UI you can’t ignore

-My feature isn’t getting enough traffic, let’s promote it on the home page.
-But the home page is crowded as it is.
-Nah, it’s okay, we can shuffle stuff around a bit and make some room in the top right.
-But we can’t just keep pushing stuff into every nook and cranny we find!
-Well, why not?
-Because it clutters the screen and gets in people’s way.-If they don’t want to read it, they don’t have to, they can just ignore it. Like I do with your emails!

If this sounds familiar, it means you’ve probably been involved with a digital product or two. At first glance, this “they don’t have to” approach does make some sense. After all, we’re all grownups, many of us have been in committed relationships, meaning that we’ve mastered the fine art of selective hearing — so why not selective viewing?

That’s probably the question that drove Barbara and Charles Eriksen, two experimental psychologists in the University of Illinnois at Urbana-Champaign, to dig a little deeper. In 1974, they brought students into their lab and sat them in front of a tachistoscope, which is basically a monitor that allows really fine control over the duration of the exposure, and looks like this:

I realize this may come as a shock, but no, it was not, in fact, designed by Apple.

Next to the device there was a small lever, and the instructions went as follows: “You are going to be presented with a string of 7 letters at a time. Focus on the middle letter only. In case it’s an H or a K, move the lever to the left. If it’s an S or a C, move it to the right”.

For the sake of randomness, half of the subjects were given opposite instructions in terms of left and right.

So the subjects were presented with a string of seven letters at a time, and had to pull the lever left or right depending on the letter in the middle — we’ll call it the “target letter”. There were three letters to each side of the target letter, and we’ll call them “noise letters”, although in later years they would come to be known as “flankers”. There were six different ways it could go:

  1. Noise same as target — when it’s the same letter seven times.
  2. Noise response compatible — when the noise letters require the same response as the target letter
  3. Noise response incompatible — when the noise letters require the opposite response of the target letter
  4. Noise heterogeneous, similar — when the noise letters aren’t part of the target letter set, but share visual characteristics with the current target letter (i.e. made up of straight lines in the case of H and K, or have rounded features in the case of C and S)
  5. Noise heterogeneous, dissimilar — when the noise letters have visual characteristics opposite to those of the target letter.
  6. Target alone — this version was used as the control group and appeared in two variations — either mixed with the other conditions, or as a separate sequence of 12 trials straight.

Just to be perfectly clear on the idea of the experiment. The Eriksen’s assumption is that if people can in fact ignore irrelevant signals, they will be able to ignore the “noise” letters, and their response to the target letter won’t be affected by them. In order to test whether noise letters do affect response time, they arranged for half of them to require the same response as the target letter, while the other half required the opposite response. So if people cannot filter out the distraction, their brains would be getting conflicting signals in the “incongruent” cases, which should have an adverse effect on their performance.

The experimenters also varied the spacing between the letters, which could be short, medium or long. The response times (in milliseconds) were recorded for all the correct responses, which produced the following chart:

The conditions 1–5 correspond to the list above, where 1 is “noise same as target” and 5 is “noise heterogeneous — dissimilar”.

A statistical analysis of the data produced the following findings:

  1. There was no significant difference between noise identical to the target and noise similar to the target. Meaning that it doesn’t matter whether a K is surrounded by H’s or by other K’s.
  2. The response time for the congruent conditions (1,2) was significantly shorter than the one for the incongruent condition (3). Meaning that it was much easier to reach the right decision for a K flanked by K’s or H’s than by K flanked by S’s or C’s.
  3. The larger the distance between the letters, the lower the response time. Meaning that the further apart the letters are, the easier it is to distinguish between them and reach the right decision.
  4. The mixed control condition (single-letter trials mixed between 7-letter trials) yielded longer response times than the blocked control condition (single-letter trials one after the other). Researchers postulated that this may be explained by an inhibitory tactic assumed by the subjects, where they tried to prevent themselves from reacting instinctively in the mixed conditions and slowing themselves down — a mechanism unnecessary in the blocked condition.
  5. When the noise letters weren’t mapped to a congruent response but did look similar to letters which were, response time were shorter than in cases where the noise letters looked similar to the opposite target letters. Which basically supports and generalizes finding number 2.

Or, in other words:

  • Even when you don’t have to search for the target, the number of displayed items slows down your processing of the target.
  • The more similar the noise to the target, the harder it becomes to distinguish between the two (shocking!).
  • Distance between items affects performance, because it’s easier to distinguish between items which are farther apart.

And the bottom line is that we cannot really ignore stuff that we see, even if we know that it interferes with the task at hand.

This of course has immediate practical implications on everyday UX work — which are all common best practice, but it’s nice to be reminded of the experimental foundations of the field:

  • The more items you place on your screen, the more difficult it becomes to process each one — even when searching for it is not an issue. People can’t decide to “ignore” items in their field of vision. This is directly related to Hick’s Law, although they’re not the same.
  • To alleviate this problem, the irrelevant items should not look similar to relevant items. This is what the Gestalt Laws are all about.
  • Let your UI breathe. The less crammed it is, the easier it is to distinguish between all your UI elements — even the similar ones, let alone those that are different.

The original article was published in Perception & Psychophysics, 1974, 16(1), pp143–149, under the title “Effects of noise letters upon the identification of a target letter in a nonsearch task”, and you can download it free of charge here.