Wednesday, August 20, 2008

100th anniversary of “A New Visual Illusion of Direction”

What to notice: The letters in the word “LIFE” appear to tilt left and right. The letters are actually vertical, even though they are made up of little tilted line segments.

Press the button to put red vertical lines on the display. This way you can convince yourself that the letters are indeed aligned.

Brief Comment: The image is my reconstruction of Figure 1 from “A New Visual Illusion of Direction,” written by James Fraser in 1908.

The 100th anniversary of Fraser’s paper is worth commemorating. Many of the illustrations in the paper—like the one above—are a staple in books on illusions.

Fraser worked with two strands of fiber, one black and one white. When the two fibers are twisted together, the resulting cord looks like a series of black and white line segments, all inclined at a similar angle. Fraser referred to the line segments as “units of direction” that could make lines appear to tilt one way or the other (see the image above, for example).

The “units of direction” can also make a collection of circles appear as a spiral. In the example below (from the original paper), the image looks like a spiral, but if you click on the button to place red lines on top of the twisted cords, you can see that the image is composed of individual twisted cords that form circles.

The important point is that the twisted cords can be thought of as a global object (i.e., lines and letters) composed of local features (the line segments). The illusion occurs because the visual system receives different stories from these two sources of information: in the word “LIFE,” the global object (a series of lines) says straight, while the local features (line segments) say tilt. The visual system must create a reasonable percept from the conflicting stories.

There have been many studies that have examined how a global percept is influenced by local features. One of the most revealing is Michael Morgan and Bernard Moulden’s 1986 paper, “The Munsterburg Illusion and twisted cords,” published in Vision Research. Morgan and Moulden digitally filtered a twisted cord image to produce a new image in which the tilt of the line is physically present. That is, if you remove some of the information from the original image, you can measure the tilt in the new image with a ruler.

In some ways, then, twisted cord displays disagree with “reality” only if we are tied to the idea that the line (or the circles) are what is important for vision. If you want to read more about this, I strongly recommend Michael Morgan’s chapter on visual illusions in the book Unsolved Mysteries of the Mind, edited by Vicki Bruce (here is a Google sample from the chapter).

A complete pdf of Fraser’s original paper can be downloaded at this link (the pdf file is 3.5 Mb).

Thursday, August 7, 2008

Thin lines can stop the perception of "winking"

Here is an illusion from Shapiro, Charles, and Shear-Heyman, “Visual illusions based on single-field contrast asynchronies”

What to notice: You are looking at two rectangles that change from dark blue to bright yellow and back again. The colors of the rectangles are always identical to each other; that is, they blink together "yellow-blue-yellow-blue, etc." But the rectangles look as if they are not modulating together; rather, they appear to "wink" asynchronously.

If you wish to convince yourself that the two rectangles have the same color, click on and drag one rectangle to place it next to the other rectangle.

Click on the button to add thin bars on the top and bottom of the rectangles. When the bars are present, the perception of alternation disappears—the rectangles appear to modulate together.

Comments: In my first post on this blog, I presented an illusion similar to this one (the contrast asynchrony), so—you may be wondering—how is this illusion different from that one? The important point from this post is to show how little it takes to kill the perception of alternation.

The rectangles appear to modulate in phase as soon as the thin bars are added to the top and bottom of the rectangles. The lever on the side of the display allows you to adjust the length of the thin bars. See how small you can make the bars and still have the rectangles appear to modulate together. In the experiments conducted in my lab (link to abstract), the rectangles always appeared to modulate together once the bars were greater than 10 minutes of visual angle, with some effect occurring at 1 minute of visual angle (that is pretty small).

The illusion “works” because it juxtaposes two sources of information: the color of the rectangles, and the color contrast of the rectangles relative to the background. The color information modulates together, but the color contrast alternates. The effect is surprising because we don’t always take into account that our visual system can process both color and color contrast. The fact that thin edges can effectively shut down the perception of contrast is curious and will be the subject of some future posts.

Lastly, I would like to point out that the contrast asynchrony display is fundamentally different from “dynamic simultaneous contrast” displays, where the appearance of the center patch changes due to induction from the surround field. I will leave this as an open topic to return to later, but I discuss the differences between the two types of displays in the introduction to my 2008 paper “Separating color from color contrast.”

/*********google analytics*/ /*********google analytics end */