Negation and Negative Antonyms in Tlingit 1
Seth Cable
University of Massachusetts Amherst
1.
The Central Puzzle: Negation in Negative Stative Predicates
In the Tlingit language (Na-Dene; Alaska, British Columbia, Yukon), there is a small but highly
frequent set of stative, gradable antonym pairs, where the negative antonym is formed from:
(i)
The root of the positive antonym
(ii)
The negation marker tlél (hél)
(iii)
An additional (unproductive) morphological operation.
(1)
(2)
Negation in Negative Predicates of Tlingit 2
Positive Predicate
Negative Predicate
Negation of Positive
a.
yakʼéi
0CL.good
It is good.
tlél ushkʼé
NEG IRR.shCL.good
It is bad.
tlél
ukʼé
NEG IRR.0CL.good
It is not good.
b.
yaa ḵudzigei
sCL.smart
He is smart.
tlél yaa ḵushgé
NEG IRR.shCL.smart
He is dumb / foolish.
tlél yaa ḵusgé.
NEG IRR.sCL.smart
He is not smart.
c.
lix̱ éitl
lCL.lucky
She is lucky.
tlél ushx̱ éitl
NEG IRR.shCL.lucky
She is unlucky.
tlél ulx̱ éitl
NEG IRR.lCL.lucky
She is not lucky.
d.
litseen
lCL.strong
She is strong.
tlél ulcheen
NEG IRR.lCL.strong
She is weak.
tlél ultseen
NEG IRR.lCL.strong
She is not strong.
On the Morphological Changes Between Positive and Negative Antonym
a.
In (1a)-(1c), the so-called ‘verbal classifier’ (glossed ‘CL’) of the positive
predicate is changed to the ‘sh-classifier’ (glossed ‘shCL’) in the negative.
b.
In (1d), the root itself undergoes a consonant mutation from /ts/ to /tʃ/
1 For reasons of space, I have put my ample acknowledgments at the end of the handout.
2 I provide only the roughest of glosses for individual Tlingit words, which can be morphologically quite complex.
This simplification is most radical for verbs, as I provide glosses only for their lexical content and that inflectional
morphology that is important to my argumentation. In addition, I employ the following glossing abbreviations
throughout: 0CL ‘0-classifier’, 1 ‘1st person’, 3 ‘3rd person’, DUB ‘dubitative’, ERG ‘ergative’, FOC ‘focus
particle’, FUT ‘future’, IRR ‘irrealis’, lCL ‘l-classifier’, NEG ‘negation’, O ‘object’, PL ‘plural’, POSS ‘possessive’,
S ‘subject’, sCL ‘s-classifier’, shCL ‘sh-classifier’, SUB ‘subordinator’.
1 (3)
A Related Puzzle
A few negative stative (gradable) predicates are formed with negation, but contain roots
that cannot (any longer) appear in a positive form.
a.
(4)
(5)
(6)
(7)
b.
* kayahélʼk
0CL.strong(?)
Another Related Puzzle/Fact
There a few positive stative (gradable) predicates whose antonyms are (or can often be)
expressed in Tlingit by simply negating the positive predicate.
a.
(i)
naaléi
0CL.far
It is far.
(ii)
tlél
unalé
NEG IRR.0CL.far
It is near.
b.
(i)
x̱ ʼalitseen
lCL.expensive
It is expensive (dear).
(ii)
tlél
x̱ ʼeiltseen
NEG IRR.1CL.expensive
It is cheap.
The Central, Overarching Question
What is the morpho-syntax and morpho-semantics of the ‘negative predicates’ in (1)?
a.
What is their morphosyntactic structure?
b.
How does this structure get mapped to their observed meaning?
Subsequent Question
Can the answer to (5a,b) shed light on the superficially related facts in (3)-(4)?
•
What is the nature of the root hélʼk such that only (3a) is well-formed?
•
Do the negated predicates in (4a,b) actually have (strong) antonymic meanings? Or
are they semantically no more than the negation of the positive predicates?
Suggested Terminology: ‘NEG Antonyms’
•
•
•
Tlél koohélʼk
NEG IRR.0CL.strong(?)
He is weak.
The ‘Negative Predicates’ in (1)
The verbal form tlél koohélʼk in (3a)
The negated predicates in (4aii) and (4bii)
2 (8)
Outline of the Rest of Talk
a.
Section 2:
Background on methodology and the Tlingit language
b.
Section 3:
Evidence that the negation in the ‘negative predicates’ in (1) is
clausal negation (not incorporated negation)
c.
Section 4:
The interactions between ‘NEG antonyms’ and degree modifiers
d.
Section 5:
Formal analysis of the ‘negative predicates’ in (1).
Possible extensions to the ‘NEG antonyms’ in (3)-(4)
e.
Section 6:
Future directions: Cross-Polar (A)nomalies in Tlingit and English
2.
Linguistic and Methodological Background
(9)
The Tlingit Language (Lingít): A Few Bullet Points
•
Traditional language of the Tlingit people of Southeast Alaska, Northwest British
Columbia, Southwest Yukon Territory (shaded area in map below) 3
•
Member of the Na-Dene language family; distantly related to Athabaskan languages
(e.g. Navajo, Slave, Hupa). Shares the complex templatic morphology of this family.
•
Highly endangered; ≤ 200 speakers, all over 60, mostly over 70. Several fluent/nearfluent second language learners; a few raising their children in the language.
