Distributed morphology

The Distributed Morphology (engl. Distributed Morphology , abbreviated DM , including: distributed morphology ) is a theoretical concept of general linguistics , in which the morphology is distributed to the syntax and phonology. It is intended to explain various inflectional but also derivation phenomena in the languages of the world.

introduction

Distributed morphology comprises a conceptual framework of different models of the processes that occur at the interface between syntax and phonology . These models try to explain how abstract morphosyntactic and morphosemantic features are replaced by their concrete phonetic realization. The model in its original form goes back to two essays by the American linguists Alec Marantz and Morris Halle from 1993 and 1994.

According to Greg Stump's (2001) taxonomy, distributed morphology is a post-syntactic, realistic-lexical theory of morphology.

Post-syntactic means that the selection of the morpho-syntactic features interpreted at the time of insertion is based on the hierarchies of syntactic derivation . This means that the syntactic part of the morphology can access the entire higher-level structure of the terminal node currently being processed .

Realisational means that the morpho-syntactic features are independent entities of the structure. In morphology, these features are realized by replacing them with phonological features. In contrast to this are the incremental theories, in which the morphology only introduces the features into the structure.

Lexical means that the information, which morphosyntactic features are replaced by which phonological, is stored morphemically as independent entities in the mental lexicon . Thus, for example, the association [ Plural ↔ / -s /] would be an entry in the mental lexicon of an English speaker. These stored in the lexicon entries which morpho with chains phonological features associating be in the DM as vocabulary items (such as: vocabulary entries , in other VI ), respectively. In inferential theories, however, the root of a word with specific rules with a separately stored in the lexicon word form , the sensitive for certain characteristics associated.

The sequence of a syntactic derivation according to the DM model proceeds as follows: First, a syntactic structure is generated on abstract morphosyntactic categories and characteristics. After completion of the syntactic derivation, the terminal symbols of the resulting structure are replaced by chains of phonological symbols (that is, sounds or the representation of sounds). This linking of syntactic categories and their phonological equivalents is commonly referred to as morphology. In the DM, the morphology itself is not viewed as an independent component of language, but as a process that is distributed across the levels of syntax and phonology, hence the name distributed (= distributed) morphology. The aim of the DM is to explain the principles on which this replacement takes place.

Distributed morphology should be understood less as an independent theory, but rather as a class of approaches that share a number of concepts. The three basic elements to which each DM derivative accesses are Late establishment (Engl. Late insertion ) lower specification (Engl. Underspecification ) and syntactic structure all the way down (Engl. Syntactic structure all the way down ). These are explained in more detail in the following section.

Basic concepts

The following three concepts are believed to underlie Distributed Morphology.

Late inauguration

The principle of the late establishment (Engl. Late insertion ) says that replacing morphosyntactic properties and characteristics is carried out by their phonological correspondences late in the derivation. In theory, the insertion takes place after completion of the syntactic structure. In the T-model of generative grammar theory , the point in time of insertion is placed at the beginning of the PF path, so in the DM-model the morphology is seen as an interface phenomenon between syntax and phonology.

Underspecification

The principle of under-specification results from a number of assumptions which are not themselves specific for the DM and are partly as old as modern theoretical linguistics itself. Only when they are combined do they form an essential basic component of any DM-like theory. These assumptions include the subsets principle, the principle of feature decomposition and the concept of specificity.

The subsets principle

The insertion, i.e. the replacement of the morphosyntactic and morphosemantic features with their phonological equivalents, takes place after the derivation of the syntactic structure, taking into account the so-called subsets principle. This principle states that a VI is inserted into a terminal node as soon as the set of features on this node are a superset of the features specified in the VI. The characteristics on the VI are said to be underspecified .

For example, if a terminal node has the characteristics [1. Person , singular , active , present tense , indicative ], a marker / e /, which for the characteristics [1. Person, Singular, Active, Indicative] is specified, can be used in this terminal node, since its set of features is a subset of the set of features on the terminal node.

One effect of the subsets principle is that syncretisms can be systematically derived from it. Syncretisms are occurrences of the same marker in different morphosyntactic contexts. The above marker / e / can appear in both indicative and subjunctive contexts, as it fulfills the subsets principle in both contexts.

Feature decomposition

In theory, characteristics are usually composed of smaller units. In German , as in all Germanic languages , there is a syncretism between the verb forms of the first and third person plural (for example we / they go). Since these syncretisms could not be derived with privative characteristics , it is assumed that these characteristics have something in common. The mentioned personal characteristics have in common, for example, that they do not include the addressee (s) of a speech act (2nd person). These personal characteristics can therefore be composed as follows:

first person: [+1, -2]
second person: [-1, +2]
third person: [-1, -2]

This splitting of the features into smaller units is called feature decomposition . The marker / -en /, which in German is attached to verbs in the first and third person plural, can be specified for the characteristics [-2, plural] according to this system of characteristics. According to the subsets principle, it can be used in both first plural and third plural contexts. This type of feature composition can be applied to almost all conventional features, although the nature of the underlying features differs from approach to approach.

