The aim of this study was to analyze the relationship between errors on Standard Progressive Matrices Plus (SPM Plus) and Advanced Progressive Matrices (APM) and eductive ability. A large sample (n = 4535) of secondary school pupils aged 11 to 19 was used. For both SPM Plus and APM, the two-parameter logistic model which includes item difficulty (b) and discrimination (a) parameters was used for the eductive ability estimation. At first, analysis of responses to SPM Plus and APM items (including such response options as “a subject omitted an item” and “a subject didn’t reach an item”) was conducted using the Classical Test Theory (CTT) difficulty and discrimination power indices as well as the Item Response Theory (IRT) Option Characteristic Curves (OCCs). The analysis has confirmed that response alternatives lacking only one correct feature are more likely in the middle range of ability and from this range upward, whereas subjects of low ability often choose alternatives that are incorrect on more than one and / or more salient features. It has also been found that higher ability subjects are prone to omit an item that is too difficult for them instead of guessing an answer as opposed to lower ability subjects who may use a random guessing as a RPM solving strategy. The analysis has also shown that higher-scoring subjects have no time enough to finish all the RPM items more often than lower-scoring individuals. This finding suggests that higher-ability subjects invest more time in solving RPM items, and therefore a time limit is not suitable for assessing a high eductive ability. Next, the APM types of error analysis has revealed that the type of errors made on APM is related to the ability level of the subjects. Errors due to failure to attend to all relevant aspects of the item, i.e. incomplete correlate errors, were made by subjects of higher ability as compared to errors due to the confluence of ideas or repetition. This finding suggests that an answer of incomplete correlate type shows a relatively high eductive ability (as compared to the confluence of ideas or repetition type answer) since it requires a correct, although incomplete, reasoning.