Creative reasoning in ill-defined problem spaces operates differently from classical reasoning in well-defined spaces. To systematically compare the two in an identical knowledge domain, we applied a classical intelligence test: the Standard Progressive Matrices (SPM), in combination with two tests of creativity: the Test for Creative Thinking Drawing Production (TCT-DP) and the newly developed Creative Reasoning Task (CRT), in which participants are asked to create an SPM-like item, to two age groups (N1 = 511, 4-12y old; N2 = 205, 6-10y old). For SPM and CRT the knowledge domain consists of relationships amongst geometrical components in 3 x 3 matrices. We developed a typology for scoring the number and complexity of the relationships used in these matrices. For the SPM, we scored frequencies of relationships solved and for CRT those created, and interpreted the scores in terms of differences and similarities between classical and creative reasoning in cognitive development.