But, even if we disagree with a particular stereotype, can it still affect us? Research shows that it might. There are what are called implicit stereotypes that operate outside of conscious awareness, but can still affect us. A particularly clever methodology has been developed by Greenwald to test out our implicit biases - the underlying beliefs and associations that we might have and not even be aware of. This methodology was originally developed to examine implicit attitudes directed towards African-Americans, but has since been extended to a wide variety of areas in which people standardly discriminate against each other.
I think this is research that provides a good model of what research should be; about important things, methodologically sound, based on prior knowledge, and ultimately, creative in its development and execution. What Greenwald's test does is use our reaction time as a measure of what things tend to go together with what other things in our brains. In other words, how quickly we can perform certain tasks is a clue about the structure of our concepts.
What happens is you are asked to make a decision about a series of pictures. A photo of a face pops up - if the photo is of the face of an African American you are asked to indicate this by pressing the "e" button, and if it's the face of a European, you are to press the "i" button. After this training session, the faces are replaced with words. If a "good" word (e.g. love, glorious) pops up you press the "e", and if a "bad" word (e.g. war, hate) pops up you are to press the "i" button. Then, after this second training session, you are asked to do both tasks at once. Either a word or a face will pop up, and you have to categorize them correctly. This process is repeated, although the pairs are varied (i.e. first you might be asked to categorize (AA/good, EA/bad), then (AA/bad, EA/good)).
The critical dependent variables are your reaction times and mistakes. If you can respond more quickly when the "good" words and European American faces are paired, then this suggests that you have an automatic association of European American and good. In other words, you are quicker to respond to that particular pairing because it fits with your own unconscious attitudes. The task and your beliefs match. On the other hand, if they don't, your performance will be slower. So, we can get a sense of how strong your biases are by how much you are sped up or slowed down in your responses. Analogously, your mistakes will tend to be in one direction. If you make very few categorization mistakes for the (EA/good, AA/bad) trials, but many mistakes for the (AA/good, EA/bad) trials, then that suggests that the second pairing is harder for you because it doesn't match your internal associations.
As it turns out, most everyone, even people who are low in prejudice, tend to have biases. In fact, when the IAT was first developed, many people became upset at discovering that they had been harboring less-than-nice beliefs (that's in fact, now in part of the consent form so that people aren't surprised to find out unsavory truths about themselves). While some people argue about what these findings might mean, Greenwald and his associates make the argument that without due diligence on our part, implicit biases can indeed color our actions and impact explicit, conscious behavior.
Try the IAT for yourself.
