In Chapter 3 we reviewed associative network theories of emotion. We began with a discussion of Bower's theory (e.g., Bower, 1981; Bower & Forgas, 2000) in which emotions are represented by single nodes in a localised network. Having suggested numerous empirical and theoretical problems with this type of model (see Chapter 3 for a fuller discussion) we speculated on the advantages of a distributed network or parallel distributed processing architecture for the modelling of emotions (although see Chapter 3 for a defence of localist connectionism). We concluded that the most likely role for such an architecture was in the representation of low-level or associative processes in the generation of emotion that does not require concurrent access to the schematic model level of representation.
Before we discuss how emotions might be generated associatively in terms of our proposed model, it is important to illustrate the kind of phenomenon we are referring to. Consider the following two examples:
Peter had experienced a difficult childhood. His father had conducted a reign of terror over him and his siblings. His father was always shouting at him even when he had done nothing wrong. Peter's memories of his childhood are shrouded by the sense of fear he always felt with his father and, as an adult, Peter harbours much anger and resentment towards the way his father treated him as a child and indeed continues to treat him. He is also very intolerant of anybody shouting at him. The very act of someone shouting immediately makes him angry. Peter and his father remain in contact, though their relationship is somewhat fraught. Often Peter's father will shout at him over something inconsequential, treating him as a child again. When this occurs Peter immediately experiences two very strong emotions at the sound of his father's raised voice: the fear that permeated so much of his childhood; and intense anger that somebody is shouting at him. Peter also feels angry at the content of his father's exclamations and the fact that he is not being given the credence he deserves as an adult.
For as long as she can remember Julie has been frightened of birds. If you get her on to the subject she will say, in a somewhat embarrassed way, that she can see that there is nothing about birds to be afraid of and, indeed, she genuinely believes that birds are not really dangerous. However, as soon as she has to go anywhere near a bird feelings of intense fear sweep over her, which are completely outside her control and she has to leave the situation as quickly as possible.
We suggest that these somewhat different examples illustrate two important points about the generation of emotions. First, that certain events lead to an associative generation of emotions in contrast to the generation via schematic-level meaning we described above. So, the event of Peter's father shouting at him in Example One leads to the associative generation of fear and also of anger. In Example Two, the presence of birds for Julie leads to the associative generation of fear.
The second point is that associatively generated emotions can conflict with either: (i) emotions generated via the schematic model route; (ii) other associatively generated emotions; or (iii) non-emotional rationalisations of an event. So, in the first example above, Peter experiences a conflict between his experience of the associative generation of fear at the sound of his father shouting and his experience of the schematic generation of anger at the content and style of his father's exclamations. In addition, there is some conflict between the experience of the associative generation of fear to the fact that it is father who is shouting, and the associative generation of anger to the fact that Peter is in the situation of being shouted at (see Figure 5.5).
In the second example Julie experiences the associative generation of fear to the presence of the bird with no rational understanding of what it is she is afraid of. Again, there is a conflict caused by the associative generation of an emotion.
In sum, the proposal is that there is a second, associative route to emotion generation which differs from the generation of emotions via the schematic model level of meaning that we described earlier (Power & Dalgleish, 1999). How this associative generation of emotion occurs in terms of the proposed model is discussed next.
In addition to the representation architecture we have discussed so far (namely, schematic models, prepositional, and analogical representations) and which is outlined in Figure 5.4, we suggest a further associative level of representation that is responsible for the second route to emotion generation (see Figure 5.6). Figure 5.6 represents the central components in the model of emotion we are proposing. From this point onwards we shall refer to this architecture as the SPAARS approach (Schematic, Propositional, Analogical and Associative Representation Systems) to reflect the four key types of representation that are utilised.
Peter being SHOUTED AT by his
who treats him
LIKE A CHILD
Figure 5.5 Some of the types of emotion shouted at by his father. = conflict.)
Automatic generation of - anger at being shouted at.
< Automatic generation of fear at being shouted at by father.
_ Schematic generations of anger at being treated like a child, conflict experienced by Peter on being
We propose that within SPAARS emotions can be generated directly, via an associative level, without concurrent access to the schematic model level of meaning. To understand what we mean by this statement in more detail, it is useful to return once again to the components of the emotional experience that we outlined following the philosophical analysis in Chapter 2. These were: an event, an interpretation, an appraisal, physiological change, an action potential, and conscious awareness.
