Distribution for modeling acceptability judgements

[Matías Guzmán Naranjo]

Hi all,

I am analyzing an experiment where participants had to rate the acceptability of sentences from 1 to 7 under different conditions. The distribution looks nothing like a normal (or binomial) distribution, because there seems to be strong grammaticality judgements.  Is there any established 'best' distribution to model this kind of data? Just assuming a gaussian distribution seems to give 'realistic' coefficients and 'good' results, while something like a truncated poisson distribution gives ridiculous coefficients.

Thanks!

[Stefan Th. Gries]

I would certainly start with Gaussian and see where that takes me. I
don't think it matters that the distribution of the judgments is
non-normal - it seems to me that normality is more important for the
residuals than for the original observed values. Just my $0.02 ...