Remember, a confidence interval for a mean (average) is the mean ± (t x standard error). The value for t, as mentioned earlier is based on how confident you want to be. The value for a 95% confidence interval is 1.96. The standard error is a reflection of the amount of variability in the data. It is calculated by dividing the standard deviation by the square root of the number of study subjects. In larger studies, small amounts of variation commonly found are blunted by the average effect. Smaller studies are not robust enough so the small amounts of "noise" in the data make a relatively larger contribution to the mean. In numerical parlance, dividing by a large number makes for a smaller one (i.e., a smaller standard error). Dividing by a small number gives a large one (i.e., a large standard error). So, all things being equal, small confidence intervals come from large studies and large confidence intervals come from small studies.