36 Comments

My understanding is that inner speech doesn't correlate with much of interest today, so I would be inclined to think it's an accident. Perhaps just a question of which thoughts get wired close to the auditory part of the brain (idk much about neurology 😅).

Expand full comment

At extremes (eg schizophrenia) it correlates with quite a bit :)

I'm somewhat bearish on psychometrics' ability to measure outwardly obvious things like extroversion. So in some ways I don't think we should look to instrument correlations. That said, my guess is GFP would correlate with some inner voice instrument. Not really sure what I'd want that instrument to measure. Amount of time spent with inner voice? Tenor of inner voice? Tendency to think in images vs language? Psychometrics actually does have a survey designed to measure inner voice. But IMO it's essentially an extroversion scale. For example: "I can make impromptu speeches even on topics about which I have almost no information." (Full test available here: https://openpsychometrics.org/tests/SMS/)

Temple Grandin has a great piece where she argues that her conscious experience is very different because she thinks using images instead of pictures: https://www.grandin.com/welfare/animals.people.autism.true.consciousness.html

So different that she thinks she does not meet some philosopher's definition of conscious.

I guess I'd characterize all of this as, "reasonable to think inner speech correlates with important outcomes". In the weakest case this is defending simply because we know so little. Hard to measure; suggests broad priors.

Expand full comment

> I'm somewhat bearish on psychometrics' ability to measure outwardly obvious things like extroversion.

Really? Why? I agree that the usual method of self-rating via questionnaire has numerous limitations, but the fact that we get heritabilities in the .50s for some traits suggest there's plenty of signal to noise in common questionnaires. And if this isn't enough, what about ratings given by other people, or content analysis of written work, or even biomarkers (e.g. "How do cortisol levels change for subjects asked to introduce themselves to someone for the first time?) Near the time of his death, I remember Eysenck was developing measures of Neuroticism using dark vision and body sway.

Expand full comment

Yeah, just the usual. My day job is actually measurement of things like dementia by behavior in VR. I do believe one can get a pretty accurate picture and even make decisions using the numbers.

It's more that if someone says a correlation between a survey result and something else is small, that isn't strong evidence to me that the ostensibly measured construct is unrelated or unimportant. I think that GFP is more important than IQ (for evolutionary reasons), even though IQ usually has higher correlation with life outcomes. My answer is that personality is harder to instrumentalize than intelligence. And inner voice must be more difficult than either. Not that anybody has tried all that hard.

Expand full comment
Comment removed
Nov 3, 2022
Comment removed
Expand full comment

> The challenge of heritability rates is that they are extremely non-specific.

Well... just so long as we agree that noise has heritability 0, and no external validity. Personality scales do a fine job at predicting criminality, career interests, and plenty of other other things starting with "c," like conservatism or belief in conspiracy theories.

But I agree with you that *what* is being measured isn't always obvious. It isn't even always the same for the same instrument! We look at the personality scores of American states, or entire countries...

https://www2.psych.ubc.ca/~heine/docs/2008Mccrae-rejoinder.pdf

...and find the scales no longer do what they were designed to do with people inside a larger group.

The solution obviously is to move away from self report where possible, toward things like analysis of posts on social media or emotional analysis software, and ultimately some kind of brain scan. Sadly, I doubt I'll ever see it. Unless someone hands me a pile of money to study it, biomarkers aren't in my future. I simply make do with observer ratings. If I can avoid it, I never measure personality by self report alone; give me friendship dyads and I'm much more confident in the scores that come out at the end.

Expand full comment

It's strange that my comments were removed, the one you responded to here and another one below. As far as I know, there was nothing offensive about them. You offer a short quote of what I wrote. It's a ever so slightly critical comment, but neutrally expressed as part of intellectual dialogue. Why would that merit being censored into silence?

Expand full comment

> I'm somewhat bearish on psychometrics' ability to measure outwardly obvious things like extroversion.

Hm, any specific problems with Extraversion? My impression is that Extraversion is the personality trait that is psychometrically the strongest.

