An image of an object is altered in the process of meditation. According to the theory of Theravāda Buddhism, a sign (nimitta) of a meditation object perceived by five sense organs is modified into an internal sign grasped by consciousness, and then into a pure counterpart sign (paṭibhāganimitta). While in the ten kinds of kasiṇa-meditations one perceives the object and expands its counterpart sign throughout the entire world, the sign should not be expanded in the ten kinds of foulness (asubha) meditation and others. In this paper, investigating various interpretations of this subject found in Theravāda literature, I consider the relation between theory and practice as well as between the canon and the commentaries. According to the Visuddhimagga (Vism), in the mindfulness related to body (kāyagatāsati) and the foulness meditations, the object to be observed is spatially limited to the body itself, and no merit is found in expansion of the sign. On the one hand, Buddhaghosa attempts to justify the theory with a textual account from an unidentified source, “Sopākapañhavyākaraṇa.” On the other hand, he needs to solve canonical inconsistencies such as that in the Theragāthā (Th v. 18) referring to “filling the perception of bones on the earth” and the Dhammasaṅgaṇi (Dhs 55) reference to “having the immeasurable object”. We should note that his interpretations of those problem are not shared with the sub-commentary on the Vism (Vism-mhṭ), the commentary on the Dhs (Dhs-a) and the commentary on the Th (Th-a). Modern scholars have neglected the fact that the Vimuttimagga (Vim) has a discussion on the same topic explicitly quoting both Th v. 18, and the Dhs 55 and shows the other solution. In conclusion, we may point out that the meditation theories arising from various practices regulate the practices themselves in reverse, and that the Pāli commentaries had a function to align the Pāli canon to their doctrine.