Colors can tell us a lot about how we recognize shapes

The Beck effect is difficult to replicate online, because it involves testing reaction times. However, I think I've figured out a way to approximate the effect. This movie (Quicktime required) will show you how it works. Just follow the directions on the opening screen:

Now, which letter did you see first? Let's make this a poll:

If we manage to replicate the effect, there should be a bias in the results, which I'll explain below so everyone has a chance to try it out before learning the "answer."

In the 1960s, Jacob Beck found that when searching arrays of letters like this, viewers were faster to find the Ts tilted on their side than the standard Ls. Yet when asked simply to identify a sideways T or upright L by itself, viewers are equally fast for each letter. If we've successfully replicated it, more people should answer "T" than "L."

Interestingly, David Navon and Ruth Kimchi have found that the effect persists even when just two objects are being compared: We identify a sideways T next to an upright T faster than an L next to an upright T.

So what about identifying the T or L in when accompanied by upright Ts is different from identifying them alone? A couple of explanations have been proposed.

One explanation posits that when we're looking at a field of Ts, we're more attentive to similar shapes, so we notice the tilted Ts more rapidly. Another explanation argues that sideways Ts are actually less similar to upright Ts: since the Ls "blend in" we take longer to locate them.

Navon and Kimchi noted that both of these explanations rely on the specific shape of the letters. What if they could replicate the Beck effect without relying on shapes at all? In a 2004 study, they did just that. Instead of letters, they used colors. Viewers were shown four patches of color; three were always dark green, while the fourth could be either light green, dark green, or brown. They were told to press a corresponding button indicating which color the fourth patch was, and reaction times were measured. As a control they were also tested on single patches of each color. Here are the results:

i-c85871965181d7c0c3ed3831ee314bff-Navon.gif

When the colors were displayed by themselves, there was no significant difference in reaction time, but when they were shown among three dark green patches, the light green patches were identified significantly faster. Thus, Navon and Kimchi argue, they have demonstrated that the Beck effect applies not only to shapes, but also to colors. This makes both explanations of how the effect works problematic.

The researchers argue that the real source of the effect might be in the difference between judging between multiple possibilities and a single possibility. An analogy might be pregnancy: "Are you pregnant or not?" is an easier question to answer than "are you pregnant with a single fetus or twins?"

Navon, D. & Kimchi, R. The Beck effect is back, now in color: A demonstration. Psychonomic Bulletin & Review, 11(1), 98-103.

More like this

[This entry was originally posted in April 2007] The Beck effect is difficult to replicate online, because it involves testing reaction times. However, I think I've figured out a way to approximate the effect. This movie (Quicktime required) will show you how it works. Just follow the directions on…
Take a look at these pictures. Each picture depicts four shapes -- irregular vertical columns spanning the height of the picture. It's easy to tell which letter is on a column and which is not, right? If our readers are typical, over 90 percent would agree that a is on a column and b is not. But…
The other day, our car wouldn't start and Jim had to ask a neighbor over to help him jump-start it. There was much rushing in and out of the house looking for flashlights and other tools to help get the job done. After the neighbor left, Jim wanted to drive somewhere and couldn't find the keys.…
All of you are probably familiar with color opponency, but just in case, I'll give you a quick refresher. I'll even start with the history. In the 19th century, there were two competing theories of color vision. The first was the Young-Helmholtz theory (sometimes called the trichromatic theory),…

The "L" in the quicktime movie is a sharper font than the sideways "T", which looks as smooth as the other "T"s. That might affect the results in an undesired way.

Yes, I noticed that. I could have made schematic Ts and Ls using identical lines (I'm pretty sure this is how the original study worked), but I got lazy. I'm not submitting this for publication, just trying to give a rough idea of how the effect works.

While I noticed the sideways T first, I know I was scanning the image like I usually read, left to right top to bottom. Can the results be duplicated if the L shows up before the sideways T assuming regular scanning?

