Does Hospital Art Speed Recovery? One 1984 Study, Read Line by Line

The claim that art in hospitals speeds recovery traces to a single paper published in Science on 27 April 1984, and that paper contains no artwork of any kind: it compared surgical patients whose window looked onto a stand of trees with patients whose window looked onto a brick wall. Reading it line by line is the fastest education available in how a small, careful, interesting study becomes a forty year old load bearing citation.

The stroke unit where I run a weekly session has a painting of boats near the lift. What I noticed, over a couple of years, is that nobody talks about it as art. They use it as a landmark. Staff say turn left at the boats. Patients working out where the day room is orient by it, and one man who had lost a good deal of language used it to explain to his daughter where he wanted to sit. That is a real function, it is worth the wall space, and it needs no claim about length of stay whatsoever. I mention it because the honest case for hospital art is generally available and generally ignored in favour of a borrowed one.

Where a claim about effect appears below it carries a tier (Tier 1 supported, Tier 2 promising but limited, Tier 3 practice and anecdote), explained on reading arts in health research.

The study, in full

One paper, two pages, one hospital, and a set of restrictions that are easy to lose.

Roger Ulrich’s View Through a Window May Influence Recovery from Surgery appeared in Science, volume 224, number 4647, on 27 April 1984, at pages 420 to 421¹. The design:

• 46 patients, in 23 matched pairs (15 female, 8 male).
• Retrospective, from hospital records covering 1972 to 1981.
• One 200 bed suburban Pennsylvania hospital.
• One procedure only: cholecystectomy.
• Operations performed between 1 May and 20 October only, so that the trees were in leaf.
• Patients under 20 and over 69 excluded.

Every one of those constraints is reasonable for the study Ulrich was doing. Together they describe something very specific: gall bladder surgery, in one American hospital, in the summer months, in the 1970s, in adults of working age. That is the population the finding belongs to.

What it found

Five results, and only some of them travel.

Length of stay: 7.96 days for tree view patients against 8.70 days for wall view patients (Wilcoxon matched pairs signed ranks, T(17) = 35, z = 1.965, p = 0.025)¹. That is the headline. It is a difference of about three quarters of a day, in 23 pairs.

Nursing notes: 3.96 negative comments per patient for wall view against 1.13 for tree view (Wilcoxon, T(21) = 15, z = 3.49, p < 0.001). This is statistically the strongest result in the paper and is quoted far less than the length of stay. Positive comments favoured the tree view group and were not statistically significant.

Antianxiety drugs: no significant variation between the groups.

Minor complications: lower in the tree view group, and not statistically significant.

Analgesics: the table below, which needs its own section because of how it gets handled.

Two of those five are null or non significant, and they are the two that almost never appear in a summary. A study reporting no difference in antianxiety medication is relevant to any claim that a pleasant view reduces anxiety, and it is routinely omitted.

The analgesic table, and the two things to get right about it

Analgesic strength	Days 0 to 1, wall	Days 0 to 1, tree	Days 2 to 5, wall	Days 2 to 5, tree	Days 6 to 7, wall	Days 6 to 7, tree
Strong	2.56	2.40	2.48	0.96	0.22	0.17
Moderate	4.00	5.00	3.65	1.74	0.35	0.17
Weak	0.23	0.30	2.57	5.39	0.96	1.09

First: these are mean doses per patient across each whole period, not per day. The column heading in the paper is number of doses, and the periods are days 0 to 1, days 2 to 5 and days 6 to 7. Describing 2.48 as a daily figure for the wall view group in days 2 to 5 misstates it by a factor of four. This error is common enough that it is worth checking whenever you see the table reproduced.

Second: the weak analgesic row in the days 2 to 5 bucket runs the other way. 2.57 for wall view against 5.39 for tree view. The strong and moderate rows in that bucket favour the tree view group; the weak row favours the wall view group by more than two to one. The coherent reading is that tree view patients were stepping down to weaker drugs sooner, which is a real and interesting pattern, and it is not the same claim as taking less pain relief overall. It is almost never printed alongside the two rows above it.

The significant difference is in the days 2 to 5 bucket only (multivariate Hotelling, T squared = 13.52, F = 4.30, p < 0.01). The other two periods showed no significant difference. And only 45% of patients took any analgesic at all after the fifth day, which is worth knowing before making anything of the days 6 to 7 column¹.

Why a retrospective matched pair design limits what can be concluded

Because the allocation was not random, and everything that decided which room somebody got is a candidate explanation.

Patients were matched on characteristics the researcher could see in the records, and pairs were compared. That controls for what was matched and nothing else. Room allocation in a hospital is not random: it depends on availability at the moment of admission, on who was admitted at the same time, on ward pressure, on time of day, on whether an isolation room was needed. Any factor that influenced which room a patient got and also influenced how quickly they went home would produce exactly this result without any causal contribution from the view.

