Switch-Reference Markers in My Conlangs, Adpihi and Reptigan

[eldin raigmore]

Preliminaries 1:

Spoiler:

Some natlangs allow "clause-chaining".

In some natlangs, there's not a big distinction between co-ordinately conjoining clauses and subordinately conjoining clauses.

Some natlangs' verbs have an "independent order", for use in main clauses, and a "conjunct order", for use in "subordinatish" clauses.

A clause chain has an anchor clause in the "independent order", and one or more other clauses in the "conjunct order".

In some natlangs, the anchor clause is always the Initial clause in the chain, and all the other clauses are called Consecutive clauses.

In some natlangs, the anchor clause is always the Final clause in the chain, and all the other clauses are called Medial clauses.

Some languages have a Switch-Reference System, a set of "switch-reference markers", (usually verb-affixes, sometimes other verbal morphology, sometimes conjunctions), which tell whether and which and how a Consecutive or Medial clause's participant(s) are related to some referenced clause's participant(s).

A Consecutive clause would be marked to show whether some of its referents were related to those of some earlier clauses in the chain; possibly the last previous clause in the chain, or possibly the Initial clause, or sometimes possibly some other previous clause.

A Medial clause would be marked to show whether some of its referents were related to those of some later clauses in the chain; possibly the next clause in the chain, or possibly the Final clause, or sometimes possibly some other later clause.

The marker may encode which referent of the marked clause is related to which referent of the referenced clause; and it may encode what the relationship is.

_____________________________________________________________________________


Preliminaries 2:

Spoiler:

In these two of my conlangs (Adpihi and Reptigan), some clauses have an intransitive Subject; it is both the most controlling participant of the clause and the most affected participant of the clause. For instance, if the clause is monovalent -- has only one participant -- that participant has to be the clause's Subject.

Some clauses have an Actor; it is the most controlling participant of the clause. Actor is a kind of proto-Agent macro-role. If any bivalent or trivalent clause's most controlling participant is not also its most affected participant, then the controlling one is that clause's Actor.

Some clauses have an Undergoer; it is the most affected participant of the clause. Undergoer is a kind of proto-Patient macro-role. If any bivalent or trivalent clause's most affected participant is not also most controlling, then the affected one is the Undergoer. If you like, you can think "Object" or "Direct Object" or "Patient" whenever you see "Undergoer".

In every natlang it is sometimes necessary to say clauses that have a participant that is an argument rather than an adjunct, and that is highly involved and necessary to the nuclear verb's meaning, but is neither the most Agent-like or controlling participant, nor the most Patient-like or affected participant. Barry J. Blake calls such participants "Datives". Beatrice Primus calls them "proto-Recipients". Rick A. Morneau calls them "Focus". Dixon and Aikhenvald call them "Extended-core terms". I'll just abbreviate them as E.

In my two main conlangs, most clauses come with one of the following four arrangements of arguments (core terms):
S for monovalent, intransitive clauses;
A & U for bivalent, monotransitive clauses;
S & E for "bivalent intransitive clauses";
A & U & E for trivalent, ditransitive clauses.

About the SE clauses: In every natlang, there are two-argument verbs in which one participant is both most controlling and most affected; and the other participant, though highly involved, crucial to the verb's meaning, and part of the verb's core, is neither most controlling nor most affected.
English examples would be the verbs "seek" and "find" and "follow".

These verbs might be considered "bivalent intransitive" verbs; one might think of them as having a Subject and an Indirect Object, but no Direct Object.

For many natlangs, but probably and apparently not English, the difference between SE clauses and AU clauses is marked in the morphology or the syntax or both.

Adpihi and Reptigan do that; SE clauses don't masquerade as AU clauses.

Lots of SE clauses' English glosses would have a reflexive verb and/or an oblique argument or a prepositional-phrase argument.

"She worries about the hat", for instance, has "about the hat" as the E argument and "She" as the S argument.

