Update on the Romantic Lang
- I have been more productive in conlanging in the last month than I
think I have in the last several years put together, and most of that
in the last week and a half. I've managed to generate a vocabulary of
~250 words in less than a week (depending on what you count as a
separate lexeme... it could be as little as 20, if you're really
conservative, but I prefer the bigger estimate), and though this is
partially due to motivation to work on this particular language, it's
more the approach taken to it, I think. So, I wanted to give a little
overview of the process, and then some of the new features that have
come out of it.
I am trying to ensure that I don't think about any decision longer
than it takes to get to a point where the results of that decision are
needed to say something. So, wherever I or my fiance have existing
opinions about something, or come up with neat ideas ahead of having
to actually use them, those things get decided a-priori, which
provides a good deal or planning and guidance to the development of
the language. However, if we want to say something and don't know how,
it gets made up on the spot. So far, we've been conscientious enough
not to accidentally introduce horrible inconsistencies that way, but I
suspect that will start happening as we try to develop fluency.
Already, I introduced one bit of grammar in the dictionary, and then
noticed a couple of days later that I had been consistently saying it
"wrong", so I just changed the documentation, 'cause actual usage is
canonical. We've made one retroactive change so far, as a matter of
aesthetics- through an accident of derivational morphology, the word
for "romantic love" (a very important word in this language!) ended up
sounding just absolutely horrible to my ear, so the relevant paradigm
slot got updated.
I'm quite convinced that part of the reason this works so well is that
we've got a very powerful derivational morphology to work with right
off the bat, so it's fairly easy to think up a word that you need
without having to think up a whole new root; and, when you do think up
a whole new root, you get up to 20 additional words "for free" (though
some of those have very confusing or not-obviously-useful literal
meanings, and it takes some thought later to figure out what those
words actually should be used for; e.g., [j\eK\asa] "the good place
for cuddling" == "couch")
Among the non-pre-decided issues is pretty much everything about
phonotactics. While there are a-priori rules about how to resolve
disallowed consonant clusters that result from derivational
morphology, there are no a-priori rules about what constitutes a
disallowed cluster. Prosody and allophony are also being left up
entirely to what happens to fall out of our speech. I'm thinking in a
couple of months (or whenever we get sufficient fluency that I feel
like the data is worth something), I'll go ahead and do a formal
analysis of what the rules have turned out to be.
This langue-based approach is not something that is available to the
typical conlanger, and I feel like it's even significantly different
having two people negotiate the language than the few instances of an
individual going with the "once I write it, it's canon" approach. The
distinction between internal history and external history is
effectively dissolved- the history of whatever changes we decide to
make to the language is the same as the history of how the language
Some of the interesting things that have cropped up "on their own", or
as a result of spur-of-the-moment decisions:
The clusters /tk/ and /kt/ are disallowed in the same syllable (or
possibly the same word; for now it's ambiguous whether the relevant
environment is syllable boundaries or word boundaries). This doesn't
seem to apply to /gd/, though. That might just be due to lack of
examples, though; voiced consonants are rarer than unvoiced
Stress is not lexical, nor is strictly phonological; stress mostly
falls on the first syllable of a root. This has a tendency to break
down only for instrumental nominatives, which flip-flop between normal
and terminal stress. I'm not sure how that's going to work itself out.
There's some complex vowel allophony going on (probably a result of
the fact that we're both native English speakers), such that while
there are only 5 vowel phonemes, there are at least 12, possibly 13,
vowel phones so far. Basic allophony is conditioned by being in a
stressed open syllable (the environment for canonical pronunciation)
vs. not, but there's some additional stuff going on to distinguish
sequences of vowels that have been reduced to diphthongs from vowels
followed by underlying glide consonants.
One interesting bit of prosody is the tendency to lengthen (with no
change in quality) vowels in clause-terminal syllables. This turns out
to be a redundant marker, along with the verb-phrase-terminating
particle "des", for valency-changing operations.
In the lexicon, we've ended up with 5 different words for "love"
(loosely based on C.S. Lewis's classifications), 4 different ways to
say that you miss someone or something with shades of meaning that are
really difficult to explain in English, a set of interrogative verbs
(which allow you to ask things like "what did you do?" with a single
word), and only two interrogative pronouns, with the rest being
rendered by prepositional phrases (though I'm feeling a bit of
pressure to eventually insert some more basic interrogatives). Also,
while there are verbs meaning "to be an example of a set" and "to be
equivalent to", there's also a zero-copula construction. Since there's
no verb there, if there's no other sentential particle to mark the
start of a clause, a last-resort complementizer "at" was introduced.
Unlike the English complementizer "that", though, "at" can be used on
top-level clauses just to clarify that this is in fact a clause, and
not just a noun phrase.
Nouns have inflections for singular, plural, negative (no or none of),
and non-particular, inspired by the particular / non-particular
distinction in Blackfoot. This is similar to definiteness, but not
quite exactly the same- it depends only on the speakers knowing what
he's on about, nothing to do with the discourse.
