More replies:

Alex F: I have pondered previously this statistical approach, and have
rejected it. The issue is that the statistics do not seem to pertain
specifically to linguistically relevant causes. Firstly, there are two ways
to calculate these statistics. One is by number of speakers, and the second
is by number of languages. The first one obviously is no good because the
number of speakers is based too much on economic or military dominance by a
handful of countries to be meaningful in a linguistic sense. The second one
is better, but still suffers too much from the accidents of history rather
than other, deeper causes. For example, let us say, in some region, there
had been one language that was SVO and a half-dozen which were OSV. The SVO
culture is more warlike and eventually they enslave the populations of the
other groups and those OSV languages completely disappear. Then, the
statistics will show a rarity of OSV languages, not due to any innate human
repugnance to OSV grammars, but merely because of the military supremacy of
a single culture long ago. It is not possible to disentangle the
linguistic, or neuroanatomical, reasons for linguistic statistics from the
historical reasons. That is why I have concluded the best one can do is a
simpler procedure: Does this particular feature exist in any language? Yes
or no? My assumption has been that if the answer is “no,” which means that
no language anywhere in the world evidences that feature, then it is
“unspeakable” and should not be included in a conlang that strives for
naturalism. It seems unlikely (in my rough assessment), that the complete
absence of a linguistic feature can be due solely to historical factors,
and that therefore there must be some innate cause which can explain the
feature’s absence.

David P: The process is not the test. The test of a lightbulb is to screw
it into the socket, turn on the current, and see if it glows. The
manufacturing process does not figure. A sound manufacturing process may
indeed produce high quality lightbulbs, but the two things are
intrinsically different. In like manner, following sound diachronic
processes in the creation of a conlang may produce naturalistic languages,
but it’s not the test. There may well be other processes which can produce
naturalistic conlangs equally well, or perhaps more efficiently; and doing
dichronism right is difficult, and requires extensive linguistic knowledge.
Moreover, phonological diachronism is much more studied and developed than
grammatical dichronism. If the goal is a naturalistic conlang grammar, then
the diachronic method of construction may be of little help. Please note
I’m not saying that what you do when you create a conlang is wrong, merely
that you should not be dogmatic in insisting that it is the only way to
achieve naturalism. I don’t have a suggestion for an alternative procedure,
however. I still believe the true test is the aesthetic sense that comes
into play upon the usage of the language by the author or others. This is,
admittedly, of little help as advice to aspiring conlangers.


On Wed, Jun 3, 2015 at 4:34 PM, David Peterson <[log in to unmask]> wrote:

> Gah, I really want to jump in here and say stuff, but I really have to.
> I’m going to mention a couple of headline points:
> (1) Lakoff says conlangs aren’t languages because they lack metaphors.
> Obviously, laughably false. The issue is whether or not the conlanger has
> created metaphorical concepts in the language without unconsciously
> borrowing in metaphors from the natlangs they speak, under the notion that
> if you didn’t actually try to do something it’s not art: it’s lazy. Also,
> metaphor works on two separate levels: semantic evolution (dead metaphors)
> and daily conceptual metaphor (active). Two different things that require
> two different skillsets/mindsets.
> (2) The best way to get irregularities and other naturalistic features is
> to evolve the language, as Jörg mentioned. If the processes you use are
> sound, the result will be naturalistic, whether or not it actually exists
> in a natural language. That’s the test. Then you don’t have to worry about
> coming up with a grand list of everything that exists in every natural
> language on Earth and measuring the conlang the way one validates the HTML
> on a webpage.
> (3) It doesn’t matter if, for example, a language has [θ]. It matters how
> it got it. For example, Mexican Spanish is like most languages: it lacks
> [θ]. Castellano, however, does have [θ]—same language. It’s exceedingly
> rare, crosslinguistically, but it doesn’t really matter, because we know
> exactly how it got [θ]. In this case, neither fact is particularly
> interesting. Looking at the percentages crosslinguistically rather than the
> diachrony is a mistake, because you’re examining the end result, not the
> process. In the case of Spanish, *ts > [θ] in Castellano and *ts > [s] in
> Mexican Spanish. How common are each of those specific sound changes? Now
> you’re asking the right question.
> (4) While I agree there’s something to natural languages that makes them
> easy (or easier?) to learn, I’m not 100% sure we can pinpoint that yet. For
> example, it’s common for pronouns to be irregular in some way. Would it be
> easier to learn Esperanto if there were some irregular accusative forms
> (e.g. yo~mi, li~en)? I doubt it. The -n accusatives for pronouns may not be
> natural, and it may not be easier to learn than a natlang pronoun system,
> but I doubt it’s harder. I think it’s a good question to ask, but randomly
> sprinkling in irregularities is not the way to either create a good
> naturalistic conlang or a good auxlang.
> Awesome discussion! I’ve really been enjoying reading the entire exchange.
> David Peterson
> LCS Member Since 2007
> [log in to unmask]
> =================================
> Pre-Order The Art of Language Invention!
> (Available September 29th!)
> =================================
> > On Jun 3, 2015, at 4:09 PM, Alex Fink <[log in to unmask]> wrote:
> >
> > On Wed, 3 Jun 2015 13:43:21 -0700, Jeffrey Brown <[log in to unmask]>
> wrote:
> >
> >> Jörg R: I apologize for my lack of clarity: I didn’t mean to imply that
> a
> >> naturalistic grammar had to contain the same COMBINATION of grammatical
> >> features as a natlang, but that use of a truly novel feature, one that
> is
> >> not observed in any natlang whatsoever, will compromise the naturalism
> of
> >> the grammar (such as Fith’s stack-based grammar). I think that any
> >> (consistent) combination of natlang grammatical features ought to work.
> >
> > Re this, and re your original (2), I'd add that "consistency" is a fuzzy
> property.  To resume Joerg's example, are ergativity and VO order
> _inconsistent_? No.  So are they just another unremarkable combination? No
> again, they're pretty rare together.  That is, the combination is somewhere
> in the fuzzy periphery of "consistency", not in the centre near its
> prototype.
> >
> > I've stated my concept of _strong naturalism_ around here before, as the
> version of naturalism I aspire to, though surely don't attain.  The
> strong-naturalist strives to make conlangs whose statistics, on every
> feature and combination of features, follow the distribution of natural
> languages, given a large enough sample (this bit is necessarily a bit
> nebulous, e.g. we might have to observe the world for a million years to
> get enough data).  If 43% of natlangs are VO, then your languages should be
> VO with 43% probability.  If only 7.5% of natlangs have [T], then you'll
> have to go without [T] 92.5% of the time.  (That's one of the scores on
> which conlangers taken collectively are way out; there's way too much [T].
> Of course that's a hard type of thing to "solve"... not entirely unlike the
> problem of ensuring representation of minorities in fiction collectively,
> though of infinitely less social importance.)
> >
> >> Taliesin: But would you say that Tariana has a naturalistic grammar,
> >> despite it being a freaky monster conlang?
> >
> > I wondered if you had confused Tariana and Tariatta.
> >
> >
> >> John Q: (1) If cryptotypes are defined in the description of a conlang,
> are
> >> they still cryptotypes? Hmm…? I don’t think a constructed language can
> have
> >> cryptotypes.
> >
> > My reaction to that idea is not dissimilar to my reaction to Lakoff's "a
> conlang can't have frames": both seem to include a sort of confusion of
> levels between the description and the (imagined, typically) describendum.
> >
> > A cryptotype is simply a semantic or syntactic type of feature which
> lacks explicit _morphological_ expression.  It's not something like an
> "undiscovered" feature.  One could easily even make a sketchlang with a
> cryptotype -- you know, suppose verbs don't express an active vs. stative
> distinction in their morphology at all, but suddenly when you add an
> adverbial adjunct naming a point in time ("at noon"), statives are all of a
> sudden read inceptively while actives are read normally.  Bam, cryptotype.
> >
> >> (And I suspect that is why, in my opinion, no exolang really “feels”
> >> complete and naturalistic, even on its own terms. It’s missing a
> consistent
> >> conceptual framework which is a deducible extension of the sensory/motor
> >> apparatus of the aliens who are purported to speak it. To create a great
> >> exolang, one probably has to spend a few years beforehand working out
> the
> >> anatomy and physiology of the aliens, before tackling the conculture and
> >> the language.)
> >
> > Well said.
> >
> >> Anyway, this (conceptual metaphors and cognitive frames) is a really
> >> excellent point, and something that conlangers striving for naturalism
> >> should pay closer attention to.
> >
> > In accord with your "read a dozen natlang grammars" advice, what we
> really need is a cross-linguistic (and not overfocussed on SAE) treatment
> of which sorts of metaphors and frames are commonplace and which are rarer
> -- without this, how will we know what makes a naturalistic and not too
> calquey set of frames and metaphors?  So, have any linguists done any work
> at all like that?
> >
> > Alex