_____________________________________________________________________________

Preliminaries 3: Abbreviations used in what follows:

Spoiler:

m stands for the marked clause's participants.
r stands for the referenced clause's participants.
S stands for the intransitive Subject.
A stands for the Actor.
U stands for the Undergoer.
E stands for the E-term.
| stands for "or".
= stands for "same", "identical", "equal".
> stands for "strictly contains", "properly contains".
< stands for "is strictly contained in", "is properly contained in".
<> stands for "strictly contains or is strictly contained in", "properly contains or is properly contained in".
mS stands for the marked clause's intransitive Subject.
mA stands for the marked clause's Actor.
mU stands for the marked clause's Undergoer.
rS stands for the referenced clause's intransitive Subject.
rA stands for the referenced clause's Actor.
rU stands for the referenced clause's Undergoer.
rE stands for the referenced clause's E-term.
mS|A stands for the marked clause's S or A.
rS|A stands for the referenced clause's S or A.
rU|E stands for the referenced clause's U or E.

_____________________________________________________________________________

My Conlangs' Switch-Reference Markers:

My conlangs' clause-chains have an Initial anchor clause, and some Consecutive clauses' nuclear verbs are obligatorily marked with one to three switch-reference prefixes showing whether, and which, and how the marked clause's core arguments are related to which of the anchor clause's participants.

The relationships coded are:

• identical equality
• strict, proper containment.

If one clause's participant is a group, and the other clause's participant consists of one or some, but not all, member(s) of that group, then one of them properly contains the other.

--------------------------------------------

So-Called "Same- or Different- -Subject" Markers:

Four of the markers -- do-, re-, mi-, and -fa- -- show which of the referenced clause's participants the marked clause's Subject or Actor is related to, and how.
do- and mi- show whether and how mS|A is related to rS|A.
re- and -fa- show whether and how mS|A is related to rU|E.

do- shows that the marked clause's Subject or Actor is precisely identical to the referenced clause's Subject or Actor.

re- shows that the marked clause's Subject or Actor is precisely identical to the referenced clause's Undergoer or E term.

mi- shows that the marked clause's Subject or Actor is strictly, properly contained in the referenced clause's Subject or Actor; or vice-versa.

fa- shows that the marked clause's Subject or Actor is strictly, properly contained in, or strictly, properly contains, the referenced clause's Undergoer or E term.

--------------------------------------------

So-Called "Same- or Different- -Object" Markers:

If the marked clause is transitive or ditransitive (that is, if it has an Undergoer), four of the markers -- -so-, -la-, -ti-, and -hut- -- show which of the referenced clause's participants the marked clauses Undergoer is related to, and how.
-so- and -ti- show whether and how mU is related to rU|E.
-la- and -hut- show whether and how mU is related to rS|A.

-so- shows that the marked clause's Undergoer is precisely identical to the referenced clause's Undergoer or E term.

-la- shows that the marked clause's Undergoer is precisely identical to the referenced clause's Subject or Actor.

-ti- shows that the marked clause's Undergoer properly contains or is properly contained in the referenced clause's Undergoer or E term.

-hut- shows that the marked clause's Undergoer properly contains or is properly contained in the referenced clause's Subject or Actor.

--------------------------------------------

Ambiguity and Disambiguation:

do- and -la- are unambiguous because neither clause can have both a Subject and an Actor.

If the referenced clause is only bivalent, re- and -so- are unambiguous; but if the referenced clause has both a U and an E -- that is, if it is ditransitive -- then re- and -so- might be ambiguous. I haven't provided a way for a speaker or an addressee to disambiguate this other than just the speaker making a separate statement, maybe after the addressee asks.

mi- and -hut- are ambiguous because they don't code which clause's participant contains the other clause's participant.
This can often be disambiguated by number.
If one is singular and the other is non-singular then the nonsingular one contains the singular one.
If one is plural and the other is non-plural then the plural one contains the nonplural one.
If one is a definite number (singular or dual or trial) and the other is an indefinite number (paucal or plural) then the paucal or plural one contains the singular or dual or trial one.
And so on.
(If these conlangs had sex-based gender -- Masculine, Feminine, Neuter, and Common -- then proper containment might sometimes be disambiguated by gender. If one clause's participant is in the Common gender and the other is M or F or N, then the one in Common would have to contain the other.)
(But these two conlangs' gender-system isn't sex-based.)

