5.1 Post summary / Table of contents

Part of the Intuitive Self-Models series.

Dissociative Identity Disorder (DID) (previously known as “Multiple Personality Disorder”) involves a person having multiple “alters” (alternate identities), with different preferences and (in some cases) different names. A DID diagnosis also requires some nonzero amount of “inter-identity amnesia”, where an alter cannot recall events that occurred when a different alter was active. For example, DSM-V talks about patients “coming to” on a beach with no recollection of how they got there.

Anyway, just like trance in the previous post, DID was one of those things that I unthinkingly assumed was vaguely fictional for most of my life. Maybe that’s because my only exposure to it was in actual fiction. While the artistic merit of these fictional DID depictions spanned a broad range—from the highs of Fight Club to the oh-so-very-low lows of Me, Myself & Irene—the psychological merit was uniformly crap.

And again like trance in the previous post, my take now is that DID does in fact exist. Even the memory lapses are real, although calling them “amnesia” is an oversimplification of something more nuanced. These people aren’t just trolling us. And in light of this series (especially Post 3), my explanation for DID is the obvious one: each different “alter” (“alternate identity”) in DID corresponds to a different “homunculus” concept in the intuitive self-model.

This explanation has a lot of overlap with trance. In fact, according to DSM-V, in cultures that have spirit-possession trance rituals, DID tends to manifest as that same kind of spirit possession. What makes DID different is that the trance states are recurrent and involuntary, rather than only happening in the appropriate ceremonial context.

Seems simple, but fleshing it out will involve a lot of subtle details!

Section 5.2 talks about how the suggestion of culture and psychologists is a risk factor for DID. I argue that this is naturally expected given the discussion of probabilistic inference in §1.2, and does not imply that the condition isn’t “real”.
Section 5.3 offers an explanation of why DID strongly correlates with Borderline Personality Disorder (BPD): we all have the everyday distinction between angry-me, happy-me, sad-me, etc. These are not normally DID—they all involve the same homunculus concept, but with modifications / patches for the different emotions. However, if the emotions are very strong, as they are in BPD, the conventional self-model (one homunculus concept “patched” in different ways) becomes progressively less plausible, while the DID self-model (multiple entirely different concepts) becomes progressively more plausible.
Section 5.4 explains the relation between DID and (self-reported) childhood trauma, in a similar way.
Section 5.5 moves on to “inter-identity amnesia”, the eye-catching claim that one alter doesn’t remember things that the other alters have done. Calling it “amnesia” is an oversimplification, but I do think there’s a real phenomenon here, and that it’s closely related to trance amnesia (e.g. posthypnotic amnesia), as mentioned in the last post. I review the evidence and how I interpret it, proposing a model somewhat related to everyday phenomena like how angry memories come back more readily if you’re angry right now, or how singing a tune makes it hard to recall other tunes.

I’m not remotely an expert on DID and am happy for feedback.

5.2 Why is the suggestion of culture and psychologists a risk factor for DID?

I think it’s conventional wisdom (see e.g. here, here) that cultural expectations, and being egged on by a psychologist, are risk factors for DID, and in particular were responsible for a spike in DID cases in the USA in the 1980s-1990s.^[1]

One piece of evidence that culture plays a big role is that the manifestation of DID seems to follow cultural templates. For example, “in India [unlike the USA], the transition period as the individual shifts between alter personalities is typically preceded by sleep, a presentation that reflects common media portrayals of DID in that country”; and here in the USA “the number of alters per DID patient increased dramatically from the 1970s to the 1990s”. (Lynn 2012).

Just to be clear, if the expectations of culture and authority figures are risk factors for DID, that does not mean the DID patients are knowingly lying to fit in, or to appease their therapist. By analogy, my intuitive model of lightning says that lightning bolts are big giant static electricity sparks, whereas Julius Caesar’s intuitive model of lightning says that lightning bolts are a weapon wielded by the god Jupiter. Obviously cultural suggestions have a lot to do with this difference between me and Caesar! But at the same time, each of those is a “real” experience, not “fake”, in the §1.3.1 sense. Neither Caesar nor I acquired those beliefs out of a desire to fit in with cultural expectations, and we are both sincerely reporting how things seem to us.

Why do cultural expectations matter for DID? Because it’s an intuitive self-model. Intuitive self-models, like all intuitive models, come out of a probabilistic inference process (§1.2.2). When there are multiple possible models that issue good predictions (as in the bistable perception example of §1.2.1), then suggestions from culture or trusted authorities can do a lot to influence what happens, by helping make certain intuitive models salient and a priori plausible. Those cultural suggestions are certainly not decisive! Let’s not be crazy—the Invisible Ships Myth is in fact a myth. But they’re certainly relevant.

