 Welcome to my e-lecture, Theories of Word Meaning, which builds upon the two e-lectures sense and reference. In this overview I will discuss several frameworks for the description of the relationships between words and concepts, among them general approaches such as the word field or semantic field approach, formal approaches such as componential analysis or the use of meaning postulates and more or less cognitive approaches such as semantic networks and frame semantics. These approaches seek to solve two issues, the description of the relationships between words and concepts on the one hand and the formalization of sense relations on the other. Let us start with a general approach that seeks to capture the similarities and contrasts in an economical way. In semantic fields, word meaning is defined by relating and contrasting members of a set or field. Such areas of meaning are called word fields or semantic fields. The idea of establishing semantic fields goes back to historical studies of meaning change and changes to such fields through time. This example over here is from German where around 1200, this height, wisdom, was the head term of the field of intellectual activity. At around 1300 however it became a member of a slightly modified field. Essentially semantic fields are based on paradigmatic relations where linguistic units are contrasted or related in a particular environment. For example over here we have bees that have six legs, ants and the wasp has six legs, the bed bug has six legs and of course the fly that has also got six legs. However it has been argued that syntagmatic relations that is particular horizontal associations like in our case the item honey that is associated with bee or can fly associated with some of these items have to be incorporated as well. The items in semantic fields have related senses but they are not necessarily synonymous. In fact the meaning of a lexeme defines itself in terms of contrasts in which it stands with other lexemes in the field. Part of our understanding of fly is not ant, not bee, not termite or not wasp. The relationships among the items in a semantic field can cover a whole range of sense relations. Changes in the meaning of one lexeme may affect the meanings of other lexemes to which it is related through the network of semantic relationships within the field. Thus semantic fields are not static constructs but constantly expanding and contracting. Today semantic fields are used for example in special word books such as Roger's Thesaurus. Furthermore they have applications in the contrastive analysis of languages. Let us now turn to a more formal way of representing word meaning, the Componential Analysis approach. Like the distinctive feature approach in phonology, Componential Analysis attempts to define word meanings as complexes composed from a finite set of components which are called semantic primitives. Let us illustrate this using two concepts, girl and boy. To distinguish them we need one feature, a feature male alternatively you could have used female. So boy would be plus male and girl is minus male. These primitives like male or human are believed to be binary and universal. That is they should exist across cultures and they should hold across languages. The number of features we need depends on the number of concepts we want to describe. If we add woman and man we need another primitive for example adult where now a girl is minus adult and a man is plus adult. And if we add bachelor we need a further feature for example one that describes the marital status. As you may have realized by convention English is used as a meta language to define such primitives. Superfluous components and specifications are avoided by including a set of redundancy rules. Now redundancy rules help to avoid superfluous or redundant features by predicting automatic relationships between components. So every time we use the feature human for example we do not have to also include animate because the first of our redundancy rules predicts if something is human then it is also animate. The analysis of lexemes into meaning components not only allows us to formally define the conceptual sense of words but also provides an economical means of representing sense relations like hyponomy and incompatibility. For example a lexical item can be defined as a hyponym of another one if its feature specification contains all the positive features of the other. Thus bachelor is a hyponym of man because its feature specification contains all the positive features of man and has one additional feature. Incompatibility holds between lexical items in a set if they share a set of components but differ from each other by one or more contrasting features. Thus boy and girl are incompatible they differ in one feature that is the feature male. These basic ideas of component analysis have been extended in the work of various linguists and will be discussed in a separate e-lecture. An alternative to component analysis specifies word meaning in terms of so-called meaning postulates expressions that can be seen as explicit generalizations of relations of hyponomy. Here is an example where you see that meaning postulates define sense relations by means of logical formula. The fact that all flies are insects is shown by this formula where the upside down A stands for the universal quantifier in logic meaning for all something holds. So this formula can be read for all items X it holds if X is a fly then X is also an insect. The interesting thing about meaning postulates is that other lexical relations can also be represented using postulates. For example non-gradable autonomy dead versus alive and here we only need the symbol that makes a statement negative. It can be read as follows for all X it holds if X is dead then X is not alive or takes synonymy. If you want to show that couch and sofa are synonyms well then you need these two formulae and they can be paraphrased as if X is a couch then X is a sofa. So for all elements X it holds if X is a couch then X is a sofa and for all elements X it holds if X is a sofa then it is also a couch. Well and if you need a shorthand form if you want a shorthand form you can use the symbol for logical equivalence which means all