Is What Is Beautiful Also Usable? Thirty Years of Research on the Halo Effect

In 1995, two researchers at the Hitachi Design Center, Masaaki Kurosu and Kaori Kashimura, had 252 people rate 26 screens of an automated teller machine. Every screen contained the same buttons. Only their arrangement changed. Participants were asked whether the screen looked easy to use and then whether it looked beautiful. The correlation between the two answers reached 0.59, a link stronger than almost every ergonomic criterion (opens in a new tab) the authors had built into the screens. The people surveyed believed they were judging ease of use. Their answer mostly followed appearance.

That experiment opened thirty years of work on what researchers call the aesthetic-usability effect. For a website, the question is concrete. Does a visitor who finds a page beautiful believe it is easier to use? Does that belief survive actual use? The studies that followed measured the effect in both directions, saw it reverse under certain conditions and put numbers on what it changes in how fast people complete tasks.

Noam Tractinsky repeated the experiment in 1997 with the same screens, adapted to a different audience from Hitachi's. The correlations turned out to be very high (opens in a new tab), to the point of exceeding those of the Japanese study. The replication also ruled out a methodological artefact, the idea that two questions asked one after the other contaminate each other.

In 2000, with Adi Shoval-Katz and Dror Ikar, Tractinsky published the study that gave the effect its name, "What is beautiful is usable". One hundred and thirty-two industrial engineering students used a cash machine simulated on a computer. They worked on nine screen layouts taken from the 26 Hitachi designs, three rated beautiful, three average and three ugly according to the previous study. Actual usability varied as well. In the degraded version, the system added an average of 9 seconds of delay per task (opens in a new tab)00031-X) and some buttons did not respond to the first press. The manipulation worked. A task took 23.5 seconds in the smooth version and 37.5 seconds in the degraded one.

Before use, beauty ratings and perceived ease were correlated at 0.66. After eleven tasks, the correlation rose to 0.71. The statistical analysis produced the result that made the paper's reputation. The only factor that changed perceived ease after use was the aesthetics of the screen. The actual level of usability, responsible for a 14-second gap per task, had no measurable effect on what participants said about how easy the system was. The authors write that the most surprising finding is that "post-experimental perceptions of system usability were affected by the interface's aesthetics and not by the actual usability of the system".

The speed of that first judgment was measured in 2006 by Gitte Lindgaard and her team at the Human-Oriented Technology Lab of Carleton University in Ottawa. In a first study, 22 students rated the visual appeal of homepages shown for 500 milliseconds each, in two rounds. The mean ratings of a page from one round to the next were correlated at 0.97 (opens in a new tab). A third study showed the same pages for 50 milliseconds to one group and 500 milliseconds to another, twenty people each. The mean ratings of the two groups were correlated at 0.95. A twentieth of a second is enough for the aesthetic judgment to be formed and to be stable.

Does that early judgment say anything about actual ease? In 2015, Meinald Thielsch, Ronja Engel and Gerrit Hirschfeld compared expected usability, rated on a first impression, with experienced usability after use and with task completion rates. The expected ratings were not a valid indicator of either (opens in a new tab). Ratings given very early, like those given after ten seconds of exposure, had no significant link with what the same people went on to experience.

Marc Hassenzahl and Andrew Monk proposed an explanation of the mechanism in 2010. Reviewing 15 papers reporting 25 correlations (opens in a new tab) between perceived beauty and perceived usability, they found coefficients that varied widely from one study to the next. Their own analysis shows that the link runs entirely through an overall evaluation, the general "this is good" people attach to a product before they know it. Beauty feeds the general impression that the product is good. That impression in turn feeds perceived ease. As long as experience of use is missing, the judgment rests on what is visible. What the eye evaluates in that first instant, visual complexity and how closely the page matches the expected model, is detailed in the article on whether an immersive website is the right choice (opens in a new tab).

The word covers very different manipulations from one study to the next. In 1995, the 26 Hitachi screens contained exactly the same elements and differed only in how they were arranged. In 2012, the online shop built by Tuch and his colleagues went from beautiful to ugly through the colour of its decorative elements alone. In between, Lindgaard's second study asked 31 students to rate 50 homepages on seven pairs of adjectives. Visual appeal was linked to five of them (opens in a new tab), interesting or boring, well or poorly designed, good or bad colour, good or bad layout, imaginative or unimaginative. It was only weakly linked to the two remaining pairs, simple or complex and clear or confusing. A page judged beautiful is therefore not first of all a simple page. It is a page judged interesting and imaginative.

Morten Moshagen and Meinald Thielsch structured these dimensions in 2010 into a questionnaire that became the reference for later studies, the Visual Aesthetics of Websites Inventory. It measures four facets (opens in a new tab), simplicity, diversity, colourfulness and craftsmanship, the last term covering the perceived quality and modernity of the execution. Simplicity and diversity coexist in it. An interface can be legible and surprising at the same time. Each of the two facets contributes to the overall judgment.

An experiment run in 2006 by Tamar Ben-Bassat, Joachim Meyer and Noam Tractinsky set out to learn whether people pay for what they say they prefer. One hundred and fifty engineering students (opens in a new tab) used four versions of a data-entry application, crossing two levels of usability with two levels of aesthetics. In the questionnaires, the halo effect showed up as expected. The more beautiful systems were judged slightly easier, 4.90 against 4.66 on a 7-point scale, even though participants entered slightly more data with the less beautiful versions.

