ResearchBrisbaneOver 85 works catalogued

Research

What we know, and how we know it.

We keep a research corpus because research has a practical place in the work.

The corpus covers a range of subjects relevant to how people find, read, understand, and trust websites. Some works will be directly relevant to a particular project. Others will not.

This page is the short version. The full record shows what each work found, how much confidence we place in it, and where its findings have limits.

Works held

Over 85

The corpus currently contains over 85 works across a range of subjects relevant to how people find, read, and trust websites. Each is rated for how much confidence we place in it, and each entry says why.

Span

1955–2026

The corpus spans more than seventy years of research. Most of the works are peer-reviewed journal articles or conference papers.

Subject areas

The research ranges across information systems, human-computer interaction, cognitive psychology, information retrieval, marketing, retail, and vision science.

Bibliography

None

A bibliography is not evidence by itself.

What is in it

1955 to 2026

The corpus looks at how people decide whether to trust a website and how they choose between services.

Part of that is first impressions and visual quality, including how typical a site appears for its category and how that affects trust.

Part of it is navigation and information scent. Reading, attention, and typography are in there too, and so are response time and waiting.

There is research on forms and errors, accessibility, and interface evaluation. There is also work on cognitive load, mental models, and affordances.

Some of the research concerns small businesses and the way people arrive at websites through search.

The subjects overlap because real websites do not divide themselves neatly into academic disciplines.

What each work contributes

A citation to a paper on its own tells you where a finding came from. It does not tell you how relevant that finding is to the work at hand, or how much confidence we should place in it.

Each work in our corpus has a summary explaining what it found, how relevant it is to our work, and how much confidence we place in the evidence.

Not every work is given the same weight. Some are rated with less confidence than others, for different reasons. Preprints, undated manuscripts, and anything else that has not been peer-reviewed are marked as such. We do not treat any of these as equivalent to stronger evidence we hold.

A few findings

Six from the corpus

50 milliseconds
In one study, people were shown a webpage for just fifty milliseconds before being asked to rate its visual appearance. Their ratings were remarkably similar to those given when they were allowed to look for longer. The useful point is that people can form an impression of a site before they have had time to read it. Fifty milliseconds itself is not a magic threshold.
Lindgaard et al., 2006
89 per cent
Forty people were observed choosing health-advice websites over several months. At the initial screening stage, 89 per cent of their stated reasons for rejecting a site concerned design, while 11 per cent concerned content. It was a small qualitative study in one high-stakes subject, so the direction is more useful than the exact percentage.
Sillence et al., 2006
Over 46 per cent
In a study of 2,684 people evaluating website credibility, visual design appeared in over 46 per cent of the comments people made about what made a site credible. It was the largest single category. The study tells us that visual presentation was a substantial part of how people described credibility. It does not tell us whether those impressions became enquiries.
Fogg et al., 2003
Over 12 seconds
In one study based on real browsing rather than a laboratory task, the median time spent on a first visit was just over 12 seconds. That suggests a page has very little time to make its purpose clear. It is not a rule about how long every visitor will stay.
Weinreich et al., 2008
Size beats position
In desktop eye-tracking with twenty people, the size of an element was a stronger predictor of whether it was looked at than most of the positional measures tested. That tells us something about attention. It does not tell us what someone will do afterwards. Research can tell us what people looked at without telling us what persuaded them.
Buscher et al., 2009
Typicality matters
Research into university and banking websites found that how typical a site looked for its category affected whether people found it trustworthy. The effect was stronger for university sites than for banks. The finding is useful because it puts a limit on one of the instincts of design: being distinctive is not automatically better. A site still has to look like it belongs to the category people think they are dealing with.
Miniukovich & Figl, 2023

What the research does not say

Rules that do not hold

Some web-design rules are repeated so often that they begin to sound like facts. When we look for the evidence behind them and cannot find it, we say so.

“Everything important must be above the fold.” We have not found evidence for this as a rule.

In one study of real browsing, over 23 per cent of link clicks were on links that required a scroll first (Weinreich et al., 2008).

The first screen still matters: the site’s identity and purpose need to be clear there. But important details can sit below it.

“Fewer form fields means more enquiries.” One commonly quoted figure comes from a 2010 HubSpot post on the company’s own customer data, and its author’s reading of that data was that conversion rates fell “slightly, but not as steeply as I expected”. The rule was made afterwards, by other people. We keep forms short because shorter forms are generally simpler to complete. That is a design judgement, not a research finding.

“The button colour matters.” Size has been measured in the research we hold. Button colour has not.

So we do not treat a particular colour as a research-backed rule.

“People only read a fraction of a page.” The familiar percentages can be traced to a 1997 industry report rather than to a refereed study that establishes them as a general finding.

What the research does show is that people move on quickly: in one study the median first visit lasted just over 12 seconds. The precise percentages are not something we use as evidence.

Our own corpus has limits too.

Many of the studies use small samples. A number involve students or technical professionals rather than the people we are actually designing for.
Published research on Australian audiences is thin. The corpus holds five works with Australian samples. All five are small or descriptive studies, and none measures whether a visitor goes on to make contact. We add local work as it appears.
One study comes particularly close to our actual use case. It looked at people choosing a provider for a household problem. There were thirty participants, and they thought aloud while using a prototype. But they were not choosing a real firm’s service. That distinction matters.
Where our design decisions go beyond what the evidence supports, we label them as studio judgement. Research is useful partly because it tells you what you can reasonably claim.

What we do with it

In your project

Research helps us decide where to spend effort, and when to stop arguing from instinct and intuition.

It can influence what belongs on the first screen, how an enquiry route works, and how information is organised. It can also tell us when a familiar convention has less evidence behind it than its reputation suggests.

You will see the relevant evidence during your project. The research is written up before design begins, and it remains yours whether or not we build the site.