> Psychometrics actually does have a survey designed to measure inner voice. But IMO it's essentially an extroversion scale.

I agree that it seems more like an extraversion scale than an inner voice scale.

> Not really sure what I'd want that instrument to measure. Amount of time spent with inner voice? Tenor of inner voice? Tendency to think in images vs language?

Hmm.

My immediate thought was the loudness or crispness of the inner voice. Like for me, my inner voice is like a whisper. Meanwhile some people don't have an inner voice at all, and some have one that is very crisp and clear.

But I suppose you may be right that this is not the thing that matters, or at least not obviously so. Time spent with the inner voice seems like something that should matter more, but then that feels likely confounded with mental illness.

In studying inner voice, I guess one thing I would like to know is its factor structure. My bet would be that there's a strong general factor of ""more"" inner voice vs ""less"" inner voice, because that's what usually happens, though I would be unsure whether this general factor is real or a methods factor.

> At extremes (eg schizophrenia) it correlates with quite a bit :)

... Maybe? I'm not actually sure about this, the correlation depends not just on how many schizophrenics have strong auditory hallucinations, but also on how many nonschizophrenics do.

And regardless, the inner voices for schizophrenics seem to be for different reasons than for normal people? And I'd expect it to be the different reasons that drive the divergence between schizophrenics and normal people.

Expand full comment

>Hm, any specific problems with Extraversion? My impression is that Extraversion is the personality trait that is psychometrically the strongest.

Not particularly. In general I think personality is more important than intelligence, because intelligence is a subset of personality. But there is no personality instrument as externally valid as IQ. Personality instruments introduce a lot of noise.

>My immediate thought was the loudness or crispness of the inner voice.

My initial thoughts of a study to test the relation of personality and inner voice. Would be to make a inner voice instrument that is scored on several axis, including its unrotated first PC. Correlate these to PC1 of some general personality survey. Preregister that design. Collect data. correct for multiple comparisons. Are the effects large? I'd be happy with a 0.3 correlation.

Expand full comment

> But there is no personality instrument as externally valid as IQ. Personality instruments introduce a lot of noise.

Insofar as the issue is noise, I think this can to a large extent be solved by multi-informant data.

(I personally have a lot of additional concerns about personality, but I don't think my concerns necessarily lower the external validity if true.)

> My initial thoughts of a study to test the relation of personality and inner voice. Would be to make a inner voice instrument that is scored on several axis, including its unrotated first PC. Correlate these to PC1 of some general personality survey. Preregister that design. Collect data. correct for multiple comparisons. Are the effects large? I'd be happy with a 0.3 correlation.

My prediction for this would depend a lot on whether one includes the content of the voice or not. If one exclusively considers contentless matters like the strength of the voice, then I would be surprised by a PC1 correlation that exceeds 0.1. However, if one includes what behaviors the voice is saying one should engage in, then I could easily see the correlation exceed 0.5. I would question the direction of causality in that circumstance, though.

That said, because I suspect the direction of causality to be personality -> voice content, I sometimes wonder if one could design a more accurate personality test by including inner voice content and a bunch of other things like that. (Or rather, I hadn't thought about voice content before, but had thought about other things one could put into a personality test as alternate indicators which might be useful due to decorrelating the noise.)

Expand full comment

> Insofar as the issue is noise, I think this can to a large extent be solved by multi-informant data.

That's my preferred method. Psychologists deal with random error by asking numerous questions for the same trait, but systematic error is hard to get around without using a second rater.

Expand full comment

This is actually the question that brought me into personality structure. I was predicting Big 5 from Facebook statuses. Numbers were getting pretty good and I wondered what the ceiling was, or if there was a point that my prediction would be better than the survey "ground truth". Led me down a theoretical rabbit hole. Eventually ended up deriving the Big Five from scratch from word vectors. Structure is very similar to the original survey results so not much to update on psychology side. Though with one exception! PC1 is definitely not response bias, it also shows up in word vectors where there is no respondent.