Specificity

Due to feature decomposition and subsets, it can happen that several VIs can be used in one and the same context. For example, if one VI is specified for [+1, -pl] and another for [+1], both could be inserted in a node with the characteristics [+1, -pl, indicative]. To control such cases, one adopts the principle of specificity , which predicts which of the two markers will be used. What this looks like depends on the approach chosen. In the original version of the DM, the marker was used which is specified for most characteristics. In the above example, the first marker would be used because it is specified for two features, while the second is only for one feature. The system reaches its limits if a third VI is added, which is specified for [+1, indicative], for example. In such cases, further assumptions regarding specificity are necessary. Most approaches bring feature hierarchies into play at this point . Accordingly, the marker is used whose characteristics are higher in the hierarchy than those of the other marker. What these hierarchies look like also depends heavily on the chosen approach, but these hierarchies are usually based on typologically empirical data.

Syntactic structure all the way down

Another principle behind DM is called Syntactic Hierarchical Structure All the Way Down (English, analogously: Syntactic hierarchical structure all the way down ). It essentially says that the morphological and syntactic components of a language are based on the same hierarchical structures. The elements that both morphology and syntax work with are to be understood as discrete constituents , not as results of morphological processes.

Further concepts and background assumptions

The structure of the lexicon

The previously known concept of a mental lexicon is rejected by the authors of distributed morphology. Phenomena that were previously explained as lexical , such as the formation of compounds or derivatives , are distributed to other components in the DM. In the mental lexicon itself only the pure associations of morphosyntactically interpretable units and their meanings are stored. In order to avoid overlapping with other concepts of the mental lexicon, the DM is called an "encyclopedia" in which these associations are stored. The information stored in it is threefold: On the one hand, a semantic-conceptual meaning is associated with a so-called vocabulary item ( VI ). The VI are in turn pairs which associate morphosyntactic and / or morphosemantic features with their respective phonological counterparts. In the literature, these VIs usually take the form of context-sensitive insertion rules :

Sound correspondence ↔ chain of characteristics

As an example, the VI for the suffix is given here, which marks verbs in the first person singular active present tense in German (I go-e):

/ -ə / ↔ [1. Person, singular, active, present tense]

The notation means that the combination of features [1. Person, singular, active, present tense] is expressed by the sound / ə / and that this sound is a suffix (indicated by the symbol “-” in front of the sound). The encyclopedia entry for this morpheme therefore contains the information that the morpheme, which is realized as / -ə / and expresses the characteristics 1st person, singular, active, present tense, is associated with the meaning that the speaker alone (1st singular ) is the person performing (active) an action (verb ending ) at the time of conversation ( present tense ).

Furthermore, a distinction is made between two types of lexical entries, on the one hand the l-morphemes and on the other hand the f-morphemes. While the functional f-morphemes are fixed, the l-morphemes serve as placeholders for any lexical content. Thus, strains of nouns , verbs or adjectives (based in each case to a specific language) l-morphemes. These can be exchanged at will without changing the basic grammaticality of an expression. f-morphemes, on the other hand, are firmly linked to the grammatical construction, i.e. f-morphemes cannot easily be exchanged if the underlying characteristics are not changed. The following example should serve as an illustration:

Maria sees the man.

This sentence contains three l-morphemes, namely Maria as the stem of the subject , see , as the stem of the verb to see , and man as the stem of the object . These three can (to a certain extent) be replaced by any others without changing the grammaticality of the sentence:

Peter appreciates the hangover.

Hans reads the novel.

The other morphemes of the sentence above, -t and den , on the other hand, are firmly connected to the syntactic structure , if you change this the sentence becomes ungrammatic:

* Maria see the man

This is to make it clear that, in contrast to the l-morphemes, the f-morphemes are tied to the concrete morphosyntactic context. In addition to the inflectional affixes, the f-morphemes also include all kinds of pronouns, determiners (for example, certain and indefinite articles), adpositions and other functional categories.

The contextual, almost unlimited applicability of the l-morphemes allows you to use one and the same root in different categories, that is, a l-morpheme as "les", depending on the context into a noun (the Les ung), a verb we ( les s ), a past participle (the les end man) or the like. What a root is used as is determined by the morphosyntactic context, but does not start from the root itself. If the next neighbor of a terminal node in which the root is to be inserted, for example a tense node, the root is interpreted as a verb stem, if the next neighbor is a determinative node, the stem is interpreted as a noun, if it is a noun node as an adjective, etc.

