Frequenty Asked Questions
Below are a few questions we commonly receive from visitors to Project Implicit. The responses we provide reflect our best summary of the state of the research right now. Scientific understanding evolves over time as new data are collected, and our responses in this section have as well. If you were to ask these questions to other scientists who study these topics, many would agree with our responses, but some might not. This is a normal part of the scientific process – scientists gather evidence and aim to draw the best conclusions based on the data. Science progresses as a result of conversations and disagreements that are informed by theory and data.
An attitude is an evaluation of some concept (e.g., person, place, thing, or idea).
An attitude is an evaluation of some concept (e.g., person, place, thing, or idea). Sometimes you might be asked about your attitudes. That is what we call an explicit attitude or self-reported attitude. In this case, you deliberately think about your answer and then report it. For example, you could tell someone whether or not you like math.
On Project Implicit, we also use implicit measures (such as the IAT) to assess positive and/or negative associations, which people might be unwilling or unable to report. Implicit measures estimate a person’s implicit attitude indirectly, using things like reaction times.
A person’s explicit and implicit attitudes toward something may be similar or different, depending on a number of factors. For example, it is possible that you associate math more easily with negative words on an implicit measure even if you say that you like math.
At the simplest level, an attitude connects a group to an evaluation (e.g., good vs. bad) and a stereotype connects a group with a characteristic (e.g., height or kindness). Some examples of stereotypes could be a belief that older adults play Bingo or that tall people play basketball.
Stereotypes can be measured explicitly by asking you to deliberately think about and report your beliefs. An implicit measure of stereotypes assesses those associations indirectly, typically using reaction times.
A person’s explicit and implicit stereotypes may be similar or different, depending on a number of factors. For example, a person might believe that men and women are equally good at math and arts but might be faster to associate math with men and arts with women relative to the reverse pairing.
The Implicit Association Test (IAT) measures the strength of associations between concepts (e.g., Flowers, Insects) and evaluations (e.g., good, bad) or stereotypes (e.g., safe, dangerous). The test is built around the idea that making a response is easier when two things that are related to each other in a person’s mind share the same response key.
For example, we would say that you showed evidence of an implicit preference for Flowers relative to Insects if you responded faster during trials when Flowers were paired with Good (and Insects with Bad) compared to trials when Flowers were paired with Bad (and insects with Good).
We include the labels ‘slight’, ‘moderate’ and ‘strong’ as a way to help you think about the size of the bias based on the time it took you to respond to different trials on that IAT. These labels reflect how much faster you responded to the different pairing conditions.
For example, if you were much faster to pair Flowers + Good / Insects + Bad versus Insects + Good / Flowers + Bad, you would get feedback saying your behavior indicated a “strong automatic preference for Flowers over Insects.”
We chose these words based on scientific conventions for communicating the size of an effect. They are a way to help you think about the approximate degree of bias you demonstrated during the test.
The IAT requires you to sort words and/or pictures into categories according to a set of rules. For example, you may be asked to sort good words and pictures of insects using the E key. If you press the E key when you see a good word or a picture of an insect, this is coded as a correct response. If you press the other key, this counts as an error according to the rules of the task and you see a red ‘X’.
In order to provide meaningful feedback, we have a cut-off for the number of errors you can make during the task. If you received this feedback, it means you made enough errors during one or both pairings that your IAT score is likely not meaningful.
Yes, there is some evidence that the order in which you take the two parts of the test could influence your overall results. We refer to this as an “order effect”. However, our current understanding is that this effect is usually small. One way we try to minimize the possible order effect is by giving you more practice trials before the second pairing than we did before the first pairing.
The potential effects of order are not yet fully understood. Order effects are almost certainly larger for some tasks than for others and are likely to differ across people based on factors such as the strength of associations, ability to control one’s responses, and learning ability. We hope to know more about this in the coming years. For now, the fact that there may be order effects and that they are not yet predictable is one of several reasons that we encourage you to take your IAT feedback as an opportunity for reflection rather than as a perfect measure of your implicit biases.
Although this is one of the first things people think of to explain their results, there is actually no evidence that handedness influences IAT scores.
In order to understand why, you have to know that the IAT score is a difference between how fast you are in one sorting condition (e.g., Flowers + Good / Insects + Bad) and how fast you are in the other sorting condition (e.g., Flowers + Bad / Insects + Good).
