Creative ideas for visual stimuli needed.
February 21, 2007 8:14 AM   Subscribe

I need help brainstorming creative ideas for visual stimuli which vary independently on two dimensions.

I'm setting up a series of experiments (I'm a grad student in cognitive neuroscience) that probe visual representation. I need sets of (computer-generatable) stimuli which can vary along two independent dimensions.

My first attempt, using blob-like shapes didn't quite work right - as you may notice, there's less of a difference as you go down the left column as when you go down the right column — i.e., the dimensions interact.

I'd like to avoid using color (hue, specifically) as a dimension.
posted by dmd to Science & Nature (22 answers total) 3 users marked this as a favorite
 
Response by poster: Also, I'd like to be sampling a continuous space - i.e., I don't want the rows to be circle/square/triangle, the columns to be "one dot inside", "two dots inside", etc.
posted by dmd at 8:17 AM on February 21, 2007


Perhaps a series of two-dimensional slices of a three-dimensional object? This could be done in a 3D modeling program.

Design a suitably blobby object that tapers, spirals, or has changing curves and take slices of it.
posted by redteam at 8:22 AM on February 21, 2007


Response by poster: I like that concept - but how would it vary in independent ways? The obvious example (which I don't want to use, for technical reasons) is color, which can independently vary in hue, brightness, and saturation.
posted by dmd at 8:31 AM on February 21, 2007


Can you define your use of independence? This term has a specific mathematical definition.
posted by Blazecock Pileon at 8:35 AM on February 21, 2007


Also, please forgive me for not knowing the proper technical terms, but you might also be able to use an oscilloscope as a curve or blob generator.

I don't know exactly what mode it's called, but I've seen oscilloscopes that didn't display your typical repeating sine waves. Instead, it was sort of like a unit circle or something. There was a middle point, and any frequencies being measured made a distorted circle around this point.

The closest I've come to finding an illustration of this is here. If I knew the proper names, I'm sure I could find more.

Another thing: I once worked at a company that did a lot of digital video transfers. That is, they got videos from tapes, analog or digital, and captured them digitally in a computer. These videos naturally had a variety of different color levels in them. So, in order to have all of the videos in our collection have the same color levels, the technicians used as special machine to calibrate or correct or do some jargon word to the videos. Basically, they displayed a color key (or whatever you call those vertical color bars that come on when a small TV station goes off the air for the night) from the source video and they destination and they would try to make the digitized video's colors conform to destination's color bars. There was a circular graph, which I believe was very much like that oscilloscope that would have a curvy blob whose peaks represented varying color levels, and some crosses for targets. Basically, they would make the curvy peaks meet the targets and they would be happy.

The reason I told you that is because a device like that might be a good continuous blob generator. You could show this program a series of colors or some slowly changing colors and you would get a slowly changing blob. I do not remember what this color calibration method was called.

Again, I'm sorry for not knowing the proper jargon.
posted by redteam at 8:43 AM on February 21, 2007


I'm not 100% sure I know what you mean, but here are some ideas anyway:

line thickness

fill pattern

number of elements

These ideas mostly come from my time playing the game "Set". See this.
posted by amtho at 8:44 AM on February 21, 2007


Perhaps if you could give more info about what your research is on? Visual representation is a pretty broad area, yeah?

Is there, for instance, some reason you can't use rectangles, which can vary along x and y axes independently (and continuously)?
posted by logicpunk at 8:45 AM on February 21, 2007


So you don't want to use hue as a dimension even if it just serves to create the curve? In that case, if you used an oscilloscope, you could use sound. You've got amplitude, frequency, and timbre to work with. If timbre produces very wavy lines, perhaps you could smooth them somehow? Maybe a cutoff filter would have this effect.

A synthesizer, software or otherwise, would come in very handy here.

I'll try and think of other independent dimensions. You want independent dimensions that could easily be measured to continuously distort a circle, right?
posted by redteam at 8:53 AM on February 21, 2007


redteam probably means a Lissajous figure, which I considered posting as an answer, but I think it they come in discrete units like the dots example - because the curve has to join back up on itself.

I would be inclined to introduce a random stretching, a rotational element, or a radial frequency component (which would produce waves like those spreading out from a stone dropped in a pond, but with the wavelength of the waves being random).

But without knowing more about what you need it's hard to say.
posted by edd at 8:53 AM on February 21, 2007


I don't know what it is about AskMe that means as soon as I hit post I think of something else, but whilst the two frequency components of a Lissajous figure can't take any value if you want a closed curve, the phase component should be free, so that gives you one good dimension to choose if you used those.
posted by edd at 8:56 AM on February 21, 2007


I don't think the Lissajous figure is it, edd. I've definitely seen an oscilloscope that could be switched into a mode where everything would be measured circularly. There was no crossing of lines or anything.
posted by redteam at 8:57 AM on February 21, 2007


You could certainly perturb an ellipse to get the kind of thing redteam means then - you'd take the ellipse in polar coordinates and multiply r by the amplitude of a curve if it'd been going just across an oscilloscope trace. However, if you want a closed curve you still hit the fact that the component frequencies in your curve have to be discretised such that the thing joins back up on itself ok - like a violin string which can only have certain harmonics.

