Interesting theory based on mammalian visual processing, with significant implications not only for tabs but for any menu or list. It's also against just about every HCI standard and style guide I can find, many based on years of research and operational experience, going all the way back to MIL-STD 1472. Could they all be wrong? There’s pretty firm research showing that the further a target item is in a menu, the slower the reaction time. For example:
Nilsen EL (1991). Perceptual-motor control in human computer
interaction, Tech Rept 37. Ann Arbor, MI: The Cognitive Science and Machine Intelligence Laboratory, the University of Michigan.
Byrne MD, John R, Anderson JR, Douglass S, Matessa M (1999). Eye tracking the visual search of click-down menus. Human Factors in Computing Systems:
Proceedings of CHI 99, pp. 402–409.
What Determines Order of Attention
The first rule of visual attention is that the eye is attracted to spots of high graphic contrast. That accounts for 80% of eye-tracking findings.
The second rules is that visual attention will follow text reading order, generally left to right and top to bottom for Western readers. This accounts for the F-pattern seen for many sites (the F-pattern is also a result of sites in fact being laid out in an F-pattern, but that’s really the same thing: the point is user read and scan based on their experience).
The third rule is that users can learn to peripherally ignore certain visual patterns that they’ve learned are irrelevant or not worth the effort of studying (e.g., things that look like advertisements, even though they deliberately have high visual contrast).
The third rule is not relevant to this case here.
Your theory appears to leverage the first rule, but with a graphically uniform array of items like a row of tabs, the whole thing has high visual contrast with the rest of the page. There’s nothing about either end that particularly draws the users attention any more than the middle.
That leaves the second rule, which is what the HCI standards rely on. When graphic contrast is uniform within a block of content, such as with a collection of undifferentiated tabs, Western users will scan left-right/top-bottom.
In addition to the standards, this is what I see in eyetracking results. Within any graphically uniform block of content, you see either:
Gaze focused on the middle of the block, suggesting a quick glance at the block, and subsequent possibly preconscious decision that it’s not worth looking at more.
Gaze uniformly distributed over the block, suggesting and exhaustive study of it.
Gaze biased towards to top and/or left of the block, suggesting a reading-pattern scan.
What I don’t see is the gaze concentrated at the beginning and end of the block with a gaze hole in the middle. The heat map in your blog post is consistent with this, with every block showing one of the above three patterns and no holes. Gaze does not appear to jump to the end of lists. Indeed, the third gaze pattern is consistent with users reading the first few items, and if they don’t seem relevant, they skip all of the remaining items, including the last ones. This implies that if you put your second most important tab at the end some users may never even look at it.
Meanwhile, on the Serengeti
As for lions and zebras, if lions tend to pounce on the zebras at either end equally often, that just shows that lions don’t read. Actually, I think lions tend to target the zebra in the middle. They select the slowest looking one and then maneuver and stalk as close as they can, which usually puts the target in the middle of their visual field surrounded by more distance zebras in the background. The effect of the zebra stripes is to make it hard to continue tracking that target once all the zebras start running around in a panic. I don’t think the lion-zebra thing applies here since we thankfully don’t animate our tabs.
Users will tend to look at the current tab because it graphically contrasts with the remaining tabs (if it doesn’t you’re making your tabs wrong). If you put the second most important tab next to the first, then the user may be able to read both with a single fixation if the tab names are short enough. You can’t do better than that.
There is almost always a logical order to items in tabs, a list, or a menu. Order may be semantic, by frequency of use, by order of use, by convention (e.g., based on what other web sites do), and, if all else fails, alphabetical (under the hope that the user is looking for the same specific word you used). For example for an ecommerce page aimed at consumers the proper order is almost certainly Home, Products, Partners, Contact Us.
Where does the third most important tab go? Inside of the last tab since that’s where the user looked last? Or inside the first/current tab because that’s the closest to the most looked-at point? I don’t know, but more importantly your users won’t know. You’re disrupting their ability to strategically scan your tabs.
What about keyboard navigation? If “second most important” means that’s where users tend to go next, you’re making them tab through all the tabs taking the longest path for the most common navigation need.
Let’s Have Data
Anyway, this is all theory that can’t be resolved with debate. A pretty simple reaction time with eyetracking study should give the answer. But in the mean time I personally will stick to the HCI standards.