-fa- and -ti- may be ambiguous both for the reasons re- and -so- might be and for the reasons mi- and -hut- might be; and the remarks I said earlier about disambiguating re-, -so-, mi-, and -hut- can, with a little obvious modification, apply to -fa- and -ti-.

------------------------------------------

Obligatory Marking in Adpihi:

In Adpihi, if the situation encoded by a marker obtains, the marker is obligatory.
If no marker occurs it means either that the mS is not related to any of the r participants, or that neither the mA nor the mU is related to any of the r participants.
If do-, re-, mi-, or -fa- occurs without any of -so-, -la-, -ti-, or -hut-, it means that either there is no mU or that it is not related to any of the r participants.
If so-, la-, ti-, or -hut- occurs without any of do-, re-, mi-, or -fa-, it means that the mA is not related to any of the r participants.
If do-, mi-, -la-, or -hut- occurs without any of re-, -fa-, -so-, or -ti-, it means that either there is no rU|E or that it is not related to any of the m participants.
If re-, -fa-, -so-, or -ti- occurs without any of do-, mi-, -la-, or -hut-, it means that the rS|A is not related to any of the m participants.

------------------------------------------

Suppressed Marking in Reptigan:

For the most part, Reptigan's switch-reference marking system is like Adpihi's, except that, in Reptigan, if a clause is marked with mi- it won't also be marked with -fa-, and if a clause is marked with -ti- it won't also be marked with -hut-.

So, in Adpihi, if a clause is marked with mi- but not with -fa-, that means not only that mS|A<>rS|A, but also that there is no relationship between mS|A and rU|E. If mS|A>rS|A and mS|A>rU|E, then, in Adpihi, the clause will be marked with mifa-, not just with mi-.

But, in Reptigan, if a clause is marked with mi- but not with -fa-, that means only that mS|A<>rS|A; it means nothing about any relationship between mS|A and rU|E. If mS|A>rS|A and mS|A>rU|E, then, in Reptigan, the clause will be marked with mi-. Reptigan does not mark clauses with mifa-.

Likewise, in Adpihi, if a clause is marked with -ti- but not with -hut-, that means not only that mU<>rU|E, but also that there is no relationship between mU and rS|A. If mU>rU|E and mU>rS|A, then, in Adpihi, the clause will be marked with -tihut-, not just with -ti-.

But, in Reptigan, if a clause is marked with -ti- but not with -hut-, that means only that mU<>rU|E; it means nothing about any relationship between mU and rS|A. If mU>rU|E and mU>rS|A, then, in Reptigan, the clause will be marked with -ti-. Reptigan does not mark clauses with -tihut-.

_____________________________________________________________________________

Two Switch-Reference Markers on One Verb:

"Principle of Disjoint Reference":

In these conlangs, as in many natlangs that have highly productive valency-reducing reflexivization and reciprocalization processes (especially those that also have switch-reference-marking and clause-chaining), clauses in a clause-chain must satisfy what's called "the principle of disjoint reference".
That is, any two different referents of any single clause cannot have any common part.
If both are individuals they must be different.
If one is a group and the other is an individual, the individual must not be a member of the group.
If both are groups, they have to be disjoint; no thing can be a member of both of the groups.

This requirement is part of the reason why some pairs of the markers previously defined cannot occur together.

- - - - - - - - - - - - - - - - - - - -

Spoiler:

So,
do- encodes mS|A=rS|A
re- encodes mS|A=rU|E
mi- encodes mS|A<>rS|A
-fa- encodes mS|A<>rU|E

-so- encodes mU=rS|A
-la- encodes mU=rU|E
-ti- encodes mU<>rS|A
-hut- encodes mU<>rU|E

----------------------------------------

Two Markers Simultaneously On One Verb:

Fifteen two-marker combinations are allowed.

Any of do-, re-, mi-, or -fa- can occur with either of -so- or -ti-.
Either of re- or -fa- can occur with any of -so-, -la-, -ti-, or -hut-.
Any two of mi-, -fa-, -ti-, and -hut- can co-occur together.

