How, exactly, do sad faces affect our ability to count?

Remember this video?

A few weeks ago we used it to demonstrate that facial expressions can disrupt the perceptual system in fundamental ways. Actually, because we could only show a few short clips, we weren't able to duplicate the research results found by John Eastwood, Daniel Smilek, and Philip Merikle. But in their, more comprehensive study, although viewers were instructed only to count "upturned arcs" or "downturned arcs," when those arcs formed "faces" with negative expressions, people counted the arcs slower and less accurately.

But how exactly do facial expressions disrupt the perceptual system? Is it something automatic and irreversible? Or can we create a situation where facial expression doesn't matter?

A new team, again led by Eastwood, tackled this question with three ingenious experiments. In the first experiment, volunteers were shown dozens of movies like the one above, and again asked to count upturned and downturned arcs. But this time, there was a critical difference. In half the movies, the arcs they were supposed to count were colored red, while the other arcs were gray, like this:

The other movies were just like the example above, with no color difference in the arcs. Did coloring the arcs have an impact? Here are the results:

Clearly, coloring the arcs speeds up reaction time dramatically, but the important thing to notice is difference between positive and negative faces in each type of display. When all the arcs were the same color, viewers were significantly slower to count the arcs in negative faces. When the arcs-to-be-counted were colored red, the handicap of the negative faces disappeared. The same was true with errors: viewers made significantly more mistakes with negative faces when all the arcs were colored the same, but not when the arcs-to-be-counted were colored red.

But it's possible that seeing negative faces doesn't slow visual processing -- perhaps it only slows physical reaction time. So in a second experiment, the faces were only displayed for 500 milliseconds (this is actually more like our example above), instead of remaining onscreen until viewers responded. This way the factor of response time wasn't measure at all. But once again, viewers made significantly more mistakes with negative faces when the all arcs were colored the same, but not when the arcs-to-be-counted were colored red.

So if coloring the arcs-to-be-counted eliminates the effect of the negative face, is there any situation where a negative face trumps the color of the arcs? In their final experiment, the researchers clumped the arcs in groups of three, like this:

This time, viewers were asked to count groups of arcs -- either upturned or downturned. As before, half the time the groups-to-be-counted were colored red, and half the time they were all the same color. Here are the results:

This time, negative faces slowed counting in both cases, whether or not the groups-to-be-counted were colored differently. The difference, the researchers say, is that now viewers were perceiving the arcs as groups, rather than individual parts. Coloring the arcs doesn't help viewers to focus in on single parts that may or may not be relevant to the facial expression the arcs convey. In this case, negative faces are distracting no matter whether or not they are colored to be more prominent.

John Eastwood, Alexandra Frischen, Michael Reynolds, Cory Gerritsen, Matthew Dubins, Daniel Smilek (2008). Do emotionally expressive faces automatically capture attention? Evidence from global-local interference. Visual Cognition, 16 (2), 248-261 DOI: 10.1080/13506280701434383