Optimality theory

The Optimality Theory (English optimality theory, in other OT ) is a model of theoretical linguistics. The aim of the theory is to describe the linguistic expressions are grammatically in a single language and which are not.

The theory assumes that there are many different ways for each linguistic expression to realize this. You are playing in all these realizations in a competitive and based on the grammar of a language are gradually excluded all possibilities that do not fit this grammar. The realization of which is left at the end that meets the grammar of the best compared to all other ways, this implementation is therefore optimal in terms of the grammar.

• 2.1 A non- linguistic example
• 2.2 Examples from linguistics 2.2.1 Phonology
• 2.2.2 syntax

• 3.1 Comparison with generative grammar theories

• 5.1 evidence
• 5.2 criticism
• 5.3 Further comments

• 6.1 German Speaking
• 6.2 English spoken

Introduction

In the grammar theory, it is believed that all languages ​​of the world subject to the same principles. What can explain the theory concretely, is how the differences between these languages ​​come about and how the theory must be parameterized, that they exactly the structures derived that are grammatically in one language. The notion of grammaticality refers to the forms that actually occur in a spoken language, an ungrammatical expression would be more broadly one which does not occur in either the language or would not be understood by the speaker.

The grammar of a language is defined in the OT as an ordered set of so-called restrictions (English Constraints ). These are rules that determine what the properties are not supposed to have an expression. If an implementation has one of those " forbidden" properties, it says that it violates the corresponding limitation.

The restrictions are universal, meaning they apply to all languages. A single language - or more precisely their grammar - differs from another by the fact that these restrictions are weighted differently. The order from most to least important constraint is called ranking. In the OT are the principles to which all languages ​​are based on the restrictions that the parameter assignment would be the ranking, which is specific to any one language.

A term is referred to in the OT as input, the number of possible implementations of this expression is output or candidate set. So there are a number of candidates, including the need to choose the person you would like, in terms of the grammar of the best input for each input - so perfectly - met.

The selection of the optimal candidate is called evaluation or competition. This process essentially works as follows: In the beginning of the input, depending on the design theory, this is a deep structure, a word, the logical form of a sentence or the like to be. For this input, the candidate set is generated now, so a lot of ways that the input could be realized, ie, for example, surface structures, the phonetic form of a word, the concrete syntax or Other. Each of these candidates is characterized by the fact that he violated certain restrictions. First, all candidates will be thrown out of the competition, which violates the highest constraint. Of the remaining candidates are now thrown out, which hurt the next lower constraint and so on. This process continues until only one candidate remains, then this is the optimal candidate and represents a grammatical in a language expression.

Where the input comes to depends in large measure on the considered problem. In the case of phonology, which deals in large part to language production, the input comes for example from the mental lexicon optimized is ultimately the phonetic realization of the lexeme. In other approaches, the input be the optimal candidate of a previous evaluation, we speak here of the so-called " local optimization " (see also the section, Other Notes ). In the syntax is usually omitted entirely on an input, since you tried here to describe the structure of a language, regardless of their use. The decision whether a structure in a language is well-formed results, here only from the ranking of the constraints.

Panels

An important tool in optimality theoretic analysis are so-called panels, which are tables that are intended to illustrate the evaluation process graphically.

It is in the upper left box of the tableau of concrete input of the evaluation. In addition, the restrictions are from left to right according to their rankings listed. A notation often used in the literature for the ranking (the order ) of the restrictions is:

Where C »C means that C are ranked higher than C. In the tableaux so C would always be to the left of C.

In the first column of the tableau, the individual candidates, which were generated from the input to GEN. Injured a candidate a limitation, any injury in the corresponding field individually, each with an asterisk ( *) is indicated. If a candidate sub-optimal, that is, it does not violate a restriction, the other still in the competition befindlicher candidate or not so often violated, will be " leaving " with an exclamation mark (!) In behind the *. The decisive breach is called "fatal". As can be seen in the following example, it may happen that all candidates violate the same limitation ( This is the case in the constraint C). Since there is no optimal candidate in this case, to decide the next lower injuries. The optimal candidate is marked by the so- called " Pointing Finger ", a pointing hand (☞ ). The gray color is to highlight an additional visual aid to the suboptimal candidates.

What the two tableaux T and T is different, is alone the ranking of constraints C and C. It is clear that is optimal by rearranging these limitations the candidate CAND, although he injured a total of more restrictions than the other candidates.

