The etymology of a linguistic unit – commonly a word or a stem – is a twofold story (which is here considered backwards, in the direction of reconstruction):
Part 1 of the story thus is made up of documented history. Part 2 is reconstruction proper; it deals with the motivation of the word at the stage when it was coined.
As an example, consider the etymology of the English word hound:
An etymological dictionary is one which provides the etymology for its lemmata. Since the second part of etymology generally involves historical comparison of languages, an etymological dictionary mentions several languages.
As for lemmatization and entry structure, there are essentially two possibilities:
In the former type of etymological dictionary, loans are often excluded from lemmatization. This is justified on the grounds that part 1 of a loanword's etymology only leads back to the moment that it was borrowed – which may be a relatively short story – while part 2 of the etymology would apply to the donor language and would thus be the task of the etymological dictionary of that language.
In either type of etymological dictionary, a set of indices is usually appended, each devoted to one of the languages concerned in the dictionary and comprising those words of that language that appear in the entries.