`cognate'

[Larry Trask]

This posting will not be news to the linguists on this list, who may
prefer to stop reading at this point, unless they want to offer
criticisms of the definitions below.  But I have noticed that a
significant number of non-linguists on this list have apparently
misunderstood the sense of our technical term `cognate': many of them
appear to believe that `cognates' means something like `words of similar
form and meaning in different languages' -- which, of course, it does
not.

I hope it may be considered appropriate and helpful, therefore, if I
post the entry for `cognate' from my forthcoming dictionary of
historical and comparative linguistics.  This has not yet gone to press,
and I am still willing to accept suggestions for revisions, though I may
be unable to consider large revisions.

*************

*cognate*  1. Narrowly, and most usually, one of two or more words or
morphemes which are directly descended from a single common ancestral
form in the single common ancestor of the languages in which the words
or morphemes are found, with no borrowing.  For example, English
<father> and German <Vater> `father' are cognate, being descended from
Proto-Germanic *<fader>, and both are more distantly cognate with
Spanish <padre>, Irish <athair> and modern Greek <pateras>, all
`father', all of these being descended from PIE *<pater>.

2. Broadly, and less usually, one of two or more words which have a
single common origin but one or more of which have been borrowed.  For
example, English <jail>, Old French <jaiole> `jail', Spanish <jaula>
`cage' (Old Spanish <javola>), Basque <txabola> `hut', Occitan <cayola>
`cage' and Basque <kaiola> `cage' are all ultimately descended from an
unrecorded Latin *<caveola> `small enclosed place', but only the French,
Occitan and Spanish words are narrowly cognate: the English and Basque
words have been borrowed from Old French, Old Spanish and Occitan,
respectively.  Note: some linguists object to the use of the word in the
second sense.

3. [erroneous]  A label improperly applied to items of similar form and
meaning in languages not known to be related, when these are presented
as candidates for possible cognation.  Common among linguistic amateurs,
this objectionable usage is not unknown even among linguists, but it
should be avoided: items cannot be labelled "cognates" until a
substantial case has been made that they genuinely *are* cognate.

*************

Other entries in the dictionary, of course, stress the fallacy of
regarding mere *Anklaenge* as evidence for anything.

Larry Trask
COGS
University of Sussex
Brighton BN1 9QH
UK

larryt at cogs.susx.ac.uk