|Colin M. MacLeod, PhD|
Before reading any farther, find some paper and four colored pens (e.g., red, blue, green, and yellow). Your mission--should you decide to accept it--is to recreate Stroop's experiment.
On the first sheet, print 12 words in color in a single column. Each word must correspond to one of the pen colors (i.e., red, blue, green, and yellow), but must be printed in the wrong color (e.g., RED printed in green, then BLUE printed in red). Try to use all of the words and colors equally often, avoiding patterns. This is the incongruent condition.
On the second sheet, replace the words with rows of Xs (e.g., XXX in green, then XXXX in red), leaving everything else the same as on the first sheet. This is the control condition.
Now you need a stopwatch and a willing participant. Tell this person that the task is to name the ink colors of printed words aloud, ignoring the words themselves. This should be done as fast as possible without stopping.
Then turn over the control sheet and ask your participant to begin color naming. As soon as you hear the first color, start the watch; as soon as you hear the last color, stop the watch and record the time. Reset the watch, switch to the incongruent sheet, and repeat.
What you will find is that it takes much longer to name the colors for the incongruent condition than for the control condition. Compared with the control row of Xs, incongruent color words interfere with naming the colors. It is impossible to resist reading the words, so they get in the way, slowing you down. Try it yourself: Despite knowing the explanation, and try as you may to resist reading, you will show interference. This interference is called the Stroop effect, named after John Ridley Stroop, who first demonstrated this result in his PhD thesis in 1935. For 60 years, psychologists have continued to use his task, a hallmark measure of attention. For a cognitive psychologist, the fundamental question is always why: Why does interference occur? In 1885, James McKeen Cattell reported in his PhD thesis that people can read words faster than they can name the corresponding colors. He suggested that reading had become automatic through extended practice, whereas color naming had not. This idea led to the "speed of processing" explanation of the Stroop effect: The faster process interferes with the slower one, but not vice versa. Try reading the words in the incongruent condition; you will see that it is easy and that you experience no interference, as Stroop also showed.
For many years, the speed-of-processing explanation was the dominant one. But several years ago, Kevin Dunbar and I carried out some experiments that called this explanation into question. Our manipulation was simple: We turned the words upside down and backwards. The goal was to make people's reading of the words slower than their naming of the colors, reversing the usual pattern. We should then find that the faster colors would now interfere with reading the slower words and not vice versa, the opposite of what we usually observe. Using computers, we were able to present and time the response to each color word individually.
Our results surprised us. Contrary to the speed-of-processing prediction, we did not find that incongruent colors interfered with reading words; there was no interference at all in word reading. Moreover, we also found the normal Stroop effect: The words interfered with naming the colors. This was true despite the fact that we had succeeded in slowing word reading down dramatically, to the point where it was faster for people to name the colors than to read the rotated words.
On this basis, we argued that the speed-of-processing account was wrong. In its place, we suggested that what was important was not the speed but the balance of automaticity with which words versus colors are processed. The extensive practice we have with reading over our lives makes reading words exceedingly automatic; in contrast, color naming is relatively unpracticed and thus not very automatic. Our claim was that more automatic processes interfere with less automatic processes, and not vice versa. Greater automaticity promotes increasingly obligatory processing: In a very real sense, we cannot stop ourselves from reading words, even if we want to. Try not reading signs along the road the next time you go for a drive. Notice that you are not tempted to name the colors on the signs.
The next step was for us to test our new automaticity explanation. At its heart is the idea that practice is a major determinant of automaticity: the greater the practice, the greater the automaticity. Again, the experimental setup was straightforward. Participants had to learn to respond to each of four novel stimuli with a different name. The stimuli were simple line-drawing shapes that were easy to tell apart. The trick was in the responses: Each shape was assigned a permanent color name (pink, blue, green, or orange). We then varied the amount of practice participants had in applying these color names to the shapes.
One experiment lasted 20 days. (Needless to say, we pay our volunteers!) On Days 1, 5, and 20, there were four phases, the first two involving only one dimension: color or shape. In Phase 1, a color patch was presented on each trial, and the task was to name the color aloud as fast as possible. In Phase 2, one of the four shapes was presented (in white) on each trial, and the task was to name that shape aloud as quickly as possible. These two phases provided baseline times for the color and shape dimensions, which have been contrived to have the same names.