3 Map graphic by X’unei Lance Twitchell. Available at http://tlingit.info/.
3 (10)
Notes on the Data and Methodology
•
Unless otherwise noted, all data below were obtained through interviews with native
speakers of Tlingit (Summer 2015)
•
Four fluent elders have participated; all are residents of Juneau, AK; all are speakers
of the ‘Northern Dialect’ of Tlingit
o
o
o
o
William Fawcett (Kóoshdaak’w Éesh)
Carolyn Martin (K’altseen)
John Martin (Keihéenák’w)
One additional elder who preferred to remain anonymous
•
Interviews lasted two hours and were held in a classroom at the University of Alaska
Southeast; 2-4 elders were present at each interview.
•
Speakers were asked to translate English sentences paired with particular ‘scenarios’,
as well as to judge the ‘correctness’ (broadly speaking) of constructed Tlingit
sentences relative to those ‘scenarios’ (Matthewson 2004).
•
The scenarios were described to speakers in English, both orally and with
accompanying written text.
3.
Negation in Negative Predicates: Incorporated or Clausal?
(11)
A Natural First Perspective on the Facts in (1)
•
It’s not at all unusual for negative predicates to be derived from positive ones via
some kind of incorporated negator (11a).
•
Moreover, in some languages, this incorporated negation is lexically identical to main
clause negation (11b).
a.
English:
b.
Malay (Kroeger 2014):
(i)
tidak adil
NEG fair
unfair
•
unhappy, ineligiable, non-syntactic
(ii)
Mereka
tidak menolong
3PL
NEG help
They didn’t help us.
kami
1PL
Perhaps the difference between the negative predicates in (1) and the negation of
the positive predicates is that the former have incorporated negation?
4 (12)
(13)
An Ineffective Objection
a.
Objection:
But, in the negative predicates under (1), the negator tlél appears outside all the
inflectional morphology of the verb!
o If it were incorporated, wouldn’t the inflectional morphology appear outside
the negation?
b.
Immediate Reply:
Tlingit is a Na-Dene language. So, inflectional and derivational morphology often
show unexpected orderings (Rice 2000)…
A Better First Argument
•
a.
•
b.
•
(14)
In Tlingit, clausal negation triggers so-called ‘irrealis’ morphology on the verb, with
concomitant changes in the realization of aspect and mood inflection.
(i)
tlél
ukʼé
NEG IRR.0CL.good
It is not good.
(ii)
* tlél yakʼéi
NEG 0CL.good
Crucially, the negative predicates in (1) all obligatorily appear with this irrealis
morphology (see the ‘IRR’ in their glosses)
(i)
tlél ushkʼé
NEG IRR.shCL.good
It is bad.
(ii)
* tlél shikʼéi
NEG shCL.good
We wouldnʼt typically expect a lexically-incorporated (derivational) morpheme
to have effects upon the realization of inflectional morphology…
An Even Better, Second Argument
Just as with clear cases of clausal negation, the negator in a NEG antonym can be
separated from the rest of the verb by other material, including:
• Focus particles (15)
• Dubitatives (16)
• Indefinites (17)
(15)
Focus Particles in NEG Antonym
a.
Tlél áwé ushkʼé
NEG FOC IRR.shCL.good
It’s bad.
b.
5 Tlél áwé ulcheen
NEG FOC IRR.lCL.strong/weak
He’s weak.
(16)
Dubitatives in NEG Antonym
Tlél shákdéwé
NEG DUB.FOC
It’s probably bad.
(18)
(19)
ushkʼé.
IRR.shCL.good
Indefinites with NEG Antonym
Tlél daa sá
ushkʼé
NEG anything
IRR.shCL.good
Everything is bad.
Key Semantic Observation
•
The sentence in (17) is not only well-formed, but is claimed to contrast with that in
(18a), which contains clausal negation of the positive predicate yakʼéi ‘it is good’.
•
As reflected in their English translations, (18a) below is felicitous in scenario (18b),
while (17) is not.
a.
Tlél daa sá
NEG anything
Nothing is good.
b.
Scenario:
A bunch of food is laid out on the table. None of the food is really any good.
Some of it is just so-so (so it’s not really bad), but nothing is really good.
ukʼé
IRR.0CL.good
Third Argument: Negation of Negative Predicates
•
a.
(20)
(17)
As indicated in (19a), there is not – nor is there expected to be – any morphosyntactic incompability between clausal negation and incorporated negation.
He is not unlucky.
•
As shown in (20)-(22) below, however, it is not possible in Tlingit to directly negate
one of the negative predicates in (1).
•
The ill-formedness of (20), (21a), (22a) would follow if the negation in such negative
predicates were clausal negation…
o Many languages disallow multiple instances of clausal negation…
Negative Predicates in (1) Cannot be Directly Negated
* Tlél tlél yaa ḵushgé
NEG NEG IRR.shCL.smart
Judgment:
Ill-formed; not meaningful; does not mean ‘not dumb’
6 (21)
(22)
Negative Predicates in (1) Cannot be Directly Negated
a.
* Tlél tlél ushx̱ éitl.
NEG NEG IRR.shCL.lucky
Judgment:
Ill-formed, not meaningful, does not mean ‘not unlucky’
b.
