-ency/-ancy vs. -ence/-ance

[Arnold M. Zwicky]

my colleague Beth Levin wrote on 9/29 to forward a query:

> A computer science friend of mine wrote me the following and I
> thought you might have some insight.  Is it in part a way of
> creating a "jargon"?
>
> -----
> My German-born colleague Torsten just asked me about the following:
>
> In technical papers, people often use the words "coherency" and
> "relevancy" rather than
> "coherence" and "relevance".  Why the replacement of "ency/ancy"
> for "ence/ance"? How common is this?
> -----

the dictionaries i consulted treated -ncy as a variant of -nce, just
as beth's friend did.

the facts are complicated.  in a fair number of cases, only the -nce
version occurs: confidence, resistance, persistence, benevolence,
malevolence, insolence, independence. in some cases, only the -ncy
version occurs: consistency, malignancy, fluency, constituency.  when
there are pairs, sometimes the meanings are very clearly distinct
(emergence vs. emergency), sometimes more subtly so (dependence vs.
dependency), and sometimes there's considerable semantic overlap
(coherence vs. coherency, relevance vs. relevancy, valence vs.
valency).  when there's some overlap, the -nce version is more
frequent, but the ratio of -nce to -ncy (in raw google web hits)
varies quite a bit across the pairs:

dependence : dependency - 1.44
coherence : coherency - 1.56
valence : valency - 11.47
relevance : relevancy - 20.66

for coherence/coherency, there's a technical sense of "coherence" in
physics, but also specialized technical senses of "coherency" in
several fields that are well represented on the web: "cache
coherency" in computer science/engineering, and statistical
"coherency" (paired with "consistency", an -ncy-only formation).  as
far as i can tell, these -ncy technical uses don't vary with -nce at
all; the form of the expressions is fixed.  so the ratio is low.

but for valence/valency, the words are in competition pretty broadly,
in technical senses in chemistry, linguistics, and mathematics.  the
shorter and less complex -nce version then prevails by a wide
margin.  (i'm inclined to use "valency" in talking about the argument
structure of verbs, probably because i'm often contrasting valency
with constituency, and "constituency" is a -ncy-only formation.)

for relevance/relevancy, i find it very hard to see any
specialization, and the shorter and less complex version prevails
even more strongly.

the longer and more complex versions might have some value on their
own, however, just by virtue of their greater substance.  to some
ears they might seem weightier, more formal, more serious, more
technical.

i suspect there's some literature on this, but i don't have the
relevant part of my library to hand and so can't easily check out
sources.  anyone have any lines on the issue?

arnold, sending copies of this to beth and her friend