Passive voice stuff

[Jonlang]

So... I've spent a lot of time recently rebuilding the verbal system of my L conlang, and I'm happy (for now) with what I have. I have, however, overlooked passive voice constructions and I'm too lazy to try to go back and make entirely new passive forms. Also, from the start I've known I wanted to avoid constructions like English 'get' passives (he got shot) or the Welsh 'get' which is basically the same thing (cafodd e'i saethu - got he his shooting). Also, the other Welsh option - to have an impersonal conjugation do the job of passive voice (saethwyd ef) - is reserved for another conlang.

I decided that L would not use a feature like English for subjectless verbs, e.g. 'it is snowing, it snows' where 'it' refers to no actual entity, instead the verb by itself would do '(it) snows'. So, having thought about this, would it seem odd if this lone-verb with a noun in the accusative make a passive construction? Seeing as the accusative must be the object and there is no subject? So a structure like "(it) ate (the) cake" [eat.PST cake.ACC] would equate to "the cake was eaten"?

My other option is to simply use the passive participle, however because telicity is marked on the object (like Finnish) I think this means that both the agent and the patient of the passive would be somehow marked: the agent by the instrumental case (he is the means by which the action was done) and the patient by the accusative or partitive depending on the completeness of the action. So it would look something like [cake.acc pron.1st.SG.INST eat.PST.PTCP be.GNO]

[Travis B.]

You should really look at fluid-S systems where you have an agentive and a patientive case, with intransitive clauses taking either case depending on the agentivity/patientivity of the argument.

[Travis B.]

As for fun with alignment, one thing I have done is give each noun a default agentivity/patientivity, based on animacy/personhood, and only marked nouns for agentivity/patientivity when its agentivity/patientivity differs from its default.

[bradrn]

I decided that L would not use a feature like English for subjectless verbs, e.g. 'it is snowing, it snows' where 'it' refers to no actual entity, instead the verb by itself would do '(it) snows'. So, having thought about this, would it seem odd if this lone-verb with a noun in the accusative make a passive construction? Seeing as the accusative must be the object and there is no subject? So a structure like "(it) ate (the) cake" [eat.PST cake.ACC] would equate to "the cake was eaten"?

That’s not actually a ‘passive’, since the object remains in the accusative — a true passive would move the object into the nominative and eliminate the accusative position entirely. What you have here is just a rule that the subject can be left out of the sentence.

As for fun with alignment, one thing I have done is give each noun a default agentivity/patientivity, based on animacy/personhood, and only marked nouns for agentivity/patientivity when its agentivity/patientivity differs from its default.

That is to say, an animacy-based alignment split. Combining it with active-stative alignment is a bit weird, but apparently some Tupi–Guarani languages have something similar (e.g. Guajajara, though I don’t have complete access to the source). There’s also Georgian, but the details there are quite different to what you describe.

[zompist]

I decided that L would not use a feature like English for subjectless verbs, e.g. 'it is snowing, it snows' where 'it' refers to no actual entity, instead the verb by itself would do '(it) snows'. So, having thought about this, would it seem odd if this lone-verb with a noun in the accusative make a passive construction? Seeing as the accusative must be the object and there is no subject? So a structure like "(it) ate (the) cake" [eat.PST cake.ACC] would equate to "the cake was eaten"?

My other option is to simply use the passive participle, however because telicity is marked on the object (like Finnish) I think this means that both the agent and the patient of the passive would be somehow marked: the agent by the instrumental case (he is the means by which the action was done) and the patient by the accusative or partitive depending on the completeness of the action. So it would look something like [cake.acc pron.1st.SG.INST eat.PST.PTCP be.GNO]

You can absolutely do these things, but I'd think in terms of reducing valence or topicalization rather than "passive". You're reducing the valence of the verb (from 2 arguments to 1). This is often, though not always, accompanied by making the argument more subject-like-- e.g. moving it to subject position, or removing the accusative.

[hwhatting]

Another way to reduce valancy is using reflexives, if your languages has them. Russian reflexives often correspond to English passives:
Vse chitayut etu knigu.
all-NOM.PL read-3PL.PRES.ACT this-F.SG.ACC book-SG:ACC
Everybody reads this book.

