Perception
From CS294-10 Visualization Fa08
Lecture on Sep 10, 2008
Readings
- Perception in visualization. Healey. (html)
- Graphical perception. Cleveland & McGill. (jstor)
- 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)
Matt Gedigian - Sep 10, 2008 02:13:58 am
The jstor link didn't work for me, which article is it? Corrected now.
Ketrina Yim - Sep 10, 2008 04:23:14 pm
After reading "Perception in Visualization" and being in lecture, I was left with a couple questions.
First, is it possible to improve one's preattentiveness? Sure, different people have different preattentive processing times, but can someone actually do training (such as running the first applet for a few minutes every day--not that I would do that...) to pick up on preattentive details in a visualization faster?
Second, what exactly does "artistic properties" (see Table 1) mean?
In general, "Perception in Visualization" was quite an intriguing read. Aside from color and possibly shape, I never really thought much about what made items stand out in an image or chart before or why they did so. And learning about change blindness has left me wondering why I had so much difficulty seeing the not-so-subtle differences between the images in the movies, particularly since I've always thought of myself as being quite attentive to detail. It might have been partially due to the gray blank that flashed between the images, since spotting the difference seemed a bit easier when I managed to get the two images to show one immediately after the other in the Quicktime movies (though there was no way I could have noticed the change on the first cycle).
And does anyone else think that Tufte spends rather a lot of time saying "no" to dark grid lines?
Maneesh Agrawala - Sep 13, 2008 12:58:34 pm
Matt: I've fixed the link to the paper by Cleveland and McGill. It is the second paper you found that was published in JASA.
Ketrina: I'm pretty sure you can improve pre-attentiveness with training.
Calvin Ardi - Sep 15, 2008 04:59:14 am
@Ketrina Yes; something he emphasizes quite a bit. In particular, the sheet music example was good, but I feel the opposite is true for blank sheet music. For example, composing pieces or working on music theory would probably be done with a pencil. It's a bit easier, in my opinion, to be able to contrast the lead shade with black lines (even though the reverse would be easier to distinguish notes).
Michael So - Sep 16, 2008 01:16:35 am
I found the topic of Change Blindness to be pretty interesting in the "Perception of Visualization" reading. It is surprising how hard the intermediate gray screen makes it to detect the significant change between the two images. Without the gray screen, the change in difficulty is dramatic i.e. the change "pops out" at you. I am not sure how this interesting phenomena will help me make better visualizations. Perhaps it's the awareness that an "interruption" (such as the gray screen) would render my audience "blind" to the information I want displayed (such as significant change in an image compared to the original).
In the Tufte reading, I do like his revised visualizations as well as the ones he deemed good examples of layering and separation. Especially when compared to the visualizations that are examples of not layering and separating the data. One of the main things I will keep in mind are muted grid lines. I believe they help "pop out" the information that you are trying to display. I never really thought that dull or muted backgrounds could help avoid clutter of the data and improve cognition.
Chris - Sep 16, 2008 09:13:58 pm
Piling on to the discussion on pre-attentiveness from "Perception in Visualization," I was somewhat surprised to see that no one has mentioned synesthesia. Synesthesia is a defect (or mutant power, depending on how you want to look at it) whereby there is a pre-attentive link between stimuli. The classic example of this is that some people associate color with shapes. Shapes which, to ordinary people, would require attentive processing to distinguish appear pre-attentively distinct to some synesthetes. An illustration of a test for synesthesia can be found here. In this example, there are a number of Ses, some of which are reversed. Apparently, some synesthetes are able to pre-attentively pick out the reversed Ses.
Also, on the topic of sheet music, I agree Tufte is over-extending himself in this example. In my experience (which may be skewed because I was a Suzuki kid), one rarely reads music in in the same way that one reads text. For me, there are two phases to reading music. In the first phase, while learning something new, I'll go through the sheet music extremely attentively to learn what the notes are. In the second phase, when practicing or performing once I have (basically) memorized what the notes are, the sheet music is present only to assist my memory (e.g, I'll remember what notes to play by their vague outline and their location on the page - a semi-pre-attentive process). Because the important phase (the first phase) is extremely attentive, I think that any tweaking of visualization of sheet music is of limited utility. Further, a lot of scores (the Stravinsky in the book, for example), are very messy, so if you don't already have a pretty good idea of what the notes are beforehand, you have pretty much no chance of getting them right on the fly.
Sarah Van Wart - Sep 17, 2008 07:48:03 am
A comment on the reading: One of the things that I found particularly fascinating about Tufte’s Layering and Separation chapter was the fact that I could relate to many of the mistakes that were disected therein. In particular, I’ve been guilty of "data emprisonment" by overusing vertical and horizontal gridlines when displaying tabular data on the web. In fact, I find it interesting that in many software applications (Word, Excel, Visual Studio.NET, etc.), most of the auto formatting options that one can select heavily use gridlines for each cell in a table. Even though such options are designed to be generic and cannot be faulted for my poor formatting, perhaps this auto formatting model has inadvertently trained me to present data in a particular (and less effective) way. Given the additional clarity that selective shading and highlighting created, I will definitely be much more mindful of table formatting in the future. The medical receipt visualization was particularly impressive – it made a tedious receipt tell a story through clever gridline placement and annotation positioning.
Matt Gedigian - Sep 17, 2008 01:31:49 pm
@Michael It's amazing how easy it is to detect changes when there is no mask. Having separation between stimuli is the general case. It's only when you have them presented with no mask that your brain gets to cheat and use motion-detecting cells. That capacity evolved for useful things like avoiding predators and here we are wasting them on psych experiments.
I think one lesson to take from it isn't about the change blindness per se, but the secondary effect of how knowing what has changed forces you to see it. (This occurs with other visual phenomena too). The danger is that, the designer of a visualization has that knowledge and therefore does not experience it the same way a naive viewer would.
Yuta Morimoto
In the context of "Perception in visualization", there are many insight into preattentive and theories of preattentive processing. I am very interested in the concept of preattentive and it is significant aspect of our perception. But, how can we make use of it? When we make a figure, how to apply preattentive to it? I think that we can apply preattentive to limited simple visualized images such as a chart or graph and they work fine. Some complicated visualization are using many visual features. if we want to use preattentive to facilitate understanding of visualization, we need to decompose feature of visualization. In that situation, we can detect a feature that cause interference with other, but we can not always replace them with other feature to take advantage of preattentive. Because, replacing a visual feature may cause other interference with itself.
Jeff Bowman - Sep 17, 2008 02:44:19 pm
@Maneesh and @Ketrina: This brings up the obvious question: The preattentive variables studied seem to be universal among the studied humans. However, if you could improve the preattentive attributes through practice and studying, this brings up the "nature vs nurture" question, somewhat: Are we predisposed through evolution to sense these things, or are we predisposed through Sesame Street and stop signs and other cultural "norms"?
I remember in studying the eye that there are particular motion-sensitive cells, and I know that color is a very prominent part of the retina and visual cortex. However, if we grew up in a society where color were not as prominent, would we still have the same disposition mentioned in the first reading to favor color over shape?
I, personally, would be interested in a study about perception across different geographic study areas—Latin America, South America, Africa, Asia, and such. While perception is more-or-less universal in the homogenized First World and our common lexicon of signs and symbols. Perhaps we would be more disposed to shape if kindergarten gave us more shape blocks than color blocks, or if red lights and green lights were replaced by lights of different flashing intervals.
Who knows.
Seth Horrigan - Oct 31, 2008 01:02:04 pm
Did the experiment instead.
