Why should we measure brainwaves in order to tell whether people are paying attention? Why can't we just ask them?
There are many things you can get by asking people, so we would not advocate replacing asking people with brainwave research. However, there are lots of things you cannot get by asking people. For example, the decision to pay attention or not to a video or an advertisement, is often quite an instinctive, unconscious one, and people are not always aware of why they pay attention or not. You can ask people whether they like something - but then you are already making them pay attention to it. In the world of information overload and advertisement saturation in which we live, the key difficulty is getting people's attention in the first place. By testing content by looking for brainwaves, you can test whether it captures people's instinctive attention.
What is Inter-Subject Correlation?
Inter-Subject Correlation (ISC) is a measure of how in sync the brainwaves of different audience members watching the same video are.
What exactly do you mean by brainwaves?
Electrical signals produced by activity in the brain.
Your brain produces electrical signals?
Yes, whenever you are using your brain, it is giving off electrical signals, of varying strength.
How do you detect those electrical signals?
Through Electroencephalography (EEG), a technique whereby a researcher places electrodes on a person's head.
So what do you mean by ''in sync''?
Going up and down at the same time. Brainwave signals are in sync if the electrical signals of different people's brains are strong at the same times and weak at the same times.
What does it mean when brainwave signals are strong and weak?
The more activity in the brain, the stronger the signal.
Is there more activity in the brain if you are paying attention to a stimulus?
Yes there is.
So if you see strong signals, does that mean that the person is paying attention?
What? Why not? I thought you just said it did?
Paying attention to a stimulus means a lot of activity and strong electrical signals. But strong electrical signals resulting from a lot of activity may or may not mean that the person is paying attention to the stimulus.
Hmm. What else could those signals mean?
It could mean that they are thinking really hard about something that is totally unrelated to the stimulus, like the project they have to do for college, the presentation they have to do at work tomorrow, last night's game, or tonight, dinner. Anything. These unrelated signals are called ''noise''.
So why measure electrical signals in the brain if there is so much noise?
Because over a large sample size, noise in one subject becomes irrelevant. Also, over a large sample size, at each point in time roughly the same percentage of people will be generating noise at each second, although it will be different people at different times. So across a large enough sample size, the noise evens itself out.
But isn't it expensive to survey a large sample size of people?
Yes it is. That is why Inter-Subject Correlation (ISC) is so useful.
How is that different?
It does not measure the strength of the signal, it measures the extent to which the signals of different people watching the same video are in sync with each other. That way, noise in one person already makes the ISC go down, because it means that that person is out of sync with everybody else. So, a high ISC could not be caused by noise, even if you have only a small sample size. A study with a small sample size is a lot cheaper and a lot quicker than one with a large sample size.
OK, so there is no noise so you don't noise a large sample to even out the noise. But how do we know that ISC actually reflects whether people are paying attention?
If the stimulus is engaging, everybody is paying attention and so their brains are in sync: a lot of activity at critical moments in the video, and less activity during more chilled out moments, all according to the ups and downs of the video. If the stimulus is not engaging, people tune out, and their minds wander.
OK, but if everybody is tuned out, surely they would make a high ISC too, because they are all the same?
Because when minds wander, they wander independently, all according to each person's different thoughts rather than according to the video. So if I am thinking about last night's game and you are thinking about tomorrow night's dinner, our brainwaves go up and down totally independently of each other. So a disengaged audience produces a low ISC.
OK, makes sense, but still is there any evidence that ISC actually means anything?
Yes! A team of researchers analyzed (link) Superbowl advertisements in 2012 and 2013, and found that the ISC of 12 people watching those advertisements was 81% positively correlated with the ratings given to those advertisements by hundreds of thousands of people on USA Today / AdMeter.
What does it mean to say 81% positively correlated?
It means that advertisements that generated a higher ISC among the 12 people were more highly rated by the hundreds of thousands of people on USA Today / AdMeter, and that the relationship was so strong that, if you were looking for things that explained why different advertisements received different ratings on AdMeter, the ISC explained two thirds of that variance.
So ISC really does mean something for advertisements! What about for TV content?
The same researchers analyzed the film The Walking Dead, and found that the ISC at different points during the show was 60% positively correlated with the Nielsen ratings, a measure of how many people were watching. So again, a high ISC meant that more people were more likely to stay tuned in.
And you can really predict this by using a sample of just 12 people?
ISC negates the reason for needing a very large, expensive sample size, but we would still recommend to put the sample together with the same discipline you would a normal focus group: make sure that your sample is representative of the audience that you want to reach and has the appropriate mix of different demographics and characteristics for your product.
What else can you do with ISC?
Not just evaluate and predict responses to whole videos: you can also look within the video and see at which points people were paying attention and at which points they were tuned out.
What could you do with that information?
A number of things: for example, if they tune out towards the end of a long ad, you can reduce the length of it and save money on the media spot. Or, you can see what it was that grabbed their attention, so you can use that same concept in your next ad. As another example, you could test whether people are really paying attention at the ''branded moments'' - the parts of the ad most closely linked to your brand. It is not enough for people to like your ad, you want it to change perceptions of your brand or make them buy your product. For this, they have to associate the positive memory of the ad with your brand, so you want them to be paying attention, for example, when your logo appears.
That's interesting. Those measures that you mentioned earlier were about attention and ad-liking. Can ISC predict whether people will actually but the product as a result of an ad?
We are currently running a series of experiments to determine whether the ISC of an ad can predict the sales performance of an ad.
In the meantime, we know that ISC is a good indicator of attention. People need to pay attention to your ad in order to remember and like it, and they need to remember and like it in order to buy the product as a result. We would recommend using a study with ISC to determine whether people are paying attention, and a traditional focus group to ask whether they like it once you have made them pay attention to it.