Potential Paonese: A Reconstruction from Jack
Vance’s The Languages of Pao
Author: Logan Kearsley
MS Date: 10-14-2015
FL Date: 11-01-2015
FL Number: FL-000032-00
Citation: Kearsley, Logan. 2015. «Potential Paonese: A
Reconstruction from Jack Vance’s The
Languages of Pao.» FL-000032-00, Fiat
Lingua,
November 2015.
Copyright: © 2015 Logan Kearsley. This work is licensed
under a Creative Commons Attribution-
NonCommercial-NoDerivs 3.0 Unported License.
http://creativecommons.org/licenses/by-nc-nd/3.0/
Fiat Lingua is produced and maintained by the Language Creation Society (LCS). For more information
about the LCS, visit http://www.conlang.org/
Potential Paonese
A Reconstruction from Jack Vance’s The Languages of Pao
The Languages of Pao, by Jack Vance, explores the strong Sapir-Whorf hypothesis in the
description of a stagnant people, the Paonese, of the planet Pao. The initially uniform Paonese
culture is fractured by a massive social engineering project whose central component is the
introduction of numerous artificial languages, each designed to produce in its speakers a
tendency towards a different socially useful pursuit. Very little description is given to any of
these languages, however, with the exception of the original Paonese.
Even the original Paonese, however, is not a real language. That may seem obvious given
that it is a fictional language spoken by a fictional people, but well-developed conlangs like
Quenya, Klingon, Dothraki, Na’vi, and numerous less well-known examples can reasonably be
called “real languages” in the sense that, no matter the real or fictional status of their native
speakers, they exist in the real world; one can learn them, compose novel sentences in them, and
generally use them as a practical medium of communication. None of the languages of Pao meet
that standard; they are strictly imaginary. They exist only as ideas of languages from a fictional
world.
Of the original Paonese, however, we do have scattered words and descriptions by the
narrator, who represents a contemporary, though linguistically naïve, observer from the fictional
world. In the attempt to understand the languages of the fictional world of Pao, this puts us in a
similar position as philologists attempting to piece together the scattered fragments of a poorly
attested ancient language. There is too little known to provide a definitive description or
reconstruction. However, we may still endeavor to construct a real language that could have been
Paonese.
Purpose of this Article
The purpose of this article is not to provide a complete linguistic description, reference
grammar, or lexicon of the reconstructed Paonese language. Rather, my intent is to document the
evidence for Paonese and the reasoning behind the decisions that I made in producing a
reconstruction. To that end, while this account is significantly cleaned up compared to my
original stream-of-consciousness notes, there is still a good deal of meandering- presenting a
possibility, noting its problems, presenting possible alternatives, noting their problems, and
sometimes ending up back at the beginning again. It is my hope that, rather than being seen as a
pointless diversion, the reader will find these wanderings down blind alleys, investigating
solutions that end up getting thrown out, to be a valuable source of insight into the reconstructive
process, odd back-alleys of English pronunciation, and the reasons why certain final decisions
were made.
It is particularly important to note that the resulting “potential Paonese” is not necessarily
the Paonese. It most likely does not represent the original Paonese as it existed in the mind of
Jack Vance; and indeed, in the strictest sense that may be impossible, as I am unconvinced that
that original Paonese language ever actually existed at all beyond the scraps presented in the
book. Other equally valid reconstructions, based on different interpretations of the evidence and
different arbitrary decisions about unattested features are certainly possible.
Additionally, the purpose of this specific reconstruction is not to produce the simplest
possible language that accords with Vance’s work, with the simplest possible explanations for all
evidence. Rather, it is to develop a language which is naturalistically complex in ways that are
not obviously derived from a monoglot Anglophone perspective, and yet still can explain the
evidence in the book.
Plot Summary
At the beginning of the novel, the planet Pao is a mostly agrarian backwater world with a
culturally homogenous population ruled by an absolute monarch, known as the Panarch.
Knowing that their lack of diversity makes Pao vulnerable to external economic and military
powers, the Panarch attempts to hire a consultant from the planet Breakness, Lord Palafox, to
plan a reformation of their society. The Panarch, however, is betrayed, assassinated by his
brother, who seizes the throne, and the rightful heir, Beran, is kidnapped and taken to be raised
on Breakness as a political hostage.
After Pao is conquered and made to pay tribute to the planet Batmarsh, the new Panarch
again seeks aid from Lord Palafox, who has a plan to create castes of warriors, scientists, and
merchants by segregating Paonese society via the forcible introduction of new, mutually-
unintelligible languages, to be taught to selected populations of Paonese children. Accomplishing
this goal requires the training of a corps of language instructors from Breakness, fluent in the
new constructed languages and capable of teaching them to the Paonese. Beran infiltrates this
corps in order to return to Pao incognito, and participates in the creation of an additional pidgin
called Pastiche- a mixture of all of the constructed languages of Pao used among the instructors.
Eventually, Beran reveals himself to the people of Pao, and support for the usurping
Panarch crumbles. The warrior caste is able to repel their invaders, and Beran reclaims his title,
and the planet enters a brief golden age of wealth and progress. Displeased with the divisions
caused by Palafox’s language program, however, Beran eventually attempts to re-integrate the
castes. This results in an uprising and coup staged by the warriors. Beran convinces them,
however, that they will be unable to rule the planet alone as they share no common language
with the rest of the populace; they therefore allow him to remain in office, and agree to his
decree that every Paonese child must learn Pastiche, once again uniting the planet under a
common language and in a common culture- but now possessing the diverse linguistic attributes
of proud warriors, skilled scientists, and savvy merchants.
Overview of Attested Paonese
From what limited explicit description is given, we know that Paonese is polysynthetic,
contains no verbs or adjectives, and is composed mainly of “nouns, post-positions, and temporal
indices”. It is said to be derived from another language known as “Waydalic”, but “molded into
peculiar forms”; no further information is given about Waydalic. Paonese is also said to have no
formal comparison (which we might infer to mean “grammatical comparison”), no words for
“good”, “better”, or “best”, and no words for «prestige», «integrity», «individuality», «honor», or
«justice». In more impressionistic terms, Paonese is said to “[present] a picture of a situation
rather than describing an act”.
The lack of formal comparison combined with the absence of lexemes for “good”,
“better”, or “best” potentially has interesting Whorfian implications, and suggests that a
reconstructor should attempt to avoid introducing any means for a Paonese speaker to express
overt relative evaluation or feelings of superiority or inferiority between different situations or
people1. In a living language, it would be nearly impossible to prevent people from finding some
way of doing so if they so desired, but it says something about Paonese culture that they
apparently do not so desire- they do not “feel” the lack. A parallel can be drawn here to the
linguistic culture of the Anarresti people from Ursula K. Le Guin’s novel The Dispossessed
(1974; for comparison, The Languages of Pao was published in 1958). While practically nothing
of their language is actually known, apart from proper names, it is known that the Anarresti
culturally suppress the usage of possessive pronouns and egocentric expressions, even though the
mechanisms for such are technically available to them2. Something similar may be going on in
Paonese, where comparative constructions have not been grammaticalized due to strong social
disapproval of those who attempt to express them.
The remaining lexical requirements, while also meant to be indicative of something
interesting about Paonese culture, are, however, of least use to the reconstructive conlanger, since
it is a trivial matter to simply omit certain lexical items from the dictionary.
Only a single complete sentence of Paonese is ever given. However, there are numerous
titles and names of people and places given throughout the book which may be able to give us
some information on Paonese phonology and morphology. None are ever explicitly identified as
belonging to any particular language, but we can make inferences based on the provenance of
characters and the names of places- i.e., a native of Pao is likely to have a genuinely Paonese
name, while names of foreigners are probably not good evidence of Paonese. Similarly, cities
built for and inhabited by speakers of other languages, despite being located on Pao, may not be
good Paonese names, while pre-upheaval names of Paonese landforms and cities inhabited by
traditional Paonese, with a few obvious exceptions, almost certainly are.
