[10] Because cognate words play an important role in mutual intelligibility between languages, these figure prominently in such analyses.
The higher the percentage of cognate (as opposed to non-cognate) words in the two languages with respect to one another, the lower is their linguistic distance.
As another example, the English dish and German tisch 'table' are lexically (phonologically) similar but grammatically (semantically) dissimilar.
Cognates in related languages can even be identical in form, but semantically distinct, such as caldo and largo, which mean respectively 'hot' and 'wide' in Italian but 'broth, soup' and 'long' in Spanish.
[10] However, lexicostatistical methods, which are based on retentions from a common proto-language – and not innovations – are problematic due to a number of reasons, so some linguists argue they cannot be relied upon during the tracing of a phylogenetic tree (for example, highest retention rates can sometimes be found in the opposite, peripheral ends of a language family).