You may have heard of the Stroop effect, and you may have even seen it demonstrated. But can the Stroop effect itself be manipulated? This short demo may show that it can.
In case you're not familiar with the effect, it occurs when you try to say the color a word is printed in, while the word itself names a color (so if you see "RED" you should say "green.") Try it with these short lists. Remember, say the color the word is PRINTED in, not the color named by the word.
Which column was most difficult? Let's make this a poll:
The basic Stroop effect predicts that Columns 2 and 3 will be much harder than Column 1, since reading the name of a different color makes it more difficult for us to say the color of the letters. But if we've replicated a nuance of the effect here, Column 3 should be more difficult than Column 2. Why? It's a process called "negative priming." In Column 3, the color of the letters in a word is named in the word above it ("Black" is printed in purple, "Red"is printed in black, and so on).
Normally when we're "primed" with a word or an image, it's easier to produce that word or related words later on (so "detective" might help us remember "police" in a list like "cat, airplane, telephone, police, museum"). But in our example above, priming actually acts the opposite way, and makes the task more difficult.
A 1989 study led by Stephen Tipper (who first observed the negative priming effect) suggested that children might actually have a different response to negative primes than adults: the effect appeared to be much smaller in children. But in 2004, Verena Pritchard and Ewald Neumann found no difference in negative priming in children versus adults. Which study was correct?
To get to the bottom of the matter, Pritchard and Neumann conducted a much larger study of 244 individuals, in five separate groups ranging from 5 to 25 years old. The experimenters went into classrooms and tested kids using flash cards with lists of eleven words much like the example above. Each card had just one list, which was either a standard Stroop task like in Column 2 of our example, or a negative priming task like Column 3. The experimenter timed the respondents and tracked their number of errors. Here's a first look at the results:
The green bars show reaction times (for each word) with the standard Stroop task. The yellow bars show the reaction times when negative priming is used. For every age group, the reaction times are significantly longer with negative priming. Indeed, the difference between the Stroop reaction times and the negative priming times is significantly larger for younger kids than for adults!
What's going on here? This is the opposite of what Tipper's team found. Pritchard and Neumann point out that young kids are generally slower to react, so when you properly account for that by taking a logarithm of their reaction times, there's no difference between the scores for children and adults. Everyone they tested responded the same to negative primes, it's just that adults do the task faster.
Now arguably kids may just not be as good at reading as adults, which somehow might affect the results of this study. So in a new experiment, Pritchard and Neumann repeated the test using colored blobs instead of words:
This time the task was simply to name the color of the centered blob while ignoring the two flanking blobs. Once again, the same result was found: Both kids and adults experience negative priming, and there's no significant difference between the effect for kids and adults.
So while adults most definitely are better at the Stroop task than kids, negative priming affects kids and adults equally. Pritchard and Neumann believe this means that there's a different cognitive mechanism for negative priming versus the self-disciplined inhibition response required for the Stroop task. Negative priming may be a relatively automatic function, while inhibiting the urge to just read the word instead of naming its color may be a more conscious process.
Verena E. Pritchard, Ewald Neumann (2009). Avoiding the potential pitfalls of using negative priming tasks in developmental studies: Assessing inhibitory control in children, adolescents, and adults. Developmental Psychology, 45 (1), 272-283 DOI: 10.1037/a0014168
negative priming: I noticed it in the list and used it to help, so I actually found column 3 easier. If you look ahead in the list, you can use it to your advantage.
It's true: if people "get it," then the effect can actually be reversed -- which in some ways makes the study results more interesting.
What if you had no trouble reading any of the columns at all?
You weren't supposed to read them, you were supposed to say the color the letters were printed in (I just now made the instructions a little clearer on that point).
When you do that, are all three columns honestly equal for you? Is English your native language? If so, that's impressive!
I think why your results are skewed is because people get to look over the entier list, and then can pull out associative patterns. If you realize that it is really just a one-back test, it is a lot easier.
I found column 3 slightly easier, I think possibly because of the small bit of practice the second column gave me. Didn't notice the negative priming effect.
Very interesting! I wasn't familiar with this sort of priming, and until I read the explanation I was puzzled as to why Column 3 seemed more difficult than Column 2 (didn't notice the pattern while I was doing it).
I also found column 3 a lot easier because I spotted the pattern.
I also found the practice from the second column helped my performance on the third column.
I think column 3 would prove to be much harder if one couldnt glance the lists up front (and spot any pattern). If I was given one word at a time on a screen, say, then i think i would find myself stumbling...
I also found column three marginally easier, I think mainly because of the practice. I might have been slowed down a little early on in column three by the fact that I was noticing the pattern, but once I was aware of the pattern it actually seemed to help. (The practice effect would be easy enough to control for in a real experiment, I guess.)
