a house breed at the Universität Hamburg and local suppliers; and were maintained in same-sex groups matched for family and origin (100-200l tanks), at 25±1°C, and a 12:12 hours light:dark period. Tanks were endowed with a layer of sand (approx. 1 cm thick), an internal filter and plastic plants. Water changes (50%) were done once a week and fish were fed daily with live Artemia spp.. Water changes were done using pre-heated, aged tap water. All tanks were supplied with the same water from a common supply container, ensuring standardized water conditions for all tanks. Four days before the first boldness test, individuals were transferred to individual housing tanks (25L, 50 x 25 x 25 cm, water level = 20cm, same holding conditions as above). On the day of transfer, individuals were photographed and we measured their standard length (i.e., the length from the tip of the snout to the end of the spine) from photos using ImageJ (Schneider et al. 2012) (males: mean ± SE standard length = 5.42 ± 0.05 cm, range = [4.52, 6.25 cm], N = 54; females: mean ± SE standard length = 4.39 ± 0.04 cm, range = [3.90, 5.17 cm], N = 54). All test fish were uniquely VIE-tagged for identification (Schuett et al., 2017).
Experimental Outline
We measured boldness before breeding as activity in the presence of a computer-animated predator (see ‘Boldness tests’ for experimental details and see below for an explanation of our choice of boldness measure). We then created breeding pairs that varied regarding their similarity in average activity under simulated risk (see ‘Pairing and breeding’) and during breeding, we performed six parental care tests to assess parental activity under risk and brood guarding for each successfully reproducing breeding pair (see ‘Parental care tests’). We measured reproductive success as the likelihood to reproduce, clutch size (number of eggs laid), offspring survival and the size of offspring produced after approx. one month of care (see ‘Pairing and breeding’ for details). Boldness tests and pair formation were carried out in 5 experimental blocks with 12 males and 12 females per block, amounting to N = 60 males and females, respectively. Each experimental block took up to 10 weeks (including pre-breeding behavioral typing, pair formation, and the parental care period); and blocks were started with 2-3 weeks in between successive blocks. The total number of fish tested exceeded the number of fish used for pairing (N = 54 breeding pairs), giving us more options for the pair formation.
Prior to the experiment, we conducted a pilot study (Scherer, Godin, et al., 2017) to test how individuals react towards a predator under our specific experimental conditions and to test for the suitability of using virtual stimuli instead of live stimuli. We used a natural predator of rainbow kribs, the African obscure snakehead, Parachanna obscura. In short, we tested behavioral responses of 36 rainbow kribs towards both live predators and animated color-photographs thereof. For each focal individual, we measured the activity and mean distance to the stimulus. Both decreasing activity and keeping distance to a potential threat are well-known fear responses in fish (Broom & Ruxton, 2005; Cooper & Martín, 2016; McLean & Godin, 1989; O’Connor et al., 2015), although some fish may also cautiously “inspect” a predator to gather information (Godin & Dugatkin, 1996; Hesse et al., 2015; Pitcher et al., 1986). Time spent in an “inspection zone”, i.e., near the predator, was also recorded but it was highly correlated with the mean distance to the predator (Spearman rank correlation: rs = 0.84-0.93) and was therefore not analysed to avoid redundancy. Compared to control trials (empty aquarium and blank computer screen), individuals decreased their activity when presented with both the animated and the live predator with no difference in response between these two stimuli. Individuals decreased their distance towards the live predator compared to the empty predator tank control. But no difference in distance towards the animated predator vs. blank computer screen was observed and individuals kept a significantly larger distance to the animated predator compared to the live predator. The difference in inspection behavior may be caused by live predators having been rather motionless (and therefore potentially safe to inspect), while animated predators were swimming back and forth (and were therefore behaving potentially more suspiciously).
The results of our pilot study demonstrate a similar fear response towards both an animated and live predator, but they also highlight nuanced differences in how these two stimuli were perceived. Specifically, individuals decreasing activity in response towards both the animated and live predator indicates that the predator was perceived as potentially dangerous in both scenarios. On the other hand, individuals exhibited inspection behavior when presented with the live predator (as indicated by a lower distance to the live predator compared to the empty tank control), while no such tendency was observed when individuals were confronted with the animated predator. Taken together, activity under simulated risk seems the most robust measure to quantify boldness when using animated predators in our study species.
Boldness tests
Before breeding, all males (N = 60) and females (N = 60) were tested for their activity under simulated risk twice with five days in between trials. Boldness tests were performed following Scherer, Godin, et al. (2017). In short, individual test fish were exposed to a photograph of a predator specimen (P. obscura), which was animated to swim back and forth on a computer screen (see Scherer, Godin, et al., (2017) for details on the animation production). Test individuals were granted a 10 min acclimation period in the test tank, followed by 11 minutes of predator exposure. Recording from above (with a Sony HD-CX405), we assessed individual activity in the presence of the predator as the total number of squares visited (including revisits) for 10 min (no tracking of the first minute) using the tracking software Ethovision XT 11 (Noldus, Wageningen, The Netherlands). Therefore, test tanks (50 x 25 x 25 cm, water level = 10 cm) were divided into 8 squares each measuring 12x12 cm squares. For all trials, we used a predator specimen the test fish had not seen before (N = 4 predators, mean ± SE standard length = 19.3 ± 0.3 cm). To create breeding pairs and for analyses (excluding repeatability analysis), we used the average activity under simulated risk an individual showed over both boldness tests (mean ± SE number of squares visited: males = 63.56 ± 4.26, females = 41.07 ± 3.64). Prior to the experiment, individuals were habituated to test tanks twice for 30 min, respectively (in groups of approximately ten fish).
