["Counting" experiments with ravens using "matching-from-sample" methods. A contribution on problem-solving behavior in higher aniMALS].

The research goal is divided into two parts: 1. Detailed analysis of behavior in selection in a 6-choice matching-from-sample task. 2. Discussion of all observed preferences on the basis of the models of Sutherland and Mackintosh and of Miller, Galanter and Pribram. The experiments were run with 4 one-year-old ravens (Corvus corax); two of them were probably 2 males. The design is similar to Koehler's. In the task the correct responses were equally distributed between all stimuli and positions (random system). Although the task ("counting") was learned, a lot of relevant preferences for stimuli, positions, patterns etc. show that problem-solving in higher animals can be very complex. It could be shown that only a detailed analysis may clarify animal learning processes.