INEFFABILITY Plan of the unit: Present examples of „ineffability“ Discuss various solutions for the problem Concentrate on two solutions
Apr 01, 2015
INEFFABILITY
Plan of the unit:Present examples of „ineffability“Discuss various solutions for the problemConcentrate on two solutions
An architectural prediction
Recall : For each input, a set of candidates is generated, out of which Eval picks an optimal one
--> For every input, there must be at least one grammatical output.
This prediction of OT seems not fulfilled ...
Suffixation of -ize to an adjectival stem
rándom –> ràndomíze corrúpt –> *corruptizefóreign –> fòreigníze obscéne –>
*obscenizevápor –> vàporíze-ize can only be adjoined to a finally-stressed
stem (Raffelsiefen 1996, Kager 1999)
Raffelsiefen’s explanation: stress clash is strictly avoided. If suffixation of -ize would lead to a violation of *STRESSCLASH, the ize-verb is not realized.
Suffixation of -ize to an adjectival stem
The only viable solution is to block the output because of non-wellformedness of the output. Since the output cannot be wellformed (because stress is at the wrong place and because it cannot be moved), no optimal output arises.
Productive umlaut in German
Suffixation of the diminutive suffixes –chen and –lein to a finally stressed stem causes umlaut of the stressed vowel.
A metrically invisible schwa syllable can appear between the stressed umlauted vowel and the suffix triggering umlaut.
–chen and –lein are associated with a floating feature [front] looking for an appropriate docking place.
Productive umlaut in German a. Jahr –> Jährchen, Woche –> Wöchlein
b. Bruder –> Brüderchen, Mauer –> Mäuerchen
c. Mónat –> ?Monätchen, ?Monatchen *Mönatchen, *Mönätchen
Európa –> ?Europächen, ?Europachen, *Euröpächen
Wérmuth –> ?Wermütchen, ?Wermuthchen Wódka –> ?Wodkächen, ?Wodkachen,
*Wödkachen, *Wödkächen
Productive umlaut in German Partly unsolvable conflict in (c).
Most German speakers prefer not to diminutivize the stem, when stress is not final. An input consisting of Monat + chen is usually just not realized.
Wermüthchen with stress on ü is sometimes accepted.
Notice that speakers are unsure about their judgement of such data. They usually do not feel competent or motivate their reluctance with (irrelevant) semantic arguments (Jährchen ‘little year’ vs *Monätchen ‘little month’).
Productive umlaut in German
Similar problem as before.Stress is at the wrong place and it cannot be
moved. No output is well-formed.
This is not expected in the OT framework. The optimal candidate should be one which fulfills best the constraints. That’s it.
Prefixation of ge- to a participle
Prefixation of ge- illustrates a different kind of ineffability. This prefix is only present when the verbal root has initial stress. But the existence of a participle is independent of the presence of ge-.
Partial ineffability, then. Just the prefix is ineffable.a. geárbeitet, gegéssen, getrómmelt, [ge.[(lacht)F]PW]PW
b. gewállfahrtet, gefr´ühstückt, geóhrfeigt, gekénnzeichnetc. spazíert, trompétet, verpásst, prophezéit
Compare also:a. úntergetaucht, ángekommen (separable particle)b. überhólt (*übergehólt), entfállen (*entgefallen)
(inseparable particle)c. míssverstanden (*gemissverstanden,
*missgeverstanden, *missvergestanden), berücksichtigt (*begerücksichtigt, …) (inseparable particles)
The relevant constraint is MAX(Aff). It is just the morpheme which is not parsed, since the participle exists.
Ineffability in segmental phonology
• In each language, there is a ranking of the constraints on admissability of segments, like the following (tentative and partial for English)
*ò >> *¿ >> *Ó >> ∫ >> >> Ω >> ∫ >> etc.• [Ω] is not a canonical segment of English, but
it is realized in some environments. It is a kind of link between unallowed and allowed segments.