Types of restrictions

A restriction within the meaning of the OT is a condition that is either not met or a candidate. If a candidate does not satisfy a condition, the corresponding constraint is considered violated. It is not excluded that a restriction is violated repeatedly by a candidate, see also the example of the syntax. There are generally two types of constraints: faithfulness and markedness constraints.

Faithfulness constraints (T) relate directly to the interaction between input and candidate. In general it can be said that faithfulness constraints are always violated when a candidate characteristics differ from those of the input.

Markedness constraints (M ), however, identify features that a candidate must have in order to be optimal in a language. For each M there is this loyalty restrictions that cancel its effect. This explains why, in a language a peculiarity predominates (M »T ), while it is ungrammatical in other languages ​​(T » M).

Another type of restriction is used in the prosody and the analysis of audio languages ​​. Here lay so-called alignment constraints (literally: "Alignment Restrictions" ) determines the directions in which, for example, sounds with their corresponding segments are to be associated.

Examples

A non- linguistic example

The three men Hans, Karl and Peter want to ever buy a car. Each one will clear ideas: Hans ' car should be very economical and have a bright color, its budget amounts to € 12,000. Karl, however, want a fast car, with him, the color does not matter and he has about 20,000 € available. Peter really want to buy a blue vehicle. For him, the main thing is that it drives as it gets given the car as well as the maintenance for his rich uncle, makes money for him not matter.

However, the car dealer has a very limited range on offer:

The car dealer stated that the (hypothetical) rule of thumb: "The more horsepower a car has, the faster it is and the more expensive it is to maintain ", therefore the small car would be as " economical" to see the sports car as "fast" and so expensive car. The combination is also conventionally called "fast " to see the car and therefore " not economical". In addition, there would be no problem to reorder a model, there should be two or more customers for the same vehicle.

The decision on who buys what car is like a optimality theoretical process: each of the three men has precise ideas (input) and three models to choose from ( candidates). From the given situation can be for all three customers current restrictions postulate:

• The color should match the customer's idea ( short color)
• The vehicle should not be more expensive than the customer has money ( price)
• This truck conforms to the presentation of the customers of economy and speed ( PS)

Depending on the customer, these restrictions are weighted differently: for Hans PS is most important, followed by a bright color. The money question is last with him. He will vote for the first car, even if it does not meet its color idea, since the other two models are not efficient enough. Charles priorities are similar, even for him, the PS property is most important in terms of speed. Since his budget is limited, this limitation the color comes second to last. He will vote for the station wagon, he also referred to as " fast" and the sports car is too expensive. Peters demands on his car are weighted as follows: In the foreground is the color, the rest does not really care. He will buy the first car as it is completely consistent with his requirements.

Each of the three buyers has now bought the car, which he considers to be the most appropriate, ie that which (budget, and offer ideas ) appears to him best under the circumstances.

Examples from linguistics

In the following two examples from the linguistic part of phonology and syntax are listed.

Phonology

In the phonology of the German a phenomenon which is called devoicing exists. Thus the word song in German [li ː t] is pronounced. In the OT, however believed that the pronunciation [li: d] is a possible pronunciation of the Germans, especially since it is the underlying form / li ː d / identical. Clearly, this underlying form of the inflected forms of the word, for example in the plural [li ː. Dɐ ], where the plosive / d / is no longer at the end of a syllable and therefore is not subject to devoicing, ie is pronounced voiced.

More important than the identity between the underlying form and pronunciation, but is a limitation of pronunciation possibilities for Auslautkonsonanten: Voiced obstruents are to be avoided here. Since the identity or loyalty restriction in German is less important than the limitation of the pronunciations ( markedness constraint ), the pronunciation [li ː t] by speakers of German is preferred. In English, the faithfulness constraint is more important than the aforementioned markedness constraint. The verb lead ( lead ) has the same underlying form as the German word song. However, since there is no devoicing in this language, it is there pronounced as [li ː d] with a voiced [ d].

According to these assumptions, the following restrictions can be postulated:

• * [ sth ] $ ( markedness constraint )
• ID [ ± sth ] ( identity or loyalty restriction )

The first constraint symbolizes the devoicing. It means that a candidate's restriction is violated (indicated by the asterisk at the beginning of limitation ), if at the end of a syllable ( indicated by the symbol "$" on the right) a voiced sound emerges. This sound then the property [ sth ] has to be. The second constraint states that all sounds in terms of their voicing in input and output match, so should be identical.

