bradrn’s scratchpad

[bradrn]

Why the term Adnominal rather than Determiner for that particular series? Just curious

‘Determiner‘ in IE grammar generally refers to a syntactic slot which can be filled by a whole range of different items: articles, demonstratives, possessives, quantifiers, etc. Eŋes doesn’t have any single slot like that, so the term wouldn’t be accurate. ‘Adnominal’ is more precise.

I always interpreted "determiners" as being a figment of certain linguists' attempts to treat all languages as being dialects of English.

Yep, precisely! It’s a meaningful category in some languages (as in English), but not in most.

[quinterbeck]

Why the term Adnominal rather than Determiner for that particular series? Just curious

‘Determiner‘ in IE grammar generally refers to a syntactic slot which can be filled by a whole range of different items: articles, demonstratives, possessives, quantifiers, etc. Eŋes doesn’t have any single slot like that, so the term wouldn’t be accurate. ‘Adnominal’ is more precise.

Ah! Somehow I never clocked that. I might steal the term, as Ineru has a set of adnominal demonstratives and quantifiers that pattern with corresponding pronominal and adverbial ones. Up to now they've been labelled determiners.

[bradrn]

Nominalisations, participles and passives

This is a rather small part of the verbal system, but an important one. This is mostly due to derivation: Eŋes makes very extensive use of nominalisations and participles for lexeme-formation. Even quite basic vocabulary can be derived via nominalisation, as with e.g. naŋfŋiŋ ‘sand’ (←fŋiŋ ‘be broken’) .

Indeed, this extensive use of nominalisations is one of the most distinctive features of the Sprachbund to which Eŋes belongs. It’s aided by a notable preference towards calques rather than borrowings: e.g. anumŋun ‘Monstera deliciosa’ (← mŋun ‘with holes’) is parallel to Wēchizaŋkəŋ thēchepkɨ ‘ident.’. On the other hand, Eŋes tends not to use nominalisations for grammatical purposes to the same extent as other languages in the area.

There are three basic nominalisation strategies:

• an- forms action nouns from verbs: anwle ‘going’, anlisŋumŋun ‘seeing’, etc.
• naŋ- forms agent nouns from verbs: naŋfeys ‘one who climbs = climber’, nampasinsef ‘one which shines = light’, etc.
• anu- forms concrete nouns from adjectives: anumŋun ‘one with holes = M. deliciosa’, anusim ‘small one’, etc. (This seems transparently derived from anu ‘thing’.)

Beyond these three, there is the passive participle lse-, which applies only to underived (i.e. non-causative) transitive verbs, yielding adjectives: lsewserfin ‘built’, lselisŋumŋun ‘seen’. Several constructions are built on this:

• The passive walse- is the stative auxiliary of the participle: walsewserfin ‘be built’, etc.
• The dynamic passive malse- is the inchoative auxiliary of the participle: malselisŋumŋun ‘got seen’, etc.
• The patient noun anse- (from older anulse-) is the nominalisation of the participle: ansewserfin ‘something built = building’.

Finally, there are two less common nominalising forms, both ambiguous between an agent and patient interpretation:

• Some verbs correspond irregularly to a noun formed through vowel change, consonant addition and other internal modifications. Particularly common is a form like -eʔ placed within the verbal root: feʔ ‘place’ vs √f- ‘be located’, rweʔ ‘possession’ vs √rw- ‘have’.
• A circumfix maʔ--to gives a slightly more formal and less lexicalised noun: maʔpasinsefto ‘something which shines’, maʔwamŋunto ‘something with holes’.

The status of verbal arguments varies between constructions. Action nominals can take a subject as inalienable possessor, and be immediately followed by an object [though this will depend on me working out the possessive system, of course]. Agent nouns can on occasion take a following object, but obviously cannot take any subject. The rest of the nominalisers cannot take any arguments at all.

I haven’t yet talked about the verb core or serial verb constructions, but they also interact with these constructions in different ways. Action and agent nominalisations seem to be positioned outside the verb core: anwgiwle ‘seeming to go’ is an-[wgi-wle] ‘NMLZ-[seem-go]’, etc. On the other hand, the participle is more ambiguous, since a passive can definitely be used as a single component in an SVC: to construct an example, wasoŋwalsewserfin ‘build pile by throwing sth. down’ is [w-asoŋ]-[walse-wserfin] ‘[AUX-fall]-[PASS-build]’. Of the last two forms, irregular derivation applies only to invididual verb stems, while the circumfix can surround any verb core — making it the only way to build negated nominalisations.

