In Chapters 7 and 8, Griffiths is laying some theoretical groundwork. He finds himself caught between philosophers, psychologists, biologists, etc., all of whom have a certain perspective and theory, but not all of whom know the theories and perspectives of the other disciplines. So, he needs to explain his own theoretical apparatus.
From the start, it is a good idea to keep in mind that Griffiths speaks of "categories" when he is referring to distinct kinds of things that exist external to us, like the chemical elements, animal species, etc. By "concept," Griffiths refers to the mental thing in us which refers to something else, often external to us. So there is the category "cats" and the concept "cats." The category exists whether or not there is anyone to think or talk about cats. The concept only exists if there are agents to think or talk about cats.
First, he says something that seems obvious: he is starting out from the position that a theory of anger explains the thing to which people refer when they say "anger." That seems obvious, but it is important, because he could say something like "It does not matter what people refer to when they speak of anger, because I know that emotions are x, y, and z, and that people are so confused and mistaken about the nature of emotions that there is no point in even trying to come up with a theory that explains the thing they are referring to." That might seem obnoxious, but remember that many scientific explanations are somewhat like that.
So, by the end of the book, we should expect Griffiths to have shed light on that thing which we refer to when we say "anger."
Another obvious principle: things with the same explanation belong together, while things with a different explanation should be separated. So it might just turn out that the explanation of "anger" is so different from that of "joy" that we ought not to consider them both the same thing, namely emotion. So, by the end of Griffiths' book, we may have to redefine our concepts of emotions to keep them in line with what emotions really are. And the deciding factor will be their explanations.
Philosophers of one stripe think that it is a worthwhile project to analyze how people speak of, say, fear, and that by doing so, they can arrive at an understanding of what fear is. They think that what fear is is just whatever people mean when they say "fear." There is no need to appeal to any external reality to understand what fear is.
But other philosophers, famously Kripke and Putnam, think that my concepts are about the things they are about because "to be about" something means that there is a causal relation between a word/concept and the thing it refers to. If you want to explain fear, you appeal to an external reality, and between that external reality and the concept you have about it, there is a causal pathway. The external reality causes elements of our concepts.
The causal guys have some advantages to their theories: 1) for example, if "stars" just are what people mean when they say "star," then it makes little sense to think that an ancient astronomer is talking about the same thing when he says "stars" as a modern astronomer. In other words, transtheoretical reference is difficult if not impossible if one does not appeal to external reality. But if one can appeal to external reality, then that external reality can explain how one theory can be talking about the same thing as another radically different theory. For example, a theory that says that the stars are pinpricks in a big black globe around the earth and modern astronomy can clearly be talking about the same things. 2) the causal guys can explain how concepts are revised: for example, the discovery that biological taxa are the result of historical differentiation of species led to the revision of the concept of what birds and reptiles are. Now birds are considered descendants of reptiles. The intension and extension of the term reptile changed.
So if we want to understand fear, Griffiths thinks, we need to look both at the psychological state to which fear refers (i.e. the external reality) as well as what people mean by fear. The real nature of fear is identified by whatever theory is the best at explaining the external reality. If that leads us to reevaluate how we usually talk about fear, Griffiths thinks that is a good thing.
7.2 "Natural kinds" are usually thought of as the categories into which we divide the world when we divide it at its joints. In other words, when we categorize things with the same underlying explanations together and things with different underlying explanations separately.
The concepts, however, which we form, are a different matter. For science, we try to form concepts that refer to natural kinds. We do so chiefly because such concepts are the most projectable. By "projectable" is meant that we can take our concept of, say, "cat" and apply it to things that we have never encountered before, and it works.
Griffiths spends a good deal of time attempting to explain and defend the theory of natural kinds. I think he moves too fast and is not careful enough to explain everything in here (174-175).
7.3 Theories of Concepts Next Griffiths examines how concepts form. The classical theory was that concepts are determined by necessary and sufficient conditions. For instance, the concept of "aunt" is determined by the quality of being a parent's sister. Being a parent's sister is necessary for being an "aunt," because you can't be one if you are not. It is also sufficient, because if you are a parent's sister, then you are an aunt.
Now, it seems quite difficult to identify such clear necessary and sufficient qualities for many things. For instance, in biology, there are exceptions to every suggested definition of a species that tries to give necessary and sufficient qualities. Is a two-headed calf not a calf? Are siamese twins not human? The problems with the classical view are twofold. First, that most categories have an essential vagueness to them: their boundaries are not clear cut. Second, some objects seem to be better instances of a concept than others. So there is a certain "typicality" to concepts.
In response to those problems, people have come up with theories of concepts that say that concepts are defined by something like family resemblance. All the things to which a concept applies do not have exactly the same qualities, but they all share a family resemblance to each other. Each one has a sufficient number of the qualities that the concept has.
Research has shown that people do not necessarily have a single abstract prototype for a concept. Rather, they have a number of somewhat concrete exemplars. Thus a concept might be flexibly applicable via examplars. For instance, if I am thinking of a typical drink, I have a different exemplar in mind if I am in a bar from the one I have if I am at a health food store, and yet a different one if I am dehydrated. Thus the concept "drink" might consist of several different exemplars that are context-sensitive.
What the classical theory and all variants of the family resemblance theories have in common is that they are all formal. That is, they all see the primary job to be done by a concept as the identification and application of a rule to a set of features. The family resemblance theories also rely on similarity. But it is hard to explain how we decide what constitutes similarity when we form a concept. That difficulty lessens the worth of such theories: they do not explain concept formation well enough.