Tlél yéi
utí
tlél
ushx̱ éitl
NEG thus is
NEG IRR.shCL.lucky
He’s not unlucky. (Lit. It’s not so that he’s unlucky.)
Negative Predicates in (1) Cannot be Directly Negated
a.
* Tlél tlél
ushkʼé
NEG NEG IRR.shCL.good
Judgment:
Ill-formed, not meaningful, does not mean ‘not bad’
b.
Tlél wáa sá
utí
NEG how
IRR.is
It’s not bad. (Lit. It’s not any way)
All these facts point to the following conclusions…
(23)
Key Syntactic Conclusion
The negation found with the negative predicates in (1) is clausal negation; it is not some
form of verbally incorporated negation.
(24)
Immediately Following Conclusion
The extra morphological changes found with the negative predicates in (1) – the change
in ‘verbal classifier’ or the mutation in the root onset – are licensed by clausal negation.
(25)
Some Obvious Subsequent Questions
a.
Is it negation per se that licenses these morphological changes seen in (1)? Or,
would downward-entailment be sufficient? (i.e., are they like weak NPIs?)
b.
Can clausal negation license these morphological changes across a clauseboundary? Or, does only clause-mate negation license these changes?
7 (26)
Downward Entailment (D.E.) is Not Sufficient for Verbal Classifier Change
•
As shown by the contrast between (26a) and (26b), simply embedding the root k’éi
‘good’ in a conditional antecedent isn’t sufficient to license the shift to sh-classifier
•
As shown by the contrast between (26a) and (26c), conditional antecedents do license
the use of interrogative pronouns as (plain) indefinites (as negation does)
(cf. shenme ‘what’ in Mandarin)
•
These facts support the following picture:
o Like in many other languages (e.g., Mandarin) interrogative pronouns in
Tlingit can be used as NPIs, where they are licensed in D.E. environments
(e.g. negation, antecedent of conditionals)
o The use of an sh-classifier with the root k’éi ‘good’ in Tlingit is not
generally licensed in D.E. environments
(27)
a.
Daa sá kʼéiyi,
ax̱
éeshch
anything 0CL.good.SUB 1sg.POSS father.ERG
If anything is good, my father will eat it all up.
yax̱ ayagux̱ sax̱ áa.
FUT.eat.up
b.
* Daa sá shakʼéyi,
ax̱
éeshch
yax̱ ayagux̱ sax̱ áa.
anywhing shCL.good.SUB 1sg.POSS father.ERG FUT.eat.up
Judgment:
Ill-formed, not meaningful
c.
(i)
Aadóo sáwé
who.FOC
Who is sleeping?
tá?
0CL.sleep
Judgment: Cannot mean ‘Someone is sleeping’.
(ii)
Tlél aadóo sá
NEG anyone
Nobody is sleeping.
utá.
IRR.0CL.sleep
Higher Negation is Not Sufficient for Verbal Classifier Change
The contrast between (27a,b) shows that negation in a matrix clause does not license tuse
of the sh-classifier with kʼéi ‘good’ in a subordinate clause. Similar facts appear in (28).
a.
Yéi x̱ waajée tlél
ushkʼéyi
1sgS.think
NEG IRR.shCL.good.SUB
I think that it’s bad.
b.
* Tlél yéi x̱ wají
ushkʼéyi
NEG IRR.1sgS.think
IRR.shCL.good.SUB
Judgment:
Ill-formed, not meaningful
8 (28)
(29)
4.
Higher Negation is Not Sufficient for Verbal Classifier Change
a.
Tlél ax̱ tuwáa ushgú
NEG IRR.1sgS.want
I don’t want to be bad.
tlél
x̱ at ushkʼé.
NEG IRR.1sgS.shCL.good
b.
* Tlél ax̱ tuwáa ushgú
x̱ at ushkʼé.
NEG IRR.1sgS.want
IRR.1sgS.shCL.good
Judgment:
Ill-formed, not meaningful
Interim Summary: Negation in Negative Predicates
a.
The morphological changes found in the negative predicates in (1) – the change to
sh-classifier and the root mutation – are licensed by clause-mate clausal negation
b.
When clausal negation appears in concert with these morphological changes, the
meaning is strengthened to an antonymic concept.
(i.e., ‘not good’ < ‘bad’; ‘not smart’ < ‘dumb’)
Interactions Between ‘NEG Antonyms’ and Degree Modifiers
The formal analysis presented in Section 5 will capture the facts in (29) as well as the following
pattern regarding the interactions between ‘NEG antonyms’ (7) and degree modifiers.
(30)
Scope/Word-Order Generalization for Negation and Degree Modification
a.
In order to modify a NEG antonym with a degree modifier, the modifier must
precede the negation.
b.
In order to negate a positive predicate modified by a degree modifier, the negation
must precede the degree modifier.
As illustrated below, these generalizations are not terribly surprising…
But it’s definitely worth documenting their existence…
…And capturing them is a nice result of the proposed syntactic-semantic analysis…
9 (31)
NEG Antonyms with Degree Modifiers
The following examples illustrated degree modification of NEG antonyms, as well as
their positive antonyms.
•
a.
b.
c.
d.
e.
In all cases, the degree modifier is yáanáx̱ ‘more than’. In all cases, modification of
the NEG antonym places the degree modifier before the negator
(i)
A yáanáx̱
áwé yakʼei
3O.more.than FOC 0CL.good
This one is better than it.