Ineffability in segmental phonology
What happens to an input like /¿ Ó ∫/which is a possible input because of Richness
of the Base (which says that all inputs must be allowed) ?
The answer is probably: nothing good can emerge from such an input. It is unpronounceable and is consequently ineffable.
Ineffability in segmental phonology
All possible correspondence relations between this input and outputs are eliminated.
It is not clear how to do that.
Ineffability in Syntax:MovementConstraints on extraction typically lead
to ineffability
Adjunct Extraction out of islands
?what do you wonder how to fix*how do you wonder what to fix t*how is it time to fix the car t
Ineffability in Syntax :MovementSubject Extraction out of islands
*who do you wonder when t met Mary
Multiple Superiority violations*who came why?*why did who come?
Ineffability in Syntax :MovementThe ban against multiple questions (in
Irish, e.g.)
Cén rothar aL ghoud an gardawhich bike C stole the cop?*cé aL rinne ciadéwho C did what*cé ciadé aLrinne
Ineffability in Syntax: InterfacesIneffability also arises quite often at the
“interface” between syntax and morphology.
The syntactically derived forms can not be spelt out by the available morphological means.
Deletion in full relatives
Der Mann der wo ihn liebtthe-nom man who-nom that loves him
Der Mann ---- wo ihn liebtDer Mann den --- ihn liebt
In certain dialects of German, either the relative pronoun or Comp can be deleted (cp. Pesetsky‘s account of English and French)
Deletion in full relatives
But Case must be recoverable
der Mann den wo er liebtthe man who-acc that he loves*der Mann --- wo er liebt
Deletion in full relatives
When two Cases fall together morphologically, deletion is fine
die Frau wo er mag schläftthe women C he likes sleeps die = nom (as required by matrix)
= acc (as required by complement)
Free relatives
There is a similar constraint in free relatives in certain dialectsWer ihn kennt, liebt ihn who-nom him knows loves himwen er kennt liebt erwho-acc he knows loves hewas er sieht gefällt ihmwhat-n/a he sees pleases him
Free relatives
Ineffability may arise: *wen er kennt liebt ihnwho he knows loves him *wer ihn kennt liebt erwho him knows loves he
More conflicts at the Interface
Georgian Agreement Markers
su do io io
1 v- m- m- mi-2 g- g- gi-3 -s h- u-1 v –t gv- gv- gvi-2 -t g..t g…t gi..t3 .en h- u-
Dealing with conflicts
Plus the constraint: not more than one affix on each side ...
Morphological conflict resolution:v- disappears if it competes with a further prefix.Otherwise: no resolution
Ineffability
Ineffability is thus a reality. It is an ubiquituous phenomenon of language, but morphology seems to play a special role.
If OT cannot deal with it, ineffability would constitute a serious problem (cp. Pesetsky assessment of OT syntax).
Solutions
Input-related solution: 1. One disallows certain inputsGEN-related solutions:2. Constraining GEN3. Making GEN more liberal
Solutions
EVAL-related solution: 4. Making certain constraints unviolableThe PARSE family of solutions 5. Different degrees of Parse violations
Solutions
Architectural changes6. Comparing Inputs 7. Bidirectional Optimizations
Restricting the Scope of OT8. Pesetsky9. Interface solutions
Input related solutions
1. Certain inputs are not allowed
This is not a good solution: Inputs should be free (Richness of the base). It is the task of the constraint hierarchy to eliminate bad outputs (and inputs).
Input related solutions
The solution would have to introduce a new language specific grammar component for certain (most?) cases:
An English-German contrast*who came whywer kam warum
Input related solutions
Manipulating Inputs
Often, ineffability arises because the form that expresses the meaning „intuitively“ does not participate in the competitions.Changes in the input concept might help
Input related solutions
brotherhood, happiness *brotherity *happity--> Abstract Inputs?
*Who do you think that t came?Who do you think cameIf that is in the input, an ineffability problem might arise
Input related solutions
Likewise:Extractions in Tagalog or Kwakwala:Only the subject can be questioned ...Are the passive morphemes part of the input??