Ideas from culture can influence intuitive self-models by making certain models salient (a.k.a., increasing their prior). But these kinds of models are never adopted uncritically; they still need to issue accurate predictions.

DID is nothing special in this respect. Likewise, trance states like spirit possession and channeling (previous post) are likelier to happen when people expect them to happen for cultural or other reasons; and so too with hallucinations, such as renewalist Christian communities that normalize hearing the voice of God (coming up in Post 7).

5.3 Why is Borderline Personality Disorder (BPD) a risk factor for DID?

The way I understand BPD right now (see e.g. the Lorien Psych overview) is that it’s basically just a condition where your emotions are very much stronger than typical.

Why is BPD such a big risk factor for DID? In other words, why might strong emotions make DID more likely?

I have a theory!

Let’s start from an obvious point: we all have a thing where happy-me and sad-me and angry-me and so on are all somewhat different from each other:

They have somewhat different beliefs—maybe angry-me thinks that my parents disrespect me, and sad-me thinks that my parents are disappointed in me, and happy-me thinks that my parents love me.
They have somewhat different preferences—maybe angry-me wants to get in a fight with my parents, and sad-me wants to cry in bed, and happy-me wants to go skateboarding.
They have somewhat different habits, and somewhat different mannerisms, and somewhat different role-models, and somewhat different goals … you get the idea.

In §4.4.2, I talked about how concepts can be stretched and patched, but if you try to stretch a concept too much, then the original concept just stops applying altogether. So too with happy-me, sad-me, and angry-me.

“A baseball that is literally exactly like a carrot in every way” … is not a baseball in the first place. Thus, if you stretch a concept too far, the concept stops applying altogether. I claim that DID starts with something like that: strong emotions result in *such* a strong difference between happy-me, angry-me, traumatized-me, etc., that these splinter into separate intuitive concepts, rather than variations of the same “me”. (Image source)

If your emotions are mild, it’s easy to take the same homunculus concept “me” and patch it to be happy or sad or angry or traumatized or nurturing or whatever. But if your emotions are extremely strong, as they are in BPD, then it gets progressively less plausible to stretch a single concept into those different cases, and progressively more plausible that there are multiple irreconcilably-different homunculus concepts.

5.3.1 …So the DID alters correspond to different emotional states?

If you start with the *Inside Out* (2015) setup, and then edit it such that only one of the little homunculus characters can exist at a time (analogous to how you can’t have two different songs in your head at once), then … well it’s not *exactly* how I’m thinking about DID, but I suppose you could do worse.

Answer: Not exactly—I wouldn’t go quite that far.

My theory is that DID starts from different alters that correspond to different emotional states, and I expect that probably remains true to some extent. For example:

Typically there is said to exist a "host" personality, and several alternative personalities or "alters". … None of the personalities is emotionally well-rounded. The host is often emotionally flat, and different alters express exaggerated moods: Anger, Nurturance, Childishness, Sexiness. Because of their different affective competence, it falls to different alters to handle different social situations. Thus one may come out for lovemaking, another for playing with the kids, another for picking a fight and so on. —Humphrey & Dennett 1989

But I don’t want to over-emphasize this by claiming a perfect 1-to-1 correspondence between alters and emotional states. For example, I think DID patients these days can wind up with many dozens of “alters”, far more than they could have significantly different emotional states, I would think.

As discussed in §4.2, changing to a very different intuitive model can be hard. I presume that, after a lifetime spent in the Conventional Intuitive Self-Model (§3.2), it takes a while for a DID intuitive self-model to get established / salient / vivid—or in probabilistic inference terms (§1.2.2), for the DID intuitive self-model to get used enough to earn a high prior probability. Also, the DID intuitive self-model probably goes along with certain habits of attention-control, which need to be learned but could eventually become second-nature.

So if someone has never experienced DID at all, I think that “having strong emotions, and splitting them into alters” is a comparatively easy way for them to start experiencing DID. Whereas if someone has already been experiencing the DID intuitive self-model for years, I imagine that they would have a much lower barrier to adding additional alters, and perhaps each alter might develop more of an emotional range.

5.4 Why is childhood trauma a risk factor for DID?

There’s definitely a correlation between DID and self-reported childhood trauma. But there’s also a correlation between BPD and self-reported childhood trauma. I can’t figure out from the literature whether there’s any correlation between self-reported childhood trauma and DID beyond the effect that you’d expect from strong emotions being a mediator.^[2]

But before we get to DID, why is there a correlation between BPD and self-reported childhood trauma? I think it’s some combination of

(1) If someone has BPD, then their parents, siblings, etc. may well have also had strong emotions, including strong anger, which may have led to abuse and so on;
(2) If someone has BPD, then they probably also had strong emotions when they were a kid, which may have led to more childhood strife of various kinds, and also may have led to any given bad situation being more traumatic to them than it might have been to a different child;
(3) Maybe childhood trauma directly causes BPD somehow;
(4) Maybe BPD causes false memories of childhood trauma somehow.^[3]

In the case of BPD, my hunch is that the above list is in roughly descending order of importance / frequency / plausibility. But that’s not a very strong or informed opinion.