Several factors point to the aforementioned conclusion that the narrator is a linguistically
naïve observer: First, the narrator is unclear about the distinction between suffixes and
postpositions; this could be a matter of genuine uncertainty about the proper analysis, or failure
to clearly explain what the narrator himself actually understands well, but is equally well
explained by a naïve narrator who simply doesn’t know the difference. Second, the narrator
makes use of non-standard terminology in cases that do not present any obvious improvement for
understanding by a lay audience, such as using «volition» in place of «agent», and “agency”
where the more standard term would be “instrument”. Finally, it is known extra-fictionally that
the author, Jack Vance, had little to no formal linguistic training3, and thus cannot be expected to
have produced a narrator who is himself savvy in linguistic analysis. He did, however, travel
extensively, and thus would have been exposed to numerous foreign names and spelling
conventions which may have been reflected in his work (Williams 2006). With these
observational justifications, the assumption of a linguistically naïve narrator provides
significantly increased latitude in interpretation for the reconstructive conlanger. This is both a
blessing, in that it allows us to resolve some apparent problems by arguing that the narrator’s
descriptions were simply wrong, but also a curse, in that we lose confidence in some of the
evidence we have and thus must put more effort into original creation.
1 Thanks to John Quijada for pointing this out.
2 Intrafictionally, the Anarresti language, Pravic, is actually a conlang, voluntarily adopted by the original settlers of
Annares in order to further separate them from the culture of their homeworld, Urras, from which they were political
refugees.
3 Vance enrolled in an Army Intelligence program to learn Japanese, but washed out (Williams 2006); this may
explain where he got the idea for postpositions in Paonese, but I could find no documented evidence that he
seriously studied any other language or linguistics in general.
Attested Words
There are three Paonese titles: “Panarch”, “Medallion”, and “Ayudor”. The first two are
fairly obviously either borrowings or translations for the benefit of the reader (“Panarch” from
Greek roots meaning “ruler of all”, “Medallion” from English). This leaves only “Ayudor” as an
example of genuine Paonese. apparently meaning something like “regent”.
There are ten known Paonese personal names: “Aiello Panasper”, “Bustamonte”, “Beran
Panasper”, “Vilnis Therobon”, “Mornune”, “Hessenden Andrade”, “Est Coelho”, “Gitan
Netsko”, “Ercole Paraio”, and “Can”. Paonese names in the book follow the “given name” –
“family name” pattern, but it is unknown whether that is indicative of how names are actually
used by the Paonese, or another adaptation for the reader. Several of these names, however,
appear to be derived from real names on Earth, and might not be considered good evidence of
Paonese. “Panasper” could easily be a compound of Greek “pan” and Latin “asper”4; “Aiello” is
an Italian name (typically a surname, but conversion of surnames to given names is not unusual);
“Andrade” and “Coelho” are both Portuguese surnames; “Bustamonte” is a minor variant of
“Bustamante”, a surname originating in Spain; and “Ercole”is the Italian equivalent of Hercules.5
This leaves us with “Beran”, “Can”, “Est”, “Gitan”, “Hessenden”, “Mornune”, “Netsko”,
“Paraio”, “Therobon”, and “Vilnis” as probable good evidence for Paonese.
There are five unambiguous names of Paonese cities: “Eiljanre”, “Koroi”, “Sherifte”,
“Donaspara”, and “Spyrianthe”. In addition to the cities, we have 24 other assorted place names:
“Pergolai”, “Jhelianse Sea”, “Hylanth Littoral”, “Hylanthus Sea”, “Hyaline Gulf”, “Zelambre
Bay”, “Mervan Pond”, “Ferai”, “Viamne”, “Cantatrino”, “Moravi Inn”, “Muniment Library”,
“Plarth”, “Qurai Peninsula”, “Fraevarth”, “Sgolaph Mountains”, “Mount Droghead”, “Lido”,
“Mathiole”, “Pamalisthen”, “Vredeltope”, “Maesthgelai Peninsula”, “Rovenone Canal”, and
“Beauclare Quarter”.
“Mount Droghead” and “Beauclare Quarter” are a bit problematic; it is difficult to
imagine “head” being anything other than the homographous English word, leading us to wonder
what a “Drog” might be, and “Beauclare” is a real French surname. Perhaps this is an indication
of some French contribution to the original settlement of Pao, in addition to the Italian and
Portuguese influences, but it can hardly be considered a normal Paonese word.
We also have the names of Pao’s eight continents, whose names are transparently derived
from the basic numbers of Paonese’s base-eight numeral system: Aimand (Ai), Shraimand
(Shrai), Vidamand (Vida), Minamand (Mina), Nonamand (Nona), Dronamand (Drona), Hivand
(Hivan), and Impland (Imple).
From the one attested complete sentence, we get the words “Rhomel”, “bogal”, “mous”,
and “es”, along with various other bits of graphological material which seem to represent
exclusively bound morphemes rather than independent words.
Finally, there are a few additional miscellaneous words of probable Paonese origin:
“Mamarone”, an elite soldier; “praesens”, a “vitality word” of unclear specific meaning;
“Auriol”, Pao’s sun; “Kanetsides”, a holiday; “furze”, possibly a kind of wool (collected by
“furze-cutters” who are apparently comparable to shepherds); and, of course, “Pao” itself.
4 Rough, unrefined, sharp, or newly minted; from https://en.wiktionary.org/wiki/asper
5 As previously mentioned, Jack Vance did a great deal of world travelling, including a stay in the Italian village of
Positano, in which he wrote Vandals in the Void in 1951 (Williams 2006), and on which he based a later novel,
Strange People, Queer Notions. It is thus unsurprising that we should see several examples of specifically Italian
and other Romance names dusted off for use in The Languages of Pao.
Analysis and Reconstruction
Phonology
No attempt is ever made in the book to explain the proper pronunciation of any Paonese
words. We are left to guess at what the pronunciations may be based on the assumption that they
represent only the attempt of a native English speaker to transcribe foreign words according to
his own spelling system, with no particular thought given to consistency in transliteration.
Unfortunately, English spelling is famous for its inconsistency, and we are left with no choice but
to simply guess at many of these words. Doubly unfortunately, it is impossible to tell for sure if
Paonese may have, or been intended to have, phonemes entirely foreign to English; this
reconstruction will therefore necessarily produce a phonetic inventory that is very similar to that
of English, despite attempts to inject plausible variations from Anglophone phonology where the
evidence is ambiguous. Certain words, however have unusual features that may provide more
objective evidence for some non-Anglophone aspects of Paonese pronunciation.
The place name “Sgolaph” is unusual in its apparent juxtaposition of an unvoiced [s] and
voiced [g]. Although voicing of the
place name “Maesthgelai”. In the latter case, this could be explained by an apparent cluster
actually being broken across a syllable boundary; in the first case, however, the word-initial
environment means that
English frequently uses the grapheme
specifically employs the spelling
We have already identified some Italian influence, so this may have been the way Jack Vance
intentionally chose to spell [zg].
There are several ways to approach “Sgolaph”, but let’s first back up and discuss the
matter of voicing in general. English has a phonemic voicing distinction in all of our stops and
most fricatives (with the exception of /h/). To a native English ear, this seems pretty
straightforward- “b” is voiced, “p” is not; “d” is voiced, “t” is not; and so forth. At a phonetic
level, however, things are not so simple, and this can result in Anglophone listeners mis-
identifying the relevant phonemic distinctions in languages that do not have the same range of
allophonic variation as English, and distinguish different series of stops and fricatives by
different particular combinations of phonetic features. So, while it is tempting to simply declare
that the phonemes of Paonese correspond exactly to how an English speaker would pronounce
the Paonese words as Jack Vance wrote them7, there are good reasons to dig a little deeper.
Stopping there would result in a fairly boring phonology; and furthermore, it is simply unlikely
that a far-future language of an alien8 culture like Paonese would happen to reproduce English
phonology so closely. We can do better.
The English stops /b/, /d/, and /g/ are only fully voiced in intervocalic positions, and are
only partially voiced in initial and final positions (i.e., following or preceding silence).
