tl;dr : started out explaining a paper about LSD, visual hallucination, and fMRI; ended by ranting about science journalism. If you don’t care about LSD, skip right to the end.


Earlier last week, a new scientific study was published and received quite a lot of news coverage. The topic under investigation? Brains on LSD. Since it has brains, drugs, some foundational techniques in cognitive neuroscience, and apparently the neuroscience- equivalent of the Higgs Boson, I thought it’d be fun to do a quick dissection of the actual paper, and offer what will certainly be a much more nuanced (and hopefully more informed) perspective than some of the news coverage with overblown interpretations. But to start, I will leave you with the Wikipedia blurb,

“Lysergic acid diethylamide (LSD), also known as acid, is a psychedelic drug known for its psychological effects, which include altered thinking processes, closed- and open-eye visuals, synesthesia, an altered sense of time, and spiritual experiences. It is used mainly as a recreational drug and for spiritual reasons.”

and this article from the Guardian.

What is brain?
To sufficiently explain this study, we have to go over a few basic concepts about the brain first. In one sentence: our brains are composed of billions of brain cells (neurons), and they work via exchange of electrical and chemical signals delivered between neurons, which fulfills the roles of information transmission, computation, and storage. Of course, when I say that, it’s a very physical (and vague) description of the underlying biology, and we have very little certainty about how information (like a memory) is actually represented. But for today, we can think of the brain as a whole lot of people talking via text messages, and packets of chemical signals called neurotransmitters, are the messages flowing from one person (neuron) to another. At this point, there are two important things to understand.

1. Serotonin
Neurons, like people, come in all kinds of different personalities. Each type of neuron can only send one type of message, dictated by which neurotransmitter it releases to other neurons, while they can receive many types of different messages, meaning that they will respond to different types of neurotransmitters. Serotonin is one such neurotransmitter, and neurons all over the brain have the ability to receive this chemical and respond to it in different ways. You may have heard phrases such as “serotonin is the neurotransmitter of happiness” or “dopamine means reward”, but those are gross oversimplifications. As far as we know, there is a small cluster of serotonin neurons in the lower parts of the brain that, under normal conditions, send messages to other neurons ALL OVER the brain, to modulate their activity. The physiological effect of that on the receiving neurons is diverse and rather beyond my knowledge, and it actually goes beyond the brain to have direct and secondary effects on other parts of the body, such as the heart.

LSD, then, is a substitute for serotonin, as in when LSD is delivered to your brain, it acts just like serotonin, even though those serotonin neurons didn’t send any messages. From the point of view of the receiving neurons, they couldn’t care less. In fact, neurons EVERYWHERE throughout the brain are receiving text messages that look like this:

“LETS PARTY”

— serotonin neurons

So you know what they’re about to do - they’re gonna get turnt up. And I was just informed via Wikipedia that LSD is unique in that it not only mimics serotonin, but dopamine as well, for a subset of receiving neurons. So for those neurons, the message is loud and clear:

“PARTY WITH SEROTONIN NEURONS. TURN DOWN FOR WAT”

— dopamine neurons

(Actually, I just realized that an understanding of LSD is not at all required to understand this study, since it does not at all investigate the mechanism behind it. But still, these things are good to know.)

2. Cerebral Blood Flow (CBF)
Neurons need basic resources to stay active, so when they are communicating a lot (via “action potentials”), they need more food and oxygen delivered to them. It turns out, blood delivers food and oxygen, and someone pretty smart figured out that if you measure blood flow to an area of the brain, you can make a guess of how hard neurons in that region are working. This is the basis of functional Magnetic Resonance Imaging (fMRI) in neuroscience, and while there are several techniques that measure slightly different aspects of this signal, they are all related to changes in blood flow and blood oxygenation. The important thing to remember here is that CBF is an interpretation of neural activity, not an actual measurement. It’s like if I were to measure how much food you eat per day to infer your activities: I can tell on days that you eat a lot that you were really hungry and probably did a lot of physical activity, but I won’t be able to say exactly what you did. While this is a limitation for neuroscience research, it’s largely an assumption that we are willing to make so we can infer SOMETHING about mental processes while measuring a physical signal.

What actually happens when brain is on LSD?
Now that we know all about how brains work and how to measure brain activity, let’s get on to the paper. The experiment is simple. As with almost all cognitive neuroscience experiments, they want to measure people’s brains while they have some kind of mental experience, and see if the two have any kind of relationship. Now the title of the paper almost makes perfect sense: Neural correlates of the LSD experience revealed by multimodal neuroimaging. I like straightforward titles.