Eta kniga chitayetsya (vsemi).
this-F.SG.NOM book-SG.NOM read-3SG.PRES.REFL (all-INSTR.PL)
This book is read (by everybody).

[bradrn]

Another way to reduce valancy is using reflexives, if your languages has them. Russian reflexives often correspond to English passives:
Vse chitayut etu knigu.
all-NOM.PL read-3PL.PRES.ACT this-F.SG.ACC book-SG:ACC
Everybody reads this book.

Eta kniga chitayetsya (vsemi).
this-F.SG.NOM book-SG.NOM read-3SG.PRES.REFL (all-INSTR.PL)
This book is read (by everybody).

This is basically just a middle or mediopassive voice, which is a favourite construction of mine. There’s some weirdness associated with these: e.g. WALS notes that reflexives can be used either as intensifiers or as a middle voice, but not both (which explains why English this book reads itself makes no sense, at least with the same interpretation as Russian). Kemmer has a nice book on middle voices, which unfortunately I haven’t gotten around to reading properly yet.

[Travis B.]

As for fun with alignment, one thing I have done is give each noun a default agentivity/patientivity, based on animacy/personhood, and only marked nouns for agentivity/patientivity when its agentivity/patientivity differs from its default.

That is to say, an animacy-based alignment split. Combining it with active-stative alignment is a bit weird, but apparently some Tupi–Guarani languages have something similar (e.g. Guajajara, though I don’t have complete access to the source). There’s also Georgian, but the details there are quite different to what you describe.

It is not a simple alignment split, because underlyingly there are still just plain agentive and patientive NP's, just like in any active-stative system. The key difference is whether agentivity or patientivity are case-marked, which depends on the markedness of agentivity or patientivity (agentivity is unmarked and patientivity is marked for animate NP's, and agentivity is marked and patientivity is unmarked for inanimate NP's). I personally think it is a neat system even if it rare in natlangs.

[bradrn]

As for fun with alignment, one thing I have done is give each noun a default agentivity/patientivity, based on animacy/personhood, and only marked nouns for agentivity/patientivity when its agentivity/patientivity differs from its default.

That is to say, an animacy-based alignment split. Combining it with active-stative alignment is a bit weird, but apparently some Tupi–Guarani languages have something similar (e.g. Guajajara, though I don’t have complete access to the source). There’s also Georgian, but the details there are quite different to what you describe.

It is not a simple alignment split, because underlyingly there are still just plain agentive and patientive NP's, just like in any active-stative system. The key difference is whether agentivity or patientivity are case-marked, which depends on the markedness of agentivity or patientivity (agentivity is unmarked and patientivity is marked for animate NP's, and agentivity is marked and patientivity is unmarked for inanimate NP's). I personally think it is a neat system even if it rare in natlangs.

I never called it ‘simple’, I just said it’s an alignment split. Which it is.

[Imralu]

I decided that L would not use a feature like English for subjectless verbs, e.g. 'it is snowing, it snows' where 'it' refers to no actual entity, instead the verb by itself would do '(it) snows'. So, having thought about this, would it seem odd if this lone-verb with a noun in the accusative make a passive construction? Seeing as the accusative must be the object and there is no subject? So a structure like "(it) ate (the) cake" [eat.PST cake.ACC] would equate to "the cake was eaten"?

I don't find it odd and it definitely could equate to "the cake was eaten", but as Bradrn pointed out, although it might functionally indicate the same concept as a passive, it doesn't count as a passive, since the passive is a voice that changes the use of cases with a verb.

