Are lateralized and bold fish optimistic or pessimistic?
Data files
Nov 26, 2024 version files 30.21 KB
-
glm_meantimes.csv
3.58 KB
-
README.md
12.82 KB
-
repeatability_laterality.xlsx
13.80 KB
Abstract
Cognitive bias is defined as the influence of emotions on cognitive processes. The concept of cognitive judgement bias has its origins in human psychology but has been applied to animals over the past 2 decades. In this study, we were interested in determining if laterality and personality traits, which are known to influence learning style, might also be correlated with a cognitive bias in the three-spined sticklebacks (Gasterosteus aculeatus). We used the judgement bias test with the go/no-go procedure where fish were first trained to discriminate between a black and white card and, after reaching a minimum learning criterion, tested their response to an ambiguous card (grey). Optimistic subjects were expected to have a high expectation of reward associated with an ambiguous stimulus, whereas pessimistic subjects had a high expectation of non-reward. We used an emergence and a mirror test to quantify boldness and laterality respectively. We hypothesised that male, bolder and more strongly lateralized fish would be more optimistic than female, shy and less strongly lateralised fish. We found that males and more strongly lateralized fish were more optimistic than females and less strongly lateralized fish. In addition, bold males were more optimistic than shy males as we predicted, but females showed the opposite pattern. Finally, fish trained on the black colour card learned the training task faster than those trained on a white card. Our results indicate that both laterality and personality traits are linked to animals’ internal states (pessimistic or optimistic outlooks) which likely has broad implications for understanding animal behaviour, particularly in a welfare context.
README: Are lateralized and bold fish optimistic or pessimistic?
https://doi.org/10.5061/dryad.1ns1rn920
Description of the data and file structure
Fish were tested for personality and laterality using the emergence test as detailed below. The day after the first emergence test, fish started the judgement bias protocol. To determine the repeatability of personality, the emergence tests were conducted a second time at the end of the judgement
bias probe trials. Note we did not score laterality during the second trial.
We conducted an emergence test to assess both boldness and laterality, eliminating the need to handle the fish in the interim. We remotely opened the box and the latency to emerge (the whole head had to be out of the shelter) was scored as a measure of boldness (Brown et al. 2007).
The legend of " Duration_1: time to go out from the shelter " was used as boldness.
Laterality scores were based on which eye they used to look at either mirror by pausing the video every 5s. Specifically, we scored the number of times they watched the mirror with their left, right, and both eyes and if they were not looking at the mirror based on body orientation (sensu Sovrano et al. 1999).In the laterality index and absolute laterality we used the eye to look at the mirror and calculated the formula:
· laterality_index_1: laterality index= Right eye - left eye/ Right eye - left eye
· absolute_laterality_1: strength/absolute laterality= |Right eye - left eye/ Right eye - left eye|
After the first trial of the emergence test, The cognitive judgment bias test was divided into two phases: training and probe trials.
During the training phase (Fig. 2a), half of the fish received a food reward (bloodworm) delivered by a plastic pipette if they approached a white card within one body length whereas if they approached a black card they were not rewarded (i.e. neutral outcome). For the remaining fish, the black card was rewarded and the white card was neutral. The position of the cards (left or right side) was fixed for each fish throughout the training phase to facilitate rapid training but varied between subjects.
The legend of "· FOOD REWARD_ colour of the reward during the training, black or white." regards the colour of black and white during the training phase. It is the same as; "Group: trial is the fish during the training with the two different food rewards, so black and white card.".
The training phase continued until the fish reached the minimum learning criteria or a maximum of 50 training trials whichever occurred sooner. To reach the criteria, subjects had to select the rewarded card in 8 out of 10 trials.
In the legend:"· N_tentatives: attempts to reach the minimum learning criteria during the training to do the probe test afterwards." it is during the training phase, the number of tentatives to reach the minimum learning criteria.
While in the legend "mean_trial: average of the last five trials of the training." refers to only the last 5 trials of the training phase.
The second phase of the experiments involved conducting three probe trials on three consecutive days where fish (n = 53) were exposed to an ambiguous cue.
In the legend: "·
· CT_1: First cognitive bias test, probe test, with the ambiguous stimulus, grey colour.
· CT_2: second cognitive bias test, probe test, with the ambiguous stimulus, grey colour.
· CT_3: third cognitive bias test, probe test, with the ambiguous stimulus, grey colour.
