Banner blindness has a varied history. Common understanding is that banner ads are not seen. The first research into banner blindness determined that people had learned to avoid seeing banner ads. This was later confirmed by eye-tracking studies, which show that as people become more experienced with the web, their pattern of eye movements indicates they learn to avoid looking at banner ads. This phenomena has become known as "banner blindness." However, the fact that people do not focus on banner ads does not necessarily mean the ad is not affecting them. Just as there is research to show that unnoticed exposure to ads in the real world still affects brand recognition and response, research is emerging to show the same phenomena occurs with regard to banner ads.
While we may learn not to focus our eyes directly on an ad, it is still visible within our peripheral vision, and it has an impact from there. In November 2009, a researcher in France, Franck Largeault, took this research one step further, examining how much impact unnoticed banner ads have, and whether different banner designs have different effects. The results are stunningly clear-cut and have strong implications for ad design.
Largeault and his research
Franck Largeault's research formed his dissertation for a Master of Science at Reims Management School in France. His website provides video presentations in English and French of his research. If you're interested in this sort of thing, I recommend you watch the full 15-minute presentation. Largeault's experimental design is simple yet elegant, and his conclusions are easy to follow and unmistakably clear.
Early banner blindness studies were, in my opinion, suspect because they used ridiculously small sample sets, often no more than a dozen subjects. It is difficult to argue that an extrapolation from 12 people to the entire population is statistically valid. However, Largeault used a sample of more than 300 subjects, so his results should be statistically valid. In fact, it may be one of the largest studies into banner blindness ever.
Signal detection theory and banners
The best explanations for banner blindness come from Signal Detection Theory, which explains how people can distinguish noise from signal in a jumbled environment. The classic example is of standing at a noisy party. You are able to tune out all the other conversations so you only hear the person speaking to you. Yet, if someone mentions your name in one of those tuned-out conversations, you'll notice it instantly. Signal Detection Theory goes into the processes by which we can tune sensory input out of our consciousness, yet still detect relevant data in the "noise" we have tuned out.
The fact that we can suddenly notice important information within a stream of sensory data that we are not conscious of indicates our brains must be processing that data on some level. Experiments into sensory thresholds have shown there is no limit to how fine-tuned this processing is. Neither is there a limit to how little sensory input we can detect within the stream. Under the correct circumstances, we can detect a signal that affects only a single receptor in the eye (or ear) from amongst a mass of sensory stimulation.
Whether we detect something within the sensory stream or not is largely dependent upon psychological factors. These involve a balancing act between the penalty for missing relevant signals, the reward for correctly detecting them, and the penalty for mistakenly identifying noise as meaningful signal.
When surfing the web, the penalty for treating irrelevant content as important is fairly high. It breaks concentration, disrupts reading, makes page layout confusing, and generally makes surfing harder and slower. People on the web are in an active, goal-oriented behavior mode, unlike TV, where people are in a passive consumption mode. As such, web users typically find themselves searching for relevant data among pages littered with irrelevant content. For example, many newspaper websites allocate more screen space to advertising than to copy (just as they do in print). Thus, effective web surfing means learning to tune out irrelevant screen space to focus on the zones of interest. We see this pattern in eye-tracking studies. New users to the web tend to move their eyes over the whole screen. As their experience develops, the range of their eye movements reduces, focusing more on the center and left-hand side of the web page.
Signal Detection Theory explains these phenomena as the result of the high penalty and low reward for focusing on irrelevant content. User experience is that banners are so unlikely to be relevant that there's no point looking at them. Signal Detection Theory shows that this filtering process is very deep and very strong.
Jakub Nielsen ran a great study some years ago that demonstrated just how strong banner blindness is. He asked participants to fill out an online form. The question for each input box was placed in a 120x40 pixel banner, right next to the input box. Most people could not work out what the questions were. When told to look next to the input box, they still could not see the question. When a researcher physically pointed at the banner, the participant would often still not read it. It took an explanation along the lines of "The question is in the banner to the immediate left of the box" before people would see what was in the banner. This shows exceptionally strong motivation for banner avoidance, and that this process occurs well before the sensory input reaches consciousness.
While Signal Detection Theory explains how and why banner blindness occurs, the explanation requires that all incoming sensory data is processed by the brain to some degree. Only by processing the sensory input can the brain determine whether it should be passed through to consciousness or not. This level of processing is known as "pre-conscious processing."
Research into pre-conscious processing and memory has shown that pre-conscious processing leaves records in memory. This is how, for example, a hypnotized person can remember a past event as if they had a photographic memory, yet only recall fuzzy impressions of the same event under normal circumstances. Even though much of the data at the time the event occurred was not passed to consciousness, we know it was processed, and that it was recorded by the brain.
Research into pre-conscious processing of banner ads has shown that, even though banner ads are not consciously detected, they do involve the creation of long-term memory. Many people have assumed that if a banner is not consciously detected, it has no value. As a consequence, most online advertising is evaluated on the basis that if there is no immediate response, there is no value. On the web, immediate response means a click-through, so the click-through rate becomes the primary method of evaluating ad performance. Because click-through rate is the primary method of evaluation, it becomes the preferred basis for payment. This is helped by the fact that click-through rate is easy to measure and provides a trusted and objective assessment.
A deeper understanding of banner blindness
By the beginning of 2009, researchers had brought our conception of banner blindness to the point where we understood that people will not consciously notice banner ads, but the ad is still recorded in memory, and this memory will affect future responses to the associated brand.
Largeault's research started from this basis. He was not concerned with establishing whether banner blindness existed -- it does. Nor was he concerned with proving that ads were memorized -- other research has already proven that they are. Largeault has taken things one step further by exploring whether there are differences in how we handle textual banners, such as sponsored links, compared with primarily visual banners.
His results are very clear and unambiguous.
I'm not going to detail the actual experiment Largeault ran because he explains it very clearly in his online presentation. But his findings, and their implications, are what concern us here. While not the focus of his research, there is incidental confirmation of the pre-conscious processing and memorization of banner ads. For those who still find banner blindness difficult to accept, it is worth noting that there are now numerous experiments confirming the existence of banner blindness, and absolutely no evidence has ever been found to dispute this. Banner blindness is a fact. Similarly, there are now several experiments that confirm that banners still have a behavioral effect, and no experiments have concluded they do not.
It is important to note that Largeault's research did not treat search ads, such as Google AdWords, as banners. AdWords are not treated by people in the same way as banners. They are visually different and are presented as contextually delivered search results. We have evidence these are processed differently from banner ads delivered with non-search content. Largeault's research focused exclusively on the traditional 120x40 pixel banner. The banners used were directly relevant to test website content and conveyed the same message, one visually and one through text.
What Largeault discovered is that banners composed of artwork have a much stronger effect on memory and future recognition than banners that consist mainly of text or links. The difference is large and unmistakable: Visual banner ads have almost double the impact of textual ads.