6 Thanks to John Quijada for the suggested example.
7 It is also tempting to say that Paonese phonemes correspond exactly to the letters of Paonese words as Jack Vance
wrote them; but the fact is that Jack Vance was an English speaker, writing for English speakers, most likely using
the conventions of English orthography to convey his intended sounds. And given the weirdness of English
orthography, it is a safe assumption that neither he nor any of his readers would ever have used such simplistic
pronunciations.
8 “Alien” in the sense of “not from around here and unfamiliar to us”, not in the sense of “non-human beings from
outer space”.
Accordingly, many English phoneticians prefer the categorization of fortis (stronger airflow) vs.
lenis (weaker airflow) to distinguish /p/, /t/, /k/ from /b/, /d/, /g/, respectively (Yavaş 2011). In
particular, when preceded by an /s/, even across syllable or word boundaries boundaries, the
phonetic distinction between “voiced” and “unvoiced” English stops is purely lenis vs. fortis;
there is no difference in actual phonetic voicing in minimal pairs such as “discussed” (/diskʌst/)
vs. “disgust” (/disgʌst/) or “disperse” (/dɪspɚs/) vs. “disburse” (/dɪsbɚr/).
Given the relative weakness of the phonetic voicing distinction in English, two other
major features serve to help distinguish the fortis and lenis stops. First, vowels and nasals are
lengthened before lenis (voiced) stops relative to their pronunciations before fortis stops. Second,
lenis (voiceless) stops are aspirated in stressed syllable-initial positions. It turns out that simply
dropping the aspiration from a voiceless stop in a stressed position is sufficient for English
speakers to confuse it with a lenis (voiced) stop; the onset of voicing in a following vowel is
sufficient to move such stops into the perceptual range of allophones of the partially-voiced lenis
stop series. This can cause a great deal of difficulty for English speakers trying work with
languages that have a phonemic aspiration distinction, a fact with which I became intimately
familiar over a semester of trying to elicit linguistic data from a speaker of Eastern Armenian
If Paonese distinguishes its stop series by aspiration, this would provide us with some
flexibility in re-analyzing the voicing status of various stops in the attested corpus, but, more
importantly, would allow us to make strong inferences about where stresses must fall in order to
produce perceptions in an Anglophone listener of voiced and voiceless stops in the attested
positions.
Eastern Armenian makes a three-way distinction between fully voiced stops, voiceless
unaspirated stops, and aspirated stops. Hindi goes even further, with four series of stops covering
all combinations of voiced vs. voiceless and aspirated vs. unaspirated. There are, however, good
reasons for deciding on a two-way distinction based purely on aspiration in Paonese: mainly, the
fact that assuming only a two-way distinction with no phonemically voiced stops provides much
greater constraints on our potential guesses on other features of proper pronunciation. If we
assume that they are actually all simply unaspirated, we avoid having to make separate guesses
as to the phonetic status of every instance of an apparently-voiced stop. Additionally, we can
conclude that apparently-voiced stops cannot occur in unstressed syllables; otherwise, they
would have been properly perceived as unvoiced by the Anglophone narrator, and transcribed as
such (with a few regular exceptions; see below). This allows us to determine the stress patterns
of several words. Unfortunately, we are left guessing which unvoiced stops are aspirated or not
in unstressed syllables.
There are four words with absolutely unambiguous cases of apparently-voiced stops
occurring as the entirety of a syllable onset due to being the initial consonant of a word:
Beran
Donaspara
Gitan
bogal
The first three of these must have initial stress. The last, “bogal”, however, presents a
potential problem with two apparently-voiced stops attempting to attract the stress. If we make
the conservative syllabification assumption that a single intervocalic consonant will be grouped
as the onset of the following syllable, rather than as the coda of the preceding syllable, then we
also have four more possible examples:
Ayudor
Lido
Vredeltope
Therobon
Depending on the placement of stress in these first three words, the
in fact be in a position where /d/ and /t/ are neutralized to /ɾ/ in English; thus there is more than
one reason why a /d/ may have been misheard here, and we cannot conclude that it must be
because of following stress. Additionally, it is possible that, just as aspiration is an allophonic
variation in English unvoiced stops, voicing may be an allophonic variation in Paonese
unaspirated stops. All four of these words, plus “bogal” from the first set, have apparently-voiced
stops in intervocalic positions, a prime environment for triggering allophonic voicing. This
would producing voiced stops that are indistinguishable from other unaspirated stops to a native
Paonese speaker, but are heard, and thus written, differently by the Anglophone narrator. Since
only the second voiced stop in “bogal” can be explained by allophonic voicing, we can conclude
that the stress in that word must again fall on the initial syllable after all.
Three more words are ambiguous due to uncertainty about proper syllabification:
Hessenden Maesthgelai
Pergolai
Under the assumption that a syllable boundary falls between
has final stress, unless nasals also contribute to the environment for allophonic voicing.
Under the assumption that a syllable boundary falls between
contributes to the environment for allophonic voicing.
If we ignore the complications of neutralization and possible intervocalic voicing, we end
up with three examples of unambiguous initial stress, four of final stress, three examples of some
form of medial stress, and one indeterminate, with no obvious patterns to determine the rules for
each. The simplest conclusion in that case is that Paonese has unpredictable lexical stress. That’s
a rather unsatisfying result, as it tells us nothing useful about the rest of the attested corpus. If,
however, we allow for allophonic voicing of unaspirated stops, then we must throw out several
ambiguous examples, but we are left with four examples of initial stress and one definite and one
possible example of medial stress, where the possibly-medially-stressed words (“Pergolai” and
“Maesthgelai”) appear to share a suffix <-lai>, which provides a reasonable morphophonological
explanation for why their stress might shift.
To motivate the identification of this repeated sequence as a morphological suffix, rather
than a mere phonological coincidence, we can even make a fairly good guess at what it means.
No other words end in
Qurai. “Ai” and “Shrai” are numerals, and presumably monomorphemic, but “Qurai” is the name
of a peninsula, as is “Maesthgelai”, and “Ferai” is the name of an island. It is not unusual for
natural languages to use a single word for “island” and “peninsula”, and so the common string-
suffix may actually be a genuine morphological suffix (or compounded root) meaning
“island/peninsula”. In this case, the actual suffix may be
“Maesthegelai” only incidentally sharing an
deleted the
explaining more data all at once, we will assume a basic form of <-lai> for the suffix, and an l-
deletion rule.
9 Or antipenultimate, if
is that Vance would have been likely to use an apostrophe or some similar convention to separate the vowels if they
were meant to be separate syllables, and it is most reasonable to assume that terminal
diphthong like /aʲ/.
Now, we can return to the matter of “Sgolaph”. This word should have initial stress.
Unvoiced stops in English are not actually aspirated after /s/, but they are more fortis than
phonemically-voiced stops. It is not obvious just how much fortition a native Paonese speaker
would put into a normal unaspirated stop, so this could really go either way; a sufficiently lenis
[k] would be mis-heard as /g/.
The known Italian influence on Pao suggests that
as /z/ intentionally, and this is plausible if a voiced fricative contributes to allophonic voicing just
like nasals and vowels and /r/ might. However, while there are several instances of +stop (in
“Est” and, of course, “Sgolaph”, where the cluster must be contained in a single syllable; and in
“Netsko” and “Donaspara”, where it may be across syllable boundaries) and an
(“Sherifte”, which may be also be across a syllable boundary), there are no obvious instances of
any clusters of a voiced fricative+stop, which suggests that this should also still be treated as
an /s/, not a /z/. But that evidence is also not conclusive; it is entirely possible that clusters with
voiced fricatives are allowed, and just rare. In English, after all, we restrict voiced clusters /zd/ to
coda positions (as in “buzzed”10) or split across syllable boundaries, while all three unvoiced /s/
+stop clusters (/st/, /sp/, /sk/) can occur anywhere, but they do exist. Furthermore, while the
environments are not identical and thus there may be variation in lenis/fortis pronunciations, a
word like “Netsko” should also have come out as “Netsgo”, if we assume initial stress (so the
argument applies to “Spyrianthe”- the environment is the same, so either the
is aspirated, the
With no extrafictional evidence, I would still be more inclined towards the “lenis
pronunciation” = “misheard /sk/” explanation, because that results in a more restricted, and thus
more distinctive, phonology. However, since we know that Vance was exposed to Italian, and
borrowed Italian names, I think we must come down on the side of the fence that admits that [zg]
is the actual intended pronunciation. We thus conclude that any voiced segment, including
fricatives, can trigger allophonic voicing in unaspirated stops, and that Paonese actually allows
clusters of phonemic /zk/ (realized as phonetic [zg], and presumably /zt/ and /zp/ as well) at low
frequency.