Pairing and breeding
Four days following the boldness tests, we set up breeding pairs (N = 54) of varying behavioral contrasts in activity under simulated risk. We calculated male-female behavioral contrast as average male behavior minus average female behavior, i.e., positive values indicate that the male is more active under risk than the female and vice versa. Pairs were formed in a way that we minimized the behavioral contrast for half of the pairs, whereas the other half was maximized for within-pair behavioral contrast (varying sex of the bolder individual). To initiate breeding, we introduced the respective male and female into a breeding tank (50 x 50 cm, water level = 25 cm), equipped with half a clay pot as breeding cave, a plastic plant, a layer of sand, and an internal heater (all in a standardised position). We monitored the breeding cave for eggs daily using a small dentist mirror (diameter = 3 cm) and took a photo of clutches laid when first seen. The number of eggs was counted from the photo (mean ± SE = 190 ± 14 eggs, N = 18 out of 20 clutches, no photos for two clutches). Breeding pairs that did not produce fry within 28 days were transferred back to their home tanks and were not further used in this experiment (N = 34 males and females, respectively). Breeding pairs that did produce fry were allowed to raise their brood for 33 days (spawning = day 1). During this breeding period, we assessed parental care behavior as outlined below. Different to housing conditions, we increased the number of water changes to three times a week and reduced the amount of water being changed to approx. 30% during the breeding period (no water changes on parental care test days or the day prior to tests).
All but two breeding pairs spawned at least once, some pairs spawned two or three times (N total number of clutches laid = 83), indicating that the pairing duration of 28 days was sufficient for reproduction to occur. For breeding pairs that produced and successfully raised fry (N = 20 breeding pairs), we counted all offspring raised (mean ± SE = 68.7 ± 9.4 offspring per brood) and measured their size (mean ± SE = 1.56 ± 0.03 cm, standard length measured from photos using ImageJ, Schneider et al. (2012)) at the end of the breeding period (day 33). For statistical analyses, we averaged offspring size per brood and calculated offspring survival as the proportion of offspring raised relative to the clutch size (mean ± SE = 0.41 ± 0.05, N = 18 breeding pairs). For two breeding pairs, we found no eggs prior to fry appearing in the tank. Consequently, were not able to fully assess reproductive success for these two breeding pairs and removed them from all analyses performed on successful breeding pairs (N (final) = 18 breeding pairs). Notably, offspring survival and the absolute number of offspring raised were highly correlated (linear mixed-effect model with offspring survival as response, number of offspring raised as predictor, and male family, female family, male color morph, and experimental block as random terms; χ2 = 32.569, p < 0.001, R2 = 0.903).
Parental care tests
During the breeding period, we quantified brood guarding and parental activity in the presence of an animated intruder. We used three different intruder types: a predator (same predators as during boldness tests), a conspecific male, and a conspecific female (conspecifics are brood predators in the species). Each intruder type was used twice, i.e., each breeding pair was tested for their parental care behavior six times with three days elapsing between successive trials. For each breeding pair, we randomised the testing order for the first time each intruder type was used (first, second and third parental care test) and then repeated parental care tests in the same order (six tests in total, e.g., a breeding pair was tested in the following order: male intruder, female intruder, predator, male intruder, female intruder, predator). Parental care observations started one day after offspring became free-swimming, i.e., day 10 post spawning. Fertilised P. pulcher eggs take three days to develop into wrigglers (free embryos), which stay in the breeding cave for approximately another five days. On day 9 following fertilisation, offspring become free-swimming fry and leave the breeding cave to search for food.
To start a parental care test, we introduced a tablet (Surftab Theatre, 13.3" Full-HD-IPS display; Trekstar, Bensheim, Germany) on a standardised side of the breeding tank showing one of the three intruder types for 11 min (Scherer, Godin, et al., 2017). We video-recorded the breeding pair's response and manually assessed the activity of each parent from the videos (duration of video analysis was 10 min, starting 1 min after the start of the video). The camera (Sony HD-CX405) was positioned in front of the tank, filming from diagonally above at a distance of 1 meter. The camera angle was chosen to capture the 3-dimensionality of the tank. Prior to starting trials, fish were allowed to acclimate to the camera for 10 min. Similar to the above boldness tests, parental activity under simulated risk was assessed as the total number of squares visited (including revisits). Therefore, the breeding tank was divided into 16 squares (each measuring 12 x 12 cm) using markings alongside the vertical tank walls. Further, male and female brood guarding was quantified from snapshots, i.e., over the 10 min test period, we scored every 30 sec (21 frames in total) whether a parent was within a 6 cm distance to the brood (half a square, referring to approx. one total fish length). If so, this was scored as brood guarding (Thünken et al., 2010). Distance to the brood was assessed in 3-dimensional space, encompassing both horizontal distance on the ground and vertical distance in the water column. However, in practice, parental fish did not swim more than 6 cm above or below their offspring but all fish were mostly swimming close to the ground. Individual brood guarding was then calculated as the number of frames where the individual was attending the brood divided by all frames analysed (resulting in values ranging from 0 to 1: 0 = the parent was not attending the brood at all, 1 = the parent was with the brood throughout). All videos were analysed by the same observer (US). During video analyses, the observer was unaware of the test fish’s pre-breeding activity under simulated risk.
Male intruder sizes were matched to the male's standard length and, similarly, female intruder sizes were matched to the female's standard length (size difference ≤ 2 mm). Predator sizes were as described in the boldness test. Test fish were presented with unfamiliar intruder specimen only. The intruder type (male conspecific, female conspecific, predator) did not affect parental care behavior, regardless of focal individual personality (Supplementary information 2 “Intruder type did not affect parental care behavior”). Consequently, we averaged individual parental care behavior over all six parental care tests for statistical analyses (not repeatabilities).