Back to DID, the above mechanism question actually doesn’t matter much for this series: I think my story above, where strong emotions are the seeds of the first alters, is compatible with any of those mechanisms. In particular, if I want to tell a (3)-like story for DID, I would say: “feeling traumatized” happens to be a strong emotion, and so if strong emotions can be the seeds of the first alters, then, well, there you have it. The DID could begin with the splitting of “me” and “me-when-feeling-traumatized” into two irreconcilably different concepts, as the first two alters. And then (as above), once there are any alters, there’s a much lower barrier to additional alters popping up later on.

This kind of story seems compatible with accounts of DID that I’ve seen, for example:

Apparently normal parts (ANP) are the rational, present-oriented, and grounded parts of the individual that handle daily life or, in the case of [DID] and so multiple ANP, aspects of daily life. … Emotional parts (EP) are the parts of the personality that represent the dissociation or that contain the traumatic materials (memories of the trauma, internalized beliefs and perceptions, learned responses, etc). EP are often drawn forth by reminders of the trauma and may not experience much of everyday life. Because of this, they might be far less developed than ANP. (Source)

5.5 Amnesia

In §3.4.1, I talked about how a self-reflective explicit belief like “I’m pretty sure that the capital of Burundi is Gitega” would be conceptualized in a way that involves the homunculus; whereas an implicit belief like [implicit expectation that the bedroom doorknob is hard to twist] would be conceptualized as being unrelated to the homunculus, and rather a property of the doorknob itself.

We can generalize this idea to split up thoughts-that-involve-the-homunculus and thoughts-that-don’t. If I recall a memory of going to the carnival yesterday, that thought probably involves the homunculus. If I watched a movie last night, deeply entranced in a flow state (§4.4.1.1), and if seeing a river triggers a memory of a scene in that movie, then that thought probably doesn’t involve the homunculus. Etc.

In the case of DID, there’s not one homunculus, but several irreconcilably-different (§4.4.2) homunculi, one corresponding to each alter. And I claim that inter-identity amnesia tends to occur for memories that involve the homunculus, but not for memories that don’t.

Why should that be? Well, suppose the Alter X homunculus concept is the active one right now. And then the person sees a dog that they saw yesterday when Alter Y is active. The dog will “seem familiar”—that’s an implicit memory not involving any homunculus. But meanwhile, there’s also an autobiographical memory stored in their brain, namely M = “myself as Alter Y seeing the dog”. Normally, we would expect that the sight of the dog would summon memory M into conscious awareness via the normal auto-associative memory mechanism. But that doesn’t happen in DID. Why not?

My answer is:^[4] Alter X and Alter Y are incompatible concepts like “square circle” (§4.4.2). I currently have Alter X active, but the memory recall would require making Alter Y active. So that’s a conflict, which suppresses the auto-associative recall, analogous to how (as in the last post) you can prevent yourself from thinking about a tune by holding a different tune in your head. So as long as Alter X is active, the memory M of the dog cannot resurface. (See this comment for more gory details.)

This theory seems to predict that alters would perform at chance levels in a memory test. But in reality they perform somewhat worse than chance (see Huntjens et al. 2006)! I think I can explain that, though. Imagine we’re quizzing this person on what animals they saw yesterday. When they come to the picture of the dog, the auto-associative memory mechanism will “try” to increase the activation on memory M. But the strength of Alter X will prevent M from fully activating. However, in the meantime, there are these two mutually-contradictory thoughts duking it out below the surface. I think that spawns a transient interoceptive sensation of confusion, which gets subjectively interpreted as implausibility, which in turn is interpreted as evidence that “dog” is the wrong answer on the memory test.^[5]

Of course, another possibility is that the person will in fact remember the dog. As far as I understand, inter-identity amnesia is not universal (although if it never happens, then it wouldn’t fit the modern diagnostic criteria for DID). Lack-of-amnesia could be explained in a number of ways. Maybe these two alters happen to not be irreconcilably different concepts (§4.4.2), but rather two variations on the same concept. Maybe the alters are not so stably entrenched that it’s impossible to catch a glimpse of the Alter Y concept while Alter X is active, for whatever reason. Maybe the person has developed some subtle metacognitive attention-control strategy for extracting information from Memory M without fully summoning the Alter Y concept. I don’t know!