The typical level of fortition in Paonese unaspirated stops, while now irrelevant to the
interpretation of “Sgolaph”, does constrain the interpretation of every other instance of +stop
as aspirated or unaspirated; if the average fortition level is low, they must all be aspirated. If it is
high, aspiration is ambiguous. Low fortition thus provides greater constraints, which is
attractive- but it results in a corpus that contains no instances of followed by an unaspirated
stop, without providing any evidence that such sequences ought to be disallowed. That just feels
wrong, so I will assume a fairly uniform level of high fortition in all stops (or at least those
following /s/), so as to be able to introduce a better mixture of aspirated and unaspirated stops
into the corpus arbitrarily.
As a final note on voicing, we should address what happens to unaspirated stops on word
boundaries, where they may have a voiced segment on one side but silence on the other. For
initial stops, we could assume that they become partially voiced, in anticipation of the following
vowel; since Paonese does not distinguish voicing, there is no pressure for speakers not to make
that “error” to ease pronunciation, and eventually cement it as an allophonic rule. But, remember
that English voiced stops are themselves only partially voiced in initial positions! Thus, such a
rule in Paonese would make all initial unaspirated stops sound exactly like English voiced stops
10 Thanks to Jim Henry for the suggested example.
in all conditions, removing our justification for identifying stress. In order to preserve the
integrity of our analysis so far, we must therefore assume that this variation does not occur in
standard Paonese pronunciation, and unaspirated stops which are preceded by silent or another
unvoiced segment do remain completely devoiced.
Allophonic voicing rules for the end of a word, however, must account for the proper
perception of the names of continents, which all end in <-mand>. English /nt/ in terminal
position is not aspirated, and English speakers can still reliably tell the difference between [nt]
and [nd]. Differences in vowel length could account for that, but if we assume that Paonese
speakers do not change their vowel lengths in response to the voicing of following consonants
(and why should they? That is an English feature, not a universal human one!), that should tilt
Anglophone perception towards an accurate /nt/ instead. The simplest explanation for this
appears to be that a preceding nasal will cause voicing in a following unaspirated stop in all
positions, even if it is word-terminal and not sandwiched with other voiced sound. This implies
that the terminal
voicing. The only other example we have a pure-Paonese word ending in a stop is “Vredeltope”11
(assuming a silent
is aspirated, or a vowel, unlike a nasal, does
not trigger allophonic voicing, or both. In order to again allow more freedom in assigning
aspirations, we will assume that only nasals trigger voicing in stops that are not followed by
another voiced segment (including at the end of a word). There is some theoretical justification
for this restriction, as homorganic nasals and plosives can be be perceived as a single unit (a pre-
nasalized stop), where a vowel and following consonant would not be, Terminal unaspirated
stops may also be unreleased, although this is not strictly necessary to explain the differences in
Anglophone perception.
We can thus formulate a few general rules for Paonese phonology so far:
1. Paonese roots have default initial stress.
2. Stress may be modified by stress-attracting affixes, such as -lai.
3. Unaspirated stops undergo voicing between other voiced phonemes or after a nasal.
4.
/l/ is deleted after /r/.
“Qurai”, “Jhelianse”, and “Rhomel” provide some evidence for the possibility of
additional non-English sounds in Paonese by the simple fact that they appear to violate English
orthographic rules. It would seem an odd coincidence if “Qu” were not intended as a digraph, as
it is in English. However, that digraph, in languages that use it, is typically restricted to
representing /k/, /kw/, or /kʷ/, while other letters are already widely attested for /k/ and there is
no other evidence of a Paonese /w/. Thus, we will assume the in this case actually represents
an independent vowel, and is meant to represent some other sound. Somewhat arbitrarily,
we will assign its IPA value as a voiceless uvular stop.
The apparent digraph
which varies in realization between /h/ and /ʒ/ according to dialect. In Esperanto and the
romanization of Kurdish12, it’s also a notation for /ʒ/. If, however,
the value of plain
then that is the only attested instance of a voiced affricate, and one of only three possible
11 We also have a marginal example in “Drog” (from “Droghead”), but I’m hesitant to use that as evidence for or
against any feature of word-final stops in Paonese since the only place it appears is compounded with English.
12 Although the amount of exposure Jack Vance may have had to Esperanto or romanized Kurdish is questionable,
there isn’t a whole lot else to go on.
affricates of any kind. The other potential examples are both /ts/, but in both cases (“Netsko” and
“Kanetsides”) it is equally reasonable to assume that there is an intervening syllable break and
assume that
a cluster of voiced and voiceless fricatives at the beginning of a word would be truly odd.
Another option is to assume that Vance added an ‘h’ to the spelling simply so that readers
would be less likely to pronounce “Jhelianse” with an initial /dʒ/ (the typical English value for
we happen to have only a single attested example. The medial
pronounced simply as /ʒ/, and written differently strictly as a matter of orthographic convention.
If the
which it could contrast, there would be a stronger case for analysis as a weird non-English
phoneme of some sort; but with no other obvious leads to go on, this seems like the best
conclusion.
Handling “Rhomel” seems relatively straightforward. The initial
interpreted as an odd, but apparently fine for Paonese, sequence of two phonemes, /rh/. However,
the digraph
to indicate what was at one point a voiceless ‘r’ sound, as in “rhapsody”, from Classical Greek
ῥαψῳδία (which begins with a voiceless rho) . Thus, that is the interpretation we will adopt for
Paonese. It is tempting to posit an analogous unvoiced ‘l’ based on the name “Coelho”- but as
was noted in the overview, this is a straight borrowing from Portuguese, and so likely represents
a fossilized spelling regardless of how the Paonese actually pronounce the name, which provides
no good evidence.
We must still decide on the proper pronunciation of several other letters. As the English
[ɹ] is relatively uncommon throughout the world, we will assume a trilled [r] for the
grapheme instead. The digraph
taken as an arbitrary stylistic variation on
values, with the following exceptions:
1. The digraph “ae” will be taken to represent [e].
2. The letter will be assumed to indicate a diphthong-forming /j/ whenever it follows
another vowel.
3. The letter
4. The letter
5. The letter
in English orthography when, according to my own intuition, it sounds good. In these
cases, the
6. other instances
closed syllables13).