Some of the participants then had to bid, in real money and up to the equivalent of about 11 dollars at the time, for the system they would use in a test paid on performance. The bids followed a single criterion, actual usability. The amounts matched the performance levels each system made possible. Aesthetics no longer carried any weight.

The result separates two situations. When people are asked for an opinion, appearance contaminates the judgment of ease. When the choice costs something and performance matters, the judgment realigns with what the system lets people do. An accepted quote, a form filled in to the end or a renewed subscription are closer to the second situation.

Alexandre Tuch, Sandra Roth, Kasper Hornbæk, Klaus Opwis and Javier Bargas-Avila, of the universities of Basel and Copenhagen, took up the question in 2012 with an online shop rather than a cash machine. Four versions of the same site (opens in a new tab) were built. Usability lived in the architecture of the categories, clear and well named in one version, ambiguous in the other. Aesthetics lived in the colour of decorative elements of a professional template, without touching the legibility of the text. Eighty participants rated the site before and after a series of tasks.

Before use, the beautiful version was not judged easier. After use, perceived ease depended only on actual usability. The link did exist. It ran the other way. Participants who had struggled with the navigation then found the site less beautiful. That effect ran through their affective state, the frustration built up during the tasks. The authors sum up the situation by reversing the 2000 formula. Under these conditions, what is usable becomes beautiful.

Kai-Christoph Hamborg, Julia Hülsmann and Kai Kaspar found the same pattern in 2014, this time on a mobile phone prototype tested by 88 participants (opens in a new tab). Aesthetics changed perceived ease neither directly nor indirectly. Usability changed perceived beauty.

Why does the 2000 team find one direction and the 2012 team the other? The way usability was degraded differs. In 2000, the delays and the unresponsive buttons were faults the participant endured without being able to do anything about them. In 2012, the ambiguous categories forced the participant to search and to make mistakes. Tuch and his colleagues note that their result appears with a strong usability manipulation and a medium aesthetics manipulation. They add that a less goal-oriented task or no task at all could produce different evaluations.

The three experiments that crossed aesthetics and usability in the laboratory

The question that matters most for a website that has to convert remains. Does aesthetics change what people do and not only what they say? Don Norman argued in 2004, in Emotional Design, that an attractive object puts people in a positive affective state that broadens thinking and makes them more tolerant of small difficulties. The experiments that followed tested that hypothesis in practice. In 2009, Andreas Sonderegger and Jürgen Sauer, of the University of Fribourg, had 60 adolescents aged 14 to 17 (opens in a new tab) operate two simulated phones identical in everything but appearance. The attractive phone was judged easier. Tasks were also completed faster on it.

The same year, Morten Moshagen, Jochen Musch and Anja Göritz had 257 volunteers (opens in a new tab) search for health information on a website, in four versions crossing aesthetics and usability. Usability weighed on time and on errors. Aesthetics improved performance only in the poorly designed versions, where it partly compensated for the ergonomic faults. On the well-designed site, it gained nothing. The 2006 auction experiment even ran the other way, with slightly faster data entry on the less beautiful versions.

A preregistered meta-analysis by Marleen Schlamann, Steffen Nestler and Meinald Thielsch, published in June 2026 in the International Journal of Human-Computer Interaction (opens in a new tab), added up these contradictory results. It retains 31 studies, 234 effect sizes and 18,794 participants (opens in a new tab), accepting only studies in which users had actually perceived the aesthetic manipulation as such. The overall effect of aesthetics on performance is positive, small to medium in size, with high heterogeneity between studies. Two factors make the effect vary, the type of device used and whether the manipulation concerns typography, whose role the article on typography and conversion (opens in a new tab) already describes. The authors find no evidence that attractive design impairs performance (opens in a new tab), contrary to the idea that decoration distracts.

Most of these experiments last an hour. Andreas Sonderegger and his colleagues followed 60 phone users for two weeks (opens in a new tab), over three sessions, with devices crossing aesthetics and usability. The perceived-ease advantage the attractive device enjoyed in the first test shrank with every session. The emotions people reported followed the same slope. The authors draw a recommendation for usability tests from it, to run them over several sessions so as to neutralise this bias.

Marc Hassenzahl had observed in 2004, on skins for a music player application, that the overall judgment of quality depended more and more on perceived usability (opens in a new tab) as people used the product. The judgment of beauty, for its part, stayed attached to what the product said about its owner.

A business website is not visited once. A client comes back for a quote, a personal account or an order. The advantage granted to the first impression shrinks with each visit. What remains comes down to what the site made possible.

The two effects coexist and do not act at the same moment. Before use, aesthetics almost alone determines expected ease, in less than a tenth of a second. That expectation says nothing about the experience to come. During use, actual ease writes the lasting judgment, to the point of changing, after the fact, the perceived beauty of a site that made people struggle. When a decision involves money, appearance stops weighing on the choice.

For a website, this sets the order of requirements. A singular interface wins the first judgment and the attention that comes with it. It keeps that judgment only if the navigation, the forms and the response times bear out the visual promise. The 2026 meta-analysis adds that careful design does not harm performance and can improve it. What the first impression promises is then checked against every task completed.