The NYT piece is bad science reporting. News reporters read a science report then generalize it, and make a story that is going to catch eyes. It has Always been this way.
But the blog going on to cite the p-value.
There were five tests each, say the authors, showing a significant effect in the expected direction (literary fiction makes you more able to read other’s emotions than does reading nonliterary fiction, nonfiction, or nothing at all). But in many cases the significance levels are marginal—p values around 0.04, when the cutoff boundary is 0.05—and one value of 0.08 (sometimes psychologists use higher cutoffs like p < 0.01). Those values represent the chance of getting the observed result given that there really is no effect on empathy of reading literary stuff.The cut-off value of p is not really that important. (Waiting to see whether Andrew Gelman will talk about it.) The DOE might be more interesting. I don't think there is easy way to measure the long term effect of reading classical fiction on people's emotional intelligence. Reading is good, but to improve emotional intelligence requires much more than that.
The authors are trying to answer an important question, but they find a measurement slightly related to the real questions people are interested, and that is the area where they can research and get paper published (the sacred p-value.)
For the finding itself, with some generalization, it seems to confirm many people's observation. Like, Joseph Badaracco's new book and his Harvard class. Before I find the ungated Science report, Badaracco's Leadership in Literature is a good read.