 Linguistic encoding processes are by and large mediated by the entries in the mental lexicon, the human variant of a word store. This crucial role of the lexicon in speech generation necessitates the discussion of the internal structure of its entries. Well, and exactly these topics determine our program. We will first of all look at word stores in general and will provide some sort of taxonomy of word stores. Then we will look at the structure of lexical entries. And since a number of psycho linguists suggest that lexical entries can be separated into several parts, we have to discuss the problem of partitioning lexical entries. So let's start with the general taxonomy of word stores. Well, language is a communication system employing arbitrary symbols. These symbols which are normally words have to be stored somewhere. And the following different types of stores of word stores in general exist. Now, on the one hand we have dictionaries. Dictionaries are word stores that are primarily consulted to retrieve information about the words of a language. Today dictionaries can be found in two variants. On the one hand we have book dictionaries over here. On the other hand we have machine readable dictionaries which can either be delivered on CD-ROMs or they're contained on the web. Now the most common types of book dictionary are encyclopedic monolingual dictionaries, these bilingual dictionaries or dictionaries for special purposes such as dictionaries or synonyms for example. Machine readable dictionaries are similar however they are delivered on specific storage devices as I said such as CD-ROMs or the World Wide Web. A lexicon by contrast is the word store which is the central component of a natural language processing system. And this can be a human processing system, humans themselves, or a processing system which is part of a computer program. For example, a machine translation program. So that is this component here which is the machine again so let's use the color green. The lexicon closely interacts with the other components of grammar and it provides detailed information about the words to be produced or comprehended from phonology to conceptual. Let's list an example here, let's take the lexim C so that is the lexical entry that's why we're using capital letters here. Phonologically we know that C consists of these phonemes in present-day English. In terms of well also phonology we know that its syllable structure is just one syllable consisting of an onset, a peak and no coda, CV. We know that it is a verb so that is the morphological information word class. And syntactically we know that this verb has to be followed by a direct object in terms of a noun phrase which can itself function as a sentence. I see that he is going home. Well and conceptually we know that for example if we take Schunk's conceptual classes, his primary acts we know that it is a verb that denotes mental transfer. In language processing the central word store is the mental lexicon rather than any variant of a dictionary. The reason is quite simple. Whereas dictionaries are static and do not allow changes to their contents, the mental lexicon permits its structure to be changed. That is the addition, the extraction, the rearrangement or modification of entries, processes that are constantly at work. We know that some machine readable dictionaries allow that these days especially in machine translation programs in vocabulary, training programs and so on. Let's now look at the lexical entries themselves. All lexical entries in the mental lexicon must be subdivided into two components. First of all we have an access unit that is the item with which the mental lexicon is contacted. So the access unit then contacts the entries within the mental lexicon and the lexical specification that's the second part is the information that is associated with each lexical entry. So the information like phonology, morphology, syntax and conceptual information about each item of a language. Let's look at lexical access units first. Now here is the mental lexicon. An important issue in lexical processing concerns the nature of representation with which the mental lexicon is activated. For example in the process of linguistic encoding, language production, the lexicon is normally accessed by this sort of representation. So this is language production where we access the lexicon by a set of conceptual properties associated with the semantic primitives in the conceptual representation. So we could either say well what is an insect that has two wings that has six legs and no sting and then immediately we would activate the insect fly. Alternatively we could show a picture which is represented over here. Now if we think in terms of speech perception then of course the access unit is something completely different. Namely the acoustic representation, a bundle of acoustic properties or a set of geometrical shapes for example. So we could access the lexicon by simply writing down the word fly and so here we have by contrast of course language perception. That is the process where eventually which eventually will lead to a comprehension of what someone says to us. Let's now look at the lexical specification. When we produce language we retrieve items from the lexicon on the basis of a pre-generated conceptual structure. And this conceptual structure has to be filled with actual words. Many psycholinguists argue that the structure of lexical entries can be differentiated into two parts. Into a formal representation, so that's the first part, the form. And this formal representation includes phonological and morphological information. So for example let me write it down again, C and the syllable structure. And also of course the fact that C is a verb. And then we have a second level, the level of context related aspects, that is the lemma. So this is this part here, the lemma. And the lemma is associated with aspects like syntax, the conceptual representation and semantic information. So aspects like well let's say in this case we have the fact that C is a verb which requires an object. So typical sub-categorization frame or a conceptual representation in terms of a primitive. Which requires some arguments but that is not represented here. Form related aspects and lemma are linked by a lexical pointer. So there's this linkage here, well this lexical pointer. But is this sort of separation of a lexical entry into formal lemma really legitimate? Well that's a question that has been discussed in psycholinguistics for some time. Well even though this partitioning of a lexical entry, this sort of borderline here, even though this partitioning is still very much an open issue, a number of arguments can support this subdivision. For example language production phenomena. They particularly provide evidence that the retrieval of formal aspects from the lexicon can be independent of lemma related aspects. One piece of evidence for example comes from a very well-known phenomenon, the phenomenon of where you have something on the tip of your tongue. Well this is what happens. You know the, and then you're thinking of the word, you know the, and you have everything in your mind. You can see the object in front of your eyes, you know the concept already. You even know the syntax because you're saying something, you know the, the, and you know you must express a noun phrase. So you have it on the tip of your tongue. In other words you have accessed the syntactic and conceptual components of a lexical entry without having access to its phonetic components without having access or without being able to pronounce it. This tip of the tongue phenomenon supports the subdivision of lexical entries into form and lemma. And then of course we have speech errors. Some speech errors are closely lemma related, such some of them are form related. Just count the number of speech errors are produced within the last 15 minutes and then you know how legitimate the analysis of speech errors is to support this subdivision. Well this way of partitioning the lexical specification of an entry is not undisputed. There is evidence from the study of language disorders, for example, that some patients maintain phonological and syntactic knowledge about lexical entries in the face of a severe breakdown of semantic knowledge. So in other words they have then access to this part of the information and to this part so they're crossing the partitioning boundary. Maybe this shows a different way of partitioning and it suggests a different way of subdividing the lexical specification of an entry in our mental lexicon.