So the setup is this: inject people with a small dose of LSD, then scan their brains while they do nothing (turns out they hallucinate). Have the same people come back, inject them with salt water, then scan their brains again while they do nothing (hopefully they don’t hallucinate). Measure their brains and ask them a bunch of questions like “how much did you hallucinate” and “how much oneness did you experience with the universe”, and see if there is a relationship between their brain activity and their mental experience. Well, onto the results then!


Visual hallucination correlates with VISUAL cortex activation
Figure 1 of the paper is shown above. The first row of brains is a snapshot of the blood flow to the participants’ brains at 4 different depths (and a side view), and you can see that, when you are alive, blood flows to most parts of your brain (I thought we only used 10% of our brains???). The second row - when on acid - shows almost the exact same thing, and it’s only when you subtract the 1st row from the 2nd, do you see an increase of blood flow to a small area in the back of the head, known as the visual cortex. This is the area of the brain that is activated when we see things, except, in this case, they’re “seeing” mentally, because they’re hallucinating. I’m not sure how this compares to when people try to visually IMAGINE, without doing drugs, but that would’ve been a better control condition. The scatterplot at the bottom shows that people’s self-report of visual hallucination correlates with how much more their visual cortex is activated. In other words, the more you felt that you hallucinated, the more your brain was actually activated - pretty neat!

It’s important to bring up a subtle point here: this experiment demonstrates the neural correlate of visual hallucination (i.e.LSD experience), not of LSD itself. Because people displayed a RANGE of altered behavior, since, as you can see, a few people reported almost no changes to their degree of visual hallucination (I’d love to know if it’s because their “normal” visual experience is just the fucking bomb), it meant that LSD did not affect everyone in the same way. This will be true for the rest of the paper, in that their goal was to measure how brain activity alters in relation to the experiences LSD can trigger, but not how LSD works physiologically.


Visual hallucination correlates with increased “connectivity” of the visual cortex to other parts of the brain
Forget about what I said at beginning, THIS figure is the reason why I’m writing this post. I don’t have a particular problem with the actual figure in the paper per se (although plotting z-score is somewhat atrocious), but the way it was used in the Guardian article. In the article, the caption reads: “ A second image shows different sections of the brain, either on placebo, or under the influence of LSD (lots of orange) “. Now, if you didn’t know any neuroscience at all and was shown this picture, what would be your first guess at what’s going on here? Well, I can tell you what I thought after two years into graduate school: holy Batman, LSD makes your brain go crazy! The first time I heard about this study was, in fact, through the Guardian article, and I was shocked at how much effect LSD had in activating the entire brain. Alas, it was a good idea to read the actual paper: I was so perplexed because nowhere in there did it mention profound increase in brain activity. Well, it’s because there wasn’t. In fact, we just saw in the previous section that there was no increase in brain activity other than in a tiny region in the visual cortex. So what is this figure actually depicting then?

To understand that, I need to explain one more concept: resting state functional connectivity (RSFC). Recall earlier we said that brain areas that are working harder requires more blood to be delivered, and this signal is recorded via fMRI and fluctuates rapidly, on the order of seconds. If two areas (A and B) show similar fluctuation of activity over time, then, we would assume that there is some special relationship between them. You might say that region A is sending messages to region B to activate it in a very similar way, causing the correlated activity. But without detailed knowledge of anatomical connectivity, this is a huge assumption. Instead, we simply call these two areas “functionally connected”, and it’s a measure of correlation between the blood flow signals of these areas. If we can correlated area A with area B, we can do it for A with C, A with D, etc. In this way, we hold area A as the “seed”, use its signal as a reference, and correlate the signals of every other brain region with it to build a functional connectivity map, which is exactly what Figure 2 shows.

In the top row, our reference signal (seed) is taken from the visual cortex (purple blob), and the redness of the other areas indicate how correlated each of those areas are with it. I believe (I could be wrong) the actual color represents the z-score of the correlation, which is NOT the strength of the correlation, but how UNLIKELY that the observed correlation is due to chance. I don’t do fMRI work, so I don’t know what to criticize it too much without reading into it, and if someone can inform me of the actual reason, that would be great. But in general I don’t know why the magnitude of the correlation is not plotted instead. ANYWAY, end rant. So in the middle row, we see that when hallucinating on LSD, many more areas of the brain (almost all of it) become “functionally connected” with the visual cortex, i.e. their fluctuating activities show increased correlation in time. The bottom row is once again the difference, and the scatterplot demonstrates that the more self-reported change in visual hallucination, the more change in functional connectivity (though it’s debatable which is causing which).

Anyway, how this figure was used made me very unhappy, but I will try to wrap up the paper quickly and come back to it at the end.


