Many test items use both an image and text, but present them in a spatially separate manner. This format could potentially cause a split-attention effect in which the test taker's cognitive load is increased by having to split attention between the image and text, while mentally integrating the two sources of information. We investigated the split-attention effect in computer-based assessment by implementing: (a) "spatial integration" which embeds text information into a diagram as labels; and (b) "interactive signaling" which highlights a segment of a diagram when test takers hover their mouse over an answer option that refers to the relevant segment. Adult participants solved computer-based geometry items in which spatial integration was used as a within-subject variable and interactive signaling was used as a between-subject variable. The main findings showed that, compared with the nonintegrated and nonsignaled items: (a) spatial integration significantly increased item-solving efficiency indicated by reduced time on task, especially for test takers who had higher prior knowledge; and (b) interactive signaling increased item-solving efficiency only in the spatially integrated items. The current findings suggest that applying multimedia design principles for reducing split-attention to test item design could potentially reduce test takers' construct-irrelevant cognitive load in computer-based assessment.