My conlang Iliaqu does this, but also has a true passive.

kau

eat

Someone/something eats.
There is an eater.

kau

eat

nu

NOM.1S

I eat.

kau

eat

ni

ACC.1S

I am eaten.
Someone/something eats me.

kau

eat

kuani

BEN.1S

I am eaten for.
Someone/something eats something/someone for me.

he

undergo

kau

eat

nu

NOM.1S

I am eaten

It works the same for any case combination and I think every case has a corresponding verbal that raises it to the nominative, such as he above, meaning "undergo", "be a victim", "be affected". The dative-to-nominative raising verbal is no "receive".

xeva

be.visible

vu

NOM.2S

eni

DAT.1S

I (can) see you.
You are visible to me.

xeva

be.visible

vu

NOM.2S

You are/can be seen.
You are visible.

xeva-s

be.visible-EP

eni

DAT.1S

I (can) see.
Something/someone is visible to me.

no

receive

xeva

be.visible-EP

nu

NOM.1S

I (can) see.

In a lot of cases, there is no real reason to prefer one strategy over the other and the case method is used more commonly. The voice-changing-verbals are most commonly used in cases of relativising a verbal phrase to be used as an argument/adjunct.

ju

NOM.DEF.3S.ANIM

kau

eat

the one who eats
the eater

ju

NOM.DEF.3S.ANIM

xeva

visible

the one who is seen
the visible one

ju

NOM.DEF.3S.ANIM

he

undergo

kau

eat

the one who is eaten
the eatee

ju

NOM.DEF.3S.ANIM

no

receive

xeva

visible

the one who sees
the seer

This is basically just a middle or mediopassive voice, which is a favourite construction of mine. There’s some weirdness associated with these: e.g. WALS notes that reflexives can be used either as intensifiers or as a middle voice, but not both (which explains why English this book reads itself makes no sense, at least with the same interpretation as Russian). Kemmer has a nice book on middle voices, which unfortunately I haven’t gotten around to reading properly yet.

I'm so confused about why Swedish is yellow on the map. It's essentially the same as German in it's use of the 3rd person reflexive pronoun sig (German: sich) and the intensifier själv (German: selbst), with a greater tendency to use sig själv than German uses sich selbst. It is mentioned:

Another way of renewing reflexive markers is by intensifying weak reflexives. Such developments are widely attested in Romance languages, where the reflexive clitics are being replaced by the respective tonic forms in combination with intensifiers (cf. (10) from Portuguese; similarly in Swedish, Norwegian, Dutch).

(10) Portuguese

Maria

Maria

olha

looks

a

at

si

refl

mesma

intf

no espelho.

in mirror

‘Maria looks at herself in the mirror.’

So then, why is Portuguese red when Swedish, Norwegian and Dutch are yellow? And is the only reason German is red because it adds selbst a bit less often? German very often uses sich selbst (or mich selbst etc.) to emphasise genuine reflexives and just sich (or mich etc.) for middle voice, so, as far as I can see, it's essentially the same as Swedish and Portuguese and very different from English, which only has myself etc.

I also find it bizarre that this was a feature that was in the SAE (Standard Average European) grammar test we had here years ago, based on the list of SAE features on Wikipedia. Looking at the WALS map, I can't see any correlation.

[bradrn]

I'm so confused about why Swedish is yellow on the map.

I know nothing about Swedish, but WALS’s datapoints do seem to be frequently unreliable.

[hwhatting]

Another way to reduce valancy is using reflexives, if your languages has them. Russian reflexives often correspond to English passives:
(snip)

This is basically just a middle or mediopassive voice, which is a favourite construction of mine.

True; this was more about showing Jonlang alternatives to formal passives.
NB that Russian has besides the mediopassive both (1) a formal passive, but only for perfective verbs and so only usable in the past and the future, and (2) reflexive pronouns which can be used for an unambiguous reflexive construction:
(1) Kniga byla prochitana. Kniga budet prochitana.
book-NOM be-PAST.SG.F read-PAST.PASS.PRTC.PERF.SG.F. book-NOM be-FUT.3SG read-PAST.PASS.PRTC.PERF.SG.F
The book was / has been read. The book will be read.
(2) Kniga chitayet (samu) sebya.
book-NOM read-PRES.ACT.3SG (self-SG.F.ACC) REFL-ACC
The book reads itself
[using magic or AI]
The intensifier samu is not strictly necessary, but I'd expect it here in order to confirm that the default reading with its unusual revelation, of a book reading itself, is indeed intended.