You do have to rotation and axis ratio of the ellipse as continuous free parameters though, which might give dmd the independent inputs needed.
posted by edd at 9:06 AM on February 21, 2007


Response by poster: logicpunk: rectangles (which vary on width and height) is a decent example of what I'm looking for, although they come with other problems. Basically, I'm trying to elicit a similarity space from the subjects - I'd ideally like a space where the perceived similarity distance between any two neighbors is the same as between any other two neighbors. The trouble with rectangles is that people will judge a rectangle of size 1x5 to be quite similar to a rectangle of size 5x1. That screws up any chance of having a 2D, Euclidean metric space to work with.

amtho: I'm Set obsessed too :) — I discovered the game a few weeks ago while researching this project. It's helped me a lot. I'd really like more continuous properties, though.

redteam: I think you're talking about Fourier Descriptors, which is what the blobs I linked to were generated with.

redteam again: no, that's a misunderstanding! the blobs I made are just one example of some visual thing that changes in two different ways. nothing says it has to be anything to do with a circle.

blazecock: Good question and one that's a core part of my research. I think I mean that I should be able to manipulate one axis without manipulating the other axis. Or, imagine there is some tool to measure the value on axis X by looking at the image. Then, altering the value on axis Y should not change the measured value on axis X. Psychological mixing of the axes is another matter. E.g., hue and saturation are technically independent axes, but they're what's called "integral" axes - people can't separate them. Changes to one have effects on behavioral responses made to the other. On the other hand, something like shape vs. shading are "separable" axes - they're independent and they stay that way.
posted by dmd at 9:07 AM on February 21, 2007


Maybe you could think of it as a carrier plus modulator. The carrier would be a simple X,Y ellipse equation, where X and Y radii are the two independent dimensions that you vary. On top of that you have some modulating function, like a sinus wave at some harmonic frequency of the carrier. It doesn't have to be sinusoidal, any kind of repetitive signal would do, just something to give it some shape and variation.
posted by Rhomboid at 9:09 AM on February 21, 2007


Blargh. Didn't preview.
posted by Rhomboid at 9:10 AM on February 21, 2007


Maybe rotated lines? Varying on angle of rotation and line length.
posted by logicpunk at 9:19 AM on February 21, 2007


Response by poster: rotated/sized lines is definitely one of the things I'm going to use. I'm also hoping for more complicated ones, though - I want to compare representations in very early visual processing (where rotated lines are ideal) to late visual processing (where I want things where the representation might be stable even over at least some differences in scaling, rotation, or retinal position).
posted by dmd at 9:25 AM on February 21, 2007


Could you maybe just try a bunch of different photoshop filters? On one axis, you could introduce noise and in the other, maybe apply blur? If hue is the problematic part of color then you could also try varying just brightness and saturation.
posted by juv3nal at 9:32 AM on February 21, 2007


Ok, I'm way off then. Sorry about that. amtho's is pretty good, then.

Maybe line thickness/fuzziness and how curvy the object is?
Maybe how many lines cross the main line - like hatching - and curviness?

If I must bring out my fantasy oscilloscope again, perhaps with sound you could induce fuzziness or line thickness with certain timbres that affected only certain frequency ranges, while all of the frequency ranges could curve as a result of changing amplitude.

Also, I recommend downloading Winamp and playing music through their "Advanced Visualization Studio" visualization plugin. There are hundreds of presets in existence and all of their designers are trying to outdo each other in how they can represent sound. Maybe you could find some inspiration there. Maybe if you saw something you liked you could think of other dimensions besides sound to create them.
posted by redteam at 9:32 AM on February 21, 2007


Besides line thickness, maybe you could also use some kind of "porosity" as well? Think of a sharpie marker on an old sidewalk vs. a sharpie on a sheet of paper.

Also, I was thinking about how very high contrasts in fuzzy areas might or might not be seen because of our sensitivities to spatial frequencies and I ran into this beautiful page.
posted by redteam at 9:51 AM on February 21, 2007


Can you just "twist" (in 3D) the shapes you have? You can have the twist amount vary from 0 to some predefined level (and you can cap this if you don't want to introduce occlusions / cut-outs).

Some shapes twist much better than others (i.e., circles don't twist as well as non-mirrored shapes), but the shapes you have are certainly amenable to this.
posted by zpousman at 12:14 PM on February 21, 2007


Could maybe just filling your blob surfaces (at a constant luminance) make their shapes more discriminable? Just a small idea.
posted by randomDirtPattern at 4:45 AM on February 24, 2007


« Older Can I make my own delicious Aquafina water at home...   |   checklist for life practicalities? Newer »
This thread is closed to new comments.