Beth:

Yes -- in the original study the placement of the Ts and Ls was completely random. Since I had just a single trial I put the L in about the spot of the last instruction, and the T at the top/middle of the page. Maybe I should switch them after we get a hundred or so responses and see how we do then...

interesting, I think I saw the L first, as the last thing I was looking at before it started was the play button of the quicktime player. I then scanned up the image, across the L to the sideways T.

Do you think being left- or right-handed would affect the results?

Okay, the vote is now 49-11 in favor of Ts. I have replaced the movie with one which switches both the place of the letters and the order of the instructions. We'll see if this changes the results.

I did what Peter did, too.

By Speedwell (not verified) on 24 Apr 2007 #permalink

Knowing that I was not reading but scanning an array, I scanned from the top left downward and then up, then down, etc. I saw that sideways T first. Switching places might bring interesting results.

Perhaps we see the sideways T first because we notice the longer horizontal part of that T in the midst of other Ts with longer vertical parts. The L almost becomes camouflaged, then.

I was looking pretty much center of mass, as it were (speedread/skim standard technique) and the T caught my eye first.

-Mecha

I didn't scan. I sort of "defocused," and the sideways T popped right out.

By Scott Belyea (not verified) on 24 Apr 2007 #permalink

I also saw the T first. The question about location seemed relevant, I saw the L on a second look but wondered if how I scanned the image might have mattered.

I photoshopped the image and added an upside down T, I think it's also harder to spot, about like the L. The line weight difference in the L didn't seem to matter much, tried that too.

Another observation; the array of T's creates the illusion of a horizontal white banding between the top and bottom of adjacent T's. The L doesn't disrupt this as much as the sideways T, making the sideways T, more different and therefore more visible

Just occurred to me, this test might be considered a variation on the camouflage test with the spotted dog.

I saw the T first, but I didn't respond to the poll, since I'm pretty sure I clicked play after reading the last line, and my eyes had not saccaded away yet before the letters appear. Maybe there should be a fixation cross before the letters appear?

Now the T and the L have each been in each position, and the T clearly is dramatically outpolling the L in both spots. It was 49 to 11 with the T in the upper right and the L in the middle, and as of now it's 128 to 16 with the positions reversed.

Aaron's point about the T being a bit blurrier might have something to do with our result, but I think we can safely say we've replicated the Beck effect here.

oddly enough, i saw the L first. i first started scanning the beginning letters, aquainting myself with the format, de-focused and noticed both at the same time because of banding and black/white patterns. but my eyes went to the L instead of the T. couldn't tell you why tho

As I scanned the image I noticed there were two anomalies in the image. It wasn't until a closer inspection that I noticed which was which. I did see the random 'L' first, however, I also saw the 'sideways T' before I realized what either letter actually was. I went back to the randomly placed 'L' for identification before returning to the 'sideways T' for identification. I now write that off to the fact that modern written English is my first language, because the 'L' was randomized closer to the top left. Long story short, I recognized both as an anomaly before recognizing the 'L' as the first specific detail. It is quite interesting that the ratio is so off center (198 to 33, T's to L's)

This is an excellent entry, and your piece of interactivity made it especially enjoyable.

A friendly suggestion: I bet some scripting guru could whip up something that would give everyone a random placement. Just wish I was capable :)

Still, very well done! I'm not likely to forget the Beck Effect anytime soon.

I'd say that the main reason is that the last line of the instruction is where my eye fixated the most. So when I clicked the Play button the sideways T automatically became more protruding to me (it's closer to the last line of the instructions shown initially).

People! BOTH letters were displayed in BOTH locations. We got the same result BOTH times. The position of the letter clearly doesn't affect the results.

The 'T' is more of a break in the pattern. Especially if you de-focus. This could also apply to the colours. the brown being more of a departure from the pattern than the light green.

I saw the L first, and in what felt like a blink of an eye. Why? I started at the top and did the first 6 or so rows in one L to R scan. left. I guess I wonder if this "means" anything worth noting (about the way I think/perceive/etc.)?

