Network theories follow in a long tradition that can be traced back to Aristotle and which includes the British empiricist or associationist school of philosophy. In psychology, associationism underpins not only behaviourism and the basis for the laws of learning, but also psychoanalysis and Freud's development of the free association technique. Indeed, one of the most detailed and elegant network models of autobiographical memory was presented by Breuer and Freud (1895). Because it bears close resemblance to many current network models, we will begin with an outline of this model.
A modified version of Breuer and Freud's (1895) model of autobiographical memory is presented in Figure 3.6. Freud argued that one or more early traumatic events, thoughts, or wishes can form a "pathogenic nucleus" around which later associated memories become attached according to a number of rules. First, there is a linear chronological order, with the oldest memories running in sequence to the newer memories. However, there may be multiple pathways or themes that can be traced back to the original nuclear memory. Second, the themes are stratified around the nucleus such that the oldest strata are resisted: "near the nucleus we come upon memories which the patient disavows even in reproducing them" (Breuer & Freud, 1895/1974, p. 375). Third, logical chains run through the material to the nucleus; these logical linkages meet at memory nodes and may connect across themes as well as within them. Activation of one of the memory nodes leads to a spread of energy through the linkages to the nucleus and one of the ways in which this energy can be discharged is through the subsequent night's dreams (Freud, 1900/1976). Although Freud's focus was on autobiographical memory, it will be apparent from our review of more recent network models that his account is as elegant as any of the later accounts.
The main impetus for the use of networks within cognitive science came from attempts to model the relationship between linguistic concepts. One of the most influential of these was Quillian's trail-blazing PhD thesis, which was published in 1968. Whereas earlier associative networks had simply assumed that the connections between words in a network consisted of undifferentiated links of varying strength, Quillian proposed that there could be different types of links between the nodes in a network and that these links expressed key features of the semantic relationship between concepts. These labelled links enable the network to capture so-called "inten-sional" relationships between words in a way that unlabelled links do not. For example, the word "gooseberry" is associated with "fruit", with "fool", and with "play" (as in "play gooseberry"), but the type of semantic or intensional relation is very
THEME 1 THEME 2 THEME 3
THEME 1 THEME 2 THEME 3
PATHOGENIC high resistance
PATHOGENIC high resistance
Figure 3.6 An adaption of Breuer and Freud's (1895) model of autobiographical memory.
different in each case; a network that contained only unlabelled links between concepts could not express these different intensional relationships.
The semantic network that Quillian (1968; Collins & Quillian, 1969) presented dealt with the hierarchical arrangement of groups of nouns such as:
CANARY ^ BIRD ^ ANIMAL
In this network, the nodes represent concepts and the links represent the relation "ISA"; that is, a canary is a bird, and a bird is an animal. The empirical tests of the hierarchically organised nouns initially provided good support for the proposals (Collins & Quillian, 1969). However, it became evident that not all findings were supportive—for example, sentences such as "A horse is an animal" were verified faster than "A horse is a mammal" even though the former involved two links in the hierarchy and the latter only one. The problem expressed in this example is that a horse is not a prototypical mammal; amended versions of the network needed to be produced to take account of prototypicality (e.g., Collins & Loftus, 1975). Even more problematic was the fact that false statements were found to be rejected faster the farther apart two concepts were, rather than the closer together they were, and the fact that not all word categories can be organised hierarchically anyway (e.g., Johnson-Laird, Herrmann, & Chaffin, 1984). Again, it is possible to produce modifications of networks that can take account of these problems and some of these alternatives will be presented next. However, we hope that this short introduction to some of the features of semantic networks will aid the description of some of the specific network models that have been applied to emotion.
Was this article helpful?
This guide Don't Panic has tips and additional information on what you should do when you are experiencing an anxiety or panic attack. With so much going on in the world today with taking care of your family, working full time, dealing with office politics and other things, you could experience a serious meltdown. All of these things could at one point cause you to stress out and snap.