The initial idea for a possessive ended up only being used with
personal relationships at first, which left the door open to
introducing a second genitive construction and re-analysing the first
as strictly relational or associative, thus providing a distinction
between "my" as in "I actually own it" and "my" as in "I have some
relationship with it". The addition of an objective marker to handle
topicalization and valency changing operations then pretty much gave
us a genuine 4-case phrase-level case system; all of the markers are
phrase-level clitics, just like the English possessive.
As with the copula, while there is a verb for "to possess something",
the verb "to have" is more easily expressed with a zero-verb
construction and one of the possessive cases:
Mi - amriwa. -> "I am (someone's) fiance" / "I'm engaged."
Mir amriwa. -> "My fiancee."
At mir amriwa. -> "I have a fiancee."
On that note, this language seems to be absolutely clitic obsessed.
About half of all grammatical features are indicated by things that
I'm pretty sure are clitics.
Interpreting those as cases allowed for preposition-multiplexing, so
we've managed to get by so far with a grand total of 2 prepositions,
adding logically-connected meanings to them distinguished by object
Another very productive clitic is "ni" (pretty much stolen wholesale
from Russian, but with somewhat different usage). Attached to the
beginning of an interrogative phrase, it forms indefinites
("something(s)" for singulars and plurals, "nothing" for negatives,
and "anything" for non-particulars, though the semantic spaces
correspond closer to Russian pronouns than to English). Attached to a
verb phrase, it turns the whole thing into an adverbial phrase where
any interrogatives can be replaced with the equivalent "-ever" forms;
or, if there are no interrogatives present, it forms the subjective.
Res vu ki? -> "Whom do you love?"
Res vu niki. -> "You love someone (and I know who)."
Res vu nikin. -> "You love someone (I don't know who)." / *"You love anyone."
Ni res vu ki. -> "Whoever you love..."
Ni res vu niki. -> "You would love someone."
As for those valency changing operations:
The basic word order is V-S-(O1-(O2))
All verbs have lexically determined valency, but the language allows
for free pro-dropping. Since subordinate clauses typically have a null
complementizer, some indication is needed that a clause has terminated
so you don't accidentally think the next clause in the discourse is a
complement to the previous verb. That's usually handled by
clause-terminal vowel lengthening,as mentioned, or with an explicit
verb-phrase-terminal particle "des". In a totally unmarked sentence, a
single missing argument must be the subject ( V-(O1-(O2)) ), and two
missing arguments must be the subject and first object ( V-(O2) ). For
most cases, depending on the derivational class of the verb, there
will be a corresponding lower-valency form that could've been used,
but the higher-valency with pro-drop implies that referents for the
missing roles do exist and are simply unstated, while the
lower-valency verb does not.
The first object, however, may be marked with objective case; in a
marked sentence, the word order gets much more complicated, and allows
for fronting for topicalization and dropping of things out of order.
Possible arrangements include:
When an argument is fronted, this can result in ambiguity about the
clause boundary, requiring the usage of "at" to resolve that
Additionally, the reflexive can theoretically be used as a
sort-of-passive to promote second objects to subject position; this
ought to be useful since there are no verbal derivations that produce
a verb with a thematic focus as a subject, so this basically allows
for promoting focuses in ditransitive sentences, though I've never
actually used the reflexive yet.
If the original subject role is indicated by an oblique argument
(prepositional phrase), then the subject will be re-interpreted as
solely having the role indicated by the position of the reflexive.
Vadu mi vu yerrasa. -> I show you the couch.
Show me you couch.
Vadu yerrasa vu se iza mi. -> The couch is shown to you by me.
Show couch you REFL. because-of me.
Yerrasa vadu mi vug. -> *The couch*, I show you.
Couch show mi you-OBJ.
Yerrasa vadu vu se iza mi. -> *The couch* was shown to you by me.
Couch show you REFL. because-of me.
And without the oblique agent:
Yerrasa vadu vu (des). -> *The couch* catches your eye.
The reciprocal has turned out to be more common so far than the
reflexive and thus has a single-segment representation in the form of
a verb suffix.
Tense and aspect are kinda neat, too. There's a basic
perfective/imperfective distinction that's lexicalized and can be
altered with prefixes, much like Russian. However, unlike Russian, the
present-tense interpretation of a perfective is not implied future,
but rather emphatic "right now! and only once."
There is the capacity to distinguish perfect and imperfect as well,
but that turns out to be very very tensy and not aspecty at all,
unlike English; and it come with prospectives as well as
retrospectives built-in, so I've taken to calling them "spectives" in
general. A set of 5 sentence-initial particles (derived from
contractions of phrases meaning "It was that...", "It is that...", "It
will be that...") indicate spectives in different tenses, and the
actual tense of the sentence tells you whether its retrospective or
prospective. These aren't used as frequently as the English perfect
construction, though, because you get much the same effect by assuming
the the temporal reference for an embedded clause is the time of the
matrix clause, not the time of utterance (that is, after all, how the
spective particles were derived).