As far as I can tell (from brief study), my proposal in this section is compatible with everything known in the literature on DID amnesia. For example, Eich et al. 1997 and Allen and Movius 2000 both found lack of amnesia in tests of implicit memory / recognition, while Huntjens et al. 2006 suggests “so-called meta-memory problems, i.e. they are truly convinced of not knowing what happens when another identity is in control of their behaviour … may lead patients not to acknowledge correctly retrieved material”, which I think matches the fact that the mechanism specifically impacts self-reflective thoughts and memories.

5.5.1 If everyday strong emotions are a bit like mild DID, do they come with mild amnesia??

I proposed above that when Alter X is active, it interferes with the activation of memories that trigger Alter Y. That might sound like an exotic civil war happening within a brain, but it’s actually an extension of a perfectly ordinary dynamic.

I proposed above that DID starts out with alters that correspond to different emotional states—angry-me, sad-me, happy-me, anxious-me, etc. Now, when you’re angry, then you tend to be thinking about the fact that you’re angry—for example, you’ll dwell on all the reasons that you’ve been wronged, and pick fights, etc. And when you’re sad, you tend to be thinking about the fact that you’re sad—you’ll dwell on all the reasons that nobody will ever love you, and go brood, etc.^[6] Ditto with anxiety and so on. The mechanism here is, I claim, an interoceptive sensation of the mood, coupled to involuntary attention on those sensations (see here—more on this in the next post).

Does this connect to amnesia? Sure! If you’re angry at your parents, then (if you’re like me) you probably won’t spontaneously think of the nice thing they did for you yesterday, but you will suddenly acquire a savant-like memory of every time they have ever wronged you, practically all the way back to ancient memories of those jerks forgetting to change your poopy diaper. (It’s the dumbest superpower ever.) Likewise, if you’re anxious right now, then you probably won’t spontaneously remember examples of things going well, but will spontaneously remember examples of how everything has always been scary and awful.

(Of course you, dear reader, would never let emotions influence your judgment, memory, etc., but I’m sure you’ve seen other people in your life do those kinds of things, right?)

Why is the amnesia in DID so much more severe then this everyday phenomenon? Well, I think there’s kind of a “phase transition” when a person flips from modifiers-on-one-homunculus-concept (“happy-me” vs “angry-me”) to multiple different mutually-exclusive (§4.4.2) homunculus-concepts. If it’s the same “me”, then there’s some amnesia-inducing incompatibility between happy-me and sad-me, but it’s inherently limited; auto-associative recall can still cross that boundary. If it’s different “me” concepts altogether, then they can shut each other out far more effectively.

5.6 Conclusion

I think my DID story hangs together reasonably well, but I’m obviously not an expert. Feel free to chat in the comments.

Coming up in the next post, we’ll continue our tour of unusual intuitive self-models with “awakening”—a family of intuitive self-models that may take years of meditation practice to access.

Thanks Simon Skade, Linda Linsefors, and Justis Mills for critical comments on earlier drafts.

^{^}
For a case against DID being iatrogenic, see “Myth 4” of Brand et al. 2016. They mainly argue that the suggestion of culture and psychologists is neither necessary nor sufficient for DID. And then my response is: OK sure, but I still think it’s a strong risk factor for DID.
^{^}
I don’t think you can figure this out just from published statistics because, for example, different BPD cases have different severity, and perhaps only the most severe BPD cases are at elevated risk of DID. So looking at the set of all diagnosed BPD patients wouldn’t tell you what you need to know; I think you would need to analyze some raw dataset. But I dunno, I didn’t think about it too hard, nor did I do any lit-search.
^{^}
You can find an argument in favor of (4) in, for example, (Lynn 2012).
^{^}
This is my favorite hypothesis, but here’s another one I was playing with earlier: When Alter X is active, there may be a motivation to keep Alter X active, and not have Alter Y become active. (Not necessarily—e.g. when I’m anxious, I can remain anxious despite wanting to stop.) If that motivation is there, then any thought involving Alter Y would become demotivating / aversive, and it would get surrounded by an ugh field—any thought or action that seemed to be starting to potentially activate the Alter Y concept would itself feel aversive. So then (according to this hypothesis) the person would flinch away from the mental moves that might activate the memory M of the dog.
So that’s a hypothesis. Maybe it’s part of the story too, but I now generally don’t like it as much, for a few reasons. One of them is related to timing and associations: I think the person wouldn’t have any reason to flinch away from something that triggers the memory of the dog, until the memory had already been triggered, and then it would be too late. Another is that, having written §3.5, I now have a better appreciation of the fact that when involuntary attention drives brainstorming, that subjectively feels like “wanting”, even when it doesn’t really involve motivation (in the sense that it’s not driven by the valence mechanism). So, even though it introspectively feels like I want to remain angry when I’m angry, that might not be true in the straightforward sense that I initially thought it was.
^{^}
I’m especially unsure about the explanation in this paragraph.
^{^}
If that’s not intuitive, you can find some references near the top of this Scott Alexander post.