Tell me something I don’t already know – the challenge for neuroscience in advertising research

Graham Page, Global Director of Research & Development,

Millward Brown

Isn’t neuroscience sexy stuff?  No-one who read the May edition of Admap can have failed to be excited at the potential of brain-scanning and brain-imaging to generate better consumer insights.  It’s a tantalising prospect - a ‘clean’ read on what consumers really think and feel, untainted by interviewer bias, questionnaire design or social desirability effects. 

Given that Millward Brown is one of the world’s leading providers of survey-based market research you may be expecting the rest of this article to promptly bash ‘neuromarketing’ as hype, but that’s not the case.  We’re as excited as everyone else by this technology – it’s not in our interests to be left behind by new developments.  For that reason, over the last few years we have been actively investigating the merits of different neuroscience techniques, and this article will detail some of our findings.  In doing so our aim has been to look beyond the buzz and examine what the technology can really deliver.

There are two challenges to overcome before this technology becomes widely used in brand and advertising research:

1.      We have to demonstrate that it yields meaningful and discriminating results about brands and advertising

2.      We have to demonstrate that it adds new insights that survey-based research does not deliver – and that these extra insights pay for the cost of doing it.

Reading many of the published articles on ‘Neuromarketing’, it would be easy to believe that it’s impossible for the technology to fail either of these tests.  However we have to remember what it tells us.  Put crudely, brain-scanning/imaging tells us that the brain, or parts of the brain, ‘light up’ in response to a given stimulus.  What this means is then down to neuroscientists’ understanding of what different parts of the brain do, or what different patterns of activation represent.  Recent advances in this area have been huge, but as Tamsin Addison points out in her May article, ‘understanding the implications of [brain] changes is crude.’  As a result it’s unsurprising that practitioners such as Dr Lewis and Peter Laybourne of Neuroco acknowledge in their article that Neuroscience techniques have to be combined with more established research techniques.  Without survey-based information, interpreting neuromarketing results is incredibly difficult. 

It has also been implicitly accepted by many that survey-based techniques cannot possibly measure emotional response and consumer decision-making, because these are largely unconscious and so are not available to conscious introspection.  Therefore, the argument goes, neuroscience is the only way to get at these ‘unconscious’ events.  This simply isn’t true.  Psychologists and neuroscientists disagree a lot about the definition of emotions - but there IS extensive agreement that an emotion is a conscious experience.  You know when you feel something.  Hopefully you’re enjoying reading this article, rather than being irritated by it, but either way, you know it.  Sometimes emotions may be fleeting, and/or difficult to articulate spontaneously – but that has a very different implication.  It means that with the right prompts, introspective approaches can access how people feel.

The same goes for decision-making.  If this is an unconscious process, we’d be spending a lot of time at supermarket check-outs wondering quite why we’d bought the items in our basket.  There’s a big difference between something being difficult to articulate and impossible – and neuroscience’s proponents do themselves no favours by pretending otherwise.

Finally, there is the assumption made by a lot of neuroscientists that what people say is simply not to be trusted.  Well, yes, sometimes respondents might lie, or moderate their answers.  But to assume that this is widespread overlooks two important facts:  1) Lying takes effort and generates stress, and the extra effort entailed in making answers up is simply not worth it for most people.  2) A market research interview is not particularly threatening or important – the motivation to lie is not there, in the same way, for example, as there is for a criminal being interviewed by the police.

Clearly I’m not denying there are social desirability effects in survey data – but with the good normative databases, or by looking for changes over time in answers, these effects can be overcome. 

So it is not a given that neuroscience techniques will give more valid answers than survey-based approaches.  The right survey questions can be deeply insightful, as evidenced by the extensive validation against sales that companies such as Millward Brown have for their metrics.  This means that as with new survey-based techniques, we have to hold neuromarketing approaches to account against the two tests I outlined earlier, and this is exactly what we have done in our own exploration of these techniques.

Brainwave Science

Millward Brown’s Research & Development team recently worked with a US-based Neuroscience company, Brainwave Science, to compare brain responses to TV advertising, as measured by their approach, with the results from Millward Brown’s quantitative copy-test, Link™.

Brainwave Science offer a patented electroencephalography (EEG) technique which measures Event Related Potentials (ERPs) – the electrical signals emitted by the brain in response to a given stimulus.  Specifically their technique focuses on the differences in a brainwave response (the P300 MERMER^®) to stimuli which the subject’s brain regards as salient or important, vs stimuli which are regarded as unimportant. Put simplistically, the technique measures the ‘aha!’ response by the brain when it encounters something important (what psychologists often refer to as ‘context updating’).   To date, the main application of their approach has been in the criminal justice arena, because stimuli which a subject recognises elicit a stronger brain response than stimuli which a subject does not recognise.  This can therefore be used to ascertain if important information relating to a crime is or is not recognised by a suspect.  The scientific basis of the approach is extremely robust – and as a consequence it has been ruled admissible in US law as scientific evidence, and has been used in several high-profile cases.