Here are the allowed combinations and what they mean:

Spoiler:

doso- means: mA=rS|A and mU=rU|E.
doti- means: mA=rS|A and mU<>rU|E.
reso- means: (mA=rU and mU=rE) or (mA=rE and mU=rU).
rela- means: mA=rU|E and mU=rS|A.
reti- means: (mA=rU and mU<>rE) or (mA=rE and mU<>rU).
re-hut- means: mA=rU|E and mU<>rS|A.
mifa- means: mS|A>rS|A and mS|A>rU|E. Note that mifa- occurs only in Adpihi; in Reptigan mi- is used instead.
mi-so- means: mA<>rS|A and mU=rU|E.
mi-ti- means: mA<>rS|A and mU<>rU|E.
mihut- means: mA<rS|A and mU<rS|A.
-faso- means: (mA<>rU and mU=rE) or (mA<>rE and mU=rU).
fala- means: mA<>rU|E and mU=rS|A.
-fati- means: (mA<>rU and mU<>rE) or (mA<>rE and mU<>rU) or (mA<rU|E and mU<rU|E).
fa-hut- means: mA<>rU|E and mU<>rS|A.
-tihut- means: mU<rS|A and mU<rU|E. Note that -tihut- occurs only in Adpihi; in Reptigan -ti- is used instead.

----------------------------------------

Why are the disallowed combinations disallowed?

Spoiler:

Two items related by proper-containment <> cannot have precisely the same referent.

So do- and mi, re- and -fa-, -so- and -ti-, and -la- and -hut-, are incompatible pairs.

Spoiler:

If mA>=rS|A and mU>=rS|A then the referents of mA and mU are not disjoint because the referent of rS|A is a included in the referents of both mA and mU.
If mA>=rS|A and mU<=rS|A then mA>=mU.
If mA<=rS|A and mU>=rS|A then mA<=mU.

So we can't have both mA and mU related to rS|A unless they are both properly contained in it.
In other words, out of the markers {do-, -la-, mi-, -hut-}, the only compatible pair of markers is mi-hut-.

Spoiler:

If mS|A<=rS|A and mS|A<=rU|E then the referents of rS|A and rU|E are not disjoint because the referent of mS|A is included in the referent of both rS|A and rU|E.
If mS|A<=rS|A and mS|A>=rU|E then rS|A>=rU|E.
If mS|A>=rS|A and mS|A<=rU|E then rS|A<=rU|E.

So we can't have mS|A related to both rS|A and rU|E unless it properly contains both of them.
In other words, out of the markers {do-, re-, mi-, -fa-}, the only compatible pair of markers is mifa-.

Spoiler:

If mU<=rS|A and mU<=rU|E then the referents of rS|A and rU|E are not disjoint because the referent of mU is included in the referent of both rS|A and rU|E.
If mU<=rS|A and mU>=rU|E then rS|A>=rU|E.
If mU>=rS|A and mU<=rU|E then rS|A<=rU|E.

So we can't have mU related to both rS|A and rU|E unless it properly contains both of them.
In other words, out of the markers {-so-, -la-, -ti-, -hut-}, the only compatible pair of markers is -tihut-.

Spoiler:

(We can have both of mA and mU related to rU|E, but the interpretation is not completely free. Either one of them is related to rU while the other is related to rE (which requires that the reference clause be ditransitive), or they are both properly contained in rU|E. So each of re- and -fa- is compatible with each of -so- and -ti-. If either re- or -so- is involved, one of the marked clause's participants is related to rU while the other is related to rE. For -fati- there is also the possibility that rU|E properly contains both mA and mU.)

_____________________________________________________________________________

Three Switch-Reference Markers on One Verb:

If the marked clause is transitive or ditransitive (has both an mA and an mU) and the referenced clause is ditransitive (has an rA and an rU and an rE), there are four allowed combinations of three markers.

Either re- or fa- can go with -tihut-.
And, mifa- can go with either -so- or -ti-.

So the allowed three-marker combinations are:
retihut- which means (mA=rU and mU>rA and mU>rE) or (mA=rE and mU>rA and mU>rU);
fatihut- which means (mA<>rU and mU>rA and mU>rE) or (mA<>rE and mU>rA and mU>rU);
mifaso- which means (mA>rA and mA>rU and mU=rE) or (mA>rA and mA>rE and mU=rU); and
mifati- which means (mA>rA and mA>rU and mU<>rE) or (mA>rA and mA>rE and mU<>rU).

