Just thinking aloud here (sometimes I have to remember that this still is my blog, after all, and I can weigh in with my little two cents' worth of thoughts when I want to), I'd like to put forth a radical claim: there's actually no effective, detectable difference between the APS-C sensor size and the 4/3rds sensor size. They're simply too close together. Forum measurbators like to crow and cluck about how there are huge differences, that the larger sensor is oh-so-much better, and that the sky is falling. Hogwash. If they just looked at pictures, they couldn't tell.
As with film, a certain degree of departure is necessary before the differences begin to show up reliably in the real world. Otherwise what you have to resort to is direct comparison in controlled situations.
There are a great many parameters in photography that can be detected in direct comparison that can't be detected simply by looking at non-equivalent pictures taken under real-world conditions. This might surprise you, but if I show you some 35mm prints taken with different lenses, you won't be able to distinguish those taken with a 40mm lens from those taken with a 35mm lens, or the 100mm focal length vs. 90mm, unless it's the same subject or unless you have enough examples that you know are one or the other (that is, when viewers are asked to distinguish between pre-sorted groups of samples, their accuracy goes up somewhat). If I were to show you moderate enlargements from 645 film and 6x7 film, you couldn't tell the difference unless, again, the pictures were of the same or similar subjects, and probably not even then. If I were showing you B&W prints made on one of two papers, and you knew one of the papers was capable of a higher D-max (maximum black), you couldn't guess which paper you were looking at if I showed you only one print. Furthermore, most people can't look at two B&W prints that have different maximum blacks and tell me which one has darker blacks! That never stopped an enthusiast from declaring one paper superior to another based on a measurable difference in D-max, however. Same thing with lenses—I proved to my own satisfaction (though not to the Leicaphiles' satisfaction) that even Leicaphiles can't tell the difference between pictures taken with Leica lenses and pictures taken with other lenses...unless, again, they have before them the evidence of direct, controlled comparisons or the advantage of large, pre-sorted groups of samples. And sometimes not even then.
Anybody wanna guess how I know all these things? Actual experiments. Some better-controlled than others from a scientific-method perspective, I admit. But still.
(Please go along with the game—don't change your answer after seeing the results! I promise, you won't be graded on your answer. I'll add the answer to this post on Sunday evening.)
Now, care to guess the average forum maven's chances of looking at a digital picture, either online or in print form, and being able to guess accurately and reliably whether it was made with a 4/3rds sensor or a 1.5X sensor? I haven't done that experiment, but I'm going to go out on a limb and guess: they haven't got a chance in hell. Maybe full-time fine printers or reviewers could be able to approach statistically significant accuracy. Maybe. But even with them I'd have to see it to believe it.
With many such factors—lens contrast, maximum black in B&W papers, lens focal length, color accuracy, degree of negative enlargement, etc., etc., etc.—there has to be a certain amount of separation before differences begin to be reliably detectable without direct comparisons. There's usually a "zone" of difference where most viewers can't distinguish between two variables at all, a "zone" of confusion where some viewers can tell and others can't, or people can tell some of the time but not all of the time, and a "zone" where the differences becomes sufficiently large that almost everybody can tell almost all of the time. (A friend of mine compared 6x7 to 35mm, for instance, and found that under almost all conditions, 6x7 was clearly better—even where the 35mm result was optimized in several ways and the 6x7 samples weren't. The exception was in prints sizes that represented less than 5X or 6X enlargement from the 35mm negative.) Where focal length is concerned, research suggest that 15% is the approximate threshold of this third zone. I don't know where it is with sensor sizes, but I suspect it requires a doubling of the area of the sensor, which is two or three sizes except when taking into account the huge leap between 1/1.7" and 4/3rds—that is, between the largest digicam size and the smallest DSLR size (a 5.2X difference in area).
Sensible experimental results would never stop forum pontificators from "absolutizing" differences that they can detect, or think they can detect, in direct comparisons under controlled conditions—or from thinking that slightly better measured results will translate to real-world pictures. It's human nature to want to be on the side of "scientific proof," after all, just as it's human nature to want there to be a "winner" and a "loser" in every comparison or contest. But I suspect those same people would be dismayed by their own inability to distinguish all those differences they think are so obvious, if they were just looking at pictures.
Featured Comment by Charlie H.: "One of the fundamental tenets of psychophysics is Weber's Law, which states that the detectable difference between two magnitudes is proportional to the magnitudes themselves; e.g., humans can tell the difference between 1 oz. and 2 oz. but not between 100 lbs. and 100 lbs., 1 oz. Mike, you're a psychophysician!"
Mike replies: I knew a was a psycho-something. Thanks, Charlie.