Input related solutions
It seems obvious that a decision needs to be taken w.r.t. what is an input -- proper decisions might solve certain apparent ineffability problems ... but not all
There is no competitor to *obsenize or /¿ Ó ∫/
Input related solutions
Likewise:
how do you weep because she fixed the car t
(but see below!)
GEN related solutions
Solution 2: Constraining GENThe constraints on movement might be part of GEN(Chomsky: MLC is part of the definition of movement ...)
*who do you weep because t came
GEN related solutions
BUT:Movement contraints are rarely truly universal ... (GEN should be universal, though)
What happens to the input??*I don‘t care you weep because who came
GEN related solutions
We could also allow GEN to do more things than one would normally expect (solution 3):
Making a structure biclausal*who came why?who came, and why?
GEN related solutions
Relating free and bound relative clauses:*wen er kennt liebt ihn who he knows loves himmay be blocked ifjeder, den er kennt, liebt ihneveryone who he knows loves him is included in the candidate set.
GEN related solutions
The solution does not work in all cases (e.g. perhaps adjunct island constraints) and makes it likely that grammar gets out of control
EVAL related solutions (sol 4)
Some constraints must be obligatorily fulfilled by optimal candidates: candidates which do not fulfill them cannot be optimal. In some cases, all candidates are eliminated (usual solution).Recall: Stress clash is strictly avoided. If suffixation of -ize would lead to a violation of *STRESSCLASH, the ize-verb is not realized.
EVAL-related solution
• There is a filter called Control between Gen and Eval, consisting of unviolable constraints and blocking the formation of some words (Orgun & Sprouse 1997).
• However, *STRESSCLASH is not unviolable (compare Chìnése in English, `süßsáuer, tòt´müde in German), neither are NOUNSTRESSED ü/ö/ä (compare möblíeren) or UMLAUT (compare Frauchen).
EVAL-related solution
• So, Control cannot be the solution for ineffability in the case of -ize.
• Blocking is probably a property of the lexicon (rather than of the grammar).
EVAL related solutions
Are there truly universal constraints (In syntax, perhaps: Theta-Criterion, c-command condition on binding - but these look like interface things ...)
One leaves OT thereby ...
PARSE-related solutions (5)
Solution of Mcarthy & Prince (for Latinate suffixes), Raffelsiefen (for -ize) and many others: a candidate which violates a constraint requiring phonetic realization of morphemes is optimal.
M-PARSE: ‘Morphemes are phonetically realized.’
PARSE-related solutions
M-PARSE is ranked below the constraints requiring umlaut to occur in the right prosodic conditions. If the constraints on prosodic structure cannot be fulfilled, the morphemes are not realized and M-PARSE is violated.
M-PARSE in ineffability
Jahr+chen No unstressed
ü/ä/ö
UM
LAUT
NO
CROSSING
M-
PARSE
Jähr-chen
Jahr-chen *!
ø *!
M-PARSE in ineffability
Mónat+chen
No unstressed ü/ä/ö UMLAUT
NOCROSSING
M-PARSE
Monat-chen *!Monät-chen *!Mönat-chen *!ø *
Problems (Kager)1. Many suffixes influence the stress pattern of the
stem they adjoin to. Why do -ize or -chen don’t do that? (Not a real problem)
2. Stress clash is tolerated in English in some cases. Why not here? The same tolerance is found in German for unstressed umlauted vowels, as well as for stems suffixed with -chen and without umlaut: Frauchen, Blondchen, …. Why not Wodkachen or Wodkächen? (Not a problem either: OT alone cannot account for the different behavior of -ize and, say, -ic (remember titánic)
Summary so farThere seems to be basically two origins
of ineffabillity. First, what can be called grammatical ill-formedness. No output of a certain input can emerge as optimal, because the result always violates some high-ranking constraint which must be fulfilled in the language.
This was illustrated with *obsenize.