(ii)
A yáanáx̱
tléil ushkʼé
3O.more.than NEG IRR.shCL.good
It is worse than it.
(i)
Wéitʼaa
áwé a yáanáx̱
litseen
that.one
FOC 3O.more.than lCL.strong
That one is stronger than him.
(ii)
A yáanáx̱
áwé tlél ulcheen
3O.more.than FOC NEG IRR.lCL.strong(+RootChange)
He is weaker than him.
(i)
Ax̱ yáanáx̱
yaa ḵoodzig̱ éi
1sgO.more.than
sCL.smart
She is smarter than me.
(ii)
Héitʼaa yáanáx̱
hél yaa ḵooshgé
that.one more.than NEG IRR.shCL.smart
He’s dumber than that one.
(i)
Kúnáx̱ a yáanáx̱
naalée
very 3O.more.than 0CL.far
It’s a lot farther than that.
(ii)
A yáanáx̱
tléil unalí
3O.more.than NEG IRR.0CL.far
It’s closer than that.
(i)
Yáatʼaa
a yáanáx̱
x̱ ʼalitseen
this.one
3O.more.than lCL.expensive
This one is more expensive than that.
(ii)
A yáanáx̱
áwé ḵúnáx̱ tléil x̱ ʼeiltseen
3O.more.than FOC very NEG IRR.lCL.expensive
This one is much cheaper than that.
10 yáatʼa
this.one
yáatʼaa
this.one
(32)
Negation of Degree Modification
The following examples illustrate the negation of positive predicates modified by degree
modifiers.
• In all cases, the degree modifier appears after the negation
a.
(cf. (31a))
b.
Tlél tlax̱
ukʼé
NEG very IRR.0CL.good
Itʼs not very good.
c.
Tléil tlax̱
a yáanáx̱
x̱ ʼeiltseen
NEG very 3O.more.than IRR.lCL.expensive
Itʼs not much more expensive.
(cf. (31e))
Tlél a yáanáx̱
unalí
NEG 3O.more.than IRR.0CL.far
It’s not farther than that.
(cf. (31d))
d.
(33)
Tlél a yáanáx̱
ukʼé
NEG 3O.more.than IRR.0CL.good
Itʼs not better than that.
e.
Tlél tlax̱
unalí
NEG very IRR.0CL.far
Itʼs not very far.
f.
Hél
tlax̱
utʼá.
NEG very IRR.0CL.hot
Itʼs not very hot.
g.
Hél
a yáanáx̱
utʼá
NEG 3O.more.than IRR.0CL.hot
Itʼs not hotter than that.
Further Evidence for Semantic Contrasts in (31)-(32)
In scenario (33a), only the negation of a comparative would be true. The comparative of a
negative predicate would be false (since the food is equally good).
a.
Scenario:
b.
Tlél a yáanáx̱
ukʼé
NEG 3O.more.than IRR.0CL.good
It’s not better than that.
Judgment:
Acceptable in this scenario
A yáanáx̱
tlél ushkʼé
3O.more.than NEG IRR.shCL.good
It’s worse than that.
Judgment:
Not acceptable in this scenario
c.
We have two plates of food that are equally good.
11 (34)
Further Evidence for Semantic Contrasts in (31)-(32)
In scenario (34b), only the negation of a comparative would be true. The comparative of
a negative predicate would be false (since the hats are equally expensive).
a.
Scenario:
b.
Tléil tlax̱
a yáanáx̱
x̱ ʼeiltseen
NEG very 3O.more.than IRR.lCL.expensive
It’s not more expensive than that.
Judgment:
c.
(35)
We have two hats that are equally expensive.
Acceptable in this scenario
A yáanáx̱
tlél
x̱ ʼeiltseen
3O.more.than
NEG IRR.lCL.expensive
It’s cheaper than that.
Judgment: Not acceptable in this scenario.
Observation Regarding Negative Predicates in (4)
•
The ability for tlél x̱ eiltseen ‘cheap / not expensive’ (4b) and tlél unalí ‘near / not far’
(4a) to appear with degree modification (31d,e) suggests that they really do have
strong, antonymic readings…
o In most semantic treatments, the negation of a positive predicate
shouldn’t project the degree argument that a degree modifier would need
•
More work could/should be done, though, to somehow rule out the possibility that
(31d,e) and (34c) involve some kind of metalinguistic comparative.
o In English, it is marginally possible to say things like This restaurant is more
‘not expensive’ than that other one…
5.
A Formal Syntactic and Semantic Analysis of Tlingit NEG Antonyms
It has been independently proposed that negative adjectives in English actually contain syntactic
negation (Buering 2007, Heim 2008).
•
Under such analyses, predicates like bad in English underlyingly has the following (LF)
structure: [ NOT [ GOOD ] ]
•
The following analysis of the Tlingit NEG antonyms in (1)-(4) builds upon the tools
and insights of this prior work…
12 5.1
The Analysis: A Special, ‘Negative Concord’ Degree Operator in Tlingit
(36)
Gradable Predicates are Relations Between Degrees and Entities
a.
b.
c.