elements couch are equivalent with all elements sofa. The general idea of this more or less formal treatment of meaning is that word meaning can be stated in terms of entailment relations a fundamental relation in propositional logic. Compared with componential analysis meaning postulates are a weaker account of lexical meaning. They can be formulated without making assumptions about semantic atoms or universality and they can be used to provide deliberately incomplete specifications of lexical items. Since fundamental relations such as if something is an insect it is also an animal are stated elsewhere. This makes them applicable in large areas of vocabulary where an exhaustive componential decomposition of each lexical item would not be feasible. Let us now continue with some popular cognitive approaches towards the formalization of word meaning. One of them is the so-called semantic network. Semantic networks have a relatively long tradition. They became known through work in biology and psychology and were originally thought of as primarily representing nominal concepts. In such simple semantic networks concepts are represented by nodes here by means of boxes by the way the L letter L stands for lexeme. And the interrelationships between the nodes or concepts by means of so-called arcs these arrows over here. The nodes in the network stand for concepts that is entities events and states of the domain. Arcs represent the interrelations between concept nodes. There can be various such relationships for example there is the is a taxonomy or the is a relationship which means for example a bee is an insect. Or arcs can express the has or has proper relation in which case here in our example we could express the relationship that a bee has wings. Now wings three means a particular types of wing a fully arbitrary number here. And then we have the instance of relationship where we can express the fact that wings these particular wings of the type three are instances of a more general notion of wings. In a complex semantic network these relations are either indicated in the network or they follow from the underlying theory. Now here is our bee again with its particular relations bee it is an insect so here we have the is a relation and it has wings. Now here are our particular wings as I said fully arbitrary wings of the type three and these wings are instances of these more general wings which are in turn means of flight. One reason for the relative popularity of semantic networks is the elegance with which deductions can be made. In order to deduce that a bee has a head now a bee is defined as an animal that has wings and that has a sting. Now in order to deduce that a bee has a head all that is necessary is to trace up the hierarchy assuming that attributes associated with higher nodes such as insect are also valid for lower nodes. So the attributes associated with insect as well defined head has three pairs of legs are now moved downwards to bee or in other words bee inherits the attributes of insect as well as insect inherits the attributes of animal. So a bee then has skin has a head has well defined head has three pairs of legs has wings and has a sting all these are attributes of the concept of bee by means of feature inheritance. Despite their elegance semantic networks are problematic in many ways the main problem concerns their size even if just a fraction of the set of human concepts is presented. A gigantic database will be the result a more economic way of organizing human concepts uses frames. Frames are conceptual structures that provide a background of beliefs experiences or practices against which the meaning of words can be understood. They organize stereotype knowledge about an entity about an event or about a situation in a container structure consisting of attribute value pairs. Now here is a bee frame with three attribute value pairs where the is a attribute triggers a superordinate frame the insect frame and has access to his attributes too. Now the attributes can also have default values such as particular colors sizes or actions these default values are kept until there is contradictory evidence. Each leg seem is assumed to evoke a particular frame which is networked with all sorts of additional frames. Thus frames also involve the notion of inheritance in that subframes can inherit elements and attributes from their parent frames. So bee once more inherits from its parent frame insect the fact that it has six legs. An extended form of the frame is referred to as script. Scripts encode stereotype event sequences representing activities like buying a car or opening a bank account or eat in a restaurant. Now the script that described such an activity is essentially an event frame that specifies a number of defaults. A goal for example the goal could be that you want to eat something. The objects that are part of the script the so-called properties, tables, chairs, menu, money and the person involved the roles and then some preconditions as well as expected results. Like in a script of a movie each script consists of a number of scenes. In our restaurant script these scenes are entering, ordering, eating and paying in this particular sequence. And if you wish this could even be more specific and involve additional scenes such as leaving the restaurant. Each scene is specific concerning the actions within it. For example the transfer of money from the customer to the waiter in the paying scene. Thus the complexity of frames can vary from representations of small static scenes or states of affairs to complete scenarios with a dynamic time-based structure. Well this e-lecture could not discuss all details about all theories of word meaning. For example we didn't mention approaches such as the reductive paraphrase approach or did not discuss the details of componential analysis including its advantages and disadvantages. These aspects can be found in the respective VLC e-learning units where additional literature links provide you with further details. However as an overview this e-lecture may suffice and I hope that I managed to have shown not only the problems of coming to grips with word meaning but also the varieties of approaches that are available. See you again.