--------------------------------------------------

There are two other sets of three markers, namely {mi-, -fa-, -hut-} and {mi-, -ti-, -hut-}, such that each two members of the set are compatible with each other.

But mi-hut- requires that mA<rS|A and mU<rS|A,
while mifa- requires that mA>rS|A and -tihut- requires that mU>rS|A.

So even though mifa-, mi-hut-, and -fahut- are allowable, mifahut- is not;
and even though miti-, mi-hut-, and -tihut- are allowable, mitihut- is not.

_____________________________________________________________________________

No four-marker combinations are allowed; because there are no sets of four markers each three of which are allowed together. In particular, mifatihut- is disallowed even though mifati- and -fatihut- are both allowed; because mifa-hut- and mi-tihut- are both disallowed.

[Testyal]

When you say 'actor', do you mean' agent'?

[eldin raigmore]

When you say 'actor', do you mean' agent'?

it is the most controlling participant of the clause. Actor is a kind of proto-Agent macro-role. If any bivalent or trivalent clause's most controlling participant is not also its most affected participant, then the controlling one is that clause's Actor.

[Testyal]

Guess I should read a bit more before I ask questions.

[Micamo]

Fascinating stuff, Eldin. Quick question though: Why do you choose to call your arguments Actor and Undergoer rather than the more standard terms Agent and Patient? Does U have different properties from what's usually termed the Patient?

[eldin raigmore]

Fascinating stuff, Eldin.

:-D

Does U have different properties from what's usually termed the Patient?

Maybe; sometimes; see below.

Why do you choose to call your arguments Actor and Undergoer rather than the more standard terms Agent and Patient?

Because I was looking for protoroles or macroroles or hyperroles.
See http://www.google.com/search?q=dowty+proto-roles and http://www.google.com/search?q=semantic ... ce+grammar and http://www.google.com/search?q=semantic ... les+kibrik. Also see http://www.google.com/search?q=Jackendo ... ction+tier and http://www.google.com/search?q=Actor+Undergoer.

In a clause in Adpihi or Reptigan, the argument with the most of Dowty's proto-Agent properties is the Actor and the one with the most of Dowty's proto-Patient properties is the Undergoer. If the same argument has both the most proto-Agent properties and the most proto-Patient properties, it is the (intransitive) Subject.
The "Actor" and "Undergoer" terminology is borrowed from Foley and Van Valin and La Polla and their Role and Reference Grammar (RRG).
Among the arguments that aren't S or A or U, if there are any, the one that is most involved -- that has the highest total of Dowty's proto-role properties (proto-Agent plus proto-Patient), is Primus's proto-Recipient.
In describing the grammar of Adpihi I call that the Extended core term.

The thing is, an Agent has an established meaning which the Actor won't always satisfy; "A prototypical agent is conscious, acts with volition (on purpose), and performs an action that has a physical, visible effect."

Likewise, a Patient has an established meaning which the Undergoer won't always satisfy.

And the next most-involved argument may not exactly be a Recipient.

For any clause that has two or more arguments, either one of them is what I called a Subject, or one of them is what I called an Actor and another is what I called an Undergoer.

For any clause that has a Subject and also at least one more argument, or any clause that has an Actor and an Undergoer and at least one other argument, (one of) the remaining argument(s) is what I called an Extended-core term.
I borrowed that Extended-core terminology from Dixon & Aikhenvald.

A highly-involved participant which is neither most controlling nor most affected is called "dative" in Barry J. Blake's book on Case (see this).
It is called "focus" in Rick Morneau's work, for instance http://www.eskimo.com/~ram/lexical_semantics.html#S4_5, http://www.eskimo.com/~ram/lexical_semantics.html#S28_0, http://www.eskimo.com/~ram/vocabulary_design.html.

[Micamo]

Oh, I think I get it. So your Actor role includes things like:

"The flood destroyed thousands of homes."

