Complex residences and sociality: How coral structure and social environment influence occupation patterns in Gobiodon in Aquaria
Data files
Jul 25, 2025 version files 386.94 KB
-
choiceMF.csv
11.29 KB
-
choiceMFS.csv
13.37 KB
-
choiceS.csv
4.12 KB
-
R_Code.R
9.33 KB
-
R_Output.pdf
327.98 KB
-
README.md
8.26 KB
-
switch_glm2adults.csv
1.80 KB
-
switch_glm2r23.csv
1.99 KB
-
switch_glm2r23sub.csv
820 B
-
test.csv
7.99 KB
Abstract
Habitat size and complexity are two of many important characteristics that can have profound effects individual on survival, growth and reproduction. However, little is known about the influence these characteristics may have on sociality and how this may be altered in response to environmental stress. Amongst marine fishes, coral dwelling gobies are highly specialised cryptobenthic fish that live almost exclusively between the branches of Acropora corals, on which they rely for survival and reproduction. The study investigated whether grouping patterns and habitat choice decisions of a facultatively social species, Gobiodon quinquestrigatus, were influenced by habitat complexity and size. Corals structures were 3D printed with eco-friendly polymers to create high and low complexity Acropora structures allowing coral complexity and size to be precisely manipulated. Replicated trials consisted of three sequential 24-hour treatments: 1) breeder pair only, 2) breeder pair with the addition of a subordinate, and 3) breeder pair and subordinate with a reduced size of the high complexity coral structure. All individuals were observed more frequently in the high complexity coral structure. Females chose the high complexity structure most frequently, with males and subordinates less likely to choose the high complexity structure. Breeders were more likely to choose the high complexity coral structure even when the size of the high complexity coral structure was reduced. Breeders were observed switching more frequently in their first round (breeder pair only) compared to the latter rounds when the subordinate was added and when the coral size was reduced. Breeder males switched marginally more than breeder females. However, subordinates performed the most switches consistently throughout the two rounds they were present. These results highlight the dynamic relationship between sociality and structural complexity in a habitat specialist reef fish, with implications for our understanding of social maintenance in response to climate change and habitat loss.
https://doi.org/10.5061/dryad.dv41ns28f
Description of the data and file structure
The study investigated whether grouping patterns and habitat choice decisions of a facultatively social species, Gobiodon quinquestrigatus, were influenced by habitat complexity and size. Corals’ structures were 3D printed with eco-friendly polymers to create high and low complexity Acropora structures, allowing coral complexity and size to be precisely manipulated. Replicated trials consisted of three sequential 24-hour treatments: 1) breeder pair only, 2) breeder pair with the addition of a subordinate, and 3) breeder pair and subordinate with a reduced size of the high complexity coral structure.
The data included:
File | Info |
---|---|
R_Code.R | The R code file that can be used to run the analysis in R. |
R_Output.pdf | A PDF of the code and the outputs that were generated when the analysis was conducted for investigating and cross-referencing. |
choiceMFS.csv | Data file containing the choices of the male and female individuals throughout the conducted trials. Utilised in the generalised multinomial linear mixed models. Trial variable is the number of the trial (1 to 16). Age is adult or subordinate. Sex.Mref, Sex.Fref and Sex.Sref are three columns that allow for the different sexes to be the reference category. Round is which of the rounds/treatments this row belongs to. Choice.highref, choice.nochoiceref, and choice.lowref are three columns that allow for each of the location categories to be used as the reference category for the multinomial analysis. |
choiceMF.csv | Data file containing the choices of the male and female individuals throughout the conducted trials. Utilised in the generalised multinomial linear mixed models. Trial variable is the number of the trial (1 to 16). Age is adult or subordinate. Round is which of the rounds/treatments this row belongs to. Choice.highref, choice.nochoiceref, and choice.lowref are three columns that allow for each of the location categories to be used as the reference category for the multinomial analysis. |
choiceS.csv | Data file containing the choices of the subordinate individuals throughout the conducted trials. Utilised in the generalised multinomial linear mixed models. Trial variable is the number of the trial (1 to 16). Age is adult or subordinate. Round is which of the rounds/treatments this row belongs to. Choice.highref, choice.nochoiceref, and choice.lowref are three columns that allow for each of the location categories to be used as the reference category for the multinomial analysis. |