In investigating the utility of the approach to advertising research, we were interested to see if different parts of a TV commercial would elicit brainwave responses of different magnitudes – thereby indicating which parts of the copy respondents deemed most important, and which lodged most successfully in their brains. 

The commercial we used was a Canadian ad for Vim Cream bathroom cleaner.  To obtain a ‘clean’ read we conducted fieldwork in the US, where the ad had never aired, and where the brand is not sold, thus meaning responses were unaffected by existing memories of the ad or brand.  The ad was chosen because it is emotionally very powerful, and fabulously executed, thus hopefully allowing us to see how the brain responds to emotionally charged copy.  A description of the ad does not do it justice, but is nonetheless necessary to understand some of the results which follow (scenes from the ad are shown in figure 1).

The ad opens with a sad-faced little girl looking at her mother on the other side of a glass screen.  The girl raises her palm to the glass, and her mother, clad in an orange overall, presses her hand against the other side. 

‘When are you going to get out of here?’ asks the girl.

‘In a while’ replies the mother, distraught.  ‘I gotta go’ she adds, turning away.

The camera cuts back to reveal that rather than being in prison, the mother is in fact in a bathtub, and the glass screen was the shower screen.  The mother is scrubbing the bath.

‘I love you momma!’  shouts the girl, her hands against the screen.

The mother looks over her shoulder.  ‘I love you too, baby!’ she shouts, then turns back to her task with a sob.

We then see Vim Cream being applied to a sponge, followed by a split screen of various surfaces being ineffectively cleaned, followed by a Vim-cream soaked sponge cleaning up rapidly, with a voiceover: ‘Spend less time cleaning tough dirt.’

The ad ends with a packshot, tagline and voiceover: ‘Vim Cream – cleans the tough stuff, easily.’

We showed the ad to 34 bathroom cleaner users, and subsequently measured their brainwave responses to specific scenes from the ad. This was done by re-showing the scenes depicted in Figure 1 in a random order, and simultaneously measuring the electrical signals emitted by the brain via 6 electrodes attached to respondents’ scalps.   Intermixed with these scenes were scenes from European bathroom cleaner ads which respondents had not seen (thus providing a baseline of ‘definitely unknown’ stimuli) and pictures of George Bush and John Kerry (thus providing a reference point for ‘definitely known’ stimuli).  Each respondent was shown the scenes multiple times, thus giving us a robust number of observations across the sample.  All respondents also completed a verbal questionnaire about the ad, half prior to brainwave measurement, and half subsequent to brainwave measurement.  The nature of the verbal questionnaire entailed re-showing the ad 4 times, so half the sample had seen the ad once when their brainwaves were measured, and the other half had seen it 5 times.  For analysis, we focused on the group who had seen the ad more frequently, as we felt this was closer to the repeated exposure to the ad that would occur if it were on air (though both groups responded very similarly).

Figure 2 shows the average brain response to the three kinds of stimuli described above.  It clearly shows that the scenes from the ad elicited a stronger brain response than the ‘definitely unknown’ scenes from European ads, and of a similar scale to the ‘definitely known’ stimuli.

More interesting, however, was the difference in brain response to different scenes.  This is shown in Figure 3. The scenes which elicited the strongest brain responses tended to be the high-points of the interaction between the mother and daughter – the ‘hands on the glass’ sequence, and in particular the scenes where the mother and daughter say they love each other.  The scenes involving the product demonstration and brand message evoke a much weaker response.  This suggests the focus for consumers is on the story between the mother and daughter, rather than the brand or it’s key message.

Clearly, on this ad, the brainwave measurement yielded discriminating results with good face-validity – it passes our first test.  But how does this compare to the results from a more established copy-testing methodology?

A comparison to the ‘Link™ ’ copy-test

To generate a comparison between approaches we conducted a full TV Link test on the ad, using a separate sample of bathroom cleaner purchasers, at the same time and in the same geographical region, as the Brainwave Science test.  TV Link™ is Millward Brown’s proprietary copy-testing system, and uses a survey-based approach.  Respondents are shown the ad twice, and answer a series of introspective questions about their reaction to the copy.  This generates a series of metrics including likely memorability of the copy, rational and emotional message take-out, and motivational power.  It also includes diagnostic measures, which tell us which parts of the copy are the focus of attention for consumers, and which influenced their responses to the ad most strongly.

