Increases in turbidity reduce territory defense and exploration in an East African cichlid (Neolamprologus pulcher)

The code is meant to be run in Rmarkdown with separate chunks for separate analyses.

Subjects and holding conditions

We used cichlids (N. pulcher) that originate from wild individuals caught in Lake Tanganyika near Mpulungu, Zambia, in the years 1999, 2006, 2009 and 2020. In the laboratory they were maintained in breeding groups and stock holding tanks at a temperature range of 26 to 27 °C with a 13:11 h light:dark photoperiod. The tanks were equipped with commercial air driven sponge filters (XY-2822), 1-2 heaters, and the bottom was covered with a layer of sand (grainsize 0.1 - 0.5 mm). The water was treated with 0.4g sera GH/KH-plus and 0.067g Sodium bicarbonate per liter before use. All fish were fed once a day, six days per week, using commercial food flakes (Tetra). On experimental days, we fed the experimental subjects after the last observations.

Figure 1: The experimental set-up in sub-divided tanks with two compartments during the mirror test (left) and the exploration test (right).

Experimental procedure

We haphazardly selected 36 individuals (18 females, mean ± SD standard length: 5.54 ± 0.72 cm; 18 males, 5.41 ± 0.53 cm) from various stock tanks. We placed each of them into one of two compartments of 222 L tanks (45x110x45 cm, Figure 1). An opaque PVC sheet divided the compartments. Each compartment included one flower pot half which the fish quickly accepted as the center of their new territory indicated by the fish frequently using it as refuge. All fish were placed in the experimental compartments 48 hours before the start of the experiment to ensure that they had fully settled in their new environment.

To investigate the effect of turbidity on cichlid behavior, we adopted a within-subject experimental design. Each of our experimental subjects was tested in the control condition (turbidity mean of two measurements before the mirror assay = 0.61  NTU ± 0.23 SD, range: 0.33 - 1.15; and before the exploration assay = 0.68 NTU ± 0.34 SD, range: 0.3 – 2.2) and in a treatment where we used bentonite clay to increase the turbidity (mean before the mirror assay = 4.73 NTU ± 0.86 SD, range = 2.8-6.3; and before the exploration assay = 3.15 NTU ± 0.43 SD, range: 2.30 – 4.05). To achieve this level of turbidity, we suspended 4 grams of bentonite clay in 200 ml of water and added this suspension to the experimental tanks. A pilot trial indicated that adding this suspension initially elevated the turbidity to 6 NTU and that subsequent sedimentation led to a gradual decrease of turbidity to 2-3 NTU over the course of one hour. In the control condition, we added 200 mL of water to the tanks without suspending the bentonite clay prior to adding the water, to have the same chemical properties and disturbance in both turbid and clear treatment. Half of all subjects underwent the treatment first and one day later the control whereas in the other half, we reversed this order. To measure turbidity, we collected water samples in 50 mL falcon tubes for each individual before each behavioural assay and measured turbidity with a WTW Lab Turbidity Meter Turb 550 IR.

We tested each subject for their investment in territory defense by using a mirror assay (Balzarini et al., 2014) and for their tendency to explore the territory by using an exploration test (Figure 1, Schurch & Heg, 2010a). We started the mirror test 10 minutes after adding the turbidity suspension or the control water by placing a 15 x 22 cm mirror inside the tank. Subsequently, we conducted 10 minutes of all-occurrence sampling of aggressive display behaviors (head down, frontal display, lateral display, S-bend) and of overt attacks (ramming, biting, bow-swim) against the mirror (Balzarini et al., 2014; Culbert & Balshine, 2019; Reddon et al., 2019). After a break of 15 minutes, we added 5 new flower pots to the territory of the cichlid and started immediately to record their exploratory behavior for 10 minutes. During these 10 minutes, we recorded each time the focal fish entered or left a pot and subsequently derived the latency to leave the first pot, the number of unique pots visited, the total number of pots visited, the time spent outside pots, mean distance per pot jump, and total distance covered. All observations in both assays were carried out by the same observer sitting 1m from the tank, logging all events in real time on a laptop running BORIS event-logger (Friard & Gamba, 2016).

Statistical analysis

The different display behaviors and overt aggressive behaviors were combined into two separate response variables: the number of restrained aggressive behaviors (sum of head down, frontal display, lateral display, S-bend) and the number of overt attacks (sum of ramming, biting, bow-swim). We used paired non-parametric Wilcoxon signed rank tests to test for behavioral differences between control and treatment. To see if sex affected the behavioral response, we calculated the difference in behaviors between treatment and control for each individual. We then used Wilcoxon signed rank tests to assess if the responses differed between males and females. As latency we defined the time it took for a fish to exit the pot at the start of the exploration test. This variable could only be calculated for the 15 fish that were inside a pot at the onset of the observation in both the treatment and the control and we present the data for these cases. To score exploratory tendency besides the number of visited pots and number of unique pot visits, we calculated the total covered distance while exploring based on the sequence of visited pots (Figure S1). The level of turbidity in both the control and treatment varied due to how much the previously added clay had sedimented, or the currently added clay kept being suspended. To see if this difference in magnitude of turbidity change was associated with the strength of the treatment effect, we performed a linear regression for each response variable. Because the data of one fish in exploration test got lost the sample size is N=35 for this test.

All data manipulation, statistical analyses, and figure creation, were performed using the software R Version 4.3.3 (R Core team 2024) with packages tidyverse (Wickham et al 2019) and ggpubr (Kassambara 2023).