Mental Lexicon
The mental lexicon ( Latin mens , "thinking", "mind", "spirit" and ancient Greek neuter λεξικόν , lexikón , "concerning the word", from λέξις , léxis , "speech", "word" and the associated verb λέγειν , légein , “Collect”, “speak”, “[read]”), also the inner lexicon , is a generic term for the way in which the brain organizes vocabulary and the meaning of individual words. A lexicon is a level of description that is used to codify the vocabulary of a language, insofar as its forms and meanings cannot be derived from the general regulations of the language system . Jean Aitchison (* 1938) worked out the term “mental lexicon”, “human word store” in 1987.
background
The psycholinguistic model grew out of the need to understand how words and their meanings are represented in the brain. The vocabulary of an adult native speaker ranges from 3,000 to about 200,000 words. As a speaker, he can produce an average of three words per second and as a reader he can recognize and understand words just as quickly.
The model tries to represent in what way the mental lexicon
- is organized internally
- creates, stores and accesses the word entries in pairs of information for word and meaning
- First represents and forwards the entries for understanding when reading and listening
Further goals are the delimitation of the inner lexicon as subject knowledge from world knowledge and mental grammar and the elaboration of the mutual references.
Current research focuses are
- the process of word recognition (see also logos )
- the dissolution of polysemy and the semantic ambiguity inherent in every spoken language
- the integration of the lexical information in the environment on sentence and discourse level
- the storage of entire word families
- the structure of the vocabulary in (also foreign) language acquisition
- the representation of parts of speech
- and the handling of semantic prototypes , new words and non-words
in the inner lexicon.
The simultaneous advancement of technology and the development of ever finer measuring methods resulted in significant progress. The more precise measurement of ever smaller time units as well as the refined semantic priming of combined representations make it possible in the experiment to demonstrate in real time the activation (not only) of lexical contents (which run unconsciously) in the brain of the test subjects and also to show it. The entire experimental setup is called the on-line procedure .
The model of the mental lexicon advances to the central aspect of psycholinguistics, since all of the topics it researches are directly linked to the basic unit word. After all, all linguistic structural levels ( phonology , morphology , syntax , semantics , pragmatics ) relate to the individual word ( lexeme ).
The inner lexicon is a long-term memory for words and their meaning. The seat of the inner lexicon is long-term memory , since the information must necessarily be available over a longer period of time.
Word knowledge is based on two components - on the one hand on the lemma , which describes the semantic meaning and syntax category of the word, and on the other hand on the lexeme. Behind the lexeme is the knowledge of the word form, which describes the sound shape, morphological structure and its spelling. When storing and retrieving an entry in the brain, several pieces of information are required at the same time: the meaning of the word (semantic information), the rules for forming a sentence (syntactic information), the sound of the word ( auditory image as phonological information) and the appearance of the word ( Orthographic information).
The lexical processes ( word recognition and word formation ) run differently. What they have in common is a phase of selection from the various competition candidates found. When an entry is called up, the activation is first spread in the internal network, followed by the finding of possible candidates, from which the final term is selected in the third step.
In contrast to the usual lexicon, the entries are not sorted alphabetically. Querying frequently used words is faster than retrieving rarely used words.
The more neighboring words there are when reading, the slower the processing. ( Neighborhood effect ). Similar words (“MAUT”, “MOUSE”, “RAUS”) and non-words are recognized less quickly ( word superiority effect ).
The brain can query the meanings of words in a number of ways; the search key is sound, spelling or context. The brain is flexible (and probably organized as a neural network Spreading Activation Network ) and processes the flow of information to a high degree in parallel. With word recognition, there are two parallel access methods to the information sought in the mental lexicon ( dual route model ).
To research the neural basis of the mental lexicon, test persons with damage to the brain who show selective deficits in the processing of lexical content (e.g. Wernicke aphasia , semantic dementia or deep dyslexia ) are regularly consulted . Together with modern imaging processes and methods of electrophysiology (EKP), Elizabeth Warrington initially succeeded in narrowing down the brain regions for storing living beings and objects. Hannah Damasio's study was able to localize the zones for people, animals and tools. The empirically determined assumptions could be validated by developed imaging methods (for evaluating the PET data) .