To show the whole verbal system in context, here’s a non-exhaustive list of the most common stems derived from the single transitive verb root √ndiʔs-ʔmum ‘related to writing, drawing’:

• Iterative verb: ndiʔssiʔmum ‘write. handwrite’
  • With auxiliary: wandiʔssiʔmum ‘write with, dictate’
    • Agent noun: naŋwandiʔssiʔmum ‘orator’
    • Participle: *lsewandiʔssiʔmum ‘dictated-to’ (rare)
      • Patient noun: ansewandiʔssiʔmum ‘transcriber’
  • Action noun: anndiʔssiʔmum ‘action of writing’
  • Agent noun: naŋndiʔssiʔmum ‘writer, scribe’
  • Participle: lsendiʔssiʔmum ‘written, handwritten’
    • Dynamic passive: malsendiʔssiʔmum ‘get written’
    • Patient noun: ansendiʔssiʔmum ‘writing, handwriting’
• Atelic verb: ndiʔsoʔmum ‘draft, plan’
  • Action noun: anndiʔsoʔmum ‘planning’’
  • Agent noun: naŋndiʔsoʔmum ‘planner’
  • Participle: lsendiʔsoʔmum ‘drafted, unfinished’
    • Patient noun: ansendiʔsoʔmum ‘a draft, plan’
• Stative verb: *ndiʔsŋuʔmum (not attested)
• Intensive verb: ndiʔsniʔmum ‘engrave’
  • With auxiliary: wandiʔsniʔmum ‘engrave using’
    • Participle: lsewandiʔsniʔmum ‘for engraving’
      • Patient noun: ansewandiʔsniʔmum ‘engraving chisel’
  • Agent noun: naŋndiʔsniʔmum ‘engraver’
  • Participle: lsendiʔsniʔmum ‘engraved’
    • Passive: walsendiʔsniʔmum ‘be engraved’
    • Dynamic passive: malsendiʔsniʔmum ‘get engraved’
    • Patient noun: ansendiʔsniʔmum ‘engraving’
• Punctual: *ndiʔsmoʔmum (not attested)
• Irregular noun: ndiʔsiʔmiʔm ‘pen, writing implement’

[Man in Space]

Can maʔ--to be used for abstract nouns or is it more strictly for concrete objects/persons?

[bradrn]

Can maʔ--to be used for abstract nouns or is it more strictly for concrete objects/persons?

It simply means ‘one which verbs’ / ‘one which is verbed’. So it’s usually concrete, but with the right verb it could be more abstract.

[bradrn]

Additionally, the vowel-final paradigm itself has a slight irregularity, namely that a few roots take -k in the atelic rather than a null suffix (as with ndok above, √ndo+k).

Minor change to this: I realised the null suffix could sometimes result in forbidden hiatus, as in e.g. √wsi-ap ‘related to shooting’. It seems reasonable for such cases to insert a semivowel, as wsiyap ‘shoot at’.

[Man in Space]

Additionally, the vowel-final paradigm itself has a slight irregularity, namely that a few roots take -k in the atelic rather than a null suffix (as with ndok above, √ndo+k).

Minor change to this: I realised the null suffix could sometimes result in forbidden hiatus, as in e.g. √wsi-ap ‘related to shooting’. It seems reasonable for such cases to insert a semivowel, as wsiyap ‘shoot at’.

If you're feeling a little adventurous, why not have a stop appear in this position? e.g., *ia > ica or iɟa. It'd basically be very specific glide fortition.

[bradrn]

Additionally, the vowel-final paradigm itself has a slight irregularity, namely that a few roots take -k in the atelic rather than a null suffix (as with ndok above, √ndo+k).

Minor change to this: I realised the null suffix could sometimes result in forbidden hiatus, as in e.g. √wsi-ap ‘related to shooting’. It seems reasonable for such cases to insert a semivowel, as wsiyap ‘shoot at’.

If you're feeling a little adventurous, why not have a stop appear in this position? e.g., *ia > ica or iɟa. It'd basically be very specific glide fortition.

Ooh, very nice idea! /tʃ/ is extremely rare at the moment, so this would be a good way to get more of it. I’d need to think about how to extend ths to the other V_V environments, though.

(What’s more, this situation actually comes from *k→∅ originally, so it would be lenition, not fortition!)

[bradrn]

The verb core

The ‘verb core’ is what I’ll call that inner section of the verb which comes after the preverb. It starts, optionally, with a negator, which is followed by one or more verbal lexemes: divided into true verbs, and ‘verboids’ (more on those later).

Negation

The negative marker, if present, is the first element of the verb core. It displays the most extensive allomorphy of any category in Eŋes, having four forms: ma-, maʔ-, mo-, moʔ-. Its form is lexically determined by the immediately following morpheme:

• The vast majority of verbals take ma-, including all forms derived with the auxiliary: e.g. mafŋiŋ ‘not broken’, mawoʔŋ ‘not exist’, mawenar ‘not be hot’. (Note this is distinct from the homophonous auxiliary form: maraw ‘become good’, mamaraw ‘not become good’.)
• Most vowel-initial verbals regularly take maʔ- due to hiatus rules (which I’ve already described): e.g. maʔasoŋ ‘not low’, maʔasanmarnir ‘not sit’. A few take this form even though they begin with a consonant: most notably maʔfa- ‘not rising’.
• The form mo- is quite rare: it is most commonly encountered in moysa- ‘not coming’.
• moʔ- is even rarer: aside from its use in hiatus (like maʔ-), the only common verbs which take this form are those in √mu-.

Making this somewhat simpler, the choice depends on the verb root, not the stem. Thus, for instance, all verbs in √mu- are negated with moʔ-: as in moʔmun ‘not shaking’, moʔmuŋ ‘not confused’, moʔmus ‘not drawing’.

