Originally posted by normhead One of the reasons I take reviews so lightly are the lack of double blind testing. There was a taste test done with beer many years ago, where the participants were all brand loyal drinkers who swore their beer was better than any other and that they could tell the difference. When given a taste of the 5 top selling brews they came it at exactly random choice levels. They couldn't tell the difference. The evidence is pretty much overwhelming, people can't make these kinds of decision accurately. Even a guy like Klaus over at PZ, can't make these kinds of decision accurately, even though he has the benefit of having 2D tests to help form his opinion.
If there's one thing you learn it's strength of conviction about accuracy does not mean one's opinion is more accurate. In fact those who have the strongest opinions are often those who have excluded the most relevant information in their research, focusing on one narrow aspect. Buy reducing a lens to an MTF number it's possible to ignore a lot of other relevant data, some of which are not even measurable. If psychology has taught us anything, it's that because something has some aspects that are quantifiable, doesn't mean the quantifiable aspect is the most important. It may be the most studied and tested, but that doesn't make it any more important. That just makes it the most talked about.
The one thing everyone can tell you, is if you buy the lens, can you get what you want to do done with it. Not maybe whether it's the best lens you can get for what you do, or if it will give you the most pleasing results based on your subjective analysis.
What most people want to know is if a certain lens matches their style. A reviewer probably can't tell you that.
Well put. I recently bought the Sigma 10-20mm f4. I'd have liked the constant aperture version, but for the price differential I can make "adjustments;" the variable aperture does just fine WITHIN ITS LIMITATIONS.
Wouldn't a proper double-blind test between the two Sigmas be interesting??
Sadly I'd imagine that few know what a double-blind test is. A prior post to this thread said that we see them often in this forum, I seriously doubt that! As I understand it a double-blind requires that neither the participants in the test nor the "operator" can know which images or whatever are being presented or in what order. Only the test "developer" knows which sample is which. All variables should be accounted for, in audio, for example, this would mean equalizing sound levels to something like .25 db.
Since we're unlikely to see such tests for lenses, it would be good if reviewers who furnish MTF and other hard data would suggest what variations from "perfection" are likely to be visable in actual use, and which can be safely ignored. Granted, this would be subjective, but the lens' output, a picture, is evaluated subjectively, isn't it? These "judgements" would aid the user in understanding the limitations of a lens - and ALL lenses have limitations! - and how to work around them. For example I find the 18 - 55 kit lens a good walk around user, certainly excellent in its price point.