I usually employ a technique called "Management of the Hesitant". I adopt an even-numbered scale (usually 6 values), which actually present two midpoints and, mind, an external option like "I do not know" or "I prefer not to answer" (this is not to force anyone in giving a response). Then I perform a binomial test (or chi-square test) on the numbers of the responses collected for the 3rd and 4th options (the midpoints), to see whether some polarization occurs (many tools online allows this, e.g.,
http://graphpad.com/quickcalcs/binomial1.cfm,
http://graphpad.com/quickcalcs/chisquared2/ ). Then I report the cumulative proportion of people that have chosen one of either midpoints (3rd + 4th), stating whether a significant tendency toward either the positive or the negative extreme of the scale has been detected (this is the case if the p-value associated with the binomial test is lower than .05). In so doing, I catch the hesitant ones, those who prefer to take refuge in the midpoint (in odd-numbered scales), but I have enough information to understand if any significant pattern can be observed (against the null hypothesis that hesitant people just choose a midpoint randomly). Then I collapse this information again into a cumulative midpoint proportion, to be totally compatible with other studies (adotping an odd-numbered sscale) and at the same time to distiguish these people from those who expressed a clearer attitude towards one of the extreme values. Hope this helps!
