The tricky part about this is to build a part that won't interfere with the rest of the person's outcomes. If there really is a part that stops her from doing something, and you build a part to do it, guess what's going to happen? WAR. To prevent this, we have built into the model that all the parts of the person that don't want you to build the new part become allies during the design process.

The first thing you do is identify whatever «need» it is that you are going to build the part for. For example, a woman might come in and say «Well, you know, I've been on lots and lots of diets and I never seem to lose weight. I'm just much too heavy, so I want you to put me in a trance and make food taste bad.» If she really wants that, I would recommend that you send her to one of those Schick clinics, where they will put big cakes in front of her and shock her. If she smokes, they will put her in a room full of cigarette butts and make her drink ashes, and all kinds of wonderful things.

That's a way of building a part that stops you from doing certain things. However, it doesn't take into account the secondary gain—the outcome of the problem behavior. That makes it a very difficult way to stop behaviors. It is an experiential way of going about it, and it will Work insofar as it's reinforced. Sometimes after a period of time, when the part that you have developed discovers that you're not going to get shocked any more, then it won't care if you smoke. So you might have to go back at a later time and repeat the procedure or do something else. That's a problem with building parts in that particular way. However, don't underestimate that approach, because it works. It seems a little severe and it doesn't work with everybody, but it does work; that's an important consideration. It's important to understand what goes on when people change, and to make up a metaphor or a lie to describe it that enables us to be able to make changes more elegantly.

Let's go back to our overweight client. Her expressed need is to «lose weight.» However, if you build a part whose job it is to lose weight, what's going to happen when she loses weight? She will lose some more! She may become an anorexic! So if you opened up a weight clinic and built parts to lose weight, you would end up needing another clinic down the street for anorexics. There you could build eating parts, and you could have the client switch back and forth every six months. There's nothing in her stated outcome that has anything to do with stabilizing weight.

Most people really don't understand substituting symptoms. There's one school of thought that says «Well, if you use hypnosis, then you will get symptom substitution.» My response is «Bravo! Let's deliberately substitute something and have it be something useful.»

Years ago a man wrote an article in which he described making cigarettes taste like the worst thing he could think of—cod liver oil. The client he did this with quit smoking, but he became a cod liver oil junkie! He carried a bottle of cod liver oil in his coat all day. I guess that's better than smoking. I don't know the ramifications of overdosing on fish oil. It sounds disgusting to me. I prefer to substitute symptoms that are positive.

So the really important question is «What is it that you are going to do in terms of an outcome?» If somebody comes in and says «I want to quit smoking» and you make the outcome no cigarettes, then the way you organize that person's resources to suppress that activity can have lots and lots of other outcomes that are not positive.

The question is «How can you conceptualize change work so that you avoid undesirable side effects?» When somebody comes in with a weight problem, what part are you going to build? In other words, what is going to be the outcome of the part that you build? At the moment, her need is to lose weight. But how can you do that and not have her end up an anorexic?

Ann: You could set a specified weight that she wants to weigh, and not let that part function when she gets under that weight.

Well, yes. We can put semantic conditions on when the part is to be active and when it's not. You could have the part begin to respond every time she weighs more than a certain amount. However, parts don't like to be inactive.

Man: You could get all parts to agree on the same outcome.

Try it some time! I'm serious. If somebody comes in and wants to lose weight, you try to get the part that likes candy to agree to that. His parts may all say «Well, that's a groovy outcome.» But if you get all his parts to agree that it's a great outcome, it still won't take him there. What the parts object to is the process of getting there.

Man: Could they generate alternatives?

You can have them do that, but then you're using a different reframing model. Then you're saying that the problem is a result of the interaction of the parts you have now. You could use the six–step model to do that. However, it's not very elegant, because then you have to go in and deal with a huge number of parts. The question is simply one of expedience: if you were only going to build one part, what would it do? I want you to make a distinction between the outcome—what you want to be sure happens—and the behaviors or procedures that the part uses to get the outcome. They are both important, but now I want you to specify outcomes.

Man: You need to make it more versatile, so that it can do more than one thing.

OK, but what is it going to do? What is its job?

I don't want to get you off the track. If we build a part whose function is to have somebody weigh a hundred and five pounds, that will work. That's great. That's a well–formed outcome. Now I want to ask «What are other outcomes that will work equally well?» There are lots of right answers to this question. The important thing is that you learn how to conceptualize them. Ann was on the right track. She said «These are what we don't want; this is what we do want. This is one way of getting only what we want.» The key question you have to ask yourself is «Will this give us ONLY what we want?»

Man: You could put the part in charge of «health» or «attractiveness» or some superordinate structure that includes weight.

Woman: How about a «central eating control» that takes all those factors into account in the process of deciding how much she should weigh?

Woman: I think you have to take all those needs which that part intends to gratify, and satisfy those needs in different ways.

Well, that's all true. The question I'm asking is «What is the part going to do?» If we have a part whose job is to be in charge of overall health, and we include maintaining a certain weight in that, then will we do only that? The magic word is «only.» Sometimes it may be advisable to do a lot more than the client requested. But right now let's talk about limited therapeutic change. Your answer is an accurate answer, and may be a better answer in experience, when you are treating clients. However, each job that apart has makes it harder to install that part. I want you to keep that in mind. Every extra outcome that a part has makes it more complicated for that part to function. The more limited its outcome, the easier it is to install apart. Sometimes it is better to make the extra effort to install a more complicated part in order to get a better result in the end. A part that keeps someone at a certain weight is going to be a lot easier to install than the kind of central coordinator that you're talking about, because the coordinator will have to have a lot of knowledge about what it means to be healthy, etc. It will also have a lot more behaviors, so you're more apt to get objections from other parts.

Man: How about installing a semantic cue for eating, and a motivation strategy to get her to eat only then?