I'm not entirely sure, I haven't looked at it too much. There are a lot of global things that are one way only, so I don't know specifically if you can use different arrows for different research.
Basically the way it works is like this:
You have your .tga file with a bunch of stuff thrown in (I forget which one the arrows are in). You have your .brush file that says "Hey, take this box of pixels from this file, and call it ResearchArrowRight1", and then you have your .window file that says "Oh, we have ResearchArrowRight1 now, here's where it goes".
Check it out sometime, it's not too bad to figure out. The .brush file specifies:
- name given to the object
- type (like, for buttons, there's normal, disabled, clicked, hovered over or something like that)
- .tga file to use
- pixel box within the .tga to use for the object.
The pixelbox is basically pixel coords for the image chunk to use, in the [x,y,i,k] format. X,Y is the top left corner as the starting point, and i,k are basically length and width of the box in pixels.