The influence of sample size on causal-strength judgments