During experiments at the Zoo Berlin over 2.5 years we noticed several very interesting and noteworthy things about 11-year-old female Asian elephant Anchali’s performance. She was trained to match fruits; using four different types of fruit Anchali learned to pick one from a box and choose the same type out of two different. Our first observation was that she can generalize this concept of matching same with same to various types of fruit she had not been trained with or even never tasted. She picked correctly in ~80% of the trials. To test out limitations, we offered either ¼, ½, or a full fruit in the box and she was able to match quantity, or rather fraction, within the same type of fruit with a ~80% success rate. Finally, we tested this with ¼, ½, or a full fruit of one type and ¼, ½, or a full fruit of a different type and observed a ~76% success rate in matching fractions across two different fruits. Fractions of fruits were not controlled for weight/size, it seems like Anchali has some understanding of what ¼, ½, and a full fruit is and is able to generalize this concept. It looks like her decision making is based on integration of multisensory information and like she changed her strategy during training from more visual to more olfactory/tactile. In ongoing control experiments our impression is that this effect is not due to Anchali responding to experimenters’ body language or other potentially confounding factors.