People who have better hand–eye coordination or higher cognitive ability might be generally faster to respond overall, but there is no reason to think this would be the reason they are faster for one type of pairing versus the other. Similarly, people who are right-handed may be slower to respond with their left hand, but those slower responses should occur in both pairing conditions.
The question of alliteration (i.e., category labels sharing the same initial letter) was one of the first questions that academic psychologists raised about the IAT. Although it seems intuitive that it would be easier to pair Black + Bad or Female + Family because the words share the same first letter, the evidence does not suggest that this type of alliteration is an issue. You can see one way we tested this question at Project Implicit here.
There are many reasons why you might get different results when you take the test more than once and the answer to this question might not be the same for everyone. It could mean that your attitude is highly variable, possibly due to it being a weak attitude. It could mean that your attitude changed in between tests. It could mean that you were more or less distracted or tired when you completed the test one time as compared to the other. It could also mean that the test is not working well.
There simply is not yet a scientific consensus on whether it is a problem to get two different IAT scores from the same test and if so, how big of a problem it might be. It is important for us that we are transparent about this disagreement. A lot of the scientific process consists of using data to work out disagreements, and this is one that our field has been working through for several years.
One of the questions that has been raised is whether the IAT measures something long-standing about a person that does not change much over time (like an adult’s height) or something that changes a lot due to the circumstances of the moment (like whether a person feels hungry or full).
If the test measures a trait (similar to your height), then getting different scores close in time might suggest that something is wrong with the test. However, if the test measures a state (similar to hunger), then getting different scores might be evidence that the test is, in fact, working as it should be.
We do know that “practice effects” can sometimes affect one’s score. That is, people tend to receive more neutral scores the more times they complete the IAT. This result is most likely because the person has had a chance to practice how to sort the words and/or pictures; they get faster at pairing things as they become more familiar with the task and the items that they are asked to pair.
In sum, taking an IAT should be an educational experience and an opportunity for reflection – to think about yourself and your mind in a way that you may not have before. Until we know precisely how much an individual score can tell us about a person and their future behavior, we recommend that you do not take any individual IAT score to mean something permanent about yourself, or about another person taking the test. However, this should not be taken to mean that the IAT does not or cannot measure something meaningful or important in the moment. It indicates that we should interpret any individual score with caution.
Most academic psychologists use the word ‘prejudice’ to describe people who report negative attitudes toward a social group. By this definition, showing an implicit preference for one group over another does not mean that a person is prejudiced. Some people who show this preference would also report prejudiced attitudes, while others would not. The point is that the IAT cannot indicate whether a person is or is not prejudiced because it is not an explicit or self-report measure.
The IAT attempts to assess biases that are not necessarily personally endorsed and that may even be contradictory to what one consciously believes. For this reason, we would say that your behavior on the IAT indicates that you may have an implicit preference, which may or may not be the same as the attitude you would report on an explicit measure.
In brief, taking an IAT showing that one might hold biases against, or stereotypes about, people from different groups can provide the opportunity to reflect on how to best mitigate or challenge these associations.
This question continues to generate a lot of debate among researchers in psychology. So far, the answer is that scores on the IAT can be related to behavior, but the evidence and interpretation of the evidence is mixed. Our best understanding right now is that a single IAT is unlikely to be a good predictor of a single person’s behavior at a single time point, but the IAT can predict behavior in the aggregate.
To understand what “in the aggregate” means, imagine that we have one group of 100 people who show anti-Black (pro-White) bias on the IAT and another group of 100 people who show pro-Black (anti-White) bias on the IAT. We can be reasonably confident that, over time, the group whose IAT scores show anti-Black bias will produce more anti-Black behaviors than the group whose IAT scores show pro-Black bias. Such effects have been observed in important areas such as hiring and promotion, medical treatment, and decisions related to criminal justice.
We cannot, however, be nearly as confident that any one person drawn from either group will behave in anti-Black or pro-Black ways, and especially not on one specific occasion. In other words, someone with a strong anti-Black implicit preference might choose to hire a Black employee, and someone with a pro-Black implicit preference might discriminate against a Black person in favor of a less qualified White person at a specific point in time. The reason for this is that our behavior at any given point in time is determined by many different factors, and implicit attitudes are only one of them. Therefore, based on our current knowledge of human psychology, specific behaviors of a person at a specific point in time can be extremely challenging to predict.
