Is recognition (pratyabhijñā) reliable? It is part of direct perception (pratyakṣa)? Is it a distinct instrument of knowledge?
Many Indian authors use recognition as evidence in their arguments. They may say that the same thing can be touched and then seen and that we know through recognition that it is the same thing (Nyāya). Or, they might say that the fact that we recognise things through time is evidence of the fact that there is an "I" (Nyāya). Or, they claim that this "I" can recognise its nature as identical with Īśvara (Śaivas such as Utpaladeva). Recognition might also play a role within analogy (upamāna), when we recognise the gayal as being the animal about which we heard that it was similar to a cow.
Yet, Indian authors do not generally state that recognition is a distinct instrument of knowledge. Hence, to be reliable it should be included in another instrument of knowledge. The best candidate seems to be pratyakṣa, but Rāmānujācārya, for one, says that pratyakṣa is "purely born out of the sense-faculties" (saṃskārajamātra) and that "purely" is meant to exclude recognition, which, like memory (smṛti), depends also on recollection traces (saṃskāras) (TR I).
Recognition is itself different of memory, since it does not depend only on saṃskāras (so Śālikanātha, PrP).
Hence, how can recognition be reliable? Perhaps, because a part (aṃśa) of it is perceptual and hence reliable. In the standard formula of recognition, sa eva idam ("this is the one [I cognised before]''), the perceptual part is the idam ('this'). This does not seem to entail more than the sheer perception of something. However, since perception includes, according to Mīmāṃsakas and against Buddhists, also its qualifications, the idam-part may include its resemblence with the previous saḥ. Even if notion of the saḥ depends on saṃskāras, then, the sheer fact of similarity could be perceptually established.