Do you understand less when you listen instead of reading?

On average, no. Clinton-Lisell pooled 46 studies covering 4,687 participants and found a difference of g = 0.07 in favour of reading, which was not statistically significant (Review of Educational Research, 2022). On literal comprehension the gap was effectively zero. The exception is inferential comprehension — reading between the lines — where reading kept a real advantage of g = 0.36.

Why does reading still win on inference?

The most likely reason is control rather than the eyes themselves. A reader sets the pace, stops, and goes back over a sentence; a listener travels at the speed of the recording. The same meta-analysis found reading's advantage larger when reading was self-paced, which points at pacing rather than at the modality.

Is listening as good as reading for dense technical material?

Nobody has really tested it. Singh and Alexander reviewed 32 documents and noted the shortage of studies on expository text and on adults without an identified reading difficulty — which is precisely the combination most knowledge workers are. The honest answer is that the evidence covers narrative and mixed material well and technical material poorly.

46 studies · 4,687 participants · peer-reviewed

Do you understand less when you listen?

On average, no — and this is one of the better-evidenced findings in the area. It also has one exception, which is the part worth knowing, and which nobody selling audio tools tends to mention.

The number

Clinton-Lisell pooled 46 studies covering 4,687 people and compared what they understood after reading against what they understood after listening to the same material. The overall difference was g = 0.07 in favour of reading, p = 0.23 — statistically indistinguishable from no difference at all.

On literal comprehension — what the text actually said — the gap was effectively zero (g = −0.01). If your question is “will I still know what was in it”, the answer from the evidence is yes.

Clinton-Lisell, V. (2022). Listening ears or reading eyes: a meta-analysis of reading and listening comprehension comparisons. Review of Educational Research, 92(4), 543–582. doi:10.3102/00346543211060871

The exception, which we are not going to hide

On inferential comprehension — reading between the lines, connecting two things the text never connects for you — reading kept a real advantage: g = 0.36, p = 0.02. The advantage also grew when reading was self-paced (g = 0.13).

Those two results point at the same explanation, and it is not about eyes versus ears: a reader controls the pace and a listener does not. You stop, you go back over the sentence, you sit on it. Audio moves whether you are ready or not.

Which is why the useful conclusion is not “listen instead of reading”. It is listen when the point is to take it in, read when the point is to work it out — and be able to switch without losing your place.

What this does not prove

“No significant difference” is not proof of equivalence. Rogowsky, Calhoun and Tallal found no difference across 91 adults, immediately and two weeks later (SAGE Open, 2016, doi:10.1177/2158244016669550). That is a failure to detect a difference, not a test of equivalence — a distinction the popular write-ups of that study routinely lose.
The evidence thins out on exactly our kind of material. Singh and Alexander reviewed 32 documents and found audio at least as effective for narrative and mixed texts — while noting a shortage of studies on expository text and on adults with no identified reading difficulty (Educational Psychology Review, 2022, doi:10.1007/s10648-021-09653-2). Dense technical prose, read by a working adult, is the least-studied corner of this literature. We are extrapolating into it, and we would rather say so.
Nothing here is about doing two things at once. Whether you absorb an answer better while your hands stay on the keyboard is a separate question, and the honest state of it is: no study has measured “listening to an AI answer while coding”. The theory that predicts it should work — separate channels for hearing and seeing — is a model, not an experiment.

Two numbers you will see quoted, and shouldn’t

“Handwriting beats typing for retention.” This comes from Mueller and Oppenheimer’s 2014 “the pen is mightier than the keyboard”. Its authors published a corrigendum in 2018, and direct replications in 2019 and 2021 did not reproduce the effect; it is now a standard teaching example of the replication crisis. It also compares handwriting with typing — not listening with reading — so it would not answer this page’s question even if it had held.

“It takes 23 minutes to refocus after an interruption.” The figure is misattributed, and the study usually cited for it concludes something close to the opposite. What is measurable: people take on the order of ten to fifteen minutes to return to the suspended task after an alert (Iqbal & Horvitz, CHI ’07, doi:10.1145/1240624.1240730), and among developers only about one session in ten resumes coding in under a minute (Parnin & Rugaber, Software Quality Journal, 2011, doi:10.1007/s11219-010-9104-9).

Why we built listening into SpeakApp

A long answer from an AI assistant is the case this literature covers well: prose, first reading, mostly literal. SpeakApp reads it aloud where it appears, so you can take it in without stopping what your hands are doing — and the text stays on screen, because the one place reading genuinely wins is the moment you need to go back over a line.

Is dictation faster than typing?Voice control for AI agents on Windows Download SpeakApp for Windows