Determining the likelihood of borrowing and the immediate sources of loans was constrained by the information available. Ramirez’ Tukano dictionary (1997) is by far the most extensive available dictionary of any East Tukanoan language, but it still is quite limited, and includes almost no Portuguese borrowings or names for non-native concepts. The loanword status of many Hup words in the database is thus given as ‘no evidence for borrowing’ where no Tukanoan form was available for comparison (except in the few cases where there was some other indication that the word might be borrowed). Without a good documentation of Portuguese loans into Tukano, it was also impossible in many cases to determine whether borrowings of Portuguese origin in Hup entered via Tukano or directly from Portuguese. The lack of detailed lexical data on Hup’s sister languages also made intra-family comparison difficult.

Given these limitations, the likelihood of borrowing from Tukano was determined by the following criteria. If a Tukano word nearly identical in form and meaning to the Hup word could be identified (e.g. Hup bǐ/ and Tukano bi/i ‘rat/mouse’), the degree of certainty was rated 3 (‘probably borrowed’) or 4 (‘clearly borrowed’), depending on the closeness of the match, intuitions of consultants, and whether information was available on the form of the word in Hup’s sister languages (i.e. indications that the word is not cognate were considered additional evidence of loanword status). The degree of certainty decreased where the word form and meaning were only an approximate match (e.g. Hup pí/ and Tukano yãpi ‘potato’), to 3 or 2 (‘possibly borrowed’) if Hup’s sister languages had unrelated forms, and to 2 or 1 (‘very little evidence’) where there was no comparative information available on the word in the other Nadahup languages. Finally, while existing data and the current sociolinguistic situation suggest that the direction of borrowing was consistently Tukanoan > Hup rather than vice versa, the absence of good comparative data on the Tukanoan and Nadahup families has made this impossible to establish definitively in every case.

The limited data and lack of any text record has made the dating of words in the database difficult. Relative dates were established using several criteria. First, comparison of word lists across the Nadahup family, and particularly the reconstructions in Martins (2005), were used to determine whether a given word was cognate across two or more Nadahup languages (and thus appears to reconstruct to Proto-Nadahup, Proto-Hup-Yuhup-Dâw, etc.); however, these lists included only a fraction of the words in the database, and intra-family borrowing cannot always be ruled out. Other words, particularly loanwords of Portuguese origin, are dated within a general range of 100 to 10 years ago, on the basis of their formal accommodation and the best guess at when the concept became known to Hup speakers generally (for example, Hup speakers obtained items like knives and axes through trade long before they entered into direct contact with non-Indians, whereas other things like ice, wheat bread, and canned beer are very recent and still rarely encountered).

The numerical dates assigned to these relative periods are very approximate and should be understood as little more than wild guesses. For example, the age of Proto-Nadahup is given at between -500 and 1200 CE, but even this wide range could be incorrect if the rate of change in the Vaupés is significantly slower or faster than that of other better-understood language families, such as Romance. The age of many words in the database is indicated as ‘no information’; this simply means that the Hup words have no known cognates, which is in most cases due to an unavailability of lexical data on all three sister languages is unavailable for that meaning. Since the Hup data itself is only a few years old, a date of ‘since 2000 CE’ for the first attestation of these words seems pointless, so no date is given.