How Implicit Association Testing gives unique insights into how your customers make decisions. Implicit Association Testing (IAT) gives insight into associations and beliefs before they are processed as conscious thought. IAT gives insight into the beliefs and associations that respondents don’t know about or can’t tell you about. IAT is methodologically robust and builds on neurological models for decision-making. Implicit Association Testing (IAT) measures something completely different to qualitative and quantitative market research methods.
Why IAT is different and gives unique insight into customer decision-making.
Traditional market research techniques are essential business tools for understanding how customers behave and make buying decisions. As illustrated in the figure below quantitative methods mostly give us the answer to ‘what’ and ‘how many’ questions. Qualitative methods give us some answers to the ‘why’ questions.
Quantitative What How Many description projective
Market Research Type
Questions answered
Thinking recorded
Qualitative Why projective reflective descriptive
Implicit Association Test IAT gives insight into associations and beliefs before they are processed by conscious thinking.
in-grained beliefs unprocessed associations
IAT gives insight into beliefs and associations’ respondents don’t know or can’t tell you about.
Traditional market research uses explicit measurement techniques where; 1. the respondent has already formed an opinion on the topic or is able to construct an opinion in response to the question. 2. the respondent is aware of their opinion and reasons for this opinion. 3. the respondent is willing and able to explain their opinion with the researchers.
An IAT is a computer based categorisation exercise. • IAT measures the speed at which words are categorized to one of two pairs of concepts. • The more rapid the word allocation to the concept pairing the stronger the association between the pairing.
Research with medical or other professionals does not always allow respondent the option to have ‘no opinion’ or the respondent as a professional feels they must always give ‘an opinion’. Socially desirable responses and image-management reasons also have been found to confound responses to how and who they treat in certain ways JF Dovidio AM J Pub Health May 2012.
When concepts are paired, it is quicker to allocate words or images associated with the concepts when the concepts are similar or coherent, than when they are dissonant.
What IAT measures and the neuroscience of decision-making Neuroscience of what IAT measures • The brain instantly recognizes any word or picture stimulating a large >10,000 neuronal network of memories and associations. • This recognition happens instantly and is not adapted by conscious thinking. Concept-wordimage association response times are captured by IAT. • Similar and compatible concepts placed together in the test will give faster response times than dissimilar and un-related concepts.
A neuroscientific theory about decision-making …human reasoning and decision making depend on many levels of neurobiological operation, some of which occur in mind (i.e. are conscious, overt cognitive), and some of which do not. A R Demasio, Phil Trans R Soc Lon 1996
Decision making is not mediated by the orbitofrontal cortex alone, but arises from large-scale systems that include other cortical and sub-cortical components. A Bechara, Cereb. Cortex 2000
How you can use IAT IDEA 1 Identify the relative importance of different product features / claims / positioning’s If traditional market research tells you that both efficacy and safety are important features of a new drug, Implicit Association Tests can test the relative importance of each feature on the prescribing decision. It can also test the positive or negative associations of specific words and phrases.
IDEA 2 Identify the emotional associations or likability of a product or feature. Physicians may like certain products/brands/ companies and dislike others. IAT can compare physician likes and dislikes and the emotions associated with these preferences.
IDEA 3 Discriminate between different selling situations or customer identification associations. Physicians have recurrent habits and prescribing patterns and will use a certain drug or drug class in a specific condition or type of patient. Traditional research regularly gets the answers that every patient is unique and drugs are specific for each patient. Quantitative analysis of prescribing patterns will suggest a habit or pattern. IAT can identify which condition or patients that a physician implicitly and habitually associates a drug with.
More Complicated Ideas IDEA 4 Identify how pre-exposure to images or information changes existing product or company associations. Exposure to information, marketing materials or various situational cues can change the responses or associations of a product. In this example 50% of a sample is exposed to the ‘information/image’ before or during the test. IAT can then test whether the exposure changes the implicit associations of the product or prescribing situation.
IDEA 5 Portfolio management and co-promotion Some products naturally reinforce each other when they are sold together and other products will leave contradictory memories. Implicit Association Testing can test whether the messages for two products reinforce, negate or are neutral for the other product.