Expand full comment
Comment removed
Aug 18, 2022
Comment removed
Expand full comment

Love Grandin's essay arguing that consciousness is possible without language. IMO the Jaynesian view of consciousness is less vulnerable than philosophers. He's arguing that "in the beginning was the word", the path it took to get there. A little more wiggle room than saying language is required writ large.

Expand full comment

As I recall, this and the other comment about Jaynes directly below were left by me, along with another comment about heritability. All of them were removed. Why? I find that sad and disappointing. I thought I was offering worthwhile views that weren't mean, antagonistic, trolling, or any such thing. It seemed to be friendly, easygoing, and intelligent dialogue by all involved.

Expand full comment
Comment removed
Aug 18, 2022
Comment removed
Expand full comment

Yeah, I definitely think her case is handled well by the Jaynesian view. Consciousness consists of a private mindspace and volition, not 'thinking in words'. The mindspace she describes as nonverbal could still be downstream a language-based breakdown of the bicameral mind.

Expand full comment

I'd be more interested to see what its relationship is to intelligence (g), and I'm betting that study hasn't been done.

Expand full comment

Schizophrenia is 2-3 times more common among Africans than it is in at least some other groups; Europeans, for example (most of the data come from the historically biracial America). Assuming one takes this at face value rather than attributing it to measurement error, and assuming that Jaynes's hypothesis is true, you would need some selective force to explain why the diminishment of schizophrenia relative to "normal" inner speech is more advanced in some populations than others. Schizophrenia diagnoses obviously tend to happen when the voices in your head are actually maladaptively interfering with your life, so it's not as though we can write this off as mistakenly treating benign hallucinations as pathologies. Also, if I remember correctly, schizophrenia has a paternal age effect, i.e. spermatic mutations make it significantly more likely, which seems suggestive of a simple mutational load hypothesis rather than being a consequence of any kind of strategy. So ... I dunno.

Expand full comment

Really enjoying your blog.

My thoughts are adjacent to this so I think sharing them with you might be of value, maybe you can connect the dots:

1. Game theory tells us that given a series of many repeated interactions, the optimal strategy -- the highest overall "payout" (when assuming payout for both parties matters) -- is when two agents cooperate. It would make sense that a primary factor follows along the lines of enforcing cooperation because math tells us that it's optimal, and we've been around long enough as a large number to land somewhere that is optimal *for the human race to exist* (which means highest payout for all over time).

2. The thing that I don't think is stressed enough in this post, and maybe you will expand in future posts, is that expressive language is what allows us to compare our internal voices to each other. I believe that is very significant as to why humans are so much more superior than other non-verbal animals -- it's one thing to be able to teach knowledge through verbal expression, that definitely gives us an edge having not to wait generations to learn -- but it's entirely another to be able to express, compare, challenge, and change our inner voices directly without having to "apply and die" to realize the differences.

Edit: Noting how the individual recognizes the separation between the internal voice, the words others in the "real-world" express, and the personal minimum survival instincts (by this I mean things like "thirst" which is chemical) would better explain the purpose of what Freud called "ego."

Perhaps, as people begin to spend less time with others, more time on the internet, and the world increasingly becoming image-obsessed, we are dampening our traditional inner voices more than ever while at the same time taking on increased "real-world" external expression which neither aligns with our voices nor our most basic instincts.

Or maybe it's just that the distance between those traditional inner voices and the external expression now has reached a distance that makes people question themselves (up goes anxiety) as well as increasingly become less able to simulate the minds of others (down goes trust).

You have my e-mail (sidenote: substack analytics are crazy good actually) -- feel free to reach out if you want to bounce some ideas around ever.

Expand full comment

Excellent point about game theory, it does indeed imply we should mostly cooperate. This post I relied on "Darwin said!", which is useful because he also says that cooperating is more important than anything else. That it is our niche. Hard to get the superlative from game theory.

>The thing that I don't think is stressed enough in this post, and maybe you will expand in future posts, is that expressive language is what allows us to compare our internal voices to each other.

