Perception
From Visualization Sp06
Lecture on Jan 31, 2006
Readings
- Perception in visualization. Healey. (html)
- Graphical perception. Cleveland & McGill. (jstor)
- Toward a perceptual science of multidimensional data visualization: Bertin and beyond. Green. (handout)
- Chapter 3: Layering and Separation, In Envisioning Information. Tufte.
Optional Readings
- Gestalt and composition. In Course #13, SIGGRAPH 2002. Durand. (1-up pdf) (6-up pdf)
- The psychophysics of sensory function. Stevens. (pdf)
Bryan - Jan 31, 2006 10:32:25 am
I found the discussion of "preattentive" processing from the first reading very interesting. It's something I've had a hunch about for years, having logged thousands of hours playing video games. Games always contain an interface of somesort, and often these interfaces are required to describe dozens of floating point numbers on the screen simultaneously. The success with which these quantities are visualized has a huge impact on the usability of the game. Some will just show these numbers as digits on the screen, others use partially filled bars, intensities of color or light. The clever systems seem to tap into this "preattentive" stage so that I can assess literally in a glance the state of some important game factor.
Another interesting aspect is bringing attention to change. Games often do a great job of this by flashing colors, noise, or motion in the display. The combination of change notification and efficient encoding has allowed the best of modern games to deliver huge amounts of simultaneous information in a way that can almost instantly be processed and acted upon by viewers.
Noaa - Jan 31, 2006 10:12:37 pm
In Healey's "Perception in Visualization" there's a java applet under Feature Heirarchy that's supposed to act like the applet we saw in class but instead of present/absent displays vertical or horizontal boundaries between objects. I turned on the interference and could still process the boundaries preattentively (at < 250 ms). I was wondering if the point of the applet was that the interference made the preattentive processing impossible, since making it more "difficult" seems unapplicable since preattentive shouldn't have any processing involved which could be classified as easy or difficult. So what was the point of that applet?
Maneesh - Feb 01, 2006 09:55:22 am
Noaa - I too am able to see the vertical/horizontal divisions preattentively. However it is a bit more difficult with the interference. Interestingly luminance seems to produce less interference with hue than the interference that hue produces with shape.
Todd - Feb 01, 2006 07:15:29 pm
Robert Kosara has done work on visualization techniques involving blur as a means for depicting which things are less relevant. This is very relevant to today's discussion because he claims that blur is preattentive. I find this claim to be highly plausible. If you are given a screen full of objects, most of which are blurred, and your task is to find the object that is not blurred, you will be able to find the non-blurred object very quickly.
Here is a link to his paper. http://www.kosara.net/papers/Kosara_InfoVis_2001.pdf
Yi-Tao - Feb 06, 2006 12:35:05 pm
Noaa - From the text, Healey is saying that some preattentive elements (like shape) is masked by other elements (like color). I don't think the applet is a good demo; I noticed that I was focusing on the element that was under consideration. The randomness of the interference prevents me from focusing on it. Maybe if the applet had the interference and the real element going in opposite direction (i.e., color is vertical and shape is horizontal), the applet would have been a better demo of interference.
Mehershad - Feb 06, 2006 10:34:19 pm
From the readings and lecture, the most interesting facts were about preattention. Personally, I have used luminance as a method to draw attention to more important elements in the past as I have felt it effectively leaves out the data which does not need attention. The interesting element was the number of variables which vary to draw attention. It seems that varying the right variable and the right number of variables is key. The example from the lecture where the color is changed acts as the variable and counts as 1. In that case, it was fairly easy to catch. In other cases where the shape and color were changed it become increasing harder to catch the difference. Thus, keeping the number of variables down to minimum help in quickly identifying differences but the type of variable also plays a major role. From examples encountered it would seem color does better than shape in most cases but what about those who are color blind ;). Thus, it is upto the graphician to make the appropriate calls on who is the target audience and accordingly design.
Raymond - Feb 07, 2006 01:19:59 pm
Bryan- Your comments on preattention in video games are very interesting. I believe that many recent games that I have played have been able to portray more realistic graphics but at the same time still let the player know easily which objet you can or cannot interact with. This includes uses of color, halos, contaiments, etc. I think it would be also very interesting to compare the visualization techniques used in games and see how they compare to Mackinglay's or someone else's rankings - i.e. are the highest rankings used the most, and are they even the most effective?
Nchentan - Feb 07, 2006 05:38:20 pm
In the discussion about Figure/Ground I somewhat disagree with the Principle of surrounddedness and Principle of relative size. It seems to me that people's past experience have a more prominent effect on which of the figures "pop out" first. For example, the "TIE" figure, to a person who do not know English at all, TIE will probably never be considered forground, but to others, it can be a reasonable foreground, even so, I have over heard several people mentioned that they did not see the "TIE" until it was explicitly stated in class.
Brien - Feb 08, 2006 01:45:58 pm
I think framed rectangles in Graphical Perception work well. It would be interesting to see them used on a "vertical pie chart" -- the bar chart where each bar is a percentage of the whole. Healey's query visualizer app. (video linked from paper) gets across the effects of motion on perception but is kind of overkill. It looks like it has six dimensions of data coming onto the screen. A simple radial layout in 2D might work just as well. Healey's texton example shows two textons with the same size, shape, and num. terminators. I'd like to see the effects of varying these parameters on preattentiveness. I like those classifications.
