The present study examined differential effects of whole-class discussion begin- ning with a problem, with and without answer alternatives, on its developments and cognitive consequences in the case of adding fractions with different denominators; 289 4th- and 5th-graders from six elementary schools were used as subjects. Results indicated that: (1) almost all students with or without alternatives figured out the correct answer through the discussion before being given feedback from the experi- menter; (2) in the process of the discussion the students in Group Without Alternatives made more remarks including explanations, questions and counterarguments, espe- cially those about the correct solution than students in Group With Alternatives; (3) the students in Group Without Alternatives tended to perform better in the transfer test. The inferiority of Group With Alternatives in the acquisition of transferable knowledge was interpreted as due to the lack of effective division of labor based on partisan motivation where rival answer alternatives were easily refuted in discussion. Strength and weakness of the above two types of discussion and their possible flexible uses are discussed.