[[ k’ei / good ]]
[[ tseen / strong ]]
[[ lei / far ]]
=
=
=
[ λd : λx : goodness(x) ≥ d ]
[ λd : λx : strength(x) ≥ d ]
[ λd : λx : distance(x) ≥ d ]
In order for a gradable predicate to be used as a main predicate, it must appear with the
(phonologically null) POS operator…
(37)
The POS Operator
[[ POS ]]
=
[ λP<dt> : Lc ⊂ P ]
•
Lc is the contextually determined ‘no mans land’ of non-extreme degrees.
•
For example, in the case of ‘good’, it’s those degrees of goodness that neither count
as ‘good’ nor as ‘bad’.
Another ingredient I will use is the (phonologically null) degree-relative operator, DEG-REL.
(38)
(39)
The DEG-REL Operator
a.
Semantics:
[[ DEG-REL ]] =
[ λP<dt> : P ]
b.
Syntax:
• Generated in the degree argument position of a gradable predicate.
• For type-theoretic reasons, undergoes movement, creating a syntactically
derived <d,t>-predicate (just like a relative clause operator)
Illustration: Positive Predicate Sentence
a.
Sentence:
Yak’éi ‘it is good’ (1a)
b.
LF:
[ POS [ DEG-REL1 [ 1 [ pro2 [ t1 [ yak’éi ] … ]
c.
Truth-Conditions:
Lc ⊂ { d’ : goodness(x2) ≥ d’ }
•
(39c) states that the degrees of goodness falling below x2’s degree of goodness
contain the ‘non-extreme’ degrees.
•
This could only be the case if x2’s degree of goodness is above the ‘nonextreme’ degrees, and so is among those degrees qualifying as ‘good’.
13 To capture the ‘negative predicates’ in (1), I propose the existence (in Tlingit) of an additional
degree-relative operator, NEG-REL
(40)
The NEG-REL Operator
a.
Semantics:
b.
Syntax: Same as ‘DEG-REL’, except that…
•
[[ NEG-REL ]] =
[ λP<dt> : P ]
Only licensed in the specifier of NegP; must move to SpecNegP
o In this sense, it’s a bit like a ‘negative concord’ item…
•
Furthermore, it is the only degree operator that is permitted to cross
negation and move to SpecNegP
•
Adjacency to predicate
phonological changes.
triggers
(lexically
stipulated)
morpho-
o E.g., shift to the sh-classifier; root consonantal mutation
(41)
Illustration: Negative Predicate Sentence
a.
Sentence:
tlél ushkʼé ‘it is bad’ (1a)
b.
LF:
[ POS [ NEG-REL1 [ 1 [ NEG [ pro2 [ t1 [ yak’éi ] … ]
c.
Truth-Conditions:
Lc ⊂ { d’ : ¬ (goodness(x2) ≥ d’) }
•
(41c) states that the degrees of goodness that don’t fall below x2’s degree of
goodness contain the ‘non-extreme’ degrees.
•
This could only be the case if x2’s degree of goodness falls below the ‘nonextreme’ degrees, and so is among those degrees qualifying as ‘bad’.
Side-Note:
• Movement of NEG-REL to SpecNegP will create a syntactically derived <d,t> predicate
•
Thus, to be used a main predicate, this structure must (again) be complement to a POS
operator…
•
Consequently, we must allow the POS operator in Tlingit the option of being
generated above NegP…
14 (42)
(43)
Comparison: Negation of a Positive Predicate
a.
Sentence:
tlél ukʼé ‘it is not good’ (1a)
b.
LF:
[ NEG [ POS [ DEG-REL1 [ 1 [ pro2 [ t1 [ yak’éi ] … ]
c.
Truth-Conditions:
¬ [ Lc ⊂ { d’ : goodness(x2) ≥ d’ } ]
•
(42c) states that the degress of goodness falling below x2’s degree of goodness
don’t contain the ‘non-extreme’ degrees.
•
Thus, these truth-conditions will hold if either:
o x2’s goodness falls within the non-extreme degrees (x2 is just so-so)
o x2’s goodness falls below the non-extreme degrees (x2 is bad)
Prediction: No Negative Predicate Reading of (42a)
This analysis predicts that sentence (42a) cannot have the LF in (41b), and so cannot get
the ‘negative antonym’ reading of (41a).
•
After all, if there were a degree operator in SpecNegP, it could only be NEG-REL…
•
But if NEG-REL were generated as the argument for yak’éi, then the classifier would
shift to the sh-classifier…
This analysis also predicts the interesting contrast in meaning between (17)-(18)…
(44)
Key Prediction: NEG Antonyms and NPIs
a.
Sentence:
Tlél daa sá ushkʼé ‘Everything is bad’ (17)
b.
LF:
[ POS [ NEG-REL1 [ 1 [ NEG [ anything [ t1 [ yak’éi ] … ]
•
•
•
c.
Truth-Conditions:
daa sá ‘anything’ is in the scope of NEG
NEG-REL moves to SpecNegP
The derived degree-predicate is argument to POS
Lc ⊂ { d’ : ¬ ∃x (goodness(x) ≥ d’) }
•
(44c) states that the non-extreme degrees are contained within those degrees
of goodness such that nothing has a degree of goodness above them.
•
This could only hold if everything’s degree of goodness falls below the nonextreme degrees, and so everything qualifies as ‘bad’
15 (45)
Comparison: NPI with a Positive Predicate
a.
Sentence:
Tlél daa sá ukʼé ‘Nothing is good’ (18a)
b.