The following two panels represent the pronunciation of the words song in German ( ranking of constraints: * [ sth ] $ " ID [ ± sth ] ) and lead in English ( Ranking: ID [ ± sth ] » * [ sth ] $ ) opposite.

(Note: The devoicing relates only German plosives and fricatives, this fact has been ignored for simplicity in the postulation of constraints. )

Syntax

An example of the syntax is the explanation of different Wh - movement patterns in multiple interrogative sentences in the languages ​​of the world. It is about the position of Wh - phrases ( eg interrogative such as who, why, whose in German or why and what, in English, or more complex phrases, preceded by an interrogative such as Whose mother or Which of the many children who you mean). For example there is always only one Wh - phrase at the beginning of a (sub ) set:

In Korean, however, all Wh - phrases remain in situ, ie in the position where in a declarative sentence, the respective response would stand on the question words:

The Bulgarian hand, is a language in which all Wh - elements are moved to the beginning of the sentence:

( Note: The asterisk (*) stands for ungrammaticality, t denotes a trace, ie, the position of which was moved out of the coindexed element The index clarified here, is one which element to which track the structural representation of the expressions is here.. oversimplified. )

The following three limits are sufficient for the analysis:

• W - Crit: A W - phrase must be in a block at the beginning.
• Pur -EP: This is a constraint which punishes the emergence of more than one element between start of block and left sentence bracket. ( The precise definition is as follows: in the CP are not Mehrfachspezifizierer allowed. )
• Ökon: Disallow movement (more precisely tracks - t) in general.

The constraints are ranked yet as follows:

• German: Pur -EP »W- Krit » Ökon
• Korean: Pur -EP » Ökon » W- Crit
• Bulgarian: W - Krit » Pur -EP » Ökon

Since it is markedness constraints on all restrictions, an input is not necessary. How the candidates are generated, can be disregarded.

The selection of the optimal candidate is calculated as:

Developing the theory

The optimality theory was developed in the early 1990s by Alan Prince and Paul Smolensky. First, they declared this language-specific differences in the structuring of syllables. Then, the OT was also applied to other phonological problem questions.

Soon after work has been published that algorithms should show with which the OT can learn. The work " Optima " by Vieri Samek - Lodovici and Alan Prince shows in a very formal way regularities of the theory and derives what qualities must have candidate so that they can be optimally at all.

Since about 1995, the OT is increasingly being used in areas outside of phonology, for example, in the syntax.

Meanwhile, there are also approaches that do entirely without input and assume that the set of candidates is minimized by other processes, for example by assuming that candidates entering into a competition with only certain minimally different other candidates. (See also the example of the syntax)

Compared with generative grammar theories

In contrast to rule-based theories of grammar, the OT makes fundamentally different assumptions about the nature of constraints and candidates:

Certain similarities with the approaches of optimality theory shows the minimalist program (MP). So it is, in Chomsky (1995 ), among others, that one of two syntactic derivations ( derivations ) that can derive the same set, is preferable to that which is most economical. So two similar derivations go a competition with each other. In addition, restrictions may be violated under certain conditions in the MP if other constraints are satisfied by.

Formal Description

Restrictions ( constraints ) are generally considered in linguistics as a relation from a set of structures in a subset of the same:

Moreover, there is a constraint for every so-called layered hierarchy that places the elements in according to how often an item violates constraint. "Layered " The Adjunct means that within an order of multiple items can be classified as high, formal means

With an order relation over a set is. This means that behave two elements that are highly ordered in the same hierarchy, equal to all other elements in the order.

The maximum layer of the set of elements of is called, which are arranged at the highest in the order relation. In relation to the OT, this means that these elements the constraint less frequently injured than any other, said does not imply that these elements do not violate the Constraint. The consequence of this is that the restrictions are vulnerable.

This ratio indicates the maximum layer of back:

The amount is also referred to as favoring set of, so be the set of all structures that best meet the restriction in comparison to all other structures, thus " favors " from the constraint.

The second basic assumption of OT is that constraints are self- organized again, this order relation is called ranking (), the corresponding set of constraints is called as. Since the constraints are defined as relations, also the ranking can be regarded as a function in which the individual constraints apply according to their order on the set:

This means that the amount by höchstgerankte constraint is reduced by the items often violated as the elements which are contained. The result is a set, which in turn interacts with the next lower geranktem constraint (see concatenation of functions ), etc. This goes on until all constraints have been processed. The resulting set now consists of the structures that satisfy all constraints in terms of better rankings than all other structures and are therefore referred to as optimal. They behave identically with respect to the limitations of, so obey all constraints equally often.

