Updated 16 November 2006

From my response to a note from Martin Dale on Thosk and conlanging:

Dear Martin,

Thank you for your message and for your kind words about Thosk.  I'm glad the FAQ echoes what you've said over the years to critics of conlanging.  WE know why we do it, and what pleasure we derive from it -- odd that this should be the hardest part to convey to others, when it feels like the most important part!

I do make use of international vocabulary, as far as it goes, which isn't too far:  Thosk taksi, otel, telefon, bas, gaz, suker, sjofer, zjanre, etc.  And if there's no handy root to use, I sometimes can devise a periphrastic equivalent.  For instance, I couldn't find an IE root to indicate "disappointment" so I borrowed a Turkish idiom I happened to know and used IE roots to express it in Thosk form:  had-rud, or "desire-collapse" (from IE *ka- and *ru-). Of course, this is time-consuming, and doesn't always work.  So I do also resort to the third option you mention, and make historical changes of meaning (and by analogy, kinds of changes in meaning) serve my purpose.  The meanings of many roots in most lexicons of IE "vocabulary" seem very general, and flexible enough to suggest a range of significations.  From roots that mean break, for example, one could derive "shatter, smash, splinter, fragment" or a range of other related words as needed, extending given meanings, but not wholly violating the semantic associations of a root.  Books like Calvert Watkins' American Heritage Dictionary of Indo-European Roots (though it is Germano-centric) suggest the spread of meanings and the kinds of naturally-occuring changes in semantic fields which one can simulate in developing a root into a "new" word.  Another helpful book for showing how far afield a root can bend in meaning is Carl Buck's Dictionary of Selected Synonyms in the Principal Indo-European Languages (now fortunately in paperback).  And once a conlang takes on enough of a shape, that shape and feel can itself suggest new words that fit the sound pattern of the language, which may extend a primitive root with legitimate although perhaps never used IE endings. For instance, IE *bhr- suggests anything that protrudes, or sticks out -- and this could be a body part, a slang expression for an annoyance, a handle, a bump, a promontory, etc. -- in other words, a very wide semantic fiield of "legitimate" proto-meanings from which to work.  And of course sometimes I let all that go hang and just create a word that suits me and sounds like it belongs.