Page "False cognate" Paragraph 5

Some historical linguists presume that all languages go back to a single common ancestor.

Therefore, a pair of words whose earlier forms are distinct, yet similar, as far back as they have been traced, could in theory have come from a common root in an even earlier language, making them real cognates.

The further back in time language reconstruction efforts go, however, the less confidence there can be in the outcome.

Attempts at such reconstructions typically rely on just such pairings of superficially similar words, but the connections proposed by these theories tend to be conjectural, failing to document significant patterns of linguistic change.

Under the disputed Nostratic theory and similar theories such as that of monogenesis, some of these examples would indeed be distantly related cognates, but the evidence for reclassifying them as such is insufficient.

( Alternatively, apparent cognates in Eurasian language families far removed from each other could also be early loanwords, compare Wanderwort.

) The Nostratic hypothesis is however based on the comparative method, unlike some other superfamily hypotheses.