Are you sure you have set up column 3 right? It seems to me it would be more likely to slow down responses if the color of the font in the first word of a pair matched the named color (the thing we are supposed to be ignoring) of the succeeding word. That way we would be being primed for the wrong response. As it is, we are being primed for the right response, which ought to make the task easier, not harder (albeit primed by something we are supposed to be making an effort to ignore, which introduces scope for getting confused).
Is reading skill related to response time for either stroop or negative primes? It seems to me that one who cannot read the words will have comparable response times for all three columns, e.g., if words are presented in Polish and I cannot read Polish.
Your vote is slightly limited and therefore biased. While I found Column 2 and Column 3 more difficult than Column 1, I found no discernible difference between Column 2 and Column 3. In other words, I was perhaps not as influenced by the "negative printing". Your vote , does not provide an option for people who have the same experience as me, thus it will include only those people for whom negative printing works or people who were primed for the challenge of the Stroop Effect by Column 2 and thus found Column 3 less difficult.
When taking the test, I was proud of doing it quite well until I got stuck with the word "Orange" (printed in green) in column 3. It took me some seconds to say the right color.
Maybe that's because when I read the word orange, I think of the fruit and my mind thought of a green orange as an absurdity.
Are some words so deeply connected to notions that it can make it difficult for the mind to not think about what they represent ? And if yes, has anyone tried to make a list of those words ?
For all the doubters:
We've actually replicated the effect -- it's currently 213 to 186 in favor of Column 3 being most difficult. While your individual experience may vary for a variety of reasons, the overall effect is that the task is more difficult with negative primes.
It's a subtle effect, as the study demonstrates: It takes just 50 milliseconds longer to name the colors when negative primes are present -- that's less than a ten percent slowdown. But still, neat to see that we could replicate it in such a poorly-controlled setting.
Dave Munger :
As I mentioned before, you are subtly selecting for people who either are more susceptible to the "Stroop effect" or have alternative priming going on. Unless you factor in all the people who didn't experience the proposed effect, it is not possible to say whether whether the effect you see is coincidence, a result of your biased voting options, or really the Stroop Effect.
As you said, it is a poorly-controlled setting, but would it have been all too difficult to include a couple of more options that would really make it interesting. Alas, I must remain, humbly, a doubter.
Clearly this is a poorly controlled demonstration. No one ever claimed it was scientific. I suppose there are lots of potential explanations for our result, but it's not clear to me why people who don't experience the Stroop effect would slant the results in any direction.
The main point was to demonstrate what the experimenters are doing in the study. We clearly have shown a massive (perceived) Stroop effect, and our results are consistent with a (perceived) negative priming effect. The study scientifically demonstrates both, so whether our demo is successful or not has little to do with the validity of the study we're discussing.
Regarding your objection to a forced-choice between Column 2 and Column 3, why would our demo bias the response in a particular direction? There are lots of forced choice studies in the literature; it's an accepted practice.
Dave Munger: When you do that, are all three columns honestly equal for you? Is English your native language? If so, that's impressive!
My main difficulty this time was identifying #660066 as being "purple"; something in the vicinity of #a000a0 would have been easier. Similarly, #660000 being used for "brown" made identifying that color much more difficult. The relatively small size of the amount of color (about 2 pixel wide characters on the monitor I'm using) was also unhelpful. (Is there a size dependence as to whether color or shape determination is easier?)
I've provided odd results like that in the experiment the past, but have ADHD and am an outlier in some other cognitive measures.
Mat: I think of the fruit and my mind thought of a green orange as an absurdity.
The variety is called an Emerald Mandarin; delicious, but hard to find these days. I think they lost a lot of the trees in the mid-80s from a blight or some such.
I hate doing Stroop Interference tests. It is cognitively unpleasant to do the second to columns. I'm annoyed that I had to go and do this again for your small pool. I didn't know about the negative priming effect though which is interesting enough to make up for my irritation.
Dave Munger :
Let's leave aside the scientific validity of this demo. I'll attempt to explain why in my opinion, in this particular case the "forced choice" between Column 2 and Column 3 is causing bias.
Let's say the entire population is part of a Set U, and people who experience the Stroop Effect are Set S, and those who experience the effects of Negative Priming are Set N.
Thus, People with Choice Column 1 are :
(U - S - N + S int N),
People who choose Column 2 are : (U - N)
People who choose Column 3 are : (S + N - S int N)
With your voting choices, you're forcing a situation where
N is a subset of S, and S is a subset of U, thus forcing N = S int N artificially by this assumption - leading to an increase in the amount of people who will appear to have the Stroop Effect and a significant relative increase in the number of people who will have the effect of Negative Priming along with the Stroop Effect.
While it may truly be that N is a subset of S, and S a subset of U, a test which assumes that a priori is bound to find it in its results. I am not saying that the bias is for or against the Stroop Effect, just that the impicit assumption defeats the power of the test. I would expect that in the case of subtle phenomenon one must be a lot more careful.
Dave Munger wrote:
It takes just 50 milliseconds longer to name the colors when negative primes are present
It took a lot longer than that for me. Sometimes 3 or 4 seconds to remember the name of the colours in column 3. The only thing I can think of that might be relevant is that I'm a (self) trained rapid reader. 500wpm normal speed, can go up to 1100.