· mean_ct: average value of the three cognitive bias tests.
Where the CT_1, CT-2 and CT_3 refer to the 3 probe trials where the fish were exposed to an ambiguous cue.
Then the Mean-CT is the average of these 3 probe trials
Excel file: “glm wt mean times”
· ID: ID of the fish.
· SEX: Sex of the fish, if NA it means not recognizable.
· Duration_1: time to go out from the shelter.
· laterality_index_1: laterality index= Right eye - left eye/ Right eye - left eye
· absolute_laterality_1: strength/absolute laterality= |Right eye - left eye/ Right eye - left eye|
· FOOD REWARD_ colour of the reward during the training, black or white.
· N_tentatives: attempts to reach the minimum learning criteria during the training to do the probe test afterwards.
· CT_1: First cognitive bias test, probe test, with the ambiguous stimulus, grey colour.
· CT_2: second cognitive bias test, probe test, with the ambiguous stimulus, grey colour.
· CT_3: third cognitive bias test, probe test, with the ambiguous stimulus, grey colour.
· mean_ct: average value of the three cognitive bias tests.
· mean_trial: average of the last five trials of the training.
Excel file 2: “long format wt grey”
· ID: ID of the fish.
· FOOD REWARD_ colour of the reward during the training, black or white.
· mean_time: it is the same as mean: trial, average of the last five trials of the training.
· Group: trial are the fish during the training with the two different food rewards, so black and white card.
· Attempts: same as N_tentatives: attempts to reach the minimum learning criteria during the training to do the probe test afterwards.
· Duration_1: time to go out from the shelter.
· laterality_index_1: laterality index= Right eye - left eye/ Right eye - left eye
· absolute_laterality_1: strength/absolute laterality= |Right eye - left eye/ Right eye - left eye|
Code/Software
Data Analysis
We used the latency to emerge from the shelter during the emergence tests as our measure of boldness. Bold fish should emerge from the shelter more quickly than shy fish (Brown et al., 2007). To assess the consistency of boldness scores across the two different trials, we estimated the adjusted repeatability with 1000 parametric bootstraps. We used the “rptR” package to calculate adjusted repeatability estimates for personality (RStudio 2022.07.2+576).
During the learning phase of the cognitive bias test, we used the number of trials to reach the learning criteria as a measure of learning speed. Since all fish approached the ambiguous stimuli during the test phase, we took the mean time to approach the ambiguous stimuli over the three probe trials as our measure of cognitive bias. Optimistic fish should approach more quickly than pessimistic fish.
The data from the probe trials did not follow a normal (Gaussian) distribution, being right-skewed, so we employed Generalised Linear Models estimated using Maximum Likelihood. The response variables were modelled by implementing a Gamma probability distribution with a log link function, which handles positively skewed data. We conducted two analyses. We first examined the mean time to approach the ambiguous stimuli during the three probe trials (mean Ct). Sex, boldness (from trial 1), and the strength of lateralisation (Laterality) were entered as independent variables. We also thought that their experience during the training phase might influence mean Ct, so training colour (black or white) and the number of trials to reach the learning criteria (Attempts) were also added to the model. It might be possible that fish that reached training criteria quickly might continue to do so in the probe trails (Mean Ct).
Response variables and predictors were first rescaled (subtracting the root mean square) to improve model convergence. We fitted different models according to our hypotheses and these were then compared using the Akaike Information Criterion (AIC), which penalises increasing model complexity to avoid over-fitting. Selection of the most parsimonious model was based on Delta AICc (AIC corrected for small sample size) ≥ 2 between the best model and the second-best model, and the highest AIC weight (AICcWt), indicating the overall explanatory power among the set of models (see Tables 1 and 3). Since we built the models to test our hypotheses on selected variables from our training data set, we did not include the null model (intercept only) in the comparative analyses.