As previously noted, it is impossible to tell whether a particular unvoiced stop should be
aspirated or not in certain environments; those cases are decided arbitrarily, so as to produce a
reasonable mixture of aspirated and unaspirated examples. With issues of proper pronunciation
decided, the complete corpus of attested Paonese words with transcriptions is as follows, with
instances of arbitrary or exceptional decisions indicated in footnotes:
13 Closed syllables containing /e/ having been written with
Beran
Can
Est
Gitan
/’pe.ran/
/kʰan/
/ɛst/
/’ki.tʰan/
Netsko
Paraio
/’nɛt.sko/
/’pʰa.raʲ.o/
Donaspara
Eiljanre
Koroi
/’to.na.spʰa.ra/15
/’eʲl.ʒan.re/
/’kʰo.roʲ/
Cantatrino
Ferai
Fraevarth
Hyaline Gulf
Hylanth Littoral
Hylanthus Sea
Jhelianse Sea
Lido
Maesthgelai
Mathiole
Mervan Pond
Moravi Inn
/’kʰan.tʰa.tʰri.no/
/’fe.raʲ/
/’fre.varθ/
/’haʲ.a.lin/
/’haʲ.lanθ/
/haʲ.’lan.θus/17
/’ʒe.li.an.se/
/’li.to/
/mesθ.’ke.laʲ/
/’ma.θi.ol/
/’mɛr.van/
/’mo.ra.vi/
First Names
Last Names
Cities
Place Names
Hessenden
Mornune
Vilnis
/’hɛ.sɛn.tɛn/14
/’mor.nu.ne/
/’vil.nis/
Therobon
/’θe.ro.pon/
Sherifte
Spyrianthe
/’ʃe.rif.te/
/’spaʲ.ri.an.θe/16
Droghead
Muniment
Pamalisthen
Pergolai
Plarth
Rovenone Canal
Sgolaph
Viamne
Vredeltope
Qurai Peninsula
Zelambre Bay
/trok/
/’mu.ni.mɛntʰ/
/’pʰa.ma.lis.θen/
/pɛr.’ko.laʲ/
/plarθ/18
/’ro.ve.non/
/’zko.laf/
/’vi.am.ne/
/’vre.tɛl.top/
/qu.’raʲ/
/’ze.lampr/
14 “Hessenden” represents a particularly special case; the initial
given the conventions of English orthography, /he.sɛn.tɛn/ simply sounds indefensibly wrong. Note also that I have
chosen to interpret the doubled
the context of the second
consonant, in the absence of any other obvious phonetic significance to that choice of English orthography.
15 Here, the ‘p’ actually lands in a logical location for secondary stress- one trochaic foot after the primary stress.
Still, given the post- positioning, adding aspiration is still somewhat arbitrarily.
16 Aspiration omitted arbitrarily.
17 /-us/ arbitrarily designated as a stress-attracting suffix, because that is how I pronounced it on my first time
through the book.
18 Aspiration omitted arbitrarily.
Ai
Shrai
Vida
Mina
Aimand
Shraimand
Vidamand
Minamand
Auriol
Ayudor
bogal
es
en
furze
Kanetsides
/aʲ/
/ʃraʲ/
/’vi.ta/
/’mi.na/
/’aʲ.mant/
/’ʃraʲ.mant/
/’vi.ta.mant/
/’mi.na.mant/
/’a.u.ri.ol/
/’aʲ.u.tor/
/’po.kal/
/ɛs/
/ɛn/
/furz/
/’kʰa.nɛt.saʲtz/
Numerals
Continents
Miscellaneous
Nona
Drona
Hivan
Imple
/’no.na/
/’tro.na/
/’hi.van/
/impl/
Nonamand
Dronamand
Hivand
Impland
/’no.na.mant/
/’tro.na.mant/
/’hi.vant/
/’imp.lant/
Mamarone
mous
nli
Pao
praesens
Rhomel
ro
/’ma.ma.ron/
/’mo.us/
/nli/
/’pa.o/
/’pre.sɛnz/19
/’rr o.mɛl/
/ro/
Now that we have the whole corpus in phonemic transcription, we can start doing some
analysis to discover higher-level phonological rules and characteristics. The complete attested
consonant inventory is as follows, with total number of occurrences listed beneath each
phoneme20:
Bilabial
Labio-
dental
Dental
Alveolar
Post-
alveolar
Velar
Uvular
Glottal
Plosive
p pʰ
12 3
t tʰ
24 4
k kʰ
7 4
q
1
m
24
Nasal
Trill
Fricative
Lateral
f v
5 11
n
48
r rr
32 1
θ
9
l
21
s z
14 5
ʃ ʒ
3 2
h
6
19 Aspiration omitted arbitrarily
20 Note that, while transcriptions were of necessity done by hand, all of the statistics were automatically compiled
by a short Python script, which was a major labor-saver. Learning a little basic computer programming comes highly
recommended for the modern linguist, or conlanger.
There are a few gaps (e.g., no voiced /ð/, no aspirated /qʰ/, no devoiced /l/ to match the
/rr /), but nothing glaringly weird, so I feel pretty good about declaring that this is the complete
consonant inventory of Paonese. And while 68 words may not really be a statistically significant
sample, we can get a general feel for appropriate relative frequencies of different phonemes- e.g.,
unaspirated stops are way more common than aspirated ones, /n/ occurs about twice as often
as /m/, /q/ and /rr / each occur about once per 68 words (maybe a little more, maybe a little less)
For the vowel inventory, we have six pure vowels and three diphthongs:
ɛ
14
a
48
aʲ
15
e
19
eʲ
1
i
24
u
8
o
29
oʲ
1
Once again, there’s nothing particularly weird here, so I feel pretty good about declaring
this the complete Paonese vowel inventory, with close-to-accurate relative frequencies.
Once we have some basic phonotactic / syllable structure rules, we could plug those
frequencies into an automatic random word generator and start building a neo-Paonese
dictionary. One approach to that which would vastly simplify the work of analysis would be to
simply build a statistical model of frequencies for all attested pairs of phonemes, and all attested
triples of phonemes, and then just feed that all into the random word generator; but on such a
small corpus, that approach would be subject to severe overfitting. And besides that, it just feels
like cheating.
To help figure out the syllable structure rules, I split all the words on (my arbitrarily
decided) syllable boundaries, and then separately counted up all of the complete syllables, all of
the unique onsets, and all of the unique codas. The frequencies for onsets and codas are as
follows:
Onsets
Codas
-: 21
m: 20
n: 18
r: 15
l: 11
v: 10
t: 9
h: 6
p: 5
θ: 5
k: 4
kʰ: 4
s : 4
tr: 3
f : 2
pʰ: 2
ʃr: 2
tʰ: 2
ʒ : 2
fr: 1
nl: 1
pl: 1
pr : 1
q : 1
rr : 1
sk : 1
sp : 1
spʰ: 1
ʃ : 1
tʰr: 1
vr : 1
z : 1
zk : 1
– : 98
n : 18
nt: 8
l : 7
s : 5
r : 4
f : 2
rθ: 2
t : 2
k : 1
m : 1
mp : 1
mpl: 1
mpr: 1
ntʰ: 1
nθ : 1
nz : 1
p : 1
rz : 1
st : 1
sθ : 1
tz : 1
A few things are immediately clear: Paonese vastly prefers open syllables, with no codas.
While different from English, in which only about 40% of syllables are open, this is fairly
normal cross-linguistically, and similar to languages like Italian and Spanish (Kubozono 1995).
This may reflect some additional subconscious Italian influence on Vance’s word-creation. More
extreme cases are not uncommon- nearly 90% of all Japanese syllables are open, for example,
and numerous languages allow only CV-structured syllables, of which 100% are open by
definition.
Zero-onsets, in contrast, while certainly not rare in the Paonese corpus, are not that much
more common than some of the other single-consonant onsets. There is also a lot more variety in
possible onsets than there is in codas. Several phonemes (/ʒ/, /ʃ/, /h/, /q/ and /rr /) can only occur in
onsets. Also, there only two attested three-consonant clusters, and only in codas.
While the existence of coda consonant clusters technically place Paonese in the “complex
syllable structure” category according to the World Atlas of Language Structures, it is definitely
on the lower end, placing it close to the middle of the range of cross-linguistic variation
(Maddieson 2013). With this level of syllable structure complexity, we would expect to see
between 22 and 26 phonemic consonants in the phonetic inventory; the 20 phonemic consonants
identified above would more typically be associated with a simple CV syllable structure, but this
is still well within the range of known naturalistic variation. We can thus be fairly confident that
my intuitive choices for syllable boundary placement were reasonable.
Unlike the basic phonetic inventory, I am not willing to accept that this data represents
the complete set of all possible onsets and codas in Paonese, simply because the space of
possible two-letter combinations is much larger than the space of single segments, and this
corpus is far too small to be considered a representative sample of everything that Paonese might
allow. We can use these frequencies for certain kinds of onsets and codas to improve a-priori
word generation over what we would get from just sticking individual segments together with the
right total frequencies; but, given that we have already accepted a certain minimum complexity
in allowed Paonese syllable structure, some of these patterns should first be generalized.
