An examination of some methodological problems in testing expectancy-valence models with survey techniques