Images on personal computers - Which aspect of size to keep constant?

Question

For various reasons we're doing a reaction time study involving images that runs on unstandardized equipment: patients' personal computers (unless they don't meet certain requirements).
It's not about vision or individual differences, but group-level intraindividual change in reaction times (attention retraining), so it's not that important that participants all have the same experience as long as it's not confounded with the experimental condition.

Most reaction time studies run on standardised equipment, with a standardised distance from the monitor etc.

We can at most ask participants to keep a certain distance to the screen, but we can't ask them to buy new monitors of course.

So, three basic options arise (assuming our images are in a sufficiently high resolution):

1. Use constant number of pixels.
   Positive: We can use the same images for everyone, if we use a medium number of pixels, we can accomodate most resolutions. Also, potentially everyone sees the same amount of information (i.e. no information lost due to down-scaling).
   Negative: With high resolutions, images may look pretty small, and with large monitors, participants may usually position themselves further away from the monitor habitually, so some see "less information".

2. Use constant display size (i.e. 7 cm wide).
   Positive: On large and small monitors with different resolutions, images are the same size.
   Negative: Apparently the information used to compute actual display size can be unreliable, so we may end up with different display sizes. We'll probably need different images and people with larger monitors and higher resolutions will be able to see more details.

3. Use constant fraction of display (i.e. 70% percent wide).
   Positive: Fraction of display may do the best job of letting people perceive considering that people probably sit further away from bigger monitors.
   Negative: Need images in different sizes (dynamically computed with JS). They may see different levels of detail.

These options could be combined (i.e. constant pixels unless that is less than 50% of the monitor) and there's also the option of letting users zoom (will result in somewhat blurry images though).

Is there any research into this matter and if not, which option sounds best perception-wise?

Answer 1

I don't think that (2) works. The physical size of the screen is not available to you. You may be able to calculate the dimensions in pixels, but pixel density (dpi), pixel format and sampling pattern change from screen to screen. So you won't be able to calculate the centimeters from the pixels.

I would do what web designers have done for centuries: rely on the user to resize content to their needs.

I would provide images in high resolution, display them at an initial size* that works well on the most common screen resolutions (1366 x 768 and 1024 x 768), and provide a highly visible request to resize the image to make viewing comfortable. With Javascript you can offer fluid, continuous resizing options easily (e.g. TwinHelix or, what I prefer, YUI), allowing the users to resize web content in the same way they would resize the browser window. Or you can offer a number of preset sizes (+/- 10% etc.) and link these to different style sheets.

*A good measure for the initial size, in my opinion, is the em. You can rely on two factors with this: if the user uses the computer at its factory settings, then type size is adapted to the screen size; if the user changed the settings, then type size is adapted to his visual ability and preferences. If you give image size in em, i.e. in relation to type size, images will be scaled to the users needs.

This is, of course, no exact science, but that is what I do, when I design a website. It is the closest I manage to get to cross computer similarity in size. As soon as you read about cross browser web design, you'll lose all optimism that your goals at exactness can be achieved ;-)

---

Answer 2

2 sounds reasonable. The variability you introduce by sticking to pixels or display ratios seems like it outweighs the apparent unreliability of monitor size calculations.

Also, unlike TVs that scale the content to fit the size, most webpages do not scale -- so, when I drag a webpage from my laptop screen to my second (larger) monitor, the size stays constant. Therefore, I would predict that people do not end up sitting as far back from large screens as you might expect. Therefore, the subtended visual angle will be approximately the same if the image is the same size for everyone.

Answer 3

If I managed to understand, each participant may have a slightly different viewing experience, as long as it's constant through multiple sessions each participant has.

To achieve this, you may ask participants to sit an arm's reach from their monitor. Try to make the images span the same physical size (even though it's really hard to achieve programatically). This way, the viewing angle would be relatively constant between participants, but more importantly, would be constant for each participant.

If the experiment is dependent to the level of details they can see on the pictures, you might want to consider low-res, and large-detailed images, to eliminate the advantage of those with better monitors.

Moreover, make sure they all use desktop computers, as many might want to user their tablet computers, or even their smartphones...