LF:
[ NEG [ anything [ 2 [ POS [ DEG-REL1 [ 1 [ t2 [ t1 [ yak’éi ] … ]
•
•
•
•
c.
Truth-Conditions:
daa sá ‘anything’ is in the scope of NEG
There is no NEG-REL; only DEG-REL
The degree-predicated derived by DEG-REL is sister to POS
Since DEG-REL cannot move to SpecNegP, DEG-REL –
and thus POS – must be in scope of NEG
¬ ∃x. Lc ⊂ { d’ : goodness(x2) ≥ d’ }
•
(45c) states that there is no x such that the degrees of goodness falling below
x’s degree of goodness contain the ‘non-extreme’ degrees.
•
This could only hold if nothing has a degree of goodness higher than the ‘nonextreme’ degrees.
o Thus, nothing is definitely good
o Though some things could be merely so-so; not really bad.
Next, I introduce the following semantics for the degree phrase yáanáx̱ ‘more than’ (again,
borrowing from work by Buering (2007) and Heim (2008))…
(46)
Comparative Operator
[[ yáanáx̱ ]]
=
[ λP<dt> : [ λQ<dt> : P ⊂ Q ] ]
(47)
Illustration: Comparative with Positive Predicate
a.
Sentence:
A yáanáx̱ yak’éi ‘x2 is better than x4’ (31a)
b.
LF:
[ [ [ DEG-REL3 [ 3 [ pro4 [ t3 yak’éi ]…] yáanáx̱ ]
[DEG-REL1 [ 1 [ pro2 [ t1 yak’éi ] … ]
c.
Truth-Conditions:
{ d’ : goodness(x4) ≥ d’ } ⊂ { d’ : goodness(x2) ≥ d’ }
•
(47c) states that the degrees of goodness lower than x2’s goodness contains
the degrees of goodness lower than x4’s goodness.
•
This could only hold if x2’s degree of goodness is higher than x4’s
As shown below, this syntax-semantics predicts that yáanáx̱ ‘more than’ could take as argument
a NegP whose specifier contains NEG-REL
16 (48)
(49)
(50)
Illustration: Comparative with NEG Antonym
a.
Sentence:
A yáanáx tlél ushkʼé ‘x2 is worse than x4’ (31a)
b.
LF:
c.
Truth-Conditions:
[ [ [ NEG-REL3 [ 3 [ NEG [ pro4 [ t3 yak’éi ]…] yáanáx̱ ]
[ NEG-REL1 [ 1 [ NEG [ pro2 [ t1 yak’éi ] … ]
{ d’ : ¬(goodness(x4) ≥ d’) } ⊂ { d’ : ¬(goodness(x2) ≥ d’) }
•
(48c) states that the degrees of goodness that are above x2’s degree of
goodness contains the degrees of goodness that are above x4’s goodness.
•
This could only hold if x2’s degree of goodness is lower than x4’s
Comparison: Negation of a Positive Comparative
a.
Sentence:
Tlél a yáanáx̱ ukʼé
‘x2 is not better than x4’ (32a)
b.
LF:
[ NEG [ [ [ DEG-REL3 [ 3 [ pro4 [ t3 yak’éi ]…] yáanáx̱ ]
[DEG-REL1 [ 1 [ pro2 [ t1 yak’éi ] … ]
c.
Truth-Conditions:
¬ [{ d’ : goodness(x4) ≥ d’ } ⊂ { d’ : goodness(x2) ≥ d’ } ]
•
(49c) states that it’s not the case that x2’s goodness is higher than x4’s
•
Thus, these truth-conditions could hold if x2 and x4 were of equal goodness
Summary: the Scope/Word-Order Generalization in (30)
•
We’ve just seen how our syntax/semantic gives us an analysis of the comparative
sentences in (31) and (32)
•
Crucially, the LFs we’ve been considering accord with the surface word-order in
those sentences.
o In general, we’ve been assuming the natural view that ‘things to the left’
outscope ‘things to the right’…
•
In this way, our analysis captures the Scope/Word-Order Generalization in (30)
o In comparatives of NEG Antonyms, yáanáx̱ precedes tlél precisely because it
outscopes it.
o For exactly these reasons, tlél precedes yáanáx̱ in the negation of a (positive)
comparative.
17 5.2
Some Further Predictions and Questions
(51)
Structures Predicted to be Ill-Formed by the Analysis
Each of the following is predicted to be ill-formed under the analysis above. It is not
currently known to me whether they are in fact ill-formed (a task for this summer).
a.
Comparative Operator Appearing Before Negation of a Positive Antonym
A yáanáx̱
tlél
ukʼé
3O.more.than NEG IRR.0CL.good
(cf. (31a), (32a))
o If degree operator yáanáx̱ ‘more than’ is before negation, then it must be
scoping over negation…
o But since yáanáx̱ ‘more than’ needs a degree argument, that must mean that
tlél ukʼé is a (syntactically derived) degree predicate…
o But the only ‘degree relative’ operator that can move over tlél is NEG-REL…
o And if NEG-REL were in the structure, then the classifier with k’éi
‘good’ would have to shift to the ‘sh-classifier’
b.