Which the Agent role doesn't include? I guess I had never thought about it that way before... Would it be possible for a language to distinguish volitionary actors (agents) and non-volitionary actors ("forces") through different case markers rather than having them both share one?

[eldin raigmore]

Oh, I think I get it. So your Actor role includes things like:

"The flood destroyed thousands of homes."

Which the Agent role doesn't include? I guess I had never thought about it that way before... Would it be possible for a language to distinguish volitionary actors (agents) and non-volitionary actors ("forces") through different case markers rather than having them both share one?

I'm almost positive I remember reading of one or more natlangs that do something like that. Unfortunately I can't remember when or where or by whom or the title.

An Agent and a Force are distinguished only in that Agents are animate and Forces are inanimate.
A Recipient and a Goal are distinguished only in that Recipients are conscious of being Recipients and Goals are often not even animate.
A Causee and an Instrument are distinguished only in that Causees are capable of being Agents while Instruments are usually inanimate.

Some languages distinguish Recipients from Goals by case. English uses the preposition "to" for both, but we speakers can usually tell whether the object of "to" is a Recipient or a Goal.

I think I remember reading that some languages distinguish Causees from Instruments by case.

[Micamo]

An Agent and a Force are distinguished only in that Agents are animate and Forces are inanimate.

This seems wrong. "I fell down the stairs." The subject "I" here is non-volitional but animate.

[Trailsend]

Would it be possible for a language to distinguish volitionary actors (agents) and non-volitionary actors ("forces") through different case markers rather than having them both share one?

That's an odd question. Of course it would be. If you meant "would it be naturalistically feasible" then it's a little less trivial, but still, the answer is definitely "yes."

An Agent and a Force are distinguished only in that Agents are animate and Forces are inanimate.

This seems wrong. "I fell down the stairs." The subject "I" here is non-volitional but animate.

I'm not sure what you mean, but if we're talking about general thematic relation terminology, the "I" in your example is neither an Agent nor a Force, but probably a Patient or maybe Theme.

[eldin raigmore]

I edited my first post; I hope it's more understandable now.

_______________________________________________________________________________

An Agent and a Force are distinguished only in that Agents are animate and Forces are inanimate.

This seems wrong. "I fell down the stairs." The subject "I" here is non-volitional but animate.

I'm not sure what you mean, but if we're talking about general thematic relation terminology, the "I" in your example is neither an Agent nor a Force, but probably a Patient or maybe Theme.

What Trailsend said.

In Fluid-S languages, in fact, the difference between "jumped down" and "fell down" is shown by case-marking the subject with either an A-case or an O-case.

"the stairs", in your example, would be an E-term if an argument (or not a term at all if an adjunct). They are not affected by the fall, and also they don't control it nor abdicate any control they could have had.

[Micamo]

Yeah but fluid-S doesn't, by itself, handle transitives. The non-volitional agent of a transitive clause can't take an O-case without causing ambiguity.

[intermundi]

Yeah but fluid-S doesn't, by itself, handle transitives. The non-volitional agent of a transitive clause can't take an O-case without causing ambiguity.

How about using the receiver-case? (not sure if this happens in any natlang)

[eldin raigmore]

Yeah but fluid-S doesn't, by itself, handle transitives. The non-volitional agent of a transitive clause can't take an O-case without causing ambiguity.

If I understand what you're saying: then No, that's wrong.

---------------------------------------------

In the second place:

"Fluid-S" is a subgroup of "Split-S" aka "Split-Intransitive".

In Split-Intransitive alignment, transitive clauses have Agents treated (e.g. case-marked or agreed-with or positioned) one way (abbreviated "A" for short) and have Patients (aka Objects) treated (e.g. case-marked or agreed-with or positioned) another way (abbreviated "O" for short).

In Split-Intransitive alignment, some intransitive clauses' Subjects are treated like the A above (these are called S_A) and some intransitive clauses' Subjects are treated like the O above (these are called S_O).

In Fluid-S alignment, some intransitive verbs can be in clauses with S_A subjects and also in clauses with S_O subjects. In other words, whether a verb's subject is spoken of as Agent-like or as Patient-like is not a fixed lexical feature of the verb.

