An image capturing apparatus includes an image sensor that has a plurality of two-dimensionally arrayed pixels, each of the pixels having a first photoelectric conversion portion and a second photoelectric conversion portion, a generation unit that generates a first image signal by connecting, in a pupil divided direction, a first signal obtained by combining signals of the first photoelectric conversion portions, and generate a second image signal, in the pupil divided direction, a second signal obtained by combining signals of the second photoelectric conversion portions. In a case of combing signals of the first photoelectric conversion portions, or combining signals of the second photoelectric conversion portions, the generation unit decreases weighting of a signal of a photoelectric conversion portion in which an effect of crosstalk from a neighboring photoelectric conversion portion is large.