In the application of the theory is usually chosen as a set of structures for a limited amount of the candidate, this amount depends upon the selected specific input. As a lot of candidates here are the structures of the candidates, including the injuries they bring for each constraint with himself understood.

Evidence and Criticism

Evidence

An important argument for an OT analysis of linguistic phenomena is their parameterizability, which is the ability to derive linguistic differences with the same theoretical framework. This happens in the OT means of changes in the restriction order.

Another positive aspect of the OT is the explanation of so-called repair phenomena (also last resort; linguistic structures that are used when all alternatives would result in violations of important restrictions ) such as Hiatverhinderung in German by the establishment of the glottal plosive in initial sound vowels as in [ bə.ʔaχ.tən ] or do- support in English (compulsory establishment of the verb do negated in sentences like John did not kiss Mary or Yes / No question sentences like Did John kiss Mary? ).

In addition, the OT can be applied in many areas of linguistics. Current research in the syntax example, deal with the question of how can be embedded optimization processes into existing syntax models. Promising here is the concept of local optimization of individual Derivationsschritte within the Minimalist Program ( MP) by Noam Chomsky. In the semantics / pragmatics, the principle of bidirectional optimality has been a useful implementation.

Criticism

A major problem of the OT is the overgeneration. Assuming that there are 600 constraints, which can be freely ranked yet all of them, resulting

Possible grammars. Some authors assume in addition that can be more complex constraints by means of an operation called "local conjunction ' from a relatively small number of fundamental limitations combine, is a direct problem with this is that these combinations can be continued by potentially infinite recursion, which infinitely many restrictions - and restriction orders - would result.

A frequently voiced criticism is that it is almost impossible by definition to derive real optionality, ie with the same input and two or more candidates to make optimally with different restriction profiles. It now various possibilities have been presented to address this problem. It was suggested, for example, the assumption that one and the same language can have several restriction orders. An important tool is the so-called coupling constraints ( cf. v. a Müller (2000: Chapter 5) and in analyzes presented ). Other approaches doubt the existence of real optionality and say that different realizations of linguistic expressions imply subtle differences in meaning. This means that not the same but nevertheless grammatical expressions in a language not in the same competition with, as for example, different inputs are based. A frequently cited problem is the relatively free word order in German, often at minimal differences in meaning in terms of focus, for example, in such analyzes is returned.

A frequently mentioned, especially in the syntax criticism of the OT is the lack of grammaticality degrees. In the languages ​​of the world can be made between ungrammatical and unacceptable expressions: The sets What do you think that she has done, do you think what she did? and What do you think she did? not include for all speakers of German to the more common expressions although they should be grammatically all. This differentiation is, however, the TDC, which can filter only an optimal candidate is not possible. There are attempts to circumvent this problem, but these bring radical changes to the basic idea of ​​OT with it.

Both positive and negative affects of the phenomenon of the harmonic limit of the OT. Many candidates are blocked by other candidates, which implies that that the blocked candidate can never be optimal, no matter how the constraints are ordered. This is exactly the case when there is a candidate to another candidate, equally often injured the same limitations, but regarding at least one constraint performs better than the blocked candidate. It may happen that a supposedly grammatical structure of an ungrammatical is blocked. A zoom drawn in this context argument against the OT is the so-called opacity.

Additional comments

Due to the relatively young history of development and the rapid advancement of the theory, there are still some conceptual inconsistencies concerning individual analyzes. So you have not yet agreed in almost all areas on uniform terms of restrictions, as well as varying the nature of the inputs, especially in the syntax, from author to author and from run to run.

Some authors, for example, accept as input the logical form of a sentence. Others prefer the assumption that the input of a syntactic competition was the final sentence, the candidates are the possible phonological realizations of the same, in which various semantically empty elements (eg complementizer like " that", or copula pronouns in pro -drop languages ) are omitted or are necessary. The above- mentioned concept of Local Optimization takes as an input to a step of the derivation. The optimal candidate this optimization step is then the input for further optimization at a higher syntactic level, etc. Still others show arguments in the syntax to dispense entirely with one input.

Some linguists maintain the OT for a metatheory: A designed within the framework of a theory of grammar rule-based grammar can be transformed into an OT grammar, and vice versa. The proof that such a conversion is always possible, but is still pending.