The Theory View, however, says that we have an underlying theory which allows us to pick out which features of an object count in the process of deciding what it is. Thus if I am trying to decide whether something is a cat, I might not pay attention to whether it has a tail, because my underlying theory tells me that is not a feature that can determine whether what I am looking at is a cat. The theory view combines family resemblance with the idea that there has to be a theory present to allow us to find that family resemblance. In other words, we don't just happen to pick on the very few variable features that allow us to identify an individual human face: we have a theory that allows us to pick those few features and perceive resemblance. Even if we are not aware of that theory, we can still have it. In this sense, it is important to realize that the theory view does not require that the theory which is telling us which features are important be a consciously held theory. if we work at it, we can perhaps make the theory explicit.
7.4 So how does one decide how humans actually form concepts? Find some humans who are in the process of forming their concepts and examine them. Children!
Quine suggests that evolution has given us the ability to pick out features of objects that stand a better than average chance of being useful. Keil suggests that at first, we use our senses to determine similarity, but then later, we switch to looking for underlying structures that we may not be able to simply perceive. We start to use a theory to determine which of the perceptual features count.
In fact, when you examine children, they undergo a shift from looking at general characteristics to looking for definitional characteristics. They move from thinking that an uncle is older, male, related, etc. to thinking that the only thing that determines whether that is an uncle is whether it fits the definition "parent's brother."
You might think that that shift happens at a certain age, but it does not. In fact, it happens in different domains at different times in different children. But once a child starts that shift in a domain, say "family relations," the child rapidly shifts over all of their concepts in that domain. Think about how you learn a new area: you start with general characteristics, but then if you stick with it, you start to form concepts that don't care about just any similarities, but that pick out relevant similarities.
For traditional natural kinds, like chemical elements and animal species, children moved from thinking that it was possible to change a dog into a skunk by changing its appearance to thinking that a dog, no matter how much it looked like a skunk, was still a dog. Thus natural kinds seem to be thought of as characterized by an inner essence, not perceptible characteristics.
And children of all ages are resistant to thinking that you can change, say, a real dog into a toy dog. Thus people are resistant to allowing things to change category, and similarity does not matter much in this.
It has been found that all cultures have a concept of natural kinds for animals. Griffiths suggests that that does not mean there is an innate concept of natural kinds of animals. Rather, he thinks that all humans have similar enough resources in their environments that they happen to have formed a concept of natural kinds for animals.
Now, children do not have explicit theories of natural kinds or anything like that, but they have implicit ones. Once you make the implicit explicit, it is amenable to testing etc. and that is where science can enter the picture.
7.5 Griffiths now suggests that the natural kinds do have an inner essence, which he calls their causal homeostasis. A causal homeostasis is whatever it is that causes a thing to remain the thing that it is. It is also the thing that makes concepts projectable.
People look for a cluster of properties that a concept shares (i.e. a family resemblance), but they go beyond that. They also look for an underlying cause for that clustering (the theory view). That underlying cause is the causal homeostasis.
Any causal homeostasis can be the essence of a category. "Any theoretical structure that accounts for the projectability of a category" is an essence.
In chemistry, the homeostasis is atomic weight and internal microstructure. In biology, it is descent. So biological natural kinds do not have essences like chemicals do. They are not unchanging. But people nonetheless think of them as if they do.
Tools also have what might be called essences, but their essences are their function and the intention with which they were made. That is radically different from the essences of chemicals or animals. But nonetheless, tools, animals, and chemicals share the trait that they have a causal homeostasis.
Classifying things according to their causal homeostasis, according to natural kinds, is not the only way to classify, nor is it the only worthwhile way to classify. But it is the most legitimate way to classify things. It is the most legitimate way to classify things because it has the most explanatory power. If you know a thing's natural kind, you are in a better position to understand it as a whole than if you know any other classification it fits into.
Thus it is useful to have a classification "trees" for an ecologist, but even the ecologist would be better off knowing that "tree" is not a natural kind.
7.6 I'll skip this part. Sorry.
7.7 Social Construction and Causal Homeostasis Some people will object to all of this. They will say that there are concepts that are formed because of political goals or ethical goals. For instance, the concept "child abuse" has changed over the years because of the ethical goal of combatting an ever-wider group of bad behaviors.
Griffiths accepts that, and so he thinks that social construction does play a role in concept formation. Not all concepts are explicable via natural kinds. He never denied that anyway (see the tree example in the last section).
Griffiths suggests that Hacking's idea of "dynamic nominalism" is a good way to look at such concepts where the concept changes, and the thing it refers to changes as well. Multiple Personality Syndrome (MPS) was diagnosed at a certain time point, and after that, more and more people began to have MPS. The concept spurred on the category, which in turn led to revision of the concept, etc. Thus for a given emotion, the existence of a concept of that emotion may spur people on to have that emotion. Thus the societal beliefs about the emotion will accurately describe the phenotype of the emotion in that society.
So a concept can influence a category when the concept is socially constructed. But it should be born in mind that not all categories can be socially constructed in that way. For example, no matter how we socially construct the concept of a species of "paramecium," that species will not be affected.
The goals of politics and ethics, etc. differ from those of science in that they are not aimed primarily at knowledge. Science, however, is aimed primarily at knowledge (Griffiths would be more careful: he would say "epistemic projects" rather than knowledge).
For scientific projects, identifying the causal homeostatic mechanism underlying a category is the most legitimate way to form a concept. There may be other classifications that are useful for various other projects, and Griffiths does not want to deny that.
Emotions are problematic in that they are both the object of epistemic projects and also the object of projects like educating the young, which are more ethical in nature. They are used to form the young, and as such they are substantially socially constructed. Griffiths will try to stick to the scientific project side of emotions and avoid the socially constructed side.