This evidence sheet is published by The Built Review.
In 1986 the architecture researcher Judith Heerwagen and the ecologist Gordon Orians walked through 75 offices on the University of Washington campus and counted what hung on the walls. The windowless offices carried 195 pieces of visual material, the windowed ones 82. The content differed as well. The windowless walls carried more landscapes and fewer cityscapes than the windowed ones.
The substitutes have since become electronic. In 2025 the built-environment researcher Shuhan Liu compared a real window, an artificial window and no window at all, with heart rate variability recorded alongside cognitive tests. Against the windowless condition the artificial window raised the nLF/nHF focus ratio by about 22.2% and cut attention test times by 14.4%. On lighting comfort and acceptability it beat the real window.
The computer scientist Fariba Mostajeran built the same question in virtual reality in 2023. Thirty-nine people worked in a virtual office with plants and in one without, and scored higher on short-term memory and creativity with the plants there. Asked afterwards which office they preferred, 32 of the 39 chose the planted one and none chose the empty one. Mostajeran also passes on two of the most-quoted figures in this literature, Lohr's 12% and Nieuwenhuis's 15%. Neither original is open to us.
The most isolated test of the idea ran at the South Pole. The space medicine researcher Stijn Thoolen put 25 overwintering crew members through 16 variations of sensory stimulation at the end of their mission in 2025, varying delivery, content, duration and temperature cues. State anxiety fell by 1.7 points after nature scenes and rose by 1.0 after city scenes on a scale running from 20 to 80, with t = 4.805 and p below 0.001. Preferences beyond that split by person rather than by rule.
Michael Varkovetski's 12%, the difference between 54.33 seconds and 62.04 on an attention task, was published in an undergraduate research journal. And Nathalie Gerner's feasibility study in 2025 found no significant interaction at all between cognitive state and nature-inspired design across five established EEG ratios, working with 20 people and consumer-grade equipment whose limits Gerner names.