The link between implicit bias and behavior is fairly small on average but can vary quite greatly. The same is true for the link between explicit, or self-reported, bias and behavior. However, we do know that the relationship between implicit bias and behavior is larger in some domains than in others. Moreover, even small effects can be important. Many small effects can add up over time and create big differences at both the societal level (across lots of different people making decisions) and at the individual level (across the many decisions that one person makes).
The list of topics found here are ones we feel might be interesting to learn about. This is not meant to be a list of all of the important social issues, nor is this list meant to suggest that the selected topics are the most important to research. We are always working on updating the existing choices and adding new measures. A lot of that work happens at another part of our website where you can sign up and participate in new research that helps us expand our understanding of implicit measures and implicit attitudes. You can go to that part here.
You may also have noted that the topics chosen are fairly U.S.-centric. This is because that is where most of the researchers and visitors live. We have established collaborations with researchers in several other countries and hope to continue to add content that is particularly relevant to countries besides the United States. If you are from another country, you are welcome to check out the list our international websites to see if a country-specific website for your country is available. That list is in the bottom left corner here.
The IAT operates by categorizing words and images into one of two categories (e.g., Good/Bad; Gay/Straight). This necessarily creates binaries that we at Project Implicit do not believe to exist in such rigid forms. For example, it does not allow for people who are neither “Young” nor “Old”, “Black” nor “White”, “Gay” nor “Straight”. Nor does it allow for people who identify as being both.
We are aware of the identities that these tests are not able to highlight and affirm. It is our view that the positives of understanding the many ways in which bias operates can outweigh the negatives that come with emphasizing binary categories. Moreover, on our research website, we are continually developing new implicit measures, including ones that do not rely on binary categorizations of this kind.
We make the anonymized data collected on the Project Implicit website publicly available so that scientists, educators, journalists, and others can use these data to better understand attitudes and stereotypes. Those data are available here. We also maintain a list of published research papers that use data from the Project Implicit website.
Yes. Project Implicit is a non-profit organization and we appreciate your support. You can make a donation here.
In order to participate in a Project Implicit study, you must be 18 years of age or older. Our team is currently developing a site that will be open to participants under the age of 18 years. If you would like to be notified of when those materials are available, we encourage you to sign up for the Project Implicit newsletter here.
This is an important open question that many researchers are currently working on. Based on current research, here are some approaches to consider.
Creating new associations and challenging biases.
There is a large body of evidence suggesting that information that we encounter in the moment (for example, reading a story about a heroic Black person and an immoral White person) can, at least temporarily, shift implicit preferences. However, whether these momentary changes can translate into long-term change in implicit bias is not yet well-established.
If you want to durably change implicit preferences, a quick five-minute intervention may not be enough. Instead, you may have to become more selective about the types of information you consume in your daily life. For example, this could mean going out of your way to watch television programs and movies about people who are from groups that might be less familiar to you, or that depict people in roles that don’t fit with societal biases or stereotypes. In addition, you can work to learn more about systemic barriers that can serve to perpetuate stereotypes, biases, and inequalities in our society.
Changing the impact of biases on behavior.
An additional strategy involves changing the way that you make decisions. As a first step, it is worth reflecting on the fact that we hold biases that can influence the way we process information and how we make decisions. Instead of getting rid of these biases, we can try to make sure that they have less influence on our decisions.
For example, when making hiring decisions, you might want to blind yourself to certain types of information about candidates. If you don’t have information about the person’s gender and race, these factors cannot bias your decisions. However, this type of blinding strategy might not always work well. It is possible that you recognize the value of adding diverse voices to your organization. Therefore, instead of blinding yourself to demographic information about candidates, you might want to have access to the person’s race and gender.
Another, potentially useful, strategy can include committing to decision criteria in advance. This can help eliminate the tendency to select candidates based on gut feelings (which can be based on stereotypes about who would “fit” the role the best) and then shift the criteria to match the qualifications of that candidate. For example, studies find that those with gender biases tend to rely on whatever factor favors the male candidate — if the male candidate is more experienced, experience seems more important to them, but if the male candidate went to a more prestigious school, then that factor seems more important to them. This type of problem can be avoided by deciding in advance which of these factors is more important to you.
In addition, collecting and assessing information in systematic ways can help ensure that you don’t simply go with the person that immediately comes to mind. In fact, we followed this same procedure when we expanded the Scientific Advisory Board at Project Implicit. Instead of sending invitations to the people who were easy for us to think of (our friends and collaborators), we sent out an open call and received many applications from qualified individuals who were not on our initial list.