When comparing the choice frequency in the two geometrically correct (C, R) versus the two geometrically incorrect corners (N, F) a highly significant difference was apparent (F(1,9) = 61.714 p<0.0001), revealing more choices for the two geometrically correct corners (C,R).
When the correct corner was near the feature (animals trained on the corner with differently coloured walls as positive) there was a marginally significant difference in choice frequency between corners R, N and F (F(2,18) = 3.718 p = 0.044) with only a trend for a significant interaction with sessions (F(2,18) = 2.865 p = 0.083).
No significant difference was apparent in first choice frequency between the correct corner, C, and its geometrically equivalent corner R (F(1,9) = 1.569 P = 0.242), indicating that the disorientation procedure was effective and that bumblebees did not rely on any extra- or intra-enclosure featural cues.