Elsewhere

In many variants of the DM, the existence of a so-called elsewhere marker is assumed. This is a VI that fits all contexts currently under consideration and is used wherever no other, more highly specified VI fits. The possibility of the existence of an elsewhere marker arises from the other assumptions, which is why the subsets principle is often referred to as the elsewhere principle .

Readjustment Rules

The term “readjustment rules” refers to rules that operate again on the VIs used after the actual implementation in order to subject them to phonological changes. These rules are used in the case of morphological allophony . This is the phenomenon that a certain phonological rule is only applied in the presence of a certain morphological marker, i.e. in a not purely phonological context.

An example from the German verb inflection should serve as an illustration. The affix a verb in the second person plural is / -t / (her go -t ), which is the second person singular / st / (you go st ). With certain verbs, however, a schwa ( [ə] , written as <e>) is inserted between the stem and the ending, as in work : you work- e -st, you work- e -t. The phoneme sequence [tst] is not ungrammatic in German, cf. the word doctor speaking, [aɐ tst ] . According to the DM, the introduction of the Schwa is guaranteed by a readjustment rule that is only applied in the context of inflected verbs. Alternatively, you could say that -et or -est are the inflected endings of the second person in German and another readjustment rule deletes the -e in certain contexts.

Merger, split, amalgamation, impoverishment

A number of additional rules are adopted in the DM literature that manipulate the syntactic structure before the VIs are used.

Merger

The merger operation moves a terminal node to a different position in the hierarchy. This function is assumed to explain linearization phenomena. Such occur when the architecture of the theory provides for the licensing of a feature at a different location than where the feature is implemented. In some models of the syntax it is assumed that certain terminal nodes can only receive the value of a feature if they are in a certain position within the syntactic structure. If the case now occurs that this assumed position is different from what one perceives when listening to the finished sentence, merger operations are used and the respective terminal node is moved to the empirically observed position.

cleavage

Under cleavage (engl. Fission ) splitting a terminal node is free understood according to the principle of under specification. If a terminal node has the characteristics [a, b], two VIs can be used in the absence of a VI specified for both characteristics: one which is specified for [a] and another which is specified for [b].

While Halle and Marantz, for example, still viewed splitting as an independent operation controlled by a rule, other authors derived a concept of splitting from the subsets principle. This means that features that the most highly specified and matching VI does not itself can be used to implement less specific VIs.

merger

In DM, fusion is an operation in which two terminal nodes are merged into one. In this way, portmanteaus , for example , are explained, which are individual morphemes that implement the characteristics of several independent terminal nodes at the same time.

Impoverishment

As depletion (engl. Impoverishment ), the deletion of individual characteristics understood from the structure. Most of the time, these rules are context-sensitive. This allows a less specific VI to be used in a context that has a highly specified VI because depletion destroys the context for the more highly specified VI. Since less specific VIs fit more contexts, we talk on the application of impoverishment rules from the so-called retreat to the General Case (Engl. About return to the general case ).

Jochen Trommer regards impoverishment as the use of highly specific zero markers . He sees impoverishment as a consequence of the principle of division.

Distributed morphology and other models of grammar

Although the DM makes use of generative terminology, such as the scheme of the T-model, and some other assumptions of this, the two models are only compatible with one another to a limited extent. One of the most important differences is that all features are available during syntactic derivation, including those that are only interpreted by the morphology itself, such as features of inflection classes . In Noam Chomsky's theories, however, the syntax is devoid of any features that cannot be interpreted by the syntax itself, including inflection class features. Another difference concerns the timing of the morphological interpretation. While Chomsky prefers a pre-syntactic morphology, the morphological interpretation in the DM takes place post-syntactically.

Limits

In their introductory essay on DM from 1999, Heidi Harley and Rolf Noyer construct a “Martian” language that DM cannot explain. In this fictional language, the feature [plural] is expressed by deleting the last syllable of the word - regardless of whether this syllable belongs to an inflectional affix, the stem or a combination of several of these. In the DM, such “readjustment rules” can only affect individual vocabulary entries, but not more than one at a time. The consequence is that a language that expresses morphological features in this way cannot exist.

The pattern occurring in language, which is called subtraction , just like that of truncation , largely corresponds to this pattern, but always only affects the stem and never affixes. An example of a language with such a phenomenon is the Uto-Aztec language Papago , in which the perfect is marked by deleting the last syllable of the stem.

