A unifying framework for generalized linear mixed modeling of nominal data (unordered polytomous data and rankings) will be discussed. The models include random coefficients and/or factors varying at different levels of hierarchically structured data. Partial and tied rankings, alternative specific explanatory variables and alternative sets varying across units are accommodated. Parameter estimation proceeds by maximum likelihood and prediction of random coefficients/factors by empirical Bayes, based on adaptive quadrature as implemented in the gllamm software. The methodology is applied to party choice and rankings from the 1987-1992 panel of the British Election Study. Three levels are considered: elections, voters and constituencies.