Ignoring frequencies for now and looking only possibilities, reasonable generalizations
for onsets seem to be that an onset can be formed from
1. any individual consonant.
2. any other consonant followed by an /r/ or /l/.
3. an /s/ or /z/ followed by any stop.
Codas, in turn, can be formed from
1. any individual coda consonant,
2. any coda consonant followed by /θ/,
3. any non-fricative coda consonant followed by /z/,
4. an /s/ followed by any non-fricative coda consonant,
5. a nasal followed by a homorganic stop (/t/, /p/, /tʰ/ or /pʰ/; no /k/, because there is no
matching velar nasal),
6. any stop or homorganic nasal plus stop, followed by an /r/ or /l/ in word-final position,
where a coda consonant is any consonant other than /ʒ/, /ʃ/, /h/, /q/ or /rr /. It may seem like
the larger number of rules for codas vs. onsets contradicts the original observation that onsets
have more variety, but note that the rules for codas are actually much more restrictive, allowing
only specific special cases. Other rules that fit the data are possible, but these ones seem
reasonable and aesthetically pleasing to me.
If we then examine the set of complete syllables, it turns out that there are four vowels
that are used as complete syllables by themselves, with no onset or coda: /aʲ/ (3), /a/ (2), /o/ (2),
and /u/ (2). One of these (/aʲ/) is attested as a complete word, while the other three at least occur
in hiatus with other vowels on either side. The simplest generalization of this evidence is that any
vowel can form a complete syllable by itself.
It is theoretically possible at this point to look more extensively at the set of complete
attested syllables, and try to come up with additional rules about what onsets and codas may be
allowed with each other or with certain vowels, or what kinds of codas and onsets are allowed in
hiatus within a word. But, the data is just too sparse to make really strong conclusions, so we
might as well just say “anything goes”; once you have well-formed onsets and codas with vowels
to put in between, they can be combined however you like. There are only three inter-syllabic
restrictions which have been noted earlier, and which are restated and clarified here:
1. An /i/ cannot immediately follow any other vowel in the same word (with /a/, /e/, and /o/,
it forms diphthongs, and combinations with other vowels just won’t happen).
2. An onset /l/ is deleted if it comes in contact with a coda /r/.
3. Nasal+Stop+{/l/, /r/} codas can only occur word finally. If brought into contact with
another syllable due to suffixing or compounding, they are re-syllabified, moving the /r/
or /l/ to the onset of the next existing syllable if possible and inserting an epenthetic /a/ to
take the {/r/, /l/} onset otherwise.
The existence of /impl/ ~ /imp.lant/ suggests that maybe there should be a following-
onset-deletion rule instead of the epenthetic /a/; however, I assume that we can treat “Impland”
as an ancient fossilized form that does not necessarily reflect synchronic morphophonological
principles, and deleting onsets feels like it would produce uncomfortable levels of homophony.
The epenthetic vowel is chosen to be /a/ because /a/ is the most frequent vowel phoneme in the
corpus by nearly a factor of two, and so seems like a good choice for a “neutral” vowel.
Sentences & Morphosyntax
Although, as previously mentioned, there is only one complete sentence of Paonese, there
are two sentences glossed: the first actually provides the Paonese words, while the second has
only the free English translation and the morpheme-by-morpheme gloss. The complete sentence
is as follows:
Rhomel-en-shrai bogal-Mercantil-nli-en mous-es-nli-ro.
Statement-of-importance – in a state of readiness – two; ear-of Mercantil – in a state of
readiness; mouth – of this person here – in a state of volition
«There are two matters I wish to discuss with you.»
It is claimed that «The italicized words represent suffixes of condition.» I suspect
italicizing Statement-of-importance is simply a typographical error in the text of the book, as it is
clearly not a suffix, being the first thing in the utterance.
This first gloss gives us some insight into morphological structure and some interesting
morphemes. The word shrai for “two” we already know. Rhomel could be interpreted as a
grammatical morpheme for emphasizing the importance of an utterance, but it seems to make
more sense as a noun for “something important”. The word bogal clearly means “ear”, while
mous must be mouth. Mercantil is a proper name referring to the addressee of this utterance. We
also have the following grammatical morphemes:
-en
-ro
«readiness»
«volition» (agent)
-nli
es
genitive
«this person»
The «readiness» marker might be an existential, or perhaps a topic marker, or it could
literally mean only “ready for something”. The morphemes -en and -ro, described as suffixes of
condition, are the best matches we have so far for the expected “post-positions”; they seem to
have functional, adpositional meanings, and the fact that every noun seems to require one
matches the description of postpositions as main component of the language. However, given
that they are written attached to the previous root, and can have additional material appended to
the same orthographic word, it makes most sense to abandon the postpositional analysis for now
and instead simply call these suffixes, as explained along with the gloss as it appears in the text.
The only direct evidence we have for what kind of material can be appended to a noun-
condition complex is a simple numeral, in this case the number “two”; but, it seems reasonable to
assume that the closed class of simple numerals can probably be extended with morphemes for,
e.g., «few», «none», and «many» to allow for the expression of non-specific plurality as desired,
without having number as an obligatory grammatical category. There’s no indication of how
larger numbers are formed, however, and those often have different grammatical behavior in
natural languages.
Genitives apparently follow their heads in compounds and use a postposed genitive
marker nli.
It is particularly interesting that es, the obvious translational equivalent for «I», is glossed
as «this person here» (since we know in context that Mercantil is the addressee, we can assume
that that is the source of the pronoun “you” in the English translation). There isn’t any explicit
discussion of Paonese pronouns21, so we cannot tell whether there is some other word that means
only “I”, or any other specific personal pronouns. Given what we know of the highly communal,
non-individualistic nature of Paonese culture and the extra-fictional knowledge that the original
Paonese as well as the artificial languages that replace it are meant to reflect the strong Sapir-
Whorf hypothesis, it is reasonable to assume that Paonese in fact does not have any explicit
personal pronouns. In that case, es and other translational equivalents for English pronouns may
in fact have more in common with Japanese pronominals; i.e., while conventionally used for «I»,
es does not literally mean only “first person singular”, and can be employed as a more general
noun, most likely referring to any nearby or highly relevant person in the proper discourse
context. We know, however, that es is almost certainly exclusively animate, if not exclusively
human, and thus we can predict an animate/inanimate split in the set of conventional pronominal
words.
The second gloss we have is as follows:
«The farmer chops down a tree.»
Farmer in a state of exertion; axe agency; tree in a state of subjection to attack
Contrasting with the previous gloss, we can determine that, although we do not know its
form, there is some suffix for physical exertion as distinct from simple will, which may justify
the development of Paonese specific terminology of “volition” for the first case as opposed to a
21 There is, however, a short discussion of Breakness pronouns, Breakness being the fictional natural language of
the eponymous planet in Pao’s universe. Said discussion serves mainly to explain that the people of Breakness are so
egocentric that there’s no need to bother with an explicit word for «I».
more generic “agent”. Agency in this case, however, almost certainly refers to an instrument in
more standard terminology, and is evidence of another “suffix of condition” for which we do not
know the actual form. Further, there is a third unknown condition suffix for being subject to
attack. This seems strangely specific; it is possible that this is a misleading gloss by the naïve
narrator, but things get more interesting if we assume in this case that the unknown suffix really
means only and specifically that. That level of specificity suggests that “suffixes of condition”
are something like the lexical suffixes in native languages of the American Pacific Northwest.
Unlike lexical suffixes in Native American languages, however, these refer not to concrete
objects, but to verbal concepts. It is possible that the “suffixes of condition” may not even be a
closed class of grammatical morphemes. In contrast, they could be an open class of always-
intransitive verb-like morphemes which are merely required to occur bound in a compound with
a noun.
