A set of 12 open-ended tasks was chosen from studies by Cai, Moyer, and Wang (1999) and Silver and Lane
(1993) to assess non-routine problem-solving skills. In an open-ended format, a student is asked not only to provide an
answer but also to explain the thinking processes that led to the answer. These problems involve a variety of important
content areas, such as number sense, pattern recognition, number theory, prealgebra, ratio and proportion, estimation,
and statistics. Fig. 1 shows four sample non-routine mathematical problems. Half of these non-routine problems are
process-constrained tasks. A process-constrained task can be solved by applying a “standard algorithm”, but students
do need to figure out what algorithm to apply and actually carry out a procedure or a set of procedures in solving
the problem. For example, the Hats Averaging Task, shown in Fig. 1 is a process-constrained non-routine problem. It
assesses students’ abilities to apply the averaging algorithm flexibly in a problem situation. The other half of the nonroutine
problems are process-open tasks. A process-open task cannot be solved simply by using a standard algorithm.
A task that is process-open may not require an execution of a procedure or a set of procedures; instead it may require
an exploration of the problem situation to develop a better understanding of the problem, before work begins on a
solution. Therefore, a process-open task is set in such a way that it allows students to use alternative, acceptable solution
strategies. The Number Theory Problem and the Block Pattern Problem shown in Fig. 1 are examples of process-open
problems. Two Chinese mathematics educators and two U.S. mathematics educators were asked to independently
classify the 12 tasks into six process-constrained tasks and six process-open tasks. The four educators agreed on
the classification for almost all of the tasks, except that one Chinese educator classified one process-open task as a
process-constrained task, and one U.S. educator classified one process-constrained task as a process-open task. The
discrepancy was resolved through discussion. For this study, the Cronbach’s alpha coefficients are .77 for the Chinese
sample and .69 for the U.S. sample (Cronbach, 1990).