Willem Levelt divides the speech production process into three main stages, namely conceptualization, formulation and articulation. Code switching and transfer errors among multilingual learners indicate that the mental lexicon has its own vocabulary for each language acquired. The dictionaries are used in parallel in the preverbal phase (before speaking) until the time of articulation. The phenomena would not be possible if both dictionaries were not active during language production.
The two competing models of language processing
Hierarchically serial model
The theory of hierarchical serial models for language production goes back to Willem Levelt . In 1989, he distinguished between three different levels, the conceptual, the lexical and the articulatory. If the inner lexicon is accessed on the lexical level, this takes place in two steps: First, the word information ( lemma ) with its information on semantics and syntax is retrieved and then the associated lexeme , which provides information on the phonology and morphology of the lemma. The tongue tip phenomenon and the phenomenon of the slip of the tongue , which occur either on word swaps on the lexeme level or on sound swaps on the lemma level, are considered to be experimental evidence of this division.
The hierarchical serial model is characterized by seriality , discretion and unidirectionality . Seriality assumes that the three levels take place one after the other. The concept level is able to activate several lemmas at the word level, one of which is selected. Only after a successful selection does the selected lemma activate the corresponding information of the lexeme level, while the lemma changes from the activated to a state of rest. This also results in the need to view activation and selection as discrete processes: A process can only start when the preliminary stage (through the selection) has been completed. The requirement of unidirectionality, in turn, means that subsequent processes cannot call a predecessor, i.e. the lexeme level cannot activate the lemma level (in the sense of a top-down process or feedback).
The basis for the assumption of unidirectionality (feedfoward) were the results of an image naming task (see also semantic priming ). The test person was offered several pictures which he should name. In some cases, he was given a wrong auditory input with the picture before he could describe the picture (i.e. in the lexical phase). During the activation phase, he heard the word "goat" before the selection phase, although he was presented with the picture of a sheep. The result was a semantic interference : Since sheep and goat are semantically related, the two terms competed and the image was incorrectly named.
New research shows that the ideas of seriality and unidirectionality are not consistent. The lexeme level is activated before the lemma level is completed and a lemma has been selected. Also semantic feedback measurable, ie, a back flow of the activation of the phonological segments back at the word level associated with an increase of alertness of the corresponding lemma.
These results simplify the more complex hierarchical serial model to the interactive cascading model, which allows overlapping of the respective levels. In addition, the cascading models explain phenomena such as lexical bias, mixed error and malapropism .
Interactive activation model
The interactive activation model (interactive activation model) by Gary S. Dell (1986) is a cascading model with bidirectional processing. The basic assumption is that the spreading activation (spreading activation) in a neural carried network. It follows that at the same time the selected lemma and semantic alternatives are activated phonologically on the next level. As with the hierarchical model not discussed here, the processing parts are divided into a conceptual level , a lemma level , a phonological level and an articulatory level :
The term "node" (en: node ) is based on the connectionist assumption that knowledge is stored in the connection between the individual ( neurophysiological ) nodes of a neural network. Hierarchical models, in turn, assume that knowledge is stored as a unit. In the cascading model, knowledge is represented by several nodes and in the hierarchical serial model as exactly one node.
If the conceptual level now triggers a node and activates it, then all lexical units on the lemma level that match this linguistic concept (“dog”, “rat”, “cat”) are activated . While the selection is still in progress on the level of the lemmas, the lemmas that have already been found activate their phonological representations on the phonological level. During the selection, the found lemmas are sorted according to their value and the lemma with the highest weight becomes the target lemma. If the lemma with the highest rating was found, its phonological representation also receives the highest weighting and is articulated. Usually only one target lemma is found, otherwise phenomena such as slip of the tongue or a language change occur.
In contrast to hierarchical serial approaches, the individual semantic and phonological sub-steps of language processing overlap .
Bidirectionality (propagation in both directions) is given according to Dell, since the phonological level in turn allows signals running in opposite directions back to the semantic level. If the target lemma is selected, the semantic level remains activated and its curve rises again because the phonological level answers and sends back feedback .