In any case, these are definitely not complete verbs as they are usually understood; no
single “suffix of condition” controls the meaning of the entire sentence. Rather, the “picture of a
situation” that Paonese is said to represent is built up by contributions from all of the “suffixes of
condition” on all of the nouns in a clause. Leaving the question of whether they form an open or
closed class undecided for the moment, we will from here on adopt the term “semiverbs” to
describe these components of Paonese morphosyntax.
As an incidental cultural note, we also now know that Paonese has some word for each of
“farmer”, “tree”, and “ax”.
Tense & Aspect
In addition to nouns and postpositions, we are told that Paonese contains “temporal
indices” as a major component. Unfortunately, the available example sentences examples leave
us with no direct evidence as to what «temporal indices» might actually be; there’s nothing that
might match that description in either of the given glosses, which indicates that whatever they
are, they must be optional rather than obligatory components of a grammatical sentence. It is
unlikely that they were simply omitted from the gloss as unimportant or hard to translate.
According to the narrator, the sole point of providing the Paonese words for gloss #1 was to
demonstrate that it isn’t substantially longer than the equivalent English utterance, a purpose
which would not be properly served by lying about which words should actually be there. The
known glosses, and the general static nature of Paonese culture in light of the strong Sapir-Whorf
hypothesis, suggest that Paonese probably does not have grammatical tense. That does not,
however, rule out optional morphemes with temporal reference that could pass as “temporal
indices”. Plenty of natural languages get by just fine with no grammatical tense marking, while
still being able to express time when needed by the use of explicit adverbs (like “yesterday”,
“tomorrow”, or “right now”) or other temporal constructions. Mandarin Chinese, for example,
works this way.
Mandarin Chinese does, however, have grammaticalized aspect. Where tense tells you
where an event occurs in time, aspect relates to how an event is embedded in time. The two
concepts are frequently conflated in natural languages, both in terms of how they are marked and
how speakers conceive of their verbal systems, so it is possible that “temporal indices” also have
something to do with aspect.
Furthemore, despite the fact that Jack Vance was untrained as a linguist, the concept of
“tense” should nevertheless be familiar to any educated layman. If, therefore, he meant
“temporal indices” to refer to tense markers, we thus have to wonder why he did not just say
“tense”; this line of reasoning also points towards an aspectual interpretation. One fairly
straightforward explanation may be that he simply recognized that “tense” is not the name of a
part of speech as the word is typically used in English grammar; and rather than confusing the
reader by re-purposing it as such, he invented a novel term. But even in that case, we have to
realize that “tense” and “aspect”, while formally separate grammatical categories, are extensively
conflated in many languages, and extensively confused by linguistically-naïve English speakers.
As a slight digression, we might briefly consider what Paonese could be like if “temporal
indices” do strictly encode tense. If Paonese still had aspectual distinctions, but not in the
“temporal indices”, an alternative possibility would be that Paonese have lexical aspect, like
many Slavic languages. In a language with lexicalized aspect, verbs come in sets of multiple
words, often morphologically derived from each other but not necessarily, which all have the
same basic meaning but encode different aspects. This would be consistent with the lack of
explicit aspectual morphemes in the two attested Paonese glosses, but seems difficult to arrange
when the description of an event is spread out over multiple semiverbs- which one would control
the aspectual marking? Due to the lack of coherent verbs to carry aspectual marking in Paonese,
it seems that any aspectual marking would have to be carried on some other distinct kind of
word, which provides further evidence for an aspectual interpretation of “temporal indices”.
Finally, while at first glance the term “index” seems to me to fit more closely with the
idea of words the specify specific places in time (i.e., tenses), it is also makes sense that these
“indexes” may be meant to indicate only relative ordering of events within a single discourse,
and not necessarily absolute position with respect to the time of utterance; and that interpretation
actually fits better into the Whorfian mold, in which the Paonese language reflects the Paonese
culture, unconcerned with its placement in time. This kind of “relative tense” can be encoded in a
language’s aspectual system, orthogonal to tense, as indeed it is in English, via the perfect aspect
constructions with the auxiliary verb “have” (e.g., “to have gone”- a tenseless perfect-aspect
infinitive, as compared to “go” or “went”- the finite present and past tense forms). Due to this
fact and our knowledge of the ease with which an anglophone narrator could conflate the concept
of aspect with tense, we will therefore settle on the idea that “temporal indices” are optional
adverbial words which primarily encode aspectual information (though some indication of non-
grammaticalized tense cannot be completely ruled out).
Replacing Adjectives
The fact that Paonese is claimed to have no adjectives is comparatively unremarkable and
unproblematic. Plenty of natural languages get by with no distinct class of adjectives, making use
of nouns or verbs to fill that role. Paonese has no easily-identifiable verbs, so nouns it will be.
We already have direct evidence that multiple noun roots can be compounded together in the
presence of the genitive marker nli. Keeping with the intended polysynthetic nature of the
language, it is a small step from there to assume that there may be other combining markers to
indicate various descriptive relations between different roots in a larger compound. We can
further allow for arbitrary sequences of unmarked noun roots in which each component is
interpreted as an alternate descriptor of the same referent, taking the intersection of all of their
individual meanings.
The combining markers like nli have many of the qualities of adpositions, as semiverbs
initially seemed to. If we were to insert word boundaries after the combining markers around
noun roots (an arbitrary decision as Paonese has no morphophonological evidence to definitely
distinguish two words in hiatus from two bound components of a single compound), they would
appear exactly like postpositions, with the exception that postpositional phrases seem to follow
rather than precede their heads; this is typologically unusual, especially when the evidence of
postpositions and semiverbs together indicates the Paonese is generally head-final, but it’s hardly
impossible.
Combining markers are quite distinct from semiverbs in that they specify the relation of
nouns to other nouns, not the relation of a noun to a predicate (or “situation” in Paonese terms).
However, since English does not distinguish adpositions that modify nouns from adpositions that
modify verbs or clauses, and some languages have precisely the opposite restriction (i.e., that
adpositional phrases only or at least primarily modify clauses, not nouns)22, it is not at all
surprising that a naïve narrator could confuse the two, leading to the claim that “post-positions”
are a primary component of the language while ignoring the distinction between these combining
morphemes and the equally essential semiverbs.
This style of compounding allows for the translation of fairly complex noun phrases,
including arbitrarily long sequences of adpositional modifiers and arbitrarily long strings of
adjectives at every level. There is some room for structural ambiguity in terms of whether a
second adpositional modifier attached to a first adpositional modifier, or to the head of the whole
phrase, but that’s no different that what we already have to deal with in English23, and there’s
really no reason to ask for any more in order to get a fully functional, fully expressive human
language. While this was my final decision regarding Paonese, however, I must note that more
precise systems, with no such structural ambiguity, are possible at very little extra cost in
complexity. In particular, I have been inspired by a four-category compounding system described
by Herman Miller in a message to the CONLANG-L mailing list24, implementable in either
morphology (for building polysynthetic compound words) or syntax (for building complex noun
phrases), which allows for arbitrarily complex trees of heads and modifiers. To a native English
speaker like me, it gets a little confuddling fairly quickly, but one can imagine that perhaps a
native speaker of Inuktitut or Turkish would might not find these arbitrarily complex compounds
quite so daunting. Expanding the combinatorial system of Paonese to accommodate such things
would be a perfectly valid alternate choice for a different reconstruction.
Adverbs
Having addressed the issue of adjectives, the logical next step is dealing with adverbs.
Official descriptions of the language do not mention adverbs, but do not rule them out either. If
we are correct about the nature of “temporal indices”, those roots would be effectively adverbial
in nature. Although we have no direct evidence of other adverb roots, it is easy to argue that
Paonese sentences so far are built from nothing but adverbs; each noun+semiverb complex is
effectively an adverb which adds a little bit more to the description of a situation.
Given that starting point, we should expect the evolution over time of unanalyzable
adverb roots from what were originally idiomatic noun+semiverb complexes, filling out the class
22 Standard practice in Classical Latin pedagogy, for example, is to claim that adnominal usages of prepositions are
strictly disallowed (Wharton 2009, 184). While the true situation is somewhat more complex, adnominal
prepositions are less common in Classical Latin than are adverbials.