Particularly pertinent for this test are a series of measures we recently introduced to all tests worldwide to gather more detail on the emotional response elicited by advertising, and which elements within the copy are driving different emotional responses.  We measure the overall emotions the copy is evoking on 16 different dimensions (8 positive, 8 negative), and then ask respondents to indicate which elements within the ad generated those emotions.  Respondents do this by watching the ad again, and as they do so, they move a mouse left or right to indicate which parts of the ad made them feel more or less of the emotion they felt most strongly overall.  Looking across the sample, this enables us to produce a series of ‘Emotional Traces’ which demonstrate which elements of the ad evoke different emotions or series of emotions. (The technology is similar to that which we use to generate the ‘Interest Trace’ which some readers may be familiar with – but focuses on emotions rather than which parts of the copy respondents found more or less ‘interesting’).

Results

Overall the Link™ results suggested that the ad evoked a very strong emotional response – the ‘joke’ generated a lot of surprise, but the storyline overall generated a lot of negative emotions, particularly feeling ‘Sad,’ ‘Disappointed,’ ‘Repelled’ and ‘Unimpressed.’  Negative emotions are often crucial in funny ads to generate the humour, but in this case, the joke did not work for most US consumers, with the ad receiving a well below average enjoyment rating, and evoking verbatim responses such as ‘you end up feeling emotionally used.’  This copy may work strongly in some countries (such as Canada or the UK) but our US audience did not respond well.

This is important to understand, because we found that the elements within the ad, which evoked those negative emotions, as measured by the Emotional Trace, corresponded very strongly with the scenes within the ad which generated the strongest brainwave responses.  Figure 4 shows this.  The continuous line is the negative emotional trace, arrived at by combining the results from all respondents who said the main emotion the ad evoked was negative.  The bars show the brainwave responses to particular scenes from the ad.  We can clearly see that the two sets of results tie up very strongly.  The negative emotional trace rises as the separation of the mother and daughter becomes apparent and remains high as that part of the story progresses.  It does not dip as the ‘joke’ is revealed – instead, it kicks up further as the mother and daughter shout that they love each other.  The negative emotional trace suggests this is the peak of emotion in the ad – and this also generated the strongest brainwave response.  As the product demo and packshots are introduced, emotions fall away, as do brainwave responses.  The correlation between these results is high, r=+0.72. 

There is also strong agreement between the brainwave results, and consumers’ verbal descriptions of the ad, and with verbal reports of the elements that ‘stand out’.  In describing the ‘story’ of the ad, consumers focus very much on the interaction between the mother and daughter, and far less on the brand or the brand message.  When asked what they think will stand out when they see the ad on TV, scenes such as the mother shouting ‘I love you too!’ received very high mentions, and the correlation between the scenes cited by respondents as likely to ‘stand out’ and the brainwave responses was a very reasonable r=+0.57.  Figure 5 summarises the key relationships between Link™ and Brainwave results.  A multiple regression analysis of Link™ measures vs the brainwave results suggested two key variables were associated – the negative emotional trace and the ‘stand-out’ data, which together related to the brainwave data with a very strong r value of +0.81.

Together the Link™ and Brainwave data strongly suggests that the storyline between the mother and daughter is the key focus for US consumers, and this overrides the brand’s role in the ad.  The Link™ data further suggests that for this audience at least, the negative emotions evoked overwhelm the humour in the ad.

Conclusions

The strong relationship between the Link™ results and the Brainwave Science data provides good mutual validation for the two approaches.  It strongly suggests that the Brainwave Science data is meaningful, but equally it suggests that the introspective questions and related survey based approaches employed by the Link™ test are measuring genuine consumer responses – they do tap into what’s going on ‘in consumers’ heads,’ contrary to what is often suggested.

The key question of course is whether the Brainwave Science data passes the second test – does it tell us something new?  It certainly provides a compelling illustration of the most powerful scenes within the ad.  However it’s also clear that existing Link™ measures also highlighted these scenes, and we would have reached similar conclusions.  Furthermore, the verbal data from the Link™ tests was instrumental in understanding the nature of the emotional response to the copy – just looking at brain response alone, it would have been difficult to understand whether consumers enjoyed the story or not, or whether they got the joke or not.  Brain response data is a great indicator of overall response, but verbal data is needed to understand the valence and nuance of the response.

So where does this leave us?  This study certainly reinforces the promise of neuroscience, and of Brainwave Science’s approach in particular, because it yields meaningful and discriminating data.  However we are yet to find the incremental insights that neuroscience needs to deliver before it warrants wider adoption.  We are still exploring this area to see if we can find them, including looking at methodological refinements with Brainwave Science.  However, the evidence so far reinforces rather than undermines the validity of our survey-based techniques, like Link™.

Graham Page

Global director of Research & Development, Millward Brown

June 2005

TOP      HOME      ADVERTISING HOME