Modelling of the mean time to approach the ambiguous stimulus in the cognitive bias test started from the hypothesis of an interaction between Sex and Boldness, as they are often related, Laterality and Sex as sex differences in laterality are well known, and Laterality and card Colour to investigate a possible association between brain lateralization and colour preference (Karenina & Giljov, 2022; Lopez-Persem et al., 2020). We also included a combination of Laterality, Attempts, and training Colour, to account for possible colour bias in the learning process and increase explanatory power (see table 1). It seemed reasonable to assume that those individuals who approached the colour cards and learned the task quickly during the training phase might also continue that during the probe trials. Finally, we tested another potential interaction between Sex and Boldness, and Sex and Laterality, excluding training Colour and Attempts variables, with a reduction in model complexity. Note because the location of the coloured cards varied between individual fish, this is a true test of colour influence rather than being confounded by location. This led to five models being tested (the asterisk indicates an interaction in the following Models tested):
- Mod1 "meanCT ~ Sex + Boldness + Laterality + Attempts + Colour"
- Mod2 "meanCT ~ Sex * Boldness + Sex + Boldness +Laterality + Attempts * Colour"
- Mod3 "meanCT ~ Sex * Boldness + Sex + Boldness +Laterality * Attempts * Colour"
- Mod4 "meanCT ~ Sex + Boldness + Attempts + Laterality * Colour"
- Mod5 "meanCT ~ Sex + Boldness + Laterality + Sex*Boldness + Sex*Laterality"
Modelling of the number of attempts to reach learning criteria during the training phase was based on two possible interactions between predictor variables. In one scenario boldness, lateralisation and training colour interact to influence the number of attempts taken by a fish during training trials (Boldness * Laterality * Colour, interaction between boldness and laterality and training colour). Here, we were looking for a link between lateralization and boldness, which has been tested in other studies with diverse outcomes (Brown & Bibost, 2014; Reddon & Hurd, 2009). Alternatively, we expected interaction between Sex and Boldness, and Laterality and training Colour (Sex * Boldness + Laterality * Colour), following the same criteria used in the models above. We also included the saturated model in our AIC test, which contrasted the simplest form, lacking interactions. This approach led to four models being tested:
- Mod1 "Attempts ~ Sex + Boldness + Strength Laterality + Colour"
- Mod2 "Attempts ~ Sex + Boldness * Strength Laterality * Colour"
- Mod3 "Attempts ~ Sex * Boldness + Strength Laterality * Colour"
- Mod4 "Attempts ~ Sex * Boldness * Strength Laterality * Colour"
To determine if certain fish simply approached any card quickly, we analysed the mean time to approach (MeanTimeApproach) and the correct card (white or black depending on the fish) during the last five trials of the training phase. We tested the effect of Boldness, Laterality, Sex, Colour and their interactions. Data were log-transformed to improve model convergence. In order to account for all combinations between independent variables, we ran a multi-model inference “dredge” function (package MuMIn in R) and selected the best four performing models according to the AICc criterion (see Table 8). As AICc scores were comparable, we retained the most informative model (model 4) which included two-way interactions between laterality and personality with the colour of the reward card.
- Mod1 "Time ~ Boldness + 1"
- Mod2 "Time ~ Boldness + Colour + Boldness:Colour + 1"
- Mod3 "Time ~ Boldness + Laterality + 1"
- Mod4 "Time ~ Boldness + Laterality + Colour + Boldness:Colour + Laterality:Colour + 1"
Lastly, we analysed the time to approach the white card and black card during the training phase, based on the last 5 trials per fish, and the grey card during the probe test using a Kruskall-Wallis test followed by pair-wise comparisons.
Methods
We used the latency to emerge from the shelter during the emergence tests as our measure of boldness. Bold fish should emerge from the shelter more quickly than shy fish (Brown et al., 2007). To assess the consistency of boldness scores across the two different trials, we estimated the adjusted repeatability with 1000 parametric bootstraps. We used the “rptR” package to calculate adjusted repeatability estimates for personality (RStudio 2022.07.2 + 576) (file excel “repeatability laterality”).
During the learning phase of the cognitive bias test, we used the number of trials to reach the learning criteria as a measure of learning speed. Since all fish approached the ambiguous stimuli during the test phase, we took the mean time to approach the ambiguous stimuli over the three probe trials as our measure of cognitive bias. Optimistic fish should approach more quickly than pessimistic fish.
The data from the probe trials did not follow a normal (Gaussian) distribution, being right-skewed, so we employed Generalised Linear Models estimated using Maximum Likelihood. The response variables were modelled by implementing a Gamma probability distribution with a log link function, which handles positively skewed data. We conducted two analyses. We first examined the mean time to approach the ambiguous stimuli during the three probe trials (mean Ct). Sex, boldness (from trial 1), and the strength of lateralisation (Laterality) were entered as independent variables. We also thought that their experience during the training phase might influence mean Ct, so training colour (black or white) and the number of trials to reach the learning criteria (Attempts) were also added to the model. It might be possible that fish that reached training criteria quickly might continue to do so in the probe trails (Mean Ct).