23 E.g., in “the vine in the pot on the mantle”, is the pot on the mantle, or is the vine merely draped over the mantle,
with no clear indication of the specific location of the pot? If that seems like a distinction without a difference to
you, something that would hardly ever actually matter- well, you’re probably right. Which is why we won’t worry
about it in Paonese, either.
24 https://listserv.brown.edu/archives/cgi-bin/wa?A2=CONLANG;9a6d18ef.1306D
of distinct adverbs. “Temporal indices” may have been singled out for special mention, in
preference to a larger class of adverbs, due to a combination of factors, including the relative
difficulty of distinguishing many other adverbs from simple noun+semiverb complexes and/or a
relatively high frequency of usage compared to other semantic classes of adverbs.
Given that adjectival concepts are represented in Paonese by noun compounding, certain
adverbial concepts that would normally show up as adjective modifiers in English25 could
potentially also be handled by additional derivational morphology on noun roots.
The Unknown & The Unknowable
We are now at the very edge of what we can reasonably tie to evidence from the book.
There are many topics still to be covered and decisions to be made to fill out a complete
description of a language, but they will need to be made on a fairly arbitrary basis.
For example, how do we fill out the pronominal system? It was previously noted that
Paonese may behave much like Japanese in not having a class of strict pronouns distinct from
normal nouns; but even so, we have to ask: what kinds of words do they use for those functions?
Just running with the gloss “this person” for es, I see opportunities for three different three-way
distinctions: three persons, three animacies (human, non-human, inanimate), and three degrees of
proximity (this, that, yonder). We can assume that non-human and inanimate first person don’t
make much sense; nor does inanimate second person. That takes three cells out of the chart, but
still leaves 24 possible personal and demonstrative pronominal forms, and room for plenty of
other irregular additions in the Japanese style. Is that the best way to reconstruct Paonese
pronouns? I don’t know, but it is not contradicted by the evidence.
We also have no evidence for how complex sentences are structured. Does Paonese have
nominalized clauses? Relative clauses? Complementizers? Coordinating or subordinating
conjunctions of any kind? All of this will need to be invented from scratch.
There does not seem to be any strong reason to assume that Paonese does not have
coordinating or subordinating conjunctions. They may not be in the list of “nouns, post-positions,
and temporal indices”, but if you had to describe the major components of English, would you
include conjunctions, or prefer a list like “nouns, verbs, and adjectives”? So we might as well
posit a fairly normal array of useful conjunctions. Given the lack of verb phrases or other clear
clausal structure, however, it feels right to say that Paonese should have separate classes of
conjunctions that join noun+semiverb complexes and conjunctions that join separate clauses,
thus helping you identify where one “situation” ends and another begins. Armenian is an
example of a natural language with just such a distinction.
The decentralized nature of the Paonese clause and the way nominal arguments are
incorporated into verb complexes also make me feel like the clause structure should be relatively
flat- i.e., there should not be embedded relative clauses or nominalized complement clauses.
How, then, would those concepts be handled by Paonese, if we needed to translate them? Well,
instead of “I saw the man who had a telescope” (for example), you would just have to say “I saw
the man and he had a telescope.” That doesn’t sound like a very felicitous paraphrase in English,
because the pragmatic implications are all wrong- but there’s no reason that should have to be the
case in Paonese. For complement clauses, we can simply state the complement and then refer to
it as “that” later on; e.g., “I know you like ice cream” could be “You like ice cream, I know that.”
This is a case where an explicit complementizer could be useful to mark the boundary between
25 E.g., “very”, although that particular example may be marginal due to the need to avoid overt comparison.
clauses, but intonation or a regular conjunction would serve just as well.
All in all, I don’t think Paonese needs complementizers. If it had them, they would
probably came at the end of a clause, given the general head-final nature that we have already
observed. Perhaps some of the inter-clausal conjunctions derive from intra-clausal conjunctions,
postpositions, or semiverbs that fused with the complementizer of the previous clause in an
earlier form of the language; that is something that might be taken into account when generating
additional vocabulary. But, in the synchronic analysis, clause structure can easily be completely
flat, and this seems to better fit the intended feel of the language- not only does it suppress
explicit comparisons, even the syntax banishes any hierarchical ordering of clauses and any idea
of “subordination” to a superior.
And that brings us to the last unknown: what are all the rest of the Paonese words? How
do we fill out the vocabulary? If we were reconstructing a lost natlang, we would try to get
samples of its relatives, and do comparative analysis to figure out what the equivalent words in
this language should be. But even if we had knowledge of related languages, the number of
words that are not proper names or grammatical morphemes, for which we could try to find
reasonable correspondences, is vanishingly small- just the three of “rhomel”, “bogal”, and
“mous”, and perhaps “Ayudor”. Even including the rest of the corpus, there’s barely enough
there to plausibly isolate a couple of re-used morphemes. And we do not, in fact, have any real
knowledge of any related languages. The use of French, Italian, and Portuguese names suggests
that perhaps we should consider the use of Romance roots. But, what we do know is that Paonese
is not a direct descendent of any current Romance language; its evolution was, at the very least,
filtered through the intermediate language Waydalic, in ways that are completely unknowable;
and what little we do know suggests that any original Romance influence would be completely
unrecognizable in the language now. More likely, these Romance elements were introduced as a
superstratum influence on an already-existing Paonese, not as ancestral holdovers. But that, at
least, is something- we can plan to borrow large swaths of Romance vocabulary into Paonese
grammar for science, technology, and “high class” registers, which greatly reduces the work that
must be put into strictly a-priori vocabulary generation. Even so, if anyone else does an
independent reconstruction of a different potential Paonese, this is where it is likely to diverge
the most- even if you come up with the exact same phonetic inventory, and the exact same set of
phonotactic rules, for all of the thousands of common words that you would need to add to make
Paonese a complete usable language, you’ll just have to make them up yourself.
Acknowledgments
First of all, a reconstruction of Paonese, and an article about it, could not possibly exist
without the work of author Jack Vance to provide the raw materials. Mr. Vance died in May of
2013, while I first started working on my reconstruction in July of that same year- only a few
months too late to be able to ask questions of the only authoritative source! While my timing
was, in honesty, coincidental, some portion of this work is therefore dedicated to his memory.
I am also indebted to the reviewers who helped me figure out how to turn my jumble of
nonlinear thoughts into a reasonably well-organized document, and provided several additional
insights along the way which made it back into my reconstruction.
And finally, thanks to the CONLANG-L mailing list and all of its members in general
who have served as a great support to my love of conlangs and education in linguistics over the
years, as well as serving as a great sounding board for ideas during this and other projects.
References
Kubozono, Haruo. 1995. “Perceptual evidence for the mora in Japanese”. In Phonology and
Phonetic Evidence: Papers in Laboratory Phonology IV, 141-157. Cambridge, UK:
Cambridge University Press
Le Guin, Ursula K. 1974. The Dispossessed. New York: Harper Collins.
Maddieson, Ian. 2013. «Syllable Structure». In The World Atlas of Language Structures Online,
Dryer, Matthew S. & Haspelmath, Martin (eds.) Leipzig: Max Planck Institute for
Evolutionary Anthropology. Accessed Oct. 2015. http://wals.info/chapter/12
Wharton, David, “On the Distribution of Adnominal Prepositional Phrases in Latin Prose”,
Classical Philology 104 (2009): 184-207, doi: 10.1086/605342
Williams, David B. 2006. » Vance
Museum
2015. http
:// www
. vancemuseum
. com
– Miscellany
_ bio
/ vance
– Biographical
_1.
htm
Sketch
«. Accessed Sept.
Vance, Jack. 1958. The Languages of Pao. New York: Avalon Books
Yavaş, Mehmet. 2012. “English Consonants”. In Applied English Phonology, Second Edition, 57-
73. Chichester, UK: Blackwell Publishing
