A preliminary investigation into 'counting' in Jackdaws using the Discrimination Box as illustrated. Shadrach was trained to the buzzer as a significant cue and Meshach to light flashes. With Shadrach results, significant at the 01 level (X2), were obtained after 300 trials, and with Meshach after 400 trials. Transfer to the other modality gave results, significant at the 01 level, after 100 trials. Transfer from light to buzzer alarmed Meshach and trial sequences became so lengthy that he was discarded. A randomised schedule of 2: 3 buzzes was given to Shadrach, B door being + 3 and - 2, A door being - 3 and +2. Learning was rapid, results being significant at the 01 level after 100 trials. A randomised schedule of 3: 4 buzzes was then given keeping B door + 3 and - 4. Results were significant at the .01 level after 100 trials suggesting that the response was not being made on the basis of 'moreness'. A 2:3:4 discrimination was then given keeping B door still + 3 and -2 and - 4. Experimental breakdown occurred and a 'side response' developed. After a three month rest away from the experimental situation he was unable to relearn the original 2:3 discrimination resorting to a 'side response' which varied from 100-trial sequence to sequence. The results obtained appear to indicate that discrimination between 2 : 3 or 3:4 was based on the discrete properties of the stimulus and not on 'moreness'. Also, that the transfer between modalities is possible where the skill transferred involves discrimination of numbers of stimuli.