Do Audiobooks Count as Reading? The Comprehension Research Says Mostly Yes

The question is usually asked as a status question, is listening cheating, but there is a version of it that research can actually answer: do you understand and remember a book as well when you listen to it as when you read it?

That has been tested many times, and pooled. The short answer is that across 46 studies and 4,687 people, the difference is not reliably distinguishable from zero.

The longer answer has two conditions attached, and they are the part worth reading.

The headline finding

Clinton-Lisell''s 2022 meta-analysis in Review of Educational Research pooled 46 studies, N = 4,687, using robust variance estimation, comparing comprehension after reading against comprehension after listening to the same material.

Overall effect: g = 0.07, p = .23.

That is a trivial difference in favour of reading, and it is not statistically reliable. On the central question, the modalities are close to interchangeable.

There is a second meta-analysis by the same researcher on a related question, whether reading while listening (following the text with narration) beats reading alone. That pooled 30 studies, N = 1,945, across 62 effect sizes and found g = 0.18, 95% CI [0.04, 0.32], p = .01. A statistically detectable but, in the author''s own description, trivial benefit.

The two conditions that qualify it

An overall null can hide structure, and here it does. Two moderators came out of the 2022 analysis.

Reading wins for inference, not for facts

What was measured	Effect (g)	p
Inferential comprehension	0.36	.02
General comprehension	0.15	.05
Literal comprehension	-0.01	.93

For retrieving stated facts, listening and reading are indistinguishable. For drawing conclusions the text implies but does not state, reading had a real advantage.

Reading wins when you control the pace

When readers set their own speed, reading was slightly better (g = 0.13, p = .049). When the experimenter controlled the pace, the effect reversed direction and became non-significant (g = -0.32, p = .16).

The same pattern appears in the reading-while-listening analysis, more sharply: experimenter-paced g = 0.41, self-paced g = 0.06, p = .34, no reliable difference at all once the reader is in charge.

Put those two moderators together and a mechanism suggests itself. What reading gives you is not a better channel into the brain. It is control, the ability to slow down, stop, and go back over a sentence that did not land. That control is worth most exactly where inference is required, because inference is the thing that makes you want to reread. Take the control away, as experimenter-paced conditions do, and the advantage disappears.

An audiobook is, by default, experimenter-paced. That is the whole difference.

A clean single study

Meta-analyses pool heterogeneous work, so it is worth having one well-run individual experiment in view.

Rogowsky, Calhoun and Tallal (2016) randomly assigned 91 adults aged 25 to 40 to three conditions: audiobook only (30), e-text only (31), or both simultaneously (30). Everyone got the same portion of Unbroken and then a 48-item comprehension test, repeated two weeks later.

Condition	Immediate (of 48)	Two weeks later
Audiobook	30.9 (SD 6.7)	28.2 (SD 7.8)
E-text	31.3 (SD 5.4)	29.1 (SD 5.9)
Both	31.5 (SD 5.7)	27.6 (SD 6.5)

Immediate: F(2, 87) = 0.76, p > .05. After two weeks: F(2, 85) = 0.18, p > .05.

No difference, immediately or after a delay. Note also that the dual-modality condition, often marketed as the best of both, was not better, and after two weeks was numerically the worst of the three.

What the brain imaging does and does not add

There is one neuroimaging result that gets cited heavily here, and it is worth stating precisely because it is usually overstated.

Deniz and colleagues (2019), in the Journal of Neuroscience, scanned nine participants while they listened to and read the same narratives, more than two hours of material per modality. They found that the maps of where meaning lives in the cortex were highly correlated across the two modalities: a model trained on listening predicted the reading data and vice versa. Their conclusion was that the representation of semantic information is largely independent of the sense through which it arrives.

That is a genuinely interesting result and it fits the behavioural picture. But it is nine people. It is an existence proof that the semantic representations overlap substantially, not a population estimate, and it does not tell you anything about how well you will remember chapter four.

What this does not settle

Whether audiobooks work as well for children learning to read. A systematic review by Singh and Alexander (2022) covering 32 documents found advantages for audiobooks among younger students and, for struggling readers and English language learners, for audiobook-plus-print together. That is a different question with a different answer and this article is not it.

Whether it holds for dense or technical material. The studied texts are mostly narrative and expository passages of moderate difficulty. Nobody has run this on a statistics textbook.

Whether attention holds up over a whole book. These studies use short passages under supervision. Listening while driving, walking or washing up is divided attention, and none of this research measures that.

Singh and Alexander flag the biggest limitation of the whole area themselves: there is a paucity of studies directly comparing audiobook listening to print reading. Much of what gets pooled is laboratory listening rather than audiobooks as people actually use them.

Children are a different question with a different answer

Almost everything above concerns competent adult readers, and it would be a mistake to carry it across to someone still learning to read.

Singh and Alexander''s systematic review (2022), covering 32 documents, found a pattern that does not appear in the adult literature at all. For younger students, audiobooks alone showed advantages in the range of g = .28 to .58. For struggling readers and English language learners, audiobook combined with print produced effects ranging from g = .32 to as high as 1.67, the upper end of that being a very large effect by educational standards.

The likely reason is that for an adult, decoding print is automatic and costs nothing, so removing it saves nothing. For a child still assembling letters into words, decoding consumes most of the available capacity, and hearing the words while seeing them frees enough of it to leave room for meaning.

Which means the adult finding, that modality is roughly neutral, and the children''s finding, that audio support helps substantially, are not in conflict. They are the same principle at two different levels of skill.

So do they count?

For comprehension, on the evidence available: yes, with one qualification you can act on.

If you are listening for the story, the argument, or the information, the things most books are for, the research finds no reliable penalty. If you are listening to material where you need to work out what is being implied rather than what is being said, reading has a real advantage of about a third of a standard deviation, and the reason appears to be that reading lets you go back.

Which points at something useful, because that advantage is not intrinsic to print. It comes from control over pace. An audiobook app has a fifteen-second rewind button and a speed control. Used deliberately on the passages that do not land, they recover most of what the format costs you.

The status question, whether it counts as real reading, is not one the evidence can adjudicate, and probably not one worth having. The comprehension question has an answer, and the answer is that your ears are doing better than the people asking assume.

Modality is one of several things that turn out to matter less than people expect, see the cognitive offloading guide for the wider picture, screen reading versus paper for the same question about a different surface, and subtitles and comprehension for what happens when two channels run at once.