Comparative Operator Between Negation and a NEG Antonym
Tlél a yáanáx
ushkʼé
NEG 3O.more.than IRR.shCL.good
(cf. (31a), (32a))
o If negation is before yáanáx̱ ‘more than’, then it must be scoping over ‘more
than’…
o Consequently, there must be a ‘degree relative’ operator moving from the
degree argument position of k’éi ‘good’ to a position below yáanáx̱
o The position this ‘degree relative’ operator is moving to is not SpecNegP
(obviously), and so this degree operator is not NEG-REL
o Consequently, there should be no shift to the sh-classifier on k’éi ‘good’
c.
Negation of a Comparative of a NEG Antonym
Tlél a yáanáx
tlél
ushkʼé
NEG 3O.more.than NEG IRR.shCL.good
(cf. (32a))
o A sentence like this would end up having two clausal negations, and so should
be just as bad as (22a)
18 (52)
Intervention Effects with DegP Movement in Tlingit 4
a.
Potential Problem for the Analysis in Section 5.1
o The analysis proposes that a degree relative operator ‘NEG-REL’ moves over
negation, to create a syntactically derived antonym of the root…
o Such movement of degree-OPs over logical/quantificational expressions,
however, is generally thought to be impossible
o If we allowed such movements freely, we’d wrongly predict that Tlingit
sentences like (53b) can be construed as true in scenarios like (54c)
b.
Key Sentence:
Ldakát hás
áwé du ḵín
everyone
FOC 3O.less.than
Everyone is less tall than Bill.
has koodligéi Bill.
lCL.big
Bill
c.
Falsifying Scenario: One person (Mary) is taller than Bill.
d.
Over-Generating a True Reading of (52b) 5
(i)
Problematic LF:
[ [ [ DEG-REL2 [ 2 [ Bill [ t2 ligéi ]…] ḵín ]
[ DEG-REL1 [ 1 [ Ldakát hás [ t1 has koodligéi ]…]
§
(ii)
In the LF above, DEG-REL moves over the quantificational subject
ldakát has ‘everyone’
Predicted Truth-Conditions
{ d’ : height(Bill) ≥ d’ } ⊃ { d’ : ∀x. height(x) ≥ d’ }
§
These truth-conditions state that the degrees d’ such that everyone is
taller than d’ is contained in the degrees of height below Bill’s height.
§
These truth-conditions only require Bill to be taller than the
shortest person!
• And so they could hold if there’s a person taller than Bill!
4 I thank Rajesh Bhatt for bringing this problem to my attention, and for his helpful discussion of it.
5 This problem was first noticed by Heim (2006) for the equivalent English sentence Everyone is less tall than Bill.
19 (53)
A Not-at-All Satisfying Solution (?)
§
‘Negative concord’ items like NEG-REL are exceptionally permitted to move over
negation on their way to SpecNegP
§
And, even then, they aren’t permitted to scope over quantificational DPs
Finally, let’s consider some ways in which the analysis from Section 5.1 might address those
questions under (6)…
(54)
NEG Antonyms Whose Roots Cannot Appear without Negation
a.
Question:
What is the nature of roots like hélʼk, which can only appear in ‘NEG antonyms’?
(i)
b.
Tlél koohélʼk
NEG IRR.0CL.strong(?)
He is weak.
* kayahélʼk
0CL.strong(?)
Possible Answers:
(i)
Perhaps NEG-REL optionally causes litseen ‘strong’ to undergo a
suppletive alternation to ka-hélʼk…
(ii)
Perhaps ka-hélʼk ‘strong’ specially c-selects for NEG-REL
§
(55)
(ii)
Note that the latter analysis would raise the question of whether verb
forms like sh-k’éi should be analyzed similarly (via c-selection)…
NEG Antonyms with No Morphological Change
a.
Question:
Do the negated predicates in (4) actually have (strong) antonymic readings? If so,
how are those derived?
(i)
b.
naaléi
0CL.far
It is far.
Possible Answers:
(ii)
tlél
unalé
NEG IRR.0CL.far
It is near.
The empirical question above has yet to be answered….
But, if it turns out that those verbs do have strong, antonymic readings…
These could just be cases where where NEG-REL fails to have a morphophonological effect…
20 6.
Future Directions: Cross-Polar (A)nomalies in Tlingit?
As mentioned earlier, Buering (2007) and Heim (2008) propose that English negative gradable
adjectives underlyingly contain negation….
The primary empirical motivation for this comes from the facts in (56)-(57)…
(56)
(57)
A Curious Contrast in English: Cross-Polar (A)nomalies (Buering 2007)
a.
This board is shorter than it is wide.
b.
* This board is taller than it is narrow.
The Cross-Polar ‘Nomaly’ Generalization (Buering 2007)
In a subcomparative, the ‘polarities’ of the adjectives can differ only if the positive
adjective is in the comparative phrase.
(58)
Buering’s (2007) Analysis of Generalization (57)
Buering (2007) puts forth a treatment of (56)-(57) that rests upon the existence of certain
PF-level operations occurring in English.
a.
Pre-Spell Out Structure of (56a):
[ This board [ [ not -er ] tall ] [ than it is wide ] ]
b.
PF Form of (56a):
Short PF movement of –er creates the following structure:
[ This board [ [ not tall ] -er ] [ than it is wide ] ]
c.
Subsequent Spell-Out Rules Acting on (58b):
[ not tall ] à /short/
d.