Does anyone thing that since most people who read this blog are probably in some way self identified as "science people" that we may have been looking for patterns that other people would not have looked for (in task#3)? Would a random sample of the population have been split the same way as we readers are as the the self-reported difficulty of task #2 vs. task #3? Isn't this self- selected group automatically biased from a random sample?
BTW- the whole idea of priming, positive or negative, is fascinating. I noted clear examples of priming in political commercials last fall (ex: the McCain commercial calling Obama a celebrity began with images of Paris Hilton, Brittany Spears, and other celebs that are perceived as not so lovable by the Republican base).
I found the third to be much harder for no apparent reason at the time. My strategy for these is to focus only on the last letter of each word so I'm in no position to be identifying patterns, which I'm not supposed to be doing (so I don't try to). All I know (ideally) is what the color of the last letter is, and no clue about any pattern of words/colors in the column as a whole.
I was noticeably slower with column three. For those who quickly saw the pattern in column three and therefore found it easier, I wonder: do you find math easier than most people? If so, that would make sense to me. Personally, I've no talent for math. I always had to memorize approaches, first, then get through by rote.
Let's say the entire population is part of a Set U, and people who experience the Stroop Effect are Set S, and those who experience the effects of Negative Priming are Set N.
Due to the robustness of the phenomena in question, it is more than likely that the majority of Set U members experienced both the Stroop effect (SE) and the negative priming effect (NPE). To answer the forced-choice question about which column is experienced as being more difficult, a more accurate description of the psychological experience likely involves assessing whether the NPE is experienced as being more difficult than the Stroop effect (NPE > SE), or, whether the SE is experienced as being more difficult than the nPE (SE > NPE); not so much whether you experienced the SE, OR, the NPE. In accordance with Dave's numbers @15 and the reported research, it appears that more members of SET U experienced NPE > SE than SE > NPE.
A chi-square analysis would be needed to determine statistical significance. With the reported N, 199.5 (213+186/2) persons are expected to choose either column 2 (SE) or column 3 (NPE) if participants were choosing the columns at random. Eyeballing Dave's numbers and considering the expected N for each, it looks like a small (perhaps non-significant) effect in favor of NPE > SE.
With your voting choices, you're forcing a situation where N is a subset of S, and S is a subset of U, thus forcing N = S int N artificially by this
Yes; probably additional choices were needed to address the null hypothesis. In this instance, Control = NPE = SE (i.e., 'all columns were experienced as being equally difficulty') as suggested by the commenter @3; and/or NPE = SE ('columns 2 and 3 were experienced as being equally difficult').
Pritchard and Neumann point out that young kids are generally slower to react, so when you properly account for that by taking a logarithm of their reaction times, there's no difference between the scores for children and adults.
I am skeptical about using the log manipulation to make some important characteristics of the raw data disappear. In particular, the lifespan pattern data for the NPE above, looks the same as lifespan SE data. That is, there is an established U-shaped lifespan pattern for the SE, such that children and elderly adults show the largest SE, while young adults show the smallest SE. The pattern is also the same for reading words and naming colors (Comalli et al).
What would be interesting, is to determine whether the NPE lifespan pattern looks the same as the SE lifespan pattern, by continuing the current study with older adults (i.e., 50+). If both the NPE and SE show the same U-shaped pattern, that would suggest the two phenomena have overlapping cognitive mechanisms. It should also make it more difficult to justify using log transformation data to draw conclusions.
I guess I must've subconsciously "gotten" it. Column 3 was way easier than column 2, almost equivalent to column 1.
I find that in these lists when one of the colors/words actually matches, that always throws me off for a spell (e.g. red in blue, green in green, white in purple).
On a side note, if your squint such that you cannot read the words, it's a lot easier. ;)
I found column three to be the most difficult, however there was an obvious difference between column one and the last two columns. This "test" has always fascinated me and I have seen it performed at a young age as well as performed by high school students.For those of you looking for short cuts in the columns (ex: concentrating on the final letter of the word) that is not fair. The purpose is to see the whole word and read the color. Using short cuts just create false data. The way the test is taken must be consistent with everyone else who has taken it.
I read that all of you say pratice ,practice, practice.. Practice doesnt effect the stroop effect , it isnt important how much practice you did, it is automatic process, when you see the ink of the word , you are programmed to read the ink of the word unfortunately
In my opinion, i highly conclude that using "negative priming" is much more easier than the 2nd column. I'm a 14 year old girl and according to the calculations i'm "slower" than 25 year old adults, which is true by the fact adults have more experience in life. Yet this is a mediocre article because some of you readers suggest that "negative priming" is easier, and some of you oppose that statement. Not all humanitarians think in the same way. In my conclusion, age really does effect the outcomes of each age group.
I used this in a science fair project to back up my hypothesis!!! Sweet experiment! =]