This presentation explains and illustrates how computer-based innovative test tasks are designed in a multi-semiotic environment to effectively and comprehensibly convey meaning to students, especially English language learners (ELLs), students with learning disabilities (LDs), selected other students with disabilities, and non-identified native English speaking poor readers, and give them the tools they need to express complex thinking in mathematics and science. Research was conducted in elementary and middle school students. Research was conducted for math and science ONPAR projects with elementary and middle school students. Results from the randomized experimental study in science indicate that, when academic ability was controlled, low English proficient ELLs and non-ELLs in both grades 4 and 8 performed similarly on tasks which utilized ONPAR techniques, while there were significant differences in how the low ELL students performed on more traditional approaches. On the other hand, there were no significant differences between how non-ELL students performed on either form. Viewing these results as a whole, these findings argue that the ability of low ELL students can be measured by this method at a level that is on par with their non-ELL peers. Further, it argues that there might be flexibility in the ways ELs can be tested without advantaging them relative to scores obtained from traditional test formats used by non-ELLs. A large number of cognitive labs contributed substantively to the continued refinement of the ONPAR techniques. The experimental study in mathematics was completed in winter 2010. Preliminary results show that when adjusted for student math ability, the interaction between test type (ONPAR versus traditional) and reading level (poor vs. good) is significant with a combination of LDs and poor readers but not for native English speaking good readers, and that when controlling for math ability the LD/poor readers and good readers performed similarly on the ONPAR tasks. Correlations from the experimental study and several cognitive labs have also suggested that several of the ONPAR techniques are actually better at capturing data about complex skills and concepts than traditional methods for all students.