Also, as usual, the following verb takes its hiatus form after the negative. For instance, moysa- ‘not coming’ is the negative of isa- ‘coming’: the initial vowel has become a semivowel in hiatus.

Therefore, I will often list verbs in two principal parts: their root, and their negative. The latter shows both their hiatus form and their choice of negator (when irregular).

With regards to semantics, the main point is that the negator has scope over the entirety of the verb core, even when it contains multiple lexical verbs. This is despite the fact that it is phonologically tightly bound with the following element. If one wishes to negate an individual verb, one must either use derivation or choose a different verb.

(Also, I’ve mentioned that negation can interact with focus particles in various different ways, but that deserves its own post.)

The rest

(I need a good term for ‘the verb core aside from the negator’, but I can’t think of one…)

The rest of the verb core is composed of a sequence of verbal elements, at least one of which must be a full lexical verb. The examples we’ve seen so far have had only one or two, but in extreme cases there can be up to four or five:

Say ntofassisesyewlenfas.
[sajnˌto.fas.si.ses.jew.lenˈfas]
s-say n-to-fas-sises-ye-wlen-fas
ACC-home 2s-DEF.SG-HAB-want-return-travel-LOC
You always want to go back home.

Sey bigamwasoŋfasi?
[sej bi.ɡam.wa.soŋ.faˈsi]
sey bi-gam-w·asoŋ-fas-i
like.this 1s.Q-must-lower-stand-3s
Should I put it up like this?

Yeysaywlenaŋus.
[jej.sajw.le.naˈŋus]
ye-isay-wl·e·naŋ-u·s
again-come-stroll-talk
He came again to talk to me

The verbal morphemes making up these verb cores can be divided into two general categories. Firstly, there are the ordinary verbs which we’ve already seen. These are free forms, comprising a verbal stem extended with an aspect marker, plus optionally elements such as auxiliaries or participles.

The verb core must, at minimum, contain at least one such full verb. As in the second and third examples above, it will often contain two or more of these, to form a serial verb construction. Verb serialisation can denote actions which took place simultaneously or sequentially, or with a cause-effect relationship. I will go into more detail about the rules for verb combining in the future.

Those verbals which are not true verbs I will call verboids. This is rather lazy terminology, since it’s something of a ‘catch-all category’ — simply anything which isn’t a true verb. Verboids therefore have the opposite properties to true verbs:

• Verboids are bound forms, and are dependent on appearing with at least one free form (a lexical verb).
• Many verboids are unanalysable, whereas lexical verbs display the systematic ‘root+aspect’ structure already described.

Beyond this, many verboids share one very distinctive property: they vary between two stems, depending on the aspectual setting. Thus, for instance, ‘going’ is wal- usually, but wel- after some aspectual markers, a situation which I will write as wal~wel-. For lack of a better name, I’ll call these the ‘A’ and ‘B’ forms respectively.

Verboids span a very wide range of meanings and syntactic behaviours. They include the following categories:

• Motion/direction: wal~wel- ‘going’, isa~isay- ‘coming’, fa~fay- ‘rising’, san- ‘falling’, ye- ‘return’, mum~um- ‘moving about’
• Position/posture: asanir- ‘sitting’, fas~far- ‘standing’, eb- ‘lying’
• Manner: tob- ‘numerously’, msiŋ- ‘carefully’
• Valence-change: -pes ‘benefactive’, -fas~far ‘locative’, -muk ‘reflexive’
• Degree: -ndo ‘more’, -wis ‘as’, tob- ‘many’
• Phase: dyam- ‘starting’, diʔs- ‘finishing’
• Modality: wran~wgi- ‘seem’, rwelbuʔ~rwilbuʔ- ‘can’, fasfin~faserfin- ‘must’, gam- ‘must’, sises- ‘want’
• Aspect: (see below)

In general, we can place verboids on a continuum from ‘most verb-like’ to ‘least verb-like’. On the one hand, some verboids can behave essentially the same way as any true verb: they may even show obvious resemblance to a verb root, as with wal~wel- ‘going’ vs √wl- ‘to go’. On the other hand, aspectual verboids are fixed in position, and trigger stem alternations in other verboids, neither of which is very verb-like. Between these extremes are elements which may be more or less restricted, with their own distinctive behaviours.

A full description would be too long for this post, but I can mention a few general divisions amongst the less verb-like verboids. The most obvious is that some verboids consistently occur before the lexical verb(s), whereas others occur afterwards. A few can appear either before or after, with different meanings: for instance fas~far, which usually refers to an upright vertical position (‘standing’), can also act as a locative applicative at the end of a verb.

As was just mentioned, the most distinctive single class of verboids are the aspectuals. These verboids have a fixed position: they occur at the very beginning of the verb core, or immediately after the negator if there is one. No more than one may be present at a time. They also trigger aspectual agreement: the A stem is default, but the presence of certain aspectual markers will cause all subsequent verboids to switch to the B stem.