The levels discussed above include declarative knowledge of the individual concepts, lexemes, morphemes , phonemes and syllables . However, this does not explain the productivity of language. Productivity is expressed in turn on several levels: The knowledge of how words are linked to form sentences is syntactic productivity. Phonological productivity, on the other hand, arises from knowledge of the phonotactics of the first language and makes it possible to distinguish nonwords from "correct" words. (So Knirf would correspond to German phonotactics, while anyone would probably reject the word slmji .) Morphological productivity in turn allows the speaker to construct new words using the morphemes known to him.
Accordingly, an internal set of rules is required which knows how the individual language units can be linked to one another. This container for linguistic generative rules must also be linked to the respective processing levels of the cascading model.
While the conceptual, semantic and phonological levels represent declarative knowledge about concepts, words, morphemes, phonemes and syllables, the linguistic set of rules includes knowledge about the possible combinations of these units:
During the production of language, the cooperating rule systems create tag-mixed slots (spaces) for syntax, morphology and phonology (e.g. onset before syllable core and before final ), which in turn are filled in accordance with the theory of slot-and-filler .
The data stored in the internal lexicon knowledge content and the linguistic rules to connect the so-called setting rules (insertion rules) as a processing instance. This fills the gaps generated by the rule system with linguistic units. The intermediate system therefore needs to know which spaces can be filled with which units.
So Dell's model distinguishes between three different types of linguistic knowledge. In addition to the information that the mental lexicon stores, there are category-specific rules and additional rules for placement that combine the other two types of knowledge.
literature
- Jean Aitchison: Words in Your Mind . 1997.
- G. Dunbar: The Cognitive Lexicon. 1991.
- W. Marslen-Wilson (ed.): Lexical Representation and Process. 1989
- George A. Miller : Words. Forays into psycholinguistics. 1995
- Helmut Glück (Ed.): Metzler Lexicon Language. 2005
- Roelien Bastiaanse; Ron Zonnevald: Broca's aphasia, verbs and the mental lexicon. Brain and Language 90, 2004: 198-202.
- Rachelle Waksler: Cross-linguistic evidence for morphological representation in the mental lexicon. Brain and Language 68, 1999: 68-74.
- Thomas Pechmann: Language production for the generation of complex noun phrases. 1994
- Levelt, Willem : Speaking: From intention to articulation. 1989
- Gary S. Dell; PG O'Seaghdha: Mediated and convergent lexical priming in language production: A comment on Levelt et al. (1991) Psychological Review (1991) 98, 604-614
Web links
- Hypotheses on the organization of the mental lexicon in Turkish-German-English trilingual adolescents
- Saskia Kersten: The mental lexicon and vocabulary learning in elementary school. University of Hildesheim. 10th North German Linguistic Colloquium (2009): 66-88 Said Sahel & Ralf Vogel (eds.) 2010, pp. 66-88
Individual evidence
- ↑ Jean Aitchison: Words in the Mind. An introduction to the mental lexicon. Concepts of Linguistics and Literature Studies; translates Martina Wiese. Max Niemeyer Verlag imprint by de Gruyter, Berlin 1997, ISBN 3-4842-2056-2
- ↑ Wolfgang Börner; Klaus Vogel: Cognitive linguistics and foreign language acquisition: the mental lexicon. Volume 375 of Tübingen contributions to linguistics. Narr Francke Attempto, Tübingen 1994, ISBN 3-8233-5040-4 , p. 20
- ↑ Tongue tip phenomenon newly explained What happens in the brain when something is "on the tongue" pressetext - news agency, November 14, 2000
- ↑ Levelt, Schriefers, Vorberg, Meyer, Pechmann, Havinga (1991), Schriefers, Meyer, Levelt (1990) and electrophysiological data from van Turennout, Hagoort, Brown (1997, 1998).
- ^ Gary S. Dell: A spreading-activation theory of retrieval in sentence production. Psychological Review (1986) 93, 283-321
- ↑ ... there is some activation of phonological information during lemma access, and some activation of semantic information during phonological access. Gary S. Dell; PG O'Seaghdha: Stages of lexical access in language production. Cognition, (1992) 42: 287-314.
- ^ Because of the bottom-up as well as the top-down connections in the network, lexical retrieval is highly interactive. Nodes that participate primarily in later levels of representation can, nonetheless, influence decisions made in earlier levels via bottom-up feedback. (Gary S. Dell, 1986, p. 317)