/short/ + [er] à /shorter/
LF-Structure of (56a): Covert movement of [not –er] creates the following:
[ [ not -er ] [ 1 [ this board t1 tall ] ] [ than it is wide ] ]
§
Note that in (58d), the adjectival predicates have the same (positive) polarity
§
Under Beuring’s (2007) semantics, (58d) ends up meaning ‘this board is less
tall than it is wide’
21 (59)
Heim’s (2008) Analysis of Generalization (57)
In brief, Heim’s analysis eschews the PF-operations proposed by Buering (2008). Instead,
Heim (2008) proposes that the PF and LF of (56a) are as follows:
a.
b.
Surface Form of (56a):
[ this board [ er [ NOT long ] ] [ than it [ DEG-REL NOT ] wide ] ]
§
[ er [ NOT long ] ] à [ er [short] ] à shorter
§
The negation in ‘[ DEG-REL NOT ]’ is deleted via ellipsis under identity with
the NOT that gets spelled out in ‘short’
§
The resulting string is thus pronounced this board is shorter than it is wide
LF Form of (56a):
[ 1 [ this board t1 NOT long ] ]
[ er [ than [ NOT [ DEG-REL2 [ 2 [ it t2 NOT wide ] … ]
§
The LF in (59b) can be obtained via particular movements of er, the DEGREL and NOT
§
Under Heim’s (2008) semantics, (59b) ends up meaning ‘this board is more
short than it is narrow’
Does Tlingit offer any perspective as to which of these analyses is closer to correct?...
(60)
On Applying These Analyses to Tlingit
IF Tlingit exhibits the same contrast in (56) that English does, then…
a.
b.
Buering’s (2007) analysis is a bit of a non-starter.
§
In Buering’s syntax, the comparative operator –er never forms a syntactic or
phonological constituent with the than-phrase.
§
This is fine for English, but it won’t be exportable directly to Tlingit, where
the comparative operator yáanáx̱ ‘more than’ forms a constituent with the
standard of comparison.
Heim’s (2008) analysis is more amendable…
§
Heim’s analysis doesn’t depend on any English-specific PF rules, and it
assumes that the comparative operator does form an LF constituent with the
‘than’-phrase.
22 (61)
7.
The Upshot
In a manner of speaking…
§
Buering (2007) predicts that Tlingit should not show the cross-polar anomaly pattern
in (56), while Heim (2008) predicts that it should…
§
It would be very enlightening, then, to know which of these predictions is true!
Conclusion
§
In Tlingit, certain negative (gradable) predicates are formed from the combination of
clausal negation and a positive root
§
I’ve put forth an analysis where these predicates are formed by means of degree
operator that must move to SpecNegP (a kind of ‘negative concord Deg-OP’)
o This analysis captures an array facts regarding the ways in which these negative
predicates interact with degree operators
o This analysis also makes several further predictions which have yet to be tested
with speakers…
§
(62)
This all lends indirect support to the view that negative (gradable) predicates in English
have a similar, underlyingly negative structure.
One Major Outstanding Question
§
How can the empirical question in (55a) be answered?
§
How can we show whether tlél unalí ‘not far’ truly has a (strong) antonymic reading
meaning ‘near’?
Acknowledgements:
Deepest thanks and gratitude are owed first and foremost to Tlingit elders William Fawcett (Kóoshdaak’w Éesh),
Carolyn Martin (K’altseen), John Martin (Keihéenák’w), and one additional elder, who preferred to remain
anonymous. I am deeply grateful for all that they have taught me regarding the Tlingit language. Special thanks are
also owed to Lance Twitchell (X’unei), James Crippen (Dzéiwsh), and Rose Underhill for their crucial logistical
support of this study. Finally, I would like to dedicate this work to the memory of Dick Dauenhauer (Xwaayeenák),
who with his wife Nora Marks Dauenhauer (Keixwnéi), has done more than anyone else to advance the study and
preservation of Tlingit language and culture.
This material has benefitted greatly from presentations at the UMass Semantics Workshop and The UMass
‘Funny Languages Afternoon’ (UMAFLAA). Special thanks to Rajesh Bhatt, Elizabeth Bogal-Allbritten, Vincent
Homer, Larry Horn, Rodica Ivan, Barbara Partee, Peggy Speas, Anthony Woodbury, and Ellen Woolford, for their
helpful comments upon earlier versions of this work.
This material is based upon work supported by the National Science Foundation, under Award No. BCS1322770.
23 References
Buering, Daniel. 2007. “Cross-Polar Nomalies.” In Friedmand, T. and M. Gibson (eds)
Proceedings of Semantics and Linguistic Theory 17, pp. 37-52. Ithaca, NY: Cornell
University.
Heim, Irene. 2006. “Little.” In Gibson, M. and J. Howell (eds) Proceedings of Semantics and
Linguistic Theory 16, pp. 35-58. Ithaca, NY: Cornell University.
Heim, Irene. 2008. “Decomposing Antonyms?” In Groenn, A. (ed) Proceedings of Sinn und
Bedeutung 12. Oslo: ILOS.
Kroeger, Paul. 2014. “External Negation in Malay/Indonesian.” Language 90(1): 137-184.
Matthewson, Lisa. 2004. “On the Methodology of Semantic Fieldwork.” International Journal of
American Linguistics 70(4): 369-415.
Rice, Keren. 2000. Morpheme Order and Semantic Scope. Cambridge: Cambridge University
Press.
24