Histories of English typically list approximate proportions of words borrowed from each of the major donor languages, with a few examples, and limited comment on the likely contexts in which borrowing occurred. The figures are normally derived from the OED (first edition), filtered through the Chronological English Dictionary (1970), which takes the selective wordlist of SOED and assigns its already summarized etymologies to various large groups. A similar approach is found in many contributions to the research literature.

I will look at some of the complexities which are disguised by such approaches, and which should be flagged alongside a basic numerical analysis in any careful account:

• Etymologies are often more complex than crude summaries suggest: we find multiple inputs, dual etymologies, repeated formal or semantic borrowing from the same source.
• The focus is normally on loanwords, but loan translations and other types of semantic borrowing are also of importance (and in OED3 are more clearly flagged), although they are difficult to detect.
• The degree of phonological and morphological naturalization of loanwords, and their productivity within English, should also be taken into account, as should technical vocabularies, register, and stylistic level.

Histories typically use figures from a large wordlist, but illustrate the discussion with familiar, everyday words, which are not typically representatives of the tokens in the OED or SOED wordlists. To investigate borrowing in this component of the vocabulary, we need to look elsewhere for our wordlist: Swadesh meaning lists, defining vocabulary of learner's dictionaries, wordlists of school or learner's dictionaries, frequency lists. Such tools, combined with OED's etymologies and dates of attestation, can give a much fuller picture of how many ultimately borrowed words there are in different components of the vocabulary of contemporary English, and (a very different question) how borrowing has affected different components of the vocabulary in recent decades.

For earlier periods the available tools and data are more limited, but similar complicating factors exist, plus additional ones such as interpreting data from multilingual texts, or borrowing in situations of language shift rather than language maintenance.