The sample size of the typical ecological correlation coefficient is small and slowly declining

Larger sample sizes are desirable because they minimize sampling error. However, they are not the only desideratum, and it is unknown if sample sizes in ecology trade off with other desiderata. Here I describe the typical sample size of ecological studies reporting correlation coefficients, describe how sample sizes have changed over time, and develop a hypothesis to explain these changes. Using a database of over 16 000 correlation coefficients reported in 232 meta‐analyses covering a wide range of ecological topics, I find that the median sample size of these 16 000+ correlation coefficients is just 30 observations. This implies that the majority of ecological correlation coefficients likely have low statistical power against the null hypothesis of zero correlation. The typical sample size of an ecological correlation coefficient has decreased slowly since WW II. Decreasing sample sizes may reflect changing research practices. Recent ecology papers tend to report more correlation coefficients, with smaller sample sizes, than they did decades ago. Ecologists today may be choosing to measure more correlations among more variables than ecologists decades ago, even at the cost of smaller sample sizes per correlation. Further research is needed to identify other factors that might drive declining sample sizes of ecological correlation coefficients, and to determine if ecological sample sizes are declining more broadly. It is unclear if declining sample sizes represent a systemic problem for ecological research, given that large sample size is only one of many desiderata in scientific research, and that large sample sizes may trade off with other desiderata. But the trend towards smaller sample sizes should be recognized so that its implications can be discussed.