Sixty subjects participated in an experiment involving estimation of difficulty of items in a test of reasoning ability. The estimates were to be given both according to conventional conditions of magnitude estimations with a preassigned comparison standard and according to a modified procedure of magnitude estimation where the comparison standard was chosen individually by the subjects themselves. The test itself was administered to the subjects under standard conditions prior to the estimation procedures. When comparing the two methods of estimation used, a high correlation between estimates and a close correspondence of the modified method of magnitude estimation to the methods of ratio estimation and similarity estimation was noticed. A high correlation (r=0.90) between the rank order of items according to perceived difficulty and the item sequence was found. Furthermore, estimated difficulty could tentatively be described as a positively accelerated function of standard scores corresponding to solution frequencies. The relative increase of perceived difficulty was more pronounced for subjects with a high performance score on the test than for subjects with a poor performance score. Probable causes of the results obtained as well as possible secondary factors affecting the estimates of perceived difficulty are discussed. (Author)