In our discussion of emotion generation via schematic models we argued that the process of appraisal consists of an analysis of the interpretation of the event in terms of the individual's extant goals at the schematic model level. The point we wish to make here is that within SPAARS, emotions can be generated without this appraisal process occurring at the time of the event's occurrence. The appraisal only needs to have occurred at some time in the emotional history of the individual's experience of that event or, for a small circumscribed set of events, in the evolutionary history of the species. That is to say, the process of emotion generation can become associatively driven so that it appears as if a concurrent process of appraisal is occurring even though it is not and has in fact occurred at some time in the emotional past. In other words, the accessing of the schematic model level of meaning is "short-circuited".
Let us illustrate further by returning to the example of Peter and his father. During Peter's childhood, his father's tendency to shout at him about inconsequential things would initially have been appraised at the schematic model level and Peter would have felt afraid. Peter never knew how his father was going to react next. For example, an evening with friends could easily be ruined by a tirade from his father, resulting in his friends being asked to leave. After many repetitions of similar events the generation of fear via schematic-level meaning gradually became associatively driven. The emotion of fear was then generated as if a concurrent appraisal was occurring (that is, as if schematic-level meaning was being considered) even though it was being short-circuited.
As we have already mentioned, we propose that this second associative route to the generation of emotions occurs via a separate type of representation. In proposing such an additional representation system, two important questions need to be addressed: first, why is it not possible for part of the schematic model level to be the vehicle for associatively generated emotions? Second, if such a process does occur outside the schematic model level, why is there a need to posit a separate representational system? Would it not be possible to have direct links between events and emotional output?
In terms of the mechanisms underlying associative representations, we endorse Logan's (1988) suggestion that certain automatic processes represent single-step, direct access to past solutions in memory. These solutions, we submit, are not mere input-output connections but are representations of interrelationships between events and, in this case, emotion products (see Moors & DeHouwer, 2005, for a discussion). In a straightforward input-output model the idea of an emotion is lost. How would phobic individuals, for example, "know" that they were afraid, as opposed to angry, or even simply aroused, if their conscious experience derived merely from an awareness of the phobic object and any concurrent physiological changes that the object activated? Such an argument might be tenable if it was agreed that emotions could be distinguished reliably on the basis of physiology alone. However, it is not clear that this is the case and we have argued in Chapters 2 and 3 that the physiological systems recruited by the different emotions overlap considerably. An associatively generated emotion obtains its distinctive "feel", we propose, because it is an awareness of both a physiological change and the content of an associative representation of the interrelationships of the event with the emotion products at hand.
Granted that associatively generated emotions require more than simple input-output connections, it is incumbent on us to justify why a separate representational system is necessary. It would be functionally equivalent to locate associatively generated emotions in the schematic model level by proposing a "direct" route through that system, as it is to propose a separate associative level of representation. Indeed, the former instantiation is the architectural solution in interacting cognitive subsystems (ICS; Teasdale, 1999; Teasdale & Barnard, 1993; see Chapter 4).
The principal argument for describing certain instances of emotions as a function of a separate associative system is the clear dissociation between them and schematically generated emotions, especially in the clinical domain. Again, phobias are the clearest example; the phobic individual will experience fear or disgust at the phobic object while being able to recognise little or no rationale for such emotions. Furthermore, the fear reaction is invariable in the presence of the phobic object and cannot be modified by cognitive techniques; the phobic individual can talk ad nauseam about how there is no reason to be fearful, but be afraid nonetheless in the presence of the phobic stimulus.
As a final note it is useful to emphasise again the functional equivalence of the choices of placing associatively generated emotions either in the same or in a different system to emotions generated via schematic models. In the final analysis, we are endeavouring to provide a theoretical framework that has heuristic value in understanding emotion phenomena. It is our contention that there are clearly two routes to the generation of emotion; that this helps us to understand many aspects of the psychology of emotion; and, finally, that this understanding is facilitated by proposing separate representational systems to mediate the two routes. Clearly therefore, the dynamics of emotional life result from a potentially complex combination of schematically driven and associative emotional experience.
There are two further points we would like to make about the associative generation of emotions within SPAARS before summarising the ideas we have put forward so far. These are, first, to clarify the role of the propositional level of meaning and, second, to outline the conditions under which the associative system is recruited for emotion generation.
Was this article helpful?