The study challenged the current practices in cognitive load measurement involving complex problem solving by manipulating the presence of pictures in multiple rule-based problem-solving situations and examining the cognitive load resulting from both off-line and online measures associated with complex problem solving. Forty-eight participants were recruited from a subject pool in the Educational Psychology department in a large research university in western USA. Results showed that the presence of pictures had no effect on learners' complex problem solving as measured by the response time and accuracy. However, the online measure (eg, pupillometric measures) revealed a change in cognitive load associated with the presence of pictures in complex problem solving. The authors concluded that different measures of cognitive load may actually be measuring separate aspects of cognitive load. Discussions were made on how research on multimedia learning and cognitive load could be advanced by carefully considering multiple aspects of cognitive load and by including the use of convergent measurement techniques to capture the variations of cognitive load involved in learning.