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 why? The space defined by the irregular vertical lines is equal in both cases. The only difference between the two figures is which direction the "pointy" curves face and which direction the convex, "smooth" curves face. Yet nearly everyone agrees that areas defined by the convex curves (like those surrounding a above) are shapes, and other areas are background.
This principle, of convex curves denoting "shapes" and not "background," has been known for decades. It's one of dozens of Gestalt rules for determining what parts of the things we see go together to form shapes and what constitutes the background -- what's the figure and what's the ground. These rules can also depend on what we're interested in. Consider the view out my office window:
I might be looking at this scene because I want to turn on my lamp, in which case I'd be primarily interested in separating the shape of the lamp from my window and the trees outside. But I might want to know whether it's raining, so I'd be more interested in what's outside the window. Or I might be thinking about buying new blinds for the window, in which case I'd be looking at the slats of my blinds and how much light they do or don't allow in from outside.
It's a complicated picture, and our visual system easily breaks it into its components based on rules that have been sorted out by visual psychologists over the past century. But what is less certain is exactly how the visual system applies these rules. Do we have to be consciously thinking about what part of the picture we'd like to see? Or are some of these rules automatically applied, without us even noticing?
Ruth Kimchi and Mary Peterson showed 46 undergraduates pictures like the ones at the beginning of the post, only instead of a or b, the pictures contained grids of randomly-arranged black and white squares, like this:
These grids were shrunk down to tiny-size and then placed in front of the backgrounds with the columns/spaces, like this:
These pictures flashed by quickly (in about a half-second), and viewers had to indicate as rapidly as possible whether any of the squares changed from black to white or white to black. This was repeated dozens of times. Meanwhile, the background image -- those curvy lines defining columns or spaces between columns -- were being changed systematically between 40 different images. Sometimes the area around the grid was a column, and sometimes it was a space (as defined by the orientation of the convex curves).
So, did the figure - versus - ground distinction have any impact on the responses to the grid task? Yes, in a very interesting way. When the grids were the same, viewers were better at identifying them when the backdrop maintained the same organization of columns and spaces (remember, the backdrop always changed -- it's just that sometimes the location of the columns and spaces changed, and sometimes they stayed in the same location). When the grids changed, viewers were better identifying them when the backdrop changed. This graph summarizes the results:
Inverse efficiency is a combined measure of speed and accuracy, where higher scores are worse. So when both the backdrop and the grid changed, then viewers were better at identifying changes in the grid. When they both stayed the same, viewers were better at identifying the fact that the grid stayed the same. Yet at the end of the study, when viewers were surprised with a question about whether the last grid they saw was in a shape or the space between shapes, they couldn't answer accurately. They also couldn't accurately tell whether the organization of the backdrop had changed the last time they had seen it.
In a separate experiment where viewers were only asked about the backdrop, they could easily identify its orientation and whether it had changed.
Kimchi and Peterson say this demonstrates that we don't have to be paying attention in order to process the difference between figure and ground -- at least in this case, it's not a process that we're consciously aware of. They do point out that there are many other ways people determine the difference between the object and its background, like continuity of color, familiarity, and width of an object's base. Different methods may require different levels of attention. But in this case, attention clearly isn't needed.
Ruth Kimchi, Mary A. Peterson (2008). Figure-Ground Segmentation Can Occur Without Attention Psychological Science, 19 (7), 660-668 DOI: 10.1111/j.1467-9280.2008.02140.x
> our visual system easily breaks it into its components based on rules that have been sorted out by visual psychologists over the past century.
Is our visual system actually *based* on those rules, or are those rules a way we have of dsecribing some other (more complex or more fundamental) process.
If it is based on those rules, then how are the rulesencoded?
Personally, I don't see why one of them has to be columns and the other as background. I just see them as abstract shapes, or as "torn strips of paper".
I didn't consciously perceive them as shapes or columns either, until after the description in the text made me look again. After that, yes, it was much easier to perceive the areas enclosed by convex curves as coherent shapes.
Of course, based on the results of the study, maybe I unconsciously perceived shapes from the start!
But did you perceive them as "torn strips of paper" lying on another sheet? Or shapes on a background? Unless you just saw them as lines, you did see them as shapes of some sort. And then the question is: which is on a "column" and which is between them? Could you answer that? If you could, you were seeing some as things on a background... I suppose the question forces the perception, but it doesn't force *which* ones you see as background or shape.
Isn't it possible that the subjects were reacting, not to figure-ground relationships, but to the directionality of the intervening lines? It seems to me that there are several ways to interpret this data.
Is it possible to compare the attention-level related to reading figure-ground relationships to the attention required for recognizing words & sentences?
I'm going from the fact that when we read, we very quickly look at only parts of words and sentences and our minds automatically fill in the rest from memory.
I'm not sure that this is the same process though it seems like a reasonable assumption.