I'm a little suspicious of the collection method for the data. While you got 6500 results, when I register for most sites I have a couple of "low security" pass

I'm a little suspicious of the collection method for the data. While you got 6500 results, when I register for most sites I have a couple of "low security" passwords I use (which I rotate ever 6 months or so), so that whenever I come across a site that says my usually username is taken, I've got a good guess that I registered at some point and forgot about it (this has happens about 4-5 times a year). So, what I would be wondering is if many of the results are the same thing. I certainly wouldn't register for a survey with on of my more secure passwords (or generate a new one for it). I would also be more likely to put in a nonsense password that I won't need to remember, and will immediately forget. One thing that should probably be checked, at least before you can draw any real conclusions, is how well the user can log in. While it is true that the claim was that unobscured password fields would make for more secure passwords to be used, the reason was that an unobscured password would be easier to type and remember when logging in. If you remove that logging in feature, its hard to have any real conclusions.