I've seen some camera reviews and blogs rave over the quality of a camera because of the number of megapixels it has; I've seen others go to all lengths to deny the importance of megapixel density altogether. I have several problems with both positions because both often ignore certain elements that contribute to image sharpness, making an argument that is quite unscientific.
To accurately capture the quality of a certain aspect of digital - or any type of - photography, only the aspect being tested can change with each experiment. To make this clear, let's outline the tools and environment needed to capture a photo:
a) Subject and Lighting.
c) Film plane (sensor).
d) Camera support.
d) ISO, Shutter speed, Aperture, Focus, Distance from Subject (framing).
e) Post production (in-camera or none altogether).
g) Display (print or screen), calibration & resolution.
Rather obvious, no? If we are to effectively test the quality of a sensor against another, only the sensor should change between tests. This may be rather difficult in today's world where different products have accessories that change some or sometimes all of the other constants that would make for a scientific test, but the logic remains the same. Ken Rockwell uses this logic quite clearly in his camera testing by using the same lens, photo environment/technique and photo post-production (enlargement, cropping, etc.) in each comparison.
One aspect that most comparison-articles tend to neglect: Lens quality. A lens has 'grain' and sharpness, and if you ignore this fact in your testing, your comparison will be false - or at best, incomplete. Say, for example, you are using a not-so-sharp lens to compare sensors: If the maximum sharpness of one lens 'grain' covers more than one sensor pixel, it hardly matters how many more pixels you use to capture that (non-) sharpness, as even if the entire rest of the photo environment/technique remain the same, the result will appear the same.
Add to this neglect the method used to calculate pixel density (or a debunking thereof), and you will have a truly mislead reader. The 'megapixel' method should be ignored altogether, as it is not attainable/usable by the human mind, let alone photographers: we measure by perception and vague reckoning, and never by mathematical calculations of (dimension-undefined) area. We used to 'measure' grain density by a certain ISO 'constant' compared to the size of the film plane we were using, itself compared to the visibility of this combination in the final image (at its desired size). We have no way of isolating/including any of these elements of reason with just a single 'megapixel' number.
So, if you want to test a sensor between camera models, you're basically going to have to start ripping cameras apart and switching parts between them. Until a test of that sort occurs, I will continue to insist that pixel density does matter - but only if the rest of your camera/display environment can detect the difference too.