switch_glm2adults.csv | Data file containing the details about the switch patterns of the adults across all three rounds. Utilised in the generalised linear models for switching patterns. Group.ID is the equivalent of trial in the above, and is the group trial ID. Individual what number individual they are. Sex is the sex of the individual. Age is the age of the individual. Round is which round of the trial was recorded. Switches is the number of switches conducted by the individual in that trial. |
switch_glm2r23.csv | Data file containing the details about the switch patterns of all individuals (male, female, and subordinate) only in the second and third rounds. Utilised in the generalised linear models for switching patterns. Group.ID is the equivalent of trial in the above, and is the group trial ID. Individual what number individual they are. Sex is the sex of the individual. Age is the age of the individual. Round is which round of the trial was recorded. Switches is the number of switches conducted by the individual in that trial. |
switch_glm2r23sub.csv | Data file containing the details about the switch patterns of the subordinates only across the second and third rounds. Utilised in the generalised linear models for switching patterns. Group.ID is the equivalent of trial in the above, and is the group trial ID. Individual what number individual they are. Sex is the sex of the individual. Age is the age of the individual. Round is which round of the trial was recorded. Switches is the number of switches conducted by the individual in that trial. |
test.csv | This is the data file for the chi-squared test. No. is the number entry in the file. Sex is the sex status of the individual. Round is which round was recorded. Choice is the location of the individual during the round. |
Code/software
All analyses were conducted in R (v4.2.1) (R Core Team, 2021), using the packages ‘mclogit’ (Elff, 2022), ‘nnet’ (Venables, Ripley and Venables, 2002), ‘memisc’ (Elff, 2023), ‘sJPlot’ (Lüdecke, 2023), ‘lme4’ (Bates et al., 2015), ‘lmertest’ (Kuznetsova et al., 2017) and ‘car’ (Fox and Weisberg, 2019).
To examine whether the high complexity coral structure was chosen significantly more often than the low complexity structure and no-choice options, a chi-squared goodness of fit test was performed. All recordings 12 hours post the addition of the individuals were included in the analysis, as it was at this point that the gobies appeared to settle within the trial tank.
To examine the factors influencing host coral choice, two multinomial generalised linear mixed models were run. The first model examined the effect of sexual/social status on the choice of coral host, with group ID as a random factor. Note that the variable “sex” encompasses 3 categories, i.e., breeder male, breeder female, and subordinate non-breeder, and therefore reflects a combination of sex and status. This first model was run twice, with 1) breeder females as the reference category and 2) breeder males as the reference category, thus ensuring that comparisons between all present sexes could be observed. Only the choices made after the subordinate was added were used in the analyses, as this ensured all individuals were present at all time points used in the analysis. The second model examined the effect of trial round on host choice decisions, with group ID as a random effect. Each of the three trial rounds was represented by the recorded locations at 12 hours and 24 hours post commencement. The model was run in two stages: 1) the breeder pair only across all trial rounds, and 2) subordinates only across the second and third trial rounds. This division accommodated the fact that the subordinates were not present during the first trial round whilst also enabling comparisons between breeder and subordinate decisions to be examined. The location 12 hours post addition was used as the reference category for both status groups, respectively (breeder and subordinate), as there were very few choices made during the initial observation recorded immediately after addition.
To examine the factors influencing the number of switches, generalised linear mixed models were used with sexual/social status, trial round, and their interaction as predictors of the number of switches performed (response). The analysis was conducted in three stages: 1) breeder male and female only and during all rounds and, 2) breeder male, female and subordinate during the second and third trial rounds, and 3) subordinate during the second and third rounds (however due to overfitting, a generalised linear model without a random factor was performed due to overfitting). This design was used due to the subordinates not being present during all rounds of the trial. A random factor of group identification was included in both analyses. The ‘Poisson’ distribution family was used in both models as the number of switches is a count variable. A negative binomial distribution was initially attempted, but the Poisson model was deemed a better fit to the data based on AIC values. The models were checked for overdispersion, homogeneity, and collinearity.