Shoal cohesion (probability of splitting) and consequent effects on collective decision-times were influenced by individual boldness differences, but were not linked to majority averages in boldness. (A) The mean boldness of shoal members (5% trimmed to exclude biases by extremely bold or timid fish) had a negative, non-significant, effect on mean decision times between initial and probe trial (black line and marks), but no effect on splitting probability (grey curve and marks) as indicated by regression models (decision times: linear, probability of splitting: binomial). (B) In contrast, the variance in boldness within shoals (mean average deviation of all fish) positively predicted the probability of splitting at probe and initial trials (grey curve and marks) and the mean decision times between initial and probe trial (black line and marks). (C) The level of consistency in splitting between initial and probe trials was greater for shoals with higher variance in boldness (Zero splitting:mean MAD=0.225, one trial: mean MAD=0.279, two trials: mean MAD=1.26; ANOVA, P<0.01) and (D) shoals took longer to reach a decision if they split (split: mean=21.82±3 s.e.m., no split: mean=72.8±12 s.e.m.; Welch's t, P<0.01).