I think this is how the bicameral mind broke down. Essentially we worked it out together that we could evaluate the agents. Those that didn't work it out were told.

The ten commandments are a wonderful systematization of what the various claims bicameral agents would make. May have been the result of thousands of years of discussing inner voices.

>Perhaps, as people begin to spend less time with others, more time on the internet, and the world increasingly becoming image-obsessed, we are dampening our traditional inner voices more than ever while at the same time taking on increased "real-world" external expression which neither aligns with our voices nor our most basic instincts.

Exploring these ideas has definitely made me more cautious about the internet and what it can do to a mind. In fact, I developed these ideas only after taking a step back from the internet. Had more time to think and ruminate and got my social kicks IRL with family and sports. Incredibly different mindset than poasting. Very helpful for creativity, at least in my experience.

>You have my e-mail (sidenote: substack analytics are crazy good actually)

May take you up on that! btw, why did you pick a fox avatar? A lot of ancient stories about spirits are related to the fox...

Expand full comment

> I think this is how the bicameral mind broke down. Essentially we worked it out together that we could evaluate the agents. Those that didn't work it out were told.

The ten commandments are a wonderful systematization of what the various claims bicameral agents would make. May have been the result of thousands of years of discussing inner voices.

It would make sense that the commandments were one of the first or better attempts at trying to formalize "a single voice" -- although I still hold reservations as while I think expression allows us to compare, it may not allow us to incorporate at the same rate... in other words, I'm not sure you could "force" people to change their inner voice at an individual level (as I hold the belief that much of that "inner voice" is biologically/genetically inherited) but rather those that didn't adapt their inner voices to the rest of the society (who were subscribing) would simply not be as useful and relatively fail.

>Exploring these ideas has definitely made me more cautious about the internet and what it can do to a mind. In fact, I developed these ideas only after taking a step back from the internet. Had more time to think and ruminate and got my social kicks IRL with family and sports. Incredibly different mindset than poasting. Very helpful for creativity, at least in my experience.

>May take you up on that! btw, why did you pick a fox avatar? A lot of ancient stories about spirits are related to the fox...

I had to do a similar thing when I came to the conclusion and I spent the better parts of the last two years living out in the woods, only coming back to see friends and family occasionally. I got extremely into wildlife photograph (mostly birds) as a result... so don't read too much into the fox -- it's just one of my favourite shots I took early one morning while walking back into town. Full resolution so you can see the kill: https://ibb.co/thv9S9D

Expand full comment

>It would make sense that the commandments were one of the first or better attempts at trying to formalize "a single voice"

I don't think it's a coincidence that they emerge at the same time period as monotheism. Egyptians introduced (then rejected) the first monotheism with Aten. Freud believed that Moses was a priest from this sect: https://en.wikipedia.org/wiki/Moses_and_Monotheism

>I'm not sure you could "force" people to change their inner voice at an individual level (as I hold the belief that much of that "inner voice" is biologically/genetically inherited) but rather those that didn't adapt their inner voices to the rest of the society (who were subscribing) would simply not be as useful and relatively fail.

What about mobs, where individuals often are taken up by the wisdom/insanity of the crowd? Also interested in what you mean by genetically inherited. I can go for us inheriting an inner space that is more or less anthropomorphized (surely if dolphins had an inner space it would be strikingly different). But when Terrence Mckenna talks about tripping on psychedelics and being visited by self-transforming machine elves. I think that has more to do with him being raised with Irish folktales than his Irish stock. (Fun interview, if interested in that sort of thing: https://www.youtube.com/watch?v=uJjR3aUhsOk)

Great shot on the fox

Expand full comment

No mention of how people have been noticing Google ads are based on thoughts they had and without any chance of correlation to whatever Google's AI could infer? Same for Facebook, Apple, Microsoft, TikTok.

Silicon Valley's algorithms have long been tapping into illegal surveillance of an unprecedented kind - besides the cliques of insiders who misuse this for stalking and harassment of children, human rights defenders, activists, minorities and everyone else.

Expand full comment