After I saw how dramatically the results favored seeing the T first, I tried it again, taking a step back, and saw immediately that indeed, the sideways T is more legible in the sea of Ts than the L -- in other words, I agreed with the bias being logical wrt visual perception, even though it wasn't the case with me.

I saw the L first, and in what felt like a blink of an eye. Why? I started at the top and did the first 6 or so rows in one L to R scan. left. I guess I wonder if this "means" anything worth noting (about the way I think/perceive/etc.)?

After I saw how dramatically the results favored seeing the T first, I tried it again, taking a step back, and saw immediately that indeed, the sideways T is more legible in the sea of Ts than the L -- in other words, I agreed with the bias being logical wrt visual perception, even though it wasn't the case with me.

This was interesting. I wonder if having the play button at the top of the page would affect the results? I looked at the bottom to click the play button and then my eyes moved up the page. The sideways T is near the bottom while the L is near the top.

I only suggest this because in journalism I was taught that when reading the front page, the path the eye travels forms the shape of a "z". This pattern determines how newspapers layout the front page in regard to photos and articles.

If this is true, then it may be true for any text that the eye follows a particular path, perhaps determined by what it is looking for.

By Dixie Girl (not verified) on 01 May 2007 #permalink

There doesn't appear to be anything knew about the Navon and Kimchi (2004) study, nor their conclusion that the Beck effect can possibly be explained by judging between single and multiple possibilities. Such a conclusion was reached some time ago in a study about visual processing conducted by Treisman, Sykes, and Gelade (1977).In fact, the Treisman et al. study appears to be a cross between the Beck effect and Navon and Kimchi's (2004) work.

In this study, it was first established that a single red T "popped out" amongst an array of green T's. Further, it was established that a single red S popped out amongst an array of red T's. Thus people can find items quickly when they differ from other surrounding items in terms of shape and color.

The additional finding in Treisman et al.'s study that supports Navon and Kimchi's (2004) explanation of single vs. multiple possibilities, is when participants were asked to find a red T in an array consisting of 50% green T's and 50% red S's. The only way they could find the red T, is through examining BOTH shape and color. Not suprisingly, this experiment showed that this type of (conjunction) search takes a longer time-- because of the requirement to identify multiple aspects of an item for a successful search.

Furthermore, Treisman and Gormican's (1988) feature detection/integration theory suggests that objects have primitive features that are automatically processed by the visual system. Among these features are slant, curvature, length, brightness, and color. The fact that most people saw the "slanted" T first and not the "L" (which is not slanted), can be explained by Treisman's theory. In the example of the Beck effect provided here, the number of persons reporting seeing the (slanted) T first, and not the (non-slanted) L, is probably due to a faster primitive feature detection. The number of persons reporting the L would likely be the same as those reporting the T if both items were in the same (sideways) orientation.

References

Treisman, A.M., Sykes, M., & Gelade, G. (1977). Selective attention and stimulus integration. In S. Dornic (ed.), Attention and Performance VI (pp. 333-361). Hillsdale, NJ: Lawrence Erlbaum.

Treisman, A., & Gormican, S. (1988). Feature analysis in early vision: Evidence from search asymmetries. Psychological Review, 95, 15-48.

By Tony Jeremiah (not verified) on 01 May 2007 #permalink

I think i know just why the horizontal T was more easily spotted by majorty of the people.

As the T is placed horizontally, the the longest stem of the vertical T becomes hoqizontal. As the central stem of the T is longer than the other two( purpendicular to the central stem), the anomaly is clearly visible at an exhaustive view of the picture. In the case of the L, it almost looks like an inverted vertical half of T. I'm sure that if it were a D or R. T wouldn't have been first. Try it.

But as the case stands, there were some people who saw the L first. This can be explained by the fact that they did not take an instant complete view of the picture but scanned it part by part; so the ones who started from top got the L first.