Response variables and predictors were first rescaled (subtracting the root mean square) to improve model convergence. We fitted different models according to our hypotheses and these were then compared using the Akaike Information Criterion (AIC), which penalises increasing model complexity to avoid over-fitting. Selection of the most parsimonious model was based on Delta AICc (AIC corrected for small sample size) ≥ 2 between the best model and the second-best model, and the highest AIC weight (AICcWt), indicating the overall explanatory power among the set of models (see Tables 1 and 3). Since we built the models to test our hypotheses on selected variables from our training data set, we did not include the null model (intercept only) in the comparative analyses.
Modelling of the mean time to approach the ambiguous stimulus in the cognitive bias test started from the hypothesis of an interaction between Sex and Boldness, as they are often related, Laterality and Sex as sex differences in laterality are well known, and Laterality and card Colour to investigate a possible association between brain lateralization and colour preference (Karenina & Giljov, 2022; Lopez-Persem et al., 2020). We also included a combination of Laterality, Attempts, and training Colour, to account for possible colour bias in the learning process and increase explanatory power (see table 1). It seemed reasonable to assume that those individuals who approached the colour cards and learned the task quickly during the training phase might also continue that during the probe trials. Finally, we tested another potential interaction between Sex and Boldness, and Sex and Laterality, excluding training Colour and Attempts variables, with a reduction in model complexity. Note because the location of the coloured cards varied between individual fish, this is a true test of colour influence rather than being confounded by location. This led to five models being tested (the asterisk indicates an interaction in the following Models tested):
- Mod1 "meanCT ~ Sex + Boldness + Laterality + Attempts + Colour"
- Mod2 "meanCT ~ Sex * Boldness + Sex + Boldness +Laterality + Attempts * Colour"
- Mod3 "meanCT ~ Sex * Boldness + Sex + Boldness +Laterality * Attempts * Colour"
- Mod4 "meanCT ~ Sex + Boldness + Attempts + Laterality * Colour"
- Mod5 "meanCT ~ Sex + Boldness + Laterality + Sex*Boldness + Sex*Laterality"
Modelling of the number of attempts to reach learning criteria during the training phase was based on two possible interactions between predictor variables. In one scenario boldness, lateralisation and training colour interact to influence the number of attempts taken by a fish during training trials (Boldness * Laterality * Colour, interaction between boldness and laterality and training colour). Here, we were looking for a link between lateralization and boldness, which has been tested in other studies with diverse outcomes (Brown & Bibost, 2014; Reddon & Hurd, 2009). Alternatively, we expected interaction between Sex and Boldness, and Laterality and training Colour (Sex * Boldness + Laterality * Colour), following the same criteria used in the models above. We also included the saturated model in our AIC test, which contrasted the simplest form, lacking interactions. This approach led to four models being tested:
- Mod1 "Attempts ~ Sex + Boldness + Strength Laterality + Colour"
- Mod2 "Attempts ~ Sex + Boldness * Strength Laterality * Colour"
- Mod3 "Attempts ~ Sex * Boldness + Strength Laterality * Colour"
- Mod4 "Attempts ~ Sex * Boldness * Strength Laterality * Colour"
To determine if certain fish simply approached any card quickly, we analysed the mean time to approach (MeanTimeApproach) and the correct card (white or black depending on the fish) during the last five trials of the training phase. We tested the effect of Boldness, Laterality, Sex, Colour and their interactions. Data were log-transformed to improve model convergence. In order to account for all combinations between independent variables, we ran a multi-model inference “dredge” function (package MuMIn in R) and selected the best four performing models according to the AICc criterion (see Table 8). As AICc scores were comparable, we retained the most informative model (model 4) which included two-way interactions between laterality and personality with the colour of the reward card.
- Mod1 "Time ~ Boldness + 1"
- Mod2 "Time ~ Boldness + Colour + Boldness:Colour + 1"
- Mod3 "Time ~ Boldness + Laterality + 1"
- Mod4 "Time ~ Boldness + Laterality + Colour + Boldness:Colour + Laterality:Colour + 1"
Lastly, we analysed the time to approach the white card and black card during the training phase, based on the last 5 trials per fish, and the grey card during the probe test using a Kruskall-Wallis test followed by pair-wise comparisons.