The following table lists the more common aspectual verboids:

Verboid	Aspectual meaning	Stem triggered	Other meanings?
waniŋ-	Continuous	A	None
sanir-	Prolonged ongoing action	A	None (but asanir- ‘sitting’)
fas-	Habitual	A	fas~far- ‘standing’
gan-	‘Still’	A	None (but gam- ‘must’)
siw-	Perfective	B	None
isay-	Recent past (‘just’)	B	isa~isay- ‘coming’
wel-	Prospective	B	wal~wel- ‘going’
ye-	‘Again’	B	ye- ‘returning’

Many of these are transparently derived from other verboids (unsurprisingly, mostly motion and posture). A few possibly have more opaque connections. In the case of formal identity, the stem alternations can often help disambiguate which meaning is intended: for instance, biyewalwlesnaŋ is ‘I’ll walk back there’ but biyewelwlesnaŋ can only mean ‘I just walked there’.

[Ares Land]

Not much to comment, except that I regret there's no 'like button'

[bradrn]

Not much to comment, except that I regret there's no 'like button'

Thanks very much!

[bradrn]

Nouns and noun phrases

Eŋes is one of those languages where nouns see much less elaboration than verbs. Still, there’s some interesting stuff to say about them…

Case-marking

In the last post I said that the negative has the most complicated allomorphy of any Eŋes morpheme. I am now forced to recant: in fact, the accusative case (the only noun case in Eŋes) is more complicated. The accusative marker is placed at the very beginning of the noun phrase, with its form depending on the following morpheme in a not entirely concatenative manner:

• If the following morpheme starts with a consonant:
  • Before consonant clusters, the accusative is se-: se-mŋun ‘the eye’, dfin → se-dfin ‘hand’
  • Before voiceless consonants other than /s t/, the accusative is s-: kun → s-kun ‘the fly’, pomay → s-pomay ‘the jug’
  • Before voiced consonants other than /r/, the accusative is r- (/ʒ-/): banaʔ → r-banaʔ ‘the sky’, mec → r-mec ‘the year’
  • Initial t- or tC- become s- or se-sC-, respectively: tam → sam ‘the food’, tno → se-sno ‘the rope’
  • Initial s- or r- remains unchanged: sal ‘(the) fire’, res ’(the) yam’.
• If the following morpheme starts with a vowel, it depends on the form taken by the noun in hiatus:
  • If the hiatus form would start with /ʔ/, the accusative is s-: ap → s-ap ‘the arrow’, ansefoŋ → s-ansefoŋ ‘the place’
  • Otherwise the accusative is se- followed by the hiatus form: inab → se-nab ‘the dog’, aŋnap → se-ŋnap ‘the island’
• Pronouns can be irregular.

(Note, incidentally, that there’s only three actual forms here: r-, s- and se-. The complexity comes from their conditioning, and the nonconcatenativity.)

You’ll observe that I glossed the above forms as definite. This is because the accusative is only marked on definite NPs (including pronouns etc.). Indefinite objects never receive accusative marking.

There is another restriction on the accusative: it is limited to immediately pre-verbal objects. If an object is focussed or postposed, it loses any accusative marking which might have otherwise been present.

Thus, case-marking is a rather ‘surface-level’ feature of Eŋes, so to speak. It’s simply a marker of definite, preverbal nouns. Unlike many other languages, Eŋes displays no interesting phenomena where nouns are case-marked in unexpected ways (like accusative subjects and whatnot), or where other things depend on case-marking.

Possession

Aside from the accusative marker, a possessor NP is the only element which can come before the head noun. Ignoring elements after the head, the maximal linear order is therefore accusative—possessor—possessive.marker—head.

Eŋes distinguishes between two types of possession, depending on the possessed noun: inalienable and alienable. Starting with the former, inalienably possessed nouns include the following:

• Kinship terms: mem ‘mother’, dar ‘father’, etc.
• Orientational and relational nouns: nur ‘front’, kra ‘on top’, etc.
• Some miscellaneous nouns: not many, most notably leŋ ‘name’ and rweʔses ‘friend’

Notably, body-part terms are not inalienably possessed. This is unusual, but maintains a distinction between relational and body-part nouns, which are often homophonous: e.g. bnur ‘in front of me’ vs bmonur ‘my head’.

The marking of inalienable possession uses the same morphemes used for subject marking on verbs (the ‘subject’ column here). These morphemes are placed before the head noun (in the ‘possessive.marker’ slot mentioned above). For pronominal possessors, this is sufficient. For nominal possessors, the possessor is placed before the possessive marker: e.g. eŋes ∅-dar ‘the person’s father’, or eŋes in-dar ‘the peoples’ fathers’.

In the previous example, we see an interesting consequence of this construction — third person singular possession is zero-marked. This remains true even if the possessor is pronominal: thus dar could mean either ‘a/the father’, or ’his/her father’. We could argue that these nouns are not just inalienably possessed, but obligatorily possessed. A possessor is always implied, even if it’s simply ‘it’.

(If one really wants to add an explicit possessor, it’s always possible to use a demonstrative: e.g. naʔos dar ‘the said person’s father’.)

Nouns other than those listed above are alienably possessed. This works basically the same way as inalienable possession, but with a different set of markers:

Person	Singular	Plural
1	bgo-	baŋo-
2	ndo-	ndo-
3	(o)ŋo-	(i)ŋiŋo-

(Which I just realised I forgot to add to the pronouns post, whoops…)

Note that these are vowel-final and thus trigger the hiatus form of the possessed noun: e.g. aŋnap ‘island’ but bgoŋnap ‘my island’ (also cf. seŋnap ‘the island’ above). Additionally, the third-person markers delete their initial vowel if they immediately follow a vowel: thus dfin oŋownus ‘the man’s hand’, but anu ŋownus ‘the man’s job’.