IDEA 6 Tracking the real impact of product issues and safety alerts. Safety alerts or manufacturing issues occur for most pharmaceutical products and companies. Assessing the impact of such events and whether they have a longlasting impact on brand associations is important so that the issues are addressed in a proportionate manner. IAT can tell you whether a safety alert changes the implicit associations of the product from ‘safe’ to ‘unsafe’. Using IAT to track and monitor of these events over time gives a true picture of the impact of these issues.
Published examples of IATs Example 1 Implicit Customer understanding of Mac and PC users with concepts of self or other. This study clearly showed a difference between the way Mac and PC users conceptually understand and associate with Macs and PCs. PC users did not differentiate Macs and PCs on the concept of Self but Mac users very clearly had difficulty ( longer latency) associating PCs with self. The Self target concept included words such as me, my, mine contrasted with other, their and theirs. Brunel FF. J Cons Psych 2004, 14(4) 385-404.
Mac & Self words
PC & Self words
Response Latency ms
800 700 600 500 PC-User
MAC-User
Example 2 Obese people do not have an implicit attitude bias towards high-fat food. This study was built on the explicit findings that obese people consume more high-fat foods because the prefer them. The study hypothesized that obese people have a pre-conscious preference for high-fat foods. The hypotheses was not proven showing that obese people have more difficulty in associating high-fat foods with positive thoughts than non-obese controls. This identified a gap in the understanding of the food preferences of obese people. Roefs A. J Abnormal Psych 2002, Vol. 111, No. 3, 517–521
Control group
Obese group
Response latency ms
1400 1200 1000 800 600 400 200 0 High-fat negative
High-fat positive
IAT Reliability and Other Measures of Implicit Attitudes and beliefs. Tests such as Rorschach Inkblot Tests which require respondents to project parts of themselves by reacting to unstructured ambiguous visual stimuli, collage creation, or sentence/word completion tests require significant researcher interpretation and the rich descriptions rarely allow for accurate and specific measurement of attitudes in a population.
• • • • • • •
Reliability of Implicit Measures The following implicit measures have been compared for reliability (Cronbach alpha and test-retest correlation), correlations with explicit measures and other implicit measures and stability upon exclusion of outliers.
The reliability of the IAT and the BIAT was greater than the other measures. GNAT and ST-IAT was acceptable but these tests need more refinement to be as reliable as the IAT and BIAT. The GNAT was dependent of participants ability to perform the test correctly. The SPF, EPT and AMP were less reliable more dependent upon outlier outcomes. (Bar-Anan Y SSRN 2012)
(Brunel FF J Con Psyc 2004)
Implicit association test IAT Brief implicit association test BIAT Go/No-Go association task GNAT Single-Target implicit association test ST-IAT Sorting paired features SPF Evaluative priming task EPT Affective misattribution procedure AMP
Correlation of IAT with explicit measures IAT is especially useful in situations where market behavior and traditional explicit market research diverge. IAT will be correlated with explicit measures if the following are true: 1 – Respondents are not concerned about image management or giving the researcher the answers they think the researcher wants or are socially desirable. 2 – Respondents have a good introspective ability and understand their attitudes to the issues involved. Lack of correlation between explicit measures and observed behavior can show the need for implicit research. For example if physicians state they rarely use inhaled corticosteroids in moderate COPD but
prescription data show 50% of patients get inhaled corticosteroids it is likely that well designed IAT can uncover the reasons for the data discrepancy. Explicit measures are recommended to be used alongside IAT to help understanding the sample population and their explicit attitudes or behaviors. However because explicit measures are more likely to be altered by prior intervention it is suggested that these are complete before the IAT. The IAT has been shown to be more robust to prior intervention.
For more information or to begin developing your own IAT please contact. Dr Richard Wood Director The Implicit Testing Company +44 771 779 0253 richard.wood@implicittesting.com www.implicittesting.com Š Copyright 2012 The Implicit Testing Company Ltd