---------------------------------------------

But in the first place:

In "I fell down the stairs" I don't believe "the stairs" is an argument at all, not even an oblique one.
I believe it's merely an adjunct.
Because it's perfectly grammatical and sensible to say "someone fell down" without saying "they fell down something".

---------------------------------------------

Did I misunderstand?
Or did you?
Or what?

---------------------------------------------

How about using the receiver-case? (not sure if this happens in any natlang)

Yes, it does.

If I read this search correctly, Leista Spanish, Basque, German, and maybe Itelmen, sometimes do that. Depending on the language, that can depend on the verb or on the object or both.

If the most agent-like participant (say the controlling, volition-exercising, instigating one) is also the most patient-like participant (say the most affected one); but some other argument exists that's highly involved, even though it's neither more agent-like nor more patient-like than the other; then, in some languages, the first participant may be treated like an intransitive subject while the other argument is treated like a dative.

See Sasha Aikhenvald's and Bob Dixon's book http://www.benjamins.com/cgi-bin/t_book ... d=TSL%2046.

See also Barry J. Blake's book Case http://www.librarything.com/work/361668 or http://www.coursehero.com/textbooks/106 ... nguistics/.

---------------------------------------------

(No longer replying to the most recent previous posts;)

@Micamo et al.;

The "Look vs See" thread seems also to have stuff in it relevant to volitional acts vs nonvolitional acts.

[eldin raigmore]

I found out some time ago (feels like more than a week but less than a month) that natlangs with switch-reference markers can use them in complex sentences (ie sentences in which some clause is subordinate to another clause).

I can think of two ways this can be implemented.

I foresee difficulties unless the marked clause is contained within the referenced clause.
If the referenced clause is the subordinate clause and the main clause or matrix clause is the marked clause, I don’t see how that could work without a rule that no clause can contain more than one subordinate clause, and another rule that no subordinate clause could contain another clause subordinate to it.

So, I think, either the referenced clause is always the main clause; or,
the referenced clause is always the immediate matrix* of the marked clause.

*(A SC’s matrix is the clause in which it is embedded and upon which it is dependent.)

There’s no difference unless the marked clause is subordinate to another subordinate clause!

But if the marked clause is subordinate to a matrix which is itself subordinate to some other clause (perhaps to the main clause, or perhaps to another subordinate clause), then, I think, there must either be
a rule, for the entire language, that the referenced clause is always the main clause ; or else, instead, there must be
a rule, for the entire language, that the referenced clause is always the immediate matrix of the marked clause.

Otherwise I don’t see how the switch-reference markers can disambiguate which clause is the referenced clause!

….

This might change the truth-value of some of my previous statements in this thread.
And there may be other errors; though I’m pretty sure most readers will be able to figure out what I meant!

[Omzinesý]

), then, I think, there must either be
a rule, for the entire language, that the referenced clause is always the main clause ; or else, instead, there must be
a rule, for the entire language, that the referenced clause is always the immediate matrix of the marked clause.

An interesting note!


A[B[C]]

If B, is a same-subject clause, there is of course no problem because SS in C anyways refers to the same subject.
Some ambiguity is also allowed in languages. Very long strings of embedding don't usually appear.

[Creyeditor]

I probably didn't understand the posts in this thread well enough when they were posted. I understand them much better now, I think, so I might just try to answer your question. From the perspective of modern generative grammar (i.e. Minimalist Programm) "the referenced clause is always the immediate matrix of the marked clause." is a plausible option. The basic idea is called Locality and roughly means that dependencies should be as short as possible in order to reduce mental load. On the other hand, in earlier generative grammar Haj Ross proposed the Penthouse Principle, which basically means that their can be special rules referring to what you call "the main clause". So, anyway, just some ideas.

[Khemehekis]

On the other hand, in earlier generative grammar Haj Ross proposed the Penthouse Principle, which basically means that their can be special rules referring to what you call "the main clause". So, anyway, just some ideas.

Looked up Haj Ross, curious as to the ethnicity of someone with that name:

https://en.wikipedia.org/wiki/John_R._Ross

TIL that he invented the term "gapping"!