Position sense and kinaesthesia are mainly derived from the integration of somaesthetic and visual afferents to form a single, coherent percept. However, visual information related to the body can play a dominant role in these perceptual processes in some circumstances, and notably in the mirror paradigm. The objective of the present study was to determine whether or not the kinaesthetic illusions experienced in the mirror paradigm obey one of the key rules of multisensory integration: spatial congruence. In the experiment, the participant’s left arm (the image of which was reflected in a mirror) was either passively flexed/extended with a motorized manipulandum (to induce a kinaesthetic illusion in the right arm) or remained static. The right (unseen) arm remained static but was positioned parallel to the left arm’s starting position or placed in extension (from 15° to 90°, in steps of 15°), relative to the left arm’s flexed starting position. The results revealed that the frequency of the illusion decreased only slightly as the incongruence prior to movement onset between the reflected left arm and the hidden right arm grew and remained quite high even in the most incongruent settings. However, the greater the incongruence between the visually and somaesthetically specified positions of the right forearm (from 15° to 90°), the later the onset and the lower the perceived speed of the kinaesthetic illusion. Although vision dominates perception in a context of visuoproprioceptive conflict (as in the mirror paradigm), our results show that the relative weightings allocated to proprioceptive and visual signals vary according to the degree of spatial incongruence prior to movement onset.