Adjectives and compounds

The head noun can be immediately followed by an adjective. The noun and adjective behave for all purposes as a single unit: they have one stressed syllable (namely the ultimate), they undergo hiatus avoidance and other phonological processes, they are non-interruptable by any other element, and so on. The noun anu ‘thing’ can be used as a generic base: e.g. anu-ʔenar ‘hot thing’. (Incidentally, note how -enar takes its hiatus form here after a vowel.)

Adnominal adjectives are subject to an important restriction: a noun can be directly modified by at most one adjective. If two adjectives are really desired, a relative clause can be used: e.g. kuŋ-kbuy ‘old fly’ vs kum-bli na wakbuy ‘old, heavy fly’. (Again, note the action of phonological rules, in this case changing the final consonant of kun ‘fly’ due to nasal assimilation.) In fact, speakers may use a relative clause even when there is only one adjective: kun na wakbuy ‘fly which is old’.

(Please, don’t sue me for the contrived examples… if I had better ones, I would use them…)

An interesting consequence of the above is that adjectives are bound morphemes. In the verb complex, they require a preceding auxiliary to be verbalised. In the noun phrase, they must occur after a head noun. There is no situation in Eŋes in which an adjective may appear by itself as a free form.

Eŋes also has a handful of true noun compounds. In their phonological and grammatical properties, they behave just like the noun-adjective compounds already described — extending to being head-initial.

However, unlike the adjectival construction, noun compounding is hardly productive, and essentially limited to a number of fixed lexemes: e.g. moʔ-nir ‘place-sun’ = ‘east’, len-dfin ‘name-hand’ = ‘nickname’ And, for all their transparency, these lexemes behave in all respects as ordinary nouns, even allowing further adjectival modification: lendfiŋ-kbuy ‘old nickname’.

To express nominal modification, Eŋes instead tends to use the alienable possessive construction. This covers the same kinds of areas for which English uses noun compounds, including idiomatic and lexicalised expressions: e.g. kom oŋodaw ‘stomach truth’ = ‘hunch’, fgis oŋomŋun ‘palm.tree eye’ = ‘fruit of palm tree’.

It is important to distinguish noun compounding from singular inalienable possession, which also involves unmarked noun juxtaposition. The two have very different semantics: a noun compound is head-initial, whereas inalienable possession is head-final. Two characteristics distinguish them:

• In general, the second element of a noun compound will not be a noun which can be inalienably possessed, as is the case for nir ‘sun’ and mŋun ‘eye’ above. On the other hand, in a construction like anles dar ‘the king’s father’, dar ‘father’ is in the small set of nouns which can be inalienably possessed.
• Noun compounding shows much greater phonological integration than inalienable possession does. For instance, lendfin ‘nickname’ always shows nasal assimilation (cf. standalone leŋ ‘name’), whereas a possessive construction like dfin kra ‘the top of the hand’ may be realised as /dfinˈkra/ with no nasal assimilation.

Other nominal modifiers

Several elements can be placed after the head noun or compound:

• One of the four adnominal determiners: naʔsin ‘this’, naʔsŋan ‘that’, naʔsnan ‘which / some’, or naʔsnos ‘the said’
• A quantifier, basically limited to mog ‘many’ or tam ‘some’. (Other quantifiers exist, just elsewhere: e.g. as focus particles, nouns or in the verb complex.)
• A numeral.

For Janko’s sake, I’ll list the first few numerals:

1. yam
2. key
3. us
4. mŋa
5. tow
6. toy
7. weʔ
8. sun
9. sla
10. ndas

Generally no more than one of these will be present in a single NP — mostly because they’re largely semantically incompatible with each other. In the very rare case where two are present, this will generally comprise a numeral followed by a determiner (as in eŋes key naʔsŋan ‘those two people’, etc.).

Relative clauses

A relative clause, if present, is the last item in the NP. Relative clause construction is mostly simple, but has its subtleties.

Relative clauses are introduced by the relativiser na (otherwise a demonstrative meaning ‘which’ or ‘something’). In the simplest cases, where the relative clause is mono- or divalent, the common argument is removed from the relative clause to leave a gap:

eŋes faʔntosiwigiʔnsem res ‘the person ate the mango(s) yesterday’
⇒ Blisŋumŋun eŋes [na ___ faʔnsiwigiʔnsem res] ‘I see the person [who ate the mango yesterday]’
⇒ Blisŋumŋun res [na eŋes faʔntosiwigiʔnsem ___] ‘I see the mangos [which the person ate yesterday]’

Some subtleties here:

• A transitive relative clause must always have SVO order, unlike most transitive clauses, which may be SVO or SOV (more usually the latter).
• The definite subject marker to- (here as part of a larger preverb, fanto-) requires an overt subject within the relative clause. When the subject is relativised, the marker must be deleted.
• The main clauses here also happen to have SVO order. This is not required, but long object NPs do tend to be moved postverbally.

(From now on, I’ll focus on the NP itself and not show the main clause.)

