Even, say, the difference between "never attend church" and "sometimes attend church" (which evidently makes divorce 10% more likely)? Or how about the difference between household incomes of $85k and $110k (bars differ by 3% out of about 40%)?

But I'm glad to know that these are all results with low p-values, anyway: that adds some real confidence to these results. (Though I ought to ask: how many of these significant p-values were between, say, 0.03 and 0.05? How many different correlations were originally considered in the study before they identified these significant ones?)