Attention Restoration Theory: Thirteen Attention Measures Were Tested and Three Improved

Step away from the screen and go outside. It is the most common recommendation in digital wellbeing, it is almost always attributed to a specific psychological theory, and the theory has a systematic review that almost nobody quoting the advice has read.

The theory

Attention Restoration Theory comes from Rachel and Stephen Kaplan, set out most fully in Kaplan (1995), The restorative benefits of nature: Toward an integrative framework, in the Journal of Environmental Psychology - a paper now cited more than twelve thousand times.

The mechanism is specific and, as an idea, elegant. Deliberately concentrating on something uses directed attention, which fatigues. Natural environments contain what Kaplan called soft fascination - stimuli that hold you gently without demanding effort, such as moving water or leaves. Soft fascination occupies attention involuntarily, which lets the directed-attention system rest and recover. Urban environments, on this account, are not restorative because they demand vigilance: you have to watch for the car.

The most-cited empirical test is Berman, Jonides and Kaplan (2008) in Psychological Science 19(12), which compared walking in an arboretum with walking in a city, and viewing photographs of nature with viewing photographs of urban scenes, using a backwards digit-span task and the Attention Network Task. Their abstract states that both interventions “can improve directed-attention abilities… thus validating attention restoration theory.”

We are not printing the sample sizes for those two experiments. The publisher’s page gives the abstract but not the method, and no full text we could reach carried them. That matters more than it usually would, for a reason the next section makes clear.

The systematic review

Ohly, White, Wheeler, Bethel, Ukoumunne, Nikolaou and Garside (2016), in the Journal of Toxicology and Environmental Health, Part B, collected everything: 31 separate studies from 24 articles, using 13 distinct attention measures.

Here is what they found, and it is the number that should attach itself to this topic.

Measures showing a significant restorative effect	Measures not showing a marked beneficial effect
Digit Span Forward	Proofreading
Digit Span Backward	Necker Cube Pattern Control
Trail Making Test B	Search and Memory Task
	Sustained Attention to Response Test
	Symbol Digit Modalities Test
	and five others
3	10

Three of thirteen. And the two that carry the strongest signal - Digit Span Forward and Backward - are the family of tasks the famous 2008 study used.

On quality, the reviewers are blunt. Quality scores ranged from 22.5% to 75%, and only 7 of the 31 studies were rated high quality. Meta-analysis, they write, “was limited by small numbers of investigations, small samples.”

Their conclusion is one sentence and it is not the sentence the advice implies:

“There is uncertainty regarding which aspects of attention may be affected by exposure to natural environments.”

The theoretical objection

Effect sizes are one problem. Whether the theory explains anything is another.

Joye and Dewitte (2018), in the Journal of Environmental Psychology 59, 1-8, set out three criticisms, and they are the kind that a framework has to answer.

First, the central constructs are vague. Soft fascination, they argue, has “remained underdeveloped” and lacks “a clear operationalization” - there is no agreed way to measure how softly fascinating a scene is, which makes the core prediction hard to test.

Second, and most damaging, the framework “has failed to (adequately) test its main theoretical predictions (i.e., that nature effects are recovery effects).” Demonstrating that people do better on a task after being in a park is not the same as demonstrating that a depleted attentional resource was replenished. Almost no study establishes depletion first and recovery second; most simply compare two environments.

Third, the evolutionary story - that restoration derives from an ancient adaptive response to natural settings - has “little support”.

What is actually left standing

More than the sceptical reading suggests, and much less than the advice does.

Something real is being measured on digit-span tasks. Two of the three positive results in the systematic review are the two forms of that task, and they came up repeatedly. Digit span is a working-memory measure, so the honest summary is narrow: there is some evidence that time in natural settings improves short-term working-memory performance, and little evidence that it improves sustained attention, visual search or error detection.

Which is a slightly awkward result for the way this advice is used. The thing people want back after three hours of screens is sustained attention - the ability to stay with one thing. That is the capacity the review found least support for.

None of this is an argument against going outside. It is an argument against the specific claim that going outside restores your capacity to concentrate, and against the confident numbers that circulate with it - the twenty seconds of nature, the precise recovery windows. We looked for a study behind those and did not find one.

Two boundaries worth stating, because they are separate literatures with separate answers. How attention decays within a continuous task, and what actually reduces that decay, is measured directly and has a much firmer evidence base. And whether the resource-depletion picture that ART assumes holds up at all is the subject of the previous article in this series - where two multi-lab preregistered replications failed to detect the depletion the theory takes for granted.

That last point deserves emphasis. Attention Restoration Theory is a theory about recovering a depleted resource. Joye and Dewitte note the recovery half has not been properly tested. The depletion half, tested elsewhere at very large scale, did not replicate.

Limitations and disputes

• The founding experiment’s sample sizes are not reported here. We could reach the abstract but not the method, and the review’s own assessment of the literature is that samples were small.
• The systematic review is nine years old. Studies published since 2016 are not in its count, and some report positive results.
• Three of thirteen measures is not zero. The digit-span findings recurred, and the reviewers do not dismiss them; they say the picture is uncertain, which is different from saying it is empty.
• Quality was low across the field: only 7 of 31 studies were rated high quality, on the reviewers’ own instrument.
• Joye and Dewitte’s critique is theoretical, not an empirical refutation. They argue the predictions have not been properly tested, not that they have been tested and failed.
• We have not assessed nature and mood, stress or physical health, which are separate literatures with different and generally stronger evidence.
• No specific figure for how long any restorative effect lasts appears here, because we could not verify one.