Trivalent clauses are a bit trickier. Both objects of the verb can be relativised, but to relativise the preverbal argument, it must be replaced with the special element siw:

sam bsiwrwemisaypes inab ‘I gave the dog food’
⇒ sam [na siw bsiwrwemisaypes inab] ‘the food which I gave to the dog’
⇒ inab [na sam bsiwrwemisaypes ___] ‘the dog to which I gave food’

This isn’t the only situation where a relative clause doesn’t use gapping. When relativising on the object, an alternative to gapping is to add the object pronoun -i (singular) or -ey (plural). Thus the example from earlier could be rephrased as:

eŋes faʔntosiwigiʔnsem res ‘the person ate the mango(s) yesterday’
⇒ res [na eŋes faʔntosiwigiʔnsemey] ‘the mangos [which the person ate yesterday]’

(In this example res happens to be originally postverbal, but it could have been preverbal, too.)

So Eŋes can relativise on subjects and on objects: what about other NPs? As it happens, most NPs are either subjects or objects, so this covers most cases anyway.

The major remaining case is that of possessor NPs, which cannot be relativised. An English phrase like ‘the person [whose name is Brad]’ cannot be expressed as a single NP in Eŋes. If one does try to do so, one soon runs into difficulties: recall that the oddities of inalienable possession ensure that leŋ means both ‘name’ and ‘his name’, so there’s no good way to mark the common argument. (A sentence like *eŋes [na leŋ wgiŋ Brad] is simply ungrammatical.)

Additionally, focus constructions and nonverbal clauses cannot be relativised. Thus a sentence like eŋeymen eŋes towus ‘the person talks to some of them’, with a focalised object, cannot be relativised to *eŋes [na eŋeymen wus].

Some other miscellaneous points:

• Eŋes relative clauses are always restrictive. Thus elements such as proper nouns and demonstratives cannot head a relative clause.
• Headless relative clauses are not tolerated. As with adjectives, a generic noun like anu or eŋes can act as head instead. (Some languages allow pronouns to do this job, but of course Eŋes has no free pronouns.)
• The usual Accessibility Hierarchy would suggest that, because possessors are not relativisable, neither are standards of comparison. But Eŋes conflates standards of comparisons with direct objects, so in fact they are relativisable after all: e.g. English ‘the person [than whom I am smaller]’ ↔ eŋes [na bwsimndo ___].

[bradrn]

Definiteness and anaphora

The previous post laid out the basic structure of the noun phrase. Now, we can get to the really interesting stuff. For Eŋes NPs, this mostly involves definiteness and anaphora, and how they interact. I’ll be discussing these through three constructions: the accusative marker, the definite subject marker, and the usage of demonstratives.

Accusative marker

I’ll begin by returning to the accusative marker. In the last post I mostly focussed on its phonological behaviour, with a little syntax. This time I’ll zoom in on a different area:

You’ll observe that I glossed the above forms as definite. This is because the accusative is only marked on definite NPs (including pronouns etc.). Indefinite objects never receive accusative marking.

This is, of course, entirely too cursory: precisely what kinds of NP qualify as ‘definite’ for the purposes of the accusative marker?

Firstly, there are some NPs which could be called ‘obligatorily definite’. This category includes personal pronouns, and all nominal demonstratives aside from the indefinite series (which naturally can never receive definite marking.) Personal names and proper nouns practically always receive the accusative marker as well. The exception is if they’re used as a common noun, as in a collective or generic sense: for instance , if one knows several people named ‘Smith’ one might be able to say that Smith faʔn-blisŋumŋun ‘I saw a Smith yesterday’.

Another class of ‘obligatorily definite’ NPs is comprised of inalienably (or obligatorily possessed) nouns. In fact it’s not easy to construct a sentence where such an object isn’t definite, even in English: the best example I can come up with is ‘he has a friend’, which could be rendered into Eŋes as serweʔses rwoŋ lit. ‘[he] has [his] friend’ (but more likely expressed non-verbally, rweʔses ni ‘there is a friend [of his]’).

By contrast, some NPs are always indefinite. As mentioned, these include the indefinite series of demonstratives (naʔ etc.). This category also includes ordinary nouns used with a generic meaning: as in ‘I like dogs’, inab brwoŋses.

Outside these cases, object nouns can vary between indefinite and definite, in more or less the same way as English. Definiteness usually indicates that the object is known to both speaker and listener, for instance by having already been introduced earlier in the discourse, whereas indefiniteness indicates an unknown or new object.

All this, however, is only relevant for preverbal objects: the accusative marker cannot appear with a postverbal object, so such objects are not marked for definiteness. (That being said, such objects usually indicate new information, so are mostly indefinite anyway.)

Definite subject markers

You may recall, from way back when I described the preverb, that subject definiteness gets markers too: as the last element in the preverb, to- marks a definite singular subject, and taŋ- a definite plural one. These are both very important markers, and it’s time I gave a fuller description of their distribution.

Firstly, one of these markers is required whenever the subject is definite and a full NP. The latter point is important: an elided NP does not require a definiteness marker, being represented instead by a true subject pronoun (with which the definite subject markers are in complementary distribution).

