Several studies have shown that figural matrices can be solved with one of two strategies: (1) Constructive matching consisting of cognitively generating an idealized response, which is then compared with the options provided by the response format; or (2) Response elimination consisting of comparing the response format with the item stem in order to eliminate incorrect responses. A recent study demonstrated that employing a response format that reduces response elimination strategies results in higher convergent validity concerning general intelligence. In this study, we used the construction task, which works entirely without distractors because the solution has to be generated in a computerized testing environment. Therefore, response elimination is completely prevented. Our results show that the convergent validity of general intelligence and working memory capacity when using a test employing the construction task is substantially higher than when using tests employing distractors that followed construction strategies used in other studies. Theoretical as well as practical implications of this finding are discussed.