In the remaining two phases, the shapes were presented in color. In Phase 3, the task was to name the color ignoring the shape; in Phase 4, it was to name the shape ignoring the color. These were the test phases.
On all of the remaining days, the only task was to name shapes (Phase 2), so participants became increasingly well practiced in giving color names to the shapes. Indeed, by Day 16, one subject reported dreaming about them!
Our results fit with the automaticity explanation and dealt a further blow to the speed-of-processing account. On Day 1, naming colors showed no interference from ignored shapes (Phase 3), whereas naming shapes showed considerable interference from ignored colors (Phase 4). The new shape names were still not very automatic, so they did not influence color naming. But from pre-experimental experience, the colors were somewhat automatic, so they did interfere with naming the shapes. The automaticity of shape naming increased by Day 5, and there was now equal interference in both test phases. (Notice that the speed-of-processing account cannot explain this: Both dimensions cannot be faster!) Finally, on Day 20, shapes interfered with color naming, but colors no longer interfered with shape naming. Extensive practice had made shape naming the more automatic process.
Together, these experiments led us to revise our explanation of the Stroop effect, relying on the strength of the disposition to respond, not the speed of responding. Interference occurs because one is compelled by experience to read the word, not because reading the word is faster.
More recently, I have turned my attention to the other side of the coin: facilitation. Print the color words in their proper colors (e.g., RED in red), and you have created a congruent condition. Now name the colors, and you will find that you are a little faster than you were even in the control condition. The intuitive explanation of facilitation is that a congruent word helps you to name the color by supporting the same response, in the same way that an incongruent word hurts you by providing a conflicting response. Thus, most theories explain interference and facilitation by the same mechanism. But are they right?
In her PhD thesis in my lab, Marina Vanayan found facilitation and interference to be uncorrelated. Very recently, I have shown that practice reduces interference, but does not affect facilitation at all. These results--and others--are unexpected if facilitation and interference derive from the same source.
Penny MacDonald and I came up with an alternative explanation of facilitation: the inadvertent reading hypothesis. Very simply, we suggest that, on congruent trials, people may sometimes slip and accidentally read the word instead of naming the color. Neither the participant nor the experimenter can tell that the response was to the wrong dimension. Keep in mind that the participant is supposed to be naming colors, not reading words. Because reading words is faster than naming colors (remember Cattell?), even a few of these slips will make the average time in the congruent condition appear to be faster than that of the control condition. But, in fact, there is no true facilitation occurring.
To test our new explanation, we turned to a bilingual version of the Stroop task. Whereas the usual facilitation appeared in the same-language congruent condition (RED in red, say "red" to the color), there was actually interference in the different-language congruent condition (ROUGE in red, say "red" to the color). Essentially, translated words could be detected and removed if they were read, and doing so eliminated facilitation. The same was true in another experiment using color-related words (e.g., SKY in blue): There was interference, not facilitation, in the congruent condition for these words because reading the word could no longer masquerade as naming the color.
In another experiment, we compared a standard Stroop experiment to one with some trials mixed in where people were told to read the word rather than to name the color. The goal of this modified task was to lead people to slip and accidentally read the word more often on the color-naming trials. If we were right and facilitation derives from accidental word reading, then we should have observed more facilitation in this modified task where word reading will be more likely. That is precisely what we found.
This recent work shows that the apparent facilitation is not real. Rather than the congruent word helping you to name the color, we believe that, because the response to the word is more automatic, it sometimes is blurted out first, supplanting the naming of the color. If we are right, then there is no real facilitation, and any theory that predicts true facilitation is wrong. As it happens, virtually all existing theories make this prediction. We have explained facilitation, so now we can return to developing a theory of interference that does not also predict nonexistent true facilitation.
In sum, the Stroop task provides a rich and intriguing domain to test our ideas about the operation of attention. Indeed, it is now being widely used to study anxiety disorders and the efficacy of their treatment, and even to identify children having difficulty reading. A very simple instrument continues to be one of the most valuable ones in the cognitive toolkit.