Which subjects count as ‘definite’ for the purposes of the definite subject markers? Mostly, the same ones that can receive the accusative marker in object position. But there are two places where their behaviour diverges:

• Because the definite subject markers are in complementary distribution with the subject pronouns (or ‘Subject 1’ as I termed this series earlier), first- and second-person subjects can never receive a definite subject marker, whereas they always receive the accusative as preverbal objects.
• Conversely, while the accusative treats generic nouns as indefinite, such nouns always receive a definite subject marker. Most generic nouns can take either the singular or plural marker, though some can take only singular (e.g. many uncountable nouns).

The more interesting effects, however, occur when there is no overt subject NP. In this case, either a true subject pronoun or a definite subject marker can occur. The difference here becomes one of anaphora, not just definiteness. Essentially, the use of a third-person subject pronoun (null for singular, (i)ŋi- for plural) is taken as referring to a continuing topic of conversation. The use of a definite marker, on the other hand, implies a non-topical subject: one which was introduced earlier, but is not the topic.

This distinction could plausibly be labelled as ‘proximate’ vs ‘obviate’, similarly to that famously described for Algonquian languages. In this case, the subject pronouns would qualify as ‘proximate’, while the definiteness markers would be ‘obviate’. But there are very significant differences from the Algonquian system. Firstly, the Algonquian category is marked on all NPs, whereas Eŋes marks this category on the verb only. Secondly, in Algonquian the marking is often determined by syntactic structure (e.g. a language may require possessors to be obviate), whereas in Eŋes the marking is determined only by definiteness and pragmatics.

Arguably, one could also suggest that the Eŋes system forms an incipient switch-reference system (albeit one following an initial–medial pattern rather than the usual medial–final pattern). After all, a topical NP most often occurs in subject position, so the use of the subject pronoun could be interpreted as a same-subject marker. Similarly, a non-topical NP will rarely be subject in the previous sentence, so the definiteness marker could be considered as a different-subject marker.

This kind of logic is most clearly seen in what I call ‘close coordination’ constructions, where two clauses are juxtaposed without marking (often translatable by a preposition in English):

rMarwa bsiwlisŋumŋun sefgis nur (to)foŋ.
[ʒmaɾ.wa.bziw.lis.ŋumˈŋun sef.ɟis.nuɾ.(to).foŋ]
r-Marwa b-siw-lis·ŋu·mŋun se-fgis ∅-nur (∅-/to-)f·oŋ
ACC-Marwa 1s-PFV-see ACC-palm 3s-front (3s-/DEF.SG-)located

I saw Marwa in front of the palm tree. (lit. I saw Marwa, located at the palm tree’s front.)

Assuming that I’m the topic (or at least am non-focussed), the presence of to- would yield an implicature that it was Marwa who was in front of the palm tree. Conversely, its absence would implicate that I was the one in front of the palm tree. Of course, most usually both participants would be in the same place, in which case ∅- would be more natural. (And note that these are implicatures only, not implications: this isn’t close to being a full switch-reference system yet.)

However, all this being said, whenever the subject NP is definite, a subject definiteness marker is required — no matter whether the subject is topical, non-topical or anything else. From this point of view, all the other implications arise from the natural interpretation of an elided subject NP. At some point, these markers may yet grammaticalise further, but for now it’s still basically a marker of definiteness, with topic-tracking being secondary.

Furthermore, there are some contexts in which a definite subject marker cannot be used alone at all:

• As mentioned in the last post, this is the case in relative clauses. When the subject is relativised, the definite subject marker disappears alongside it. Conversely, if a marker is present, it can only refer to an overt NP, never to an elided one.
• Similarly, in sentences with a focus particle, a definite subject marker may never refer to the focalised NP. But unlike in relative clauses it may refer to a non-topical elided NP.

Demonstratives

Recall that there are four series of demonstratives: proximate, distal, indefinite/interrogative and anaphoric. The first two have more or less expected meanings, with the difference being distance from the speaker. But the other two are perhaps less familiar.

Starting with the anaphoric series, these forms are used to refer back to an NP earlier in the discourse — as opposed to one located physically in the surrounding space. English and many other languages combine these functions in a single form (as with ‘this’), but Eŋes keeps them separate. The nearest English-language equivalents to the Eŋes anaphoric demonstratives may be ‘said’ or ‘aforesaid’ (as in ‘the said person’ etc.), though the English terms are highly formal and rarely used in spoken conversation.

(For an example of how this might be used: if this post were translated to Eŋes, most if not all uses of ‘this’ would be translated as anaphoric nos or naʔsnos.)

Naturally, the anaphoric demonstratives are always definite. As preverbal object, they always require the accusative; as a subject they require a definite marker. Indeed, there is arguably some overlap in function with the definite subject markers: both refer back to NPs mentioned earlier, and both have a strong tendency to refer to non-topical or focussed NPs, in contrast to third-person pronouns. But the anaphoric demonstratives are far less restricted in scope, since they can be used in any position in the sentence, as nouns, adnominals or adverbials.

The final demonstrative series is the indefinite/interrogative series, covering (as one might expect) indefinite and interrogative usages. (naʔ can be translated as ‘someone/who’, na as ‘someone/what’, and so on.) The correct interpretation is decided by whether the sentence has an interrogative marker or not. Naturally, these pronouns are considered indefinite, so never get a subject or accusative marker.

