In the mid-1990s, there was an uptick in interest in the concept of implicit attitude (sometimes called implicit bias). A key motivation for this idea was the recognition that in industrialized nations like the United States, there had been a growing set of legal protections for historically marginalized groups like ethnic minorities that had led to decreases in overt acts of racism, but the overall social and economic position of members of those groups has not improved substantially.
One explanation for that observation is the prospect that people in these cultures have persistent prejudices that are not reflected in measurements of people’s overt attitudes. When asked about racist beliefs, for example, people may respond with sentiments of equality, but their actions may still be biased against group members in ways that maintain the status difference between groups.
These underlying attitudes, which are not reflected in explicit measures of prejudice, were called unconscious or sometimes implicit biases. The aim was primarily to distinguish them from biases that emerge from the application of a consciously held and applied belief. A 2026 paper by Keith Payne in the Annual Review of Psychology explored the current state of research on this topic.
Early research in this area focused on the development of tools to measure these implicit beliefs or attitudes. Many of the measures that were developed involved having people respond quickly to prompts and using their speed of reaction as a measure of the strength of relationship or association among concepts. The most widely studied of the many methods that were created was the Implicit Association Test (IAT), which had people do classification tasks interleaved. You can read more about this test and some of the concerns about its effectiveness here. For now, though, I’ll point out that an upshot of the research on techniques like this is that people’s performance on tasks like this is highly variable and ultimately does not predict much about their later behavior.
The observation that people may not score that similarly on tests of implicit associations from one administration to another (and that these tests do not predict people’s long-term behavior) led to a flurry of research exploring whether these measures reflect momentary changes in people’s engagement of biases. For example, it may be that aspects of a specific situation may drive people’s evaluation—in much the same way that a particular food may taste better when you need to eat than when you don’t. While this work is ongoing, it is unlikely that attitudes that change rapidly with context are likely to be the source of persistent prejudices.
A particularly interesting observation with these measures has led to a radically different interpretation. A number of studies reviewed in this paper have found that looking at trends across people in a region leads to greater prediction of group behaviors than looking at behavior of individuals. A study of bias measures at a country level found that countries that had a stronger relationship of males than females to science also had bigger gaps between school-age boys and girls in science performance. Similarly, within regions of a country, stronger levels of bias in implicit measures were correlated with acts of racial prejudice in those regions, including differences in overall health outcomes and overall levels of racial bias in likelihood of being stopped by the police.
The idea here is that particular ideas may be widely held within a region in ways that lead people in that region to quickly access those ideas. Those ideas may not influence the behavior of any particular individual, but on average they have their impact. That makes it difficult to use measures of implicit bias as predictors of the behavior of an individual, because it is more a marker of the prevalence of an idea in a society rather than a factor driving the specific actions someone takes.
This interpretation of these measures also suggests why many approaches to debiasing have been ineffective. After the rise of research on implicit biases, many companies spent significant amounts of money on training designed to identify and counteract implicit bias. These training programs did not have much influence on measures of bias or on behaviors. If measures of implicit bias are measuring something about the persistence of beliefs in a region, then education aimed at a small number of individuals may not have much impact.
When the idea of implicit attitudes was first raised in research, it captured a lot of the public imagination as well. The idea that societal prejudices were rooted in prejudices people were either not aware of or had no control over was powerful. As the research has matured, however, it is clear that this work captures something powerful but more mundane. There are powerful biases that operate in regions of the world. Everyone living in those areas is aware of those biases, and they influence how easily concepts related to bias can be accessed. But these powerful biases will also be complex to counteract, because they are a reflection of social forces rather than just an aggregation of individual behaviors.