Summary so far
The other cause of ineffability is explained by the comparative power of OT. Some output is not optimal because there is another one which is. Ineffability is just apparent.
Lexical blockingBidirectional optimization in syntax
Lexical blocking• Some words which are perfect from the
point of view of morphology and phonology are nevertheless non-existent. First there is the phenomenon of blocking.
Data from Aronoff:various * varietycurious * curiousityglorious glory *gloriosityspacious space *spaciosity
• The non-existence of *gloriosity and *spaciosity is explained by the fact that a non-derived form with approximately the same meaning already exists, thus glory and space.
• The fact that variety and curiosity exist is explained by the absence of a non-derived nominal with the same meaning.
Bidirectional optimization • This kind of blocking is explained by a
bidirectional optimization between input and output, as well as between output and input. If there is a better input for some output, the output cannot be optimal. Instead it is the output corresponding to the input which is chosen as optimal (it violates less faithfulness constraints).
• An output like *gloriosity sees the input glory and, since glory as output is better than *gloriosity, *gloriosity cannot emerge as optimal.
Bi-directional optimization
STEP 1: Given an input, what is the optimal way of expressing it?
STEP 2: Given an output, what is its optimal interpretation?
Bi-directional optimization
Certain possible UR-s cannot enter the lexicon ...Input/veg/ away [vek] output/vek/ optimal lexical
entry
Bi-directional optimization
An interpretation is ineffable if its optimal expression has a different optimal interpretation!
To conclude, let’s turn to a detailed syntactic example
PARSE-related solutions
Predicate argument structures as inputs may not be sufficientKiss agent = Jane
patient = who
A problem for PAS?
A similar markedness consideration seems to show that PAS are not sufficient either:what did she tell me that he boughtshe told me what he boughtSame PAS but the first sentence has one more STAY and 2 more FI violations ...
No problem for PAS!
Closer inspection reveals that the two sentence differ in terms expressible by PASwhat did she tell me that he boughtshe told me what he boughtTELL selects a wh-clause in the second example, but not in the first one!!
A more complex problem
Although the examples just discussed can be explained away, more complex structures reveal that PAS are insufficient:
There are no MLC-effects in multiple questions whenever we get a semantic difference!
A more complex problem
Who wonders who bought whatMLC respected ...
Who wonders what who boughtMLC violation fine if who gets wide scope! .... Vs.*what did who buyno semantic difference!!!
A complex view of inputs
Input = PAS + indication of final scope
who wonders what who bought
This is a perfect sentence if ParseScope >> MLC
A possible situation
PASwonder: you, wh-clausefix: MANNER: how
OBJECT: whatscope(how) = low clausescope (what) = high clause
what do you wonder how to fix
An impossible situation?
PASwonder: you, wh-clausefix: MANNER: how
OBJECT: whatscope(how) = high clausescope (what) = low clause
What is the outcome???
An impossible situation?
It would seem that we get
*how do you wonder what to fixan ungrammatical sentence. Thus we seem to have run into the ineffability problem: there is just no way of expressing the PAS+scope ...
An overlooked possibility
But what if the ban on long adjunct movement (the displacement of how) has a higher rank than Parse(Scope).Thenyou wonder how what to fixhas a better profile thanhow do you wonder what to fix
An overlooked possibility
And the ban against multiple wh-phrases/the need to fill the matrix wh-position also has a higher rank than Parse(Scope), then what do you wonder how to fixhas the best profile!
A PARSE solution
Thus, ineffability appears to have a simple solution ... there are many aspects of the input to which the output need not be faithful ... and if scope can be among these aspects, a certain meaning cannot be expressed.
Architectural Solutions (6)
The picture is not complete, however!what do you wonder how to fixmay be the optimal way of dealing with an input in which how has wide scope ... but the sentence does not have that interpretation!
Bi-directional optimization
An interpretation is ineffable if its optimal expression has a different optimal interpretation! what do you wonder how to fixInterpretation 1: respects ParseScopeInterpretation 2: violates ParseScope