A famous claim holds that a few minutes of literary fiction immediately improve the ability to infer what another person is feeling. That is too tempting a claim to leave unchecked, so several teams repeated the comparison against popular fiction. What came out when they were pooled?
Curiosity counts. Here is the answer.
Small and unstable. The original claim, which spread very widely, was that a few minutes of literary fiction immediately improve the ability to infer what another person feels, more than popular fiction does. Studies repeated exactly that comparison, and once they were pooled the overall result was not significant and the variation between studies was large. Notice what this does not say: the absence of significance is no proof that the effect is zero. It may exist and be small, or depend on the text and the reader, and in any case the claim is about a few minutes, not about a lifetime of reading.
No stable advantage. One high-powered replication came out looking much like the original, but a small meta-analysis of the together found no benefit, and the was large. So the effect may be small, text-dependent, or very sensitive to the way a test captures .
The original claim became famous because it was so strong. Not years of reading but a few minutes of literary fiction, and immediately afterwards a better ability to infer what another person feels.
The replication was done properly. A team ran a direct replication at high statistical power, meaning a sample large enough to detect the effect if it is there, rather than an improved version of the experiment. The result looked much like the original, and an examination of the significance curve suggested the original evidence was weak.
Then came the pooling. A small meta-analysis of the original study and further direct replications found no stable advantage, and showed substantial variation between studies. Not a reversed result, then, but a loss of confidence: the effect may be small, or text-dependent, or very sensitive to the way a test captures the reading of another mind.
And two jumps are worth avoiding. Absence of significance is not proof that no difference exists, and the meta-analysis itself is small while the measures are disputed. Above all, long-term reading may have real social value even if one passage does not immediately move an emotion-recognition test. Those are two different questions.
wat. One replication supported the literary advantage, and pooling the replications found no stable, significant gain in reading other minds.
wat wat
A headline that flatters reading is especially appealing, and a welcome claim has to survive replication like any other.
How do we know?
A team ran a high-powered direct replication and then pooled it with the original study and with further direct replications.
What is still unknown
The meta-analysis was small and the measures themselves are disputed; absence of significance is not proof of an absolute zero.
Learn something?
Where to next
Does a spoiler really make a story more enjoyable?literature
Does reading literary fiction make us more empathic?literature
Why would reading novels make somebody better with people?literature