On 4/9/06, Carsten Becker <[log in to unmask]> wrote:
> From: "Jim Henry" <[log in to unmask]>
> Sent: Sunday, April 09, 2006 4:27 AM
> > Neither Volapük nor Esperanto nor, as far as I know, any
> > other early auxlang had self-segregating morphology; all
> > of them had problems like the one you describe to some
> > extent (it was pretty bad in the original Volapük; Arie
> > de Jong's 1931 revision fixed some specific ambiguities
> > but didn't solve the underlying problem).
> I came across the same problem in Ayeri and have thought
> about maybe making changes to the one or the other affix,
> but eventually decided against it. Note that Ayeri is in no
> way intended to be an auxlang: Natlangs also have
> ambiguities, but nobody cares about them except there isn't
> any context given. So at least for conlangs, I see no reason
> why they shouldn't allow the one or the other ambiguity.

Of course a naturalistic conlang can and probably should
have such compound parsing ambiguities.

> I also see no problem for auxlangs in this regard, except of
> course you want to have an engelang-type auxlang.

There's a spectrum of engelanginess, from Lojban or Ithkuil
at one end to naturalistically messy artlangs at the other end.
Most of the recently developed auxlangs that aren't
euroclones have certain engelang properties, specifically
self-segregating morphology to prevent any compound
parsing ambiguities.

> I wonder,
> though, how agglutinating/fusional languages deal with
> potentially ambiguous compounds of affixes and stems.

Natural ones, or engelang/auxlang type conlangs?

In my experience, self-segregating morphology is primarily
beneficial to the learner in the early to middle stages
of learning a language.  Once youve learned it fairly
well and have a reasonably good vocabulary, apparent
possible ambiguities are rarely a problem and actual ambiguities
are fairly rare (at least in Esperanto; I haven't gotten
fluent enough in Volapük to be sure).

> In
> Ayeri, it's easy to have words that initially look like
> verbs, but that are no verbs due to not taking any
> necessary parts of a verb as suffixes like person and case
> agreement. Is that enough already? I think so. I might have
> a disambiguating look again at conditional/consequencial
> sentences. Also, verb stems should neither begin with the
> time prefixes nor end in -asa and -ong, and nouns by
> preference should not in any of the case endings as well ...

In Greek a fair number of verb roots start with epsilon (/E/),
which for other verbs is a prefix indicating tense and aspect
(exactly which tense and aspect depends on the verb
endings applied).  Those verbs lengthen the initial
epsilon to eta (/e:/) to mark those tense/aspects.  I'm not
sure if there are any verbs with a stem starting with
eta but I suspect so.

> I conclude that in my humble opinion minor ambiguities
> should be permissible if you don't go for an engelang.

Of course.  In an engelang you generally want to eliminate
all or nearly all grammatical ambiguities.
In an auxlang, I would think eliminating all compound
parsing ambiguities, and the most common grammatical
ambiguities, is sufficient.  In a naturalistic artlang, having
implausibly little ambiguity would be a flaw.

> Since natlangs aren't always 100% clear, conlangs don't
> need to be either. Some ambiguity only adds to naturalism
> I think. Anyway, natlangs usually find ways to
> disambiguate more sensible contexts by establishing an
> alternative construction expressing the same as far as
> I know.

Yes.  Almost any ambiguity is plausible in a naturalistic
artlang in the short run, but ambiguities that are especially
likely to cause confusion in real speech are likely to be
corrected sooner or later by the speakers replacing
one member of a homophonous pair, e.g., with a synonym
or paraphase.

Jim Henry