The optimum number of alternatives in multiple choice items is studied. The
new procedure applied is based on Item Response Theory and only needs a si
ngle sample. Concretely, a 5-alternative 221 items English Vocabulary test
was administered to 452 people. Their answers to the 1/2/3 worst alternativ
es were randomly reassigned to generate their hypothetical answers to items
of 4/3/2 alternatives, respectively. Changes in item parameters, test info
rmation function and ability estimation were analysed. The worst results we
re provided by 2-alternative items. As the results in the 3- and 4-alternat
ive items do hardly differ from those obtained in the original 5-alternativ
e case, and in the 3-alternative case the role of partial knowledge is less
important, 3-alternative items were considered the best choice. The effect
s of number of alternatives on ability estimation was also discussed.