Ced crossing group sizes (figure 5c). On the other hand, the

Ced crossing group sizes (figure 5c). On the other hand, the static models were inadequate at reproducing such large-scale patterns of the data (figure 5d). Therefore, on both the fine- and large-scale the dynamical models proved better at describing the decisions that produce the observed crossing behaviour. The models evaluations in(a) ?600 log2 P(data | model)/bits(b) 6 5 ?100 crossing group 4 3 2 C?C+ ?100 0 S1 S2 S3 S4 D1 D2 D3SD model 2 (d) 6 5 0.32 0.41 0.47 0.61 0.39 2 0.33 0.20 0.27 0.20 0.12 0.28 0.21 0.11 0.09 0.34 0.25 crossing group 0.18 0.14 0.06 0.02 6 6 5 4 3 2 1 0.47 0.53 2 0.23 0.28 0.49 0.11 0.14 0.25 0.51 0.06 0.06 0.15 0.26 0.48 0.04 0.04 0.06 0.12 0.29 0.45 6 3 4 5 crossing pool 6 1 0.43 0.57 0.35 0.30 0.35 0.30 0.21 0.24 0.25 0.44 0.14 0.14 0.13 0.14 0.48 0.21 0.08 0.09 0.08 0.rsif.royalsocietypublishing.org?J. R. Soc. Interface 11:(c)crossing group4 3 23 4 5 crossing pool3 4 5 crossing poolFigure 5. Large-scale and fine-scale model comparison, combined over all group sizes. (a) Log-marginal-likelihoods evaluated for the seven tested models. Model D1 (`follow last mover’) is the optimal selected model, with a large likelihood ratio compared with all other models. Within static models (S1 ?4), model S1 (`binary response’) is the best fit. Models marked as black or grey Isorhamnetin biological activity circles were respectively inconsistent or consistent in reproducing the large-scale patterns of the data (b ?d); (b) experimental results showing the proportion of time a crossing group of size n crossed the arena from the potential number of fish (crossing pool) that could have crossed (i.e. the number of fish that were initially present on the side from which the crossing was initiated.) In each case, the most probable movement is all the available fish from the pool crossing together, indicating a strong preference to follow the movements of local conspecifics. (c) Large-scale movement groups sizes obtained from simulation of the best-fit dynamic model (D1), showing consistency with the experimental pattern. (d ) Large-scale movement groups sizes obtained from simulation of the best-fit static model, S1, showing inconsistency with the experimental pattern. See the electronic supplementary material for a breakdown of results by different group size experiments and for full model details. (Online version in colour.)figure 5a are colour-coded according to their consistency with this large-scale behaviour, with grey markers indicating consistency (C? and black markers inconsistency (C2). Figure 5 shows results aggregated across different group sizes, see the electronic supplementary material, figures S1?S4 for NS-018 site group-size-specific results. Successive moves between coral patches were more likely to be in the same direction (60 ) than not (40 ). However, when the time between successive moves was more than 3.5 s crossings were more likely to be in opposite directions than expected from these averages (see the electronic supplementary material, figure S5). This provides further evidence that short-term temporal information (D1 model) is more important in driving fishes’ decisions to move between patches rather than the other forms of information described in the alternate models. We considered whether fish might switch strategies to using spatial information if none immediately followed the recent movement of a conspecific. To do this, we used the subset of data with longer intervals between successive crossings to investigate whether the static models we.Ced crossing group sizes (figure 5c). On the other hand, the static models were inadequate at reproducing such large-scale patterns of the data (figure 5d). Therefore, on both the fine- and large-scale the dynamical models proved better at describing the decisions that produce the observed crossing behaviour. The models evaluations in(a) ?600 log2 P(data | model)/bits(b) 6 5 ?100 crossing group 4 3 2 C?C+ ?100 0 S1 S2 S3 S4 D1 D2 D3SD model 2 (d) 6 5 0.32 0.41 0.47 0.61 0.39 2 0.33 0.20 0.27 0.20 0.12 0.28 0.21 0.11 0.09 0.34 0.25 crossing group 0.18 0.14 0.06 0.02 6 6 5 4 3 2 1 0.47 0.53 2 0.23 0.28 0.49 0.11 0.14 0.25 0.51 0.06 0.06 0.15 0.26 0.48 0.04 0.04 0.06 0.12 0.29 0.45 6 3 4 5 crossing pool 6 1 0.43 0.57 0.35 0.30 0.35 0.30 0.21 0.24 0.25 0.44 0.14 0.14 0.13 0.14 0.48 0.21 0.08 0.09 0.08 0.rsif.royalsocietypublishing.org?J. R. Soc. Interface 11:(c)crossing group4 3 23 4 5 crossing pool3 4 5 crossing poolFigure 5. Large-scale and fine-scale model comparison, combined over all group sizes. (a) Log-marginal-likelihoods evaluated for the seven tested models. Model D1 (`follow last mover’) is the optimal selected model, with a large likelihood ratio compared with all other models. Within static models (S1 ?4), model S1 (`binary response’) is the best fit. Models marked as black or grey circles were respectively inconsistent or consistent in reproducing the large-scale patterns of the data (b ?d); (b) experimental results showing the proportion of time a crossing group of size n crossed the arena from the potential number of fish (crossing pool) that could have crossed (i.e. the number of fish that were initially present on the side from which the crossing was initiated.) In each case, the most probable movement is all the available fish from the pool crossing together, indicating a strong preference to follow the movements of local conspecifics. (c) Large-scale movement groups sizes obtained from simulation of the best-fit dynamic model (D1), showing consistency with the experimental pattern. (d ) Large-scale movement groups sizes obtained from simulation of the best-fit static model, S1, showing inconsistency with the experimental pattern. See the electronic supplementary material for a breakdown of results by different group size experiments and for full model details. (Online version in colour.)figure 5a are colour-coded according to their consistency with this large-scale behaviour, with grey markers indicating consistency (C? and black markers inconsistency (C2). Figure 5 shows results aggregated across different group sizes, see the electronic supplementary material, figures S1?S4 for group-size-specific results. Successive moves between coral patches were more likely to be in the same direction (60 ) than not (40 ). However, when the time between successive moves was more than 3.5 s crossings were more likely to be in opposite directions than expected from these averages (see the electronic supplementary material, figure S5). This provides further evidence that short-term temporal information (D1 model) is more important in driving fishes’ decisions to move between patches rather than the other forms of information described in the alternate models. We considered whether fish might switch strategies to using spatial information if none immediately followed the recent movement of a conspecific. To do this, we used the subset of data with longer intervals between successive crossings to investigate whether the static models we.