Retrospective record review adds its own problems. The nursing notes result, which is the statistically strongest finding in the paper, depends on how nurses wrote about patients in a Pennsylvania hospital across nine years, categorised afterwards by somebody who could see which group each patient was in. That is not a criticism of the coding; it is a description of a design in which blinding was impossible.

None of this makes the study bad. It makes it a hypothesis generating study, which is what a two page paper in 1984 with 23 pairs should have been treated as.

How to grade it

Tier 3, and specifically the kind where nobody has looked properly at scale rather than the kind where a good trial reported nothing.

One retrospective matched pair study, 23 pairs, one hospital, one procedure, one season, published in 1984, never replicated at that scale, cannot support a general claim about recovery. It is not evidence that a view of trees does nothing; the result is genuinely interesting and consistent with plausible mechanisms. It is simply too small, too specific and too old to carry a claim about hospital design in general.

And it is a study about a window, not about art. There is no artwork in it. A finding about a view of trees has been used to justify paintings in corridors for four decades, which is a category drift on top of everything else, and the drift is rarely noticed because the paper is cited rather than read.

What would be needed instead

Stated so it is clear this page is not simply refusing to be satisfied.

• Prospective allocation, so that room assignment does not encode ward logistics.
• A larger sample, across several hospitals and more than one procedure.
• Outcome assessment blind to the exposure, at least for anything recorded by staff.
• Modern length of stay, since 1970s stays after gall bladder surgery bear no relation to current practice, which makes the outcome measure itself hard to transfer.
• The intervention actually under discussion, meaning artwork, if the claim is about artwork.

That last point is the one that would matter most and has been outstanding for forty years.

What hospital arts can defensibly claim instead

Several things, all of which are easier to defend than a borrowed recovery finding.

Orientation and wayfinding. Distinctive images make a building navigable in a way that signage does not, particularly for people who are unwell, frightened or losing their words. The boats by the lift do more work than any of the arrows near them.

Making waiting bearable. Hospital time is mostly waiting, for patients and for the people with them. A corridor, a waiting area or a day room that somebody chooses to sit in is an outcome in itself.

Reducing fear in specific places. Children’s outpatient departments, imaging suites, entrances. These are design problems with observable answers.

Staff and visitors. Both spend a great deal of time in these buildings and are almost never mentioned in the claims made for the artwork. Arts for carers and health staff covers where that evidence has and has not got to.

None of that requires a study to justify, in the way that a shorter stay would. The pressure to make the larger claim is easy to understand, since charitable funding competes against clinical need and it makes the place more bearable is a harder case to fund than it shortens length of stay. The sector’s own reports and case studies are gathered in one place² and are worth reading as descriptions of practice rather than as evidence of effect, a distinction worked through on creative health policy and reports.

Where hospital arts do reach Tier 1

Two claims in this setting are properly supported, and both involve music rather than visual art.

Music before surgery. Anxiety 5.72 STAI-S units lower (95% CI -7.27 to -4.17), from 26 trials and 2,051 participants³. Tier 1, with an explicit caveat: the review was published in 2013, before certainty assessment became routine, so it carries no certainty rating and none can be invented for it. See music before surgery and anxiety.

Rhythmic auditory stimulation after stroke. Gait velocity 11.34 m/min faster (95% CI 8.40 to 14.28), from 9 trials and 268 participants, at moderate quality⁴. Tier 1, quoted in the units the review used, and about walking speed only. See rhythmic auditory stimulation.

For contrast, a large but weak evidence base looks like this: the 2021 Cochrane review of music interventions in cancer care pooled 81 trials and 5,576 participants and rated its findings very low certainty⁵, which is Tier 2. Study count is the least informative number in an abstract.

And for the size of the mapped literature, which is often used to imply strength: the WHO Europe scoping review covers over 900 publications, comprising 200 plus reviews and 700 plus individual studies, with those reviews between them covering over 3,000 studies⁶. Those counts nest rather than add, and a scoping review maps a literature rather than pooling it, so it carries no tier.

The general lesson

A claim’s age and its frequency are not evidence. They are often the opposite.

A paper that has been cited for forty years without replication is telling you something about how convenient the finding was, not about how robust it is. The check that catches this takes a minute: ask what the sample size was, what the design was, when it was published, and whether anybody has repeated it. In this case the answers are 46 patients, retrospective matched pairs, 1984, and not at that scale.

The wider picture of what hospital arts programmes do, who runs them and what they are for is on arts in hospitals. The practical ethics of doing this work where the audience cannot leave is on live music on hospital wards. Nothing on this page is a reason to change, delay or decline any part of anybody’s treatment, and no arts programme substitutes for one.