In any case, definiteness doesn’t even apply to this series of demonstratives most of the time, because their most usual position is focussed before a focal particle. For indefinite usages, the focus marker acts to circumscribe the basic indefinite meaning: as with (o)ŋo for negative indefinites (naʔoŋo ‘no-one’, noŋo ‘nothing’), or nirob for free-choice indefinites (naʔnirob ‘anyone’, nanirob ‘anything’).

Meanwhile, for interrogative sentences, the focussing simply moves the question word to the beginning of the sentence:

Nani mbilisŋumŋun?
[naˈni mbi.lis.ŋumˈŋun]
na-nii mbi-lis·ŋum·ŋun
what-FOC 2s.Q-see

What do you see?

Of course, this is not mandatory: it is perfectly acceptable to say Na mbilisŋumŋun? (and note the absence of the accusative here).

[bradrn]

In a wonderful instance of ANADEW, I was pointed towards Wichman’s (2007) paper on givenness marking in Tlapanec — a system which turns out to be remarkably similar to my ‘subject definiteness markers’ in the last post. The details are somewhat different, of course, but it’s definitely a good read if you’re interested to see how this kind of topic-tracking works in practice. (The conversation quoted on pp817–819 is particularly enlightening.)

[Travis B.]

Nani mbilisŋumŋun?
[naˈni mbi.lis.ŋumˈŋun]
na-nii mbi-lis·ŋum·ŋun

Too... many... nasals...

[foxcatdog]

Never such a thing as too many nasals they are good in combination with all other types of consonants and make for great roots/words when you only include them.

[bradrn]

Two small changes. Firstly, after a discussion with zompist, I’ve decided to simplify the forms of some verb roots:

• √lis-mŋun ‘related to seeing’ becomes √ls-ŋun (taking unmetathesised aspect markers)
• √asan-rnir ‘related to sitting’ becomes √tan-rnir (retaining /a/-series markers) — but the corresponding verboid asanir- remains as is
• √igiʔ-sem ‘related to eating’ becomes √giʔ-sem (retaining vowel-final markers)

(As well as some others, but I believe these are the only ones which have been mentioned or used so far.)

Secondly, I’m contemplating romanising the glottal stop with an apostrophe rather than ⟨ʔ⟩. Usually I don’t love apostrophes, but they seem to fit with the phonotactic status of the glottal stop in Eŋes. Also ⟨ʔ⟩ feels like a very ‘heavy’ letter for such a common sound.

Trying out these changes on some previous sample sentences, we get:

Bĕlsoŋŋun res na eŋes fa’ntosiwgi’nsem.
Ndini sewŋulbu’i, banib mawa’i.

…alas, the apostrophes make it look like a b’ad fa’ntas’y lan’gu’a’ge. But it definitely feels like it gives a better impression of the ‘feel’ of Eŋes. I’ll keep it for the next few posts and then make a final decision.

(It does make some things like banaʼ look odd, mind you. And roots like √w-ʼ- look even more odd. I may reinstate ⟨ʔ⟩ for roots only, if it comes to annoy me sufficiently.)

[ophois]

One could always romanise the glottal stop <q> or even the lower-case variant of the regular symbol (on phone rn so can't type it). Or given it only occurs after vowels, perhaps mark on the vowel. I'd use an underdot.

[bradrn]

One could always romanise the glottal stop <q> or even the lower-case variant of the regular symbol (on phone rn so can't type it). Or given it only occurs after vowels, perhaps mark on the vowel. I'd use an underdot.

A few considerations:

• I used ⟨q⟩ for the last language here. I’m fond of it, but it feels even ‘heavier’ than ⟨ʔ⟩ does.
• I’ve never particularly liked lower-case ⟨ɂ⟩ — to me, it feels at the same time too similar and not similar enough to other letters. Still, it’s a possibility if all else fails.
• In large part, the apostrophe strikes me as reasonable because it feels almost like ‘marking on the vowel’. I don’t like to use a diacritic because the glottal stop regularly appears to interrupt hiatus: for instance, anuʼenar ‘hot thing’ (cf. anukbuy ‘old thing’).

[WeepingElf]

One could always romanise the glottal stop <q> or even the lower-case variant of the regular symbol (on phone rn so can't type it). Or given it only occurs after vowels, perhaps mark on the vowel. I'd use an underdot.

A few considerations:

• I used ⟨q⟩ for the last language here. I’m fond of it, but it feels even ‘heavier’ than ⟨ʔ⟩ does.
• I’ve never particularly liked lower-case ⟨ɂ⟩ — to me, it feels at the same time too similar and not similar enough to other letters. Still, it’s a possibility if all else fails.
• In large part, the apostrophe strikes me as reasonable because it feels almost like ‘marking on the vowel’. I don’t like to use a diacritic because the glottal stop regularly appears to interrupt hiatus: for instance, anuʼenar ‘hot thing’ (cf. anukbuy ‘old thing’).

Fair. I think the apostrophe makes perfect sense for the glottal stop, unless you also use it as a diacritic e.g. for ejective stops (as in many languages) or aspirated stops (as in Wade-Giles).