Other cases of non-concatenative morphology often cannot be derived in a DM approach without additional assumptions.

Explaining cases of extended exponence for certain variations of the DM approach is also problematic . Under extended Exponenz refers to the realization einunddesselben feature by more than one marker. If one assumes, like most DM approaches, that VIs “eat up” the morpho-syntactic features when they are implemented so that they are no longer available for later implementation, the occurrence of extended exponence should be excluded. There are also a number of approaches to this problem, for example context-sensitive insertion rules, secondary exponence or - in analogy to the impoverishment rules - enrichment rules that copy the multiple implemented feature in certain contexts.

A problem with the assumption of a radically underspecified Elsewhere marker in combination with the split concept based on the subsets principle is that the Elsewhere marker could potentially be used as often as desired in one and the same terminal node. To prevent this, a filter must be adopted for such cases that does not follow any of the DM principles and thus represents a potential weak point of the theory.

The strict separation of the two types of morphemes (l- and f-morphemes) leads to the problem that some features can only be available after the introduction of the l-morphemes, but these sometimes have to be taken into account in the syntax, for example Genus -Characteristics. These are inherited, for example in German, within a noun projection, i.e. gender features are available in the entire noun phrase , which can be recognized by the gender congruence of adjectives and articles with the noun . Which gender is realized in the end, however, is decided only after the introduction of the nominal stem, here too a number of non-trivial additional assumptions are necessary in order to derive the correct derivation of such a construction.

literature

Gregory T. Stump: Inflectional Morphology. A Theory of Paradigm Structure. In: Cambridge Studies in Linguistics, Volume 93. Cambridge University Press, Cambridge, 2001. ISBN 0521780470
Morris Halle and Alec Marantz: Distributed Morphology and the Pieces of Inflection. In: K. Hale & SJ Keyser (Eds.) The View from Building 20 , MIT Press, Cambridge, Mass., Pp. 111-176. 1993.
Morris Halle and Alec Marantz: Some Key Features fo Distributed Morphology. MIT Working Papers in Linguistics 21, pp. 275-288. 1994.

Web links

Distributed Morphology: Frequently Asked Questions List. Frequently asked questions about distributed morphology and answers to them on the website of Rolf Noyer (English)

Individual evidence

^ ^A ^b Heidi Harley and Rolf Noyer: Distributed Morphology. Glot International Volume 4, Number 4, Pages 3-9, 1999
↑ for example Gereon Müller: A Distributed Morphology Approach to Syncretism in Russian Noun Inflection. In: Olga Arnaudova, Wayles Browne, Maria Luisa Rivero and Danijela Stojanovic (eds.) Proceedings of FASL 12, 2004. Online (English, last accessed: July 22, 2009; PDF; 98 kB)
↑ Jochen Trommer: Morphology consuming syntax 'Resources . In: Proceedings of ESSLI Workshop on Resource Logics and Minimalist Grammars Nijmegen, 1999. Online (English, last accessed: July 22, 2009; PDF; 179 kB)
^ JJ McCarthy: Morphology, Concatenative . In: RE Asher and JMY Simpson (Eds.): The Encyclopedia of Language and Linguistics. Pergamon, Oxford 1994, pp. 2598-2600.
↑ Gereon Müller: Extended Exponence by Enrichment: Argument Encoding in German, Archi, and Timucua Ms. Uni Leipzig, 2006. Online (PDF; 136 kB), (English, last accessed: August 6, 2009)

[HarleyNoyer-1] A ^b Heidi Harley and Rolf Noyer: Distributed Morphology. Glot International Volume 4, Number 4, Pages 3-9, 1999

[2] r example Gereon Müller: A Distributed Morphology Approach to Syncretism in Russian Noun Inflection. In: Olga Arnaudova, Wayles Browne, Maria Luisa Rivero and Danijela Stojanovic (eds.) Proceedings of FASL 12, 2004. Online (English, last accessed: July 22, 2009; PDF; 98 kB)

[3] Jochen Trommer: Morphology consuming syntax 'Resources . In: Proceedings of ESSLI Workshop on Resource Logics and Minimalist Grammars Nijmegen, 1999. Online (English, last accessed: July 22, 2009; PDF; 179 kB)

[4] JJ McCarthy: Morphology, Concatenative . In: RE Asher and JMY Simpson (Eds.): The Encyclopedia of Language and Linguistics. Pergamon, Oxford 1994, pp. 2598-2600.

[5] Gereon Müller: Extended Exponence by Enrichment: Argument Encoding in German, Archi, and Timucua Ms. Uni Leipzig, 2006. Online (PDF; 136 kB), (English, last accessed: August 6, 2009)