“Ego dissolution” correlates with decreased connectivity of parahippocampus
The good thing about learning stuff is that it will maybe sometimes generalize. In this case, Figure 3 is in the exact same format as Figure 2, just with a different seed brain area and a different behavioral measure. Parahippocampus (purple blob again) is one of those areas in the brain where, if you were to ask someone what it does, they’ll either say “hell if I know”, OR they might say something stupid like “it recognizes scenery”. I’m in the former camp, mainly because it’s an understudied area, and it’s (unfortunately) involved with more abstract processes like social behavior, which you can’t really measure from a non-human mammal, or at least we don’t have a great behavioral assay for it. A brief Wikipedia search will tell you that it’s involved in schizophrenia and scene recognition, so take from it what you will. As far as I can tell, these authors basically picked it since there was prior evidence linking its activity to psychosis-like states under LSD.

“Ego dissolution”, on the behavioral side, is also a fun one, I won’t even attempt to find a definition for it, although I’m sure there was a somewhat tangible meaning in the surveys. Or, who knows, maybe it’s the kind of shit that you just have to take LSD to know what it’s like, and since the participants were all experienced drug users, they would be the ones to know. I guess if you want to find out what that’s like, drop acid (or meditate deeply, I’ve read that induces similar states.) In any case, the result here is straightforward: decreased functional connectivity, particularly between the parahippocampus and the posterior cingulate cortex (PCC, large blue blob in the bottom row), correlates with the “magnitude” of ego-dissolution. The PCC has actually been studied a fair amount in the last 5 to 10 years, in the context of awareness and conscious experiences of the world. Interestingly enough, PCC is more activated during self-related thinking and decreased during meditation.


The “LSD experience” connects resting state networks
For the previous two results, when the activity of one area of the brain is coupled with that of another, we said they were functionally connected, and you could then say that they were elements of a “network”, since they are, effectively, in sync, albeit we do not know the direction of interaction between the two areas. Taking that a step further, we can look at all the areas of the brain simultaneously, and segregate brain regions into distinct networks based on the similarity of their activity. For example, if areas A, B, and F are co-activated, while C, D, and E are co-activated, we would call these two different networks with three elements (or nodes) each. This is exactly what resting state networks (RSN) are. Basically, when people lie in the MRI scanner and do nothing, somehow, about 10 or so brain networks consistently show up, each with roughly the same functionally connected areas between different people. Thus, they were dubbed RESTING STATE networks. Nobody really knows for sure what they do or why they are there, but there are some ideas such as a network that keeps visual vigilance even when you’re not looking at anything in particular, such that when threats do appear, you have time to react. Others, such as the default mode network, are related to self- referential thinking that inevitably happens when you are told to lie in a tiny tube and do nothing. My favorite aliases for these are, for example, the “thinking about my lunch” network and “did I lock my door” network.

In this last figure, then, we see the activity of these 12 resting state networks (purple blobs) change during the experience of LSD. Firstly, only the 3 visual RSNs saw slight increase in blood flow (CBF, red bars) under LSD compared to control, once again showing that LSD does not actually increase neural activity. On the other hand, all 12 networks saw consistent decrease in network integrity , meaning the activity of members (brain areas) within each network became less similar under LSD (blue bars). Finally, the activity of different RSNs became more like each other, i.e. decreased segregation , as shown in the last checkerboard plot where each tile is the increase in activity similarity between the network in that row and column (diagonals blank since the same network is always exactly similar to itself).

All in all, the conclusion is that while tripping on acid, the visual networks are more activated and “more connected” to other brain regions, particularly in the context of visual hallucination, while these resting state networks often associated with thinking about self and rumination are “more dissolved.” I think that this is a great study (“Higgs Boson of neuroscience” might have been overstating it) and a good first step towards legitimizing drug-related brain research, one avenue that we are really not taking advantage of. Given that these people are experienced users, we’re not introducing additional risks by having them do drugs while getting their brain scanned, and it can reveal a lot of interesting things that brains under normal conditions simply cannot.


Science Communication and Journalism
With the study out of the way, I want to take this opportunity to comment on pop science. I’ve been wanting to write something about it for a year and half now, but never found the “right” thing to dig into, but since I already went through the science behind this particular study, I might as well take the opportunity to look at how news articles deliver. As always, there is a diversity of nonsense, depending on which news outlet you’re looking at. One example, headline from The Washington Post: “ Groundbreaking images show how LSD literally lights up your brain. “ While the article itself has some redeemable qualities and insights, this title is the kind of thing that makes me want to strangle a journalist. First of all, what the fuck does it mean to literally light up the brain? It certainly does not refer to the dramatic increase in neural activity under LSD, because there was none. My only interpretation for this is that this particular author saw Figure 2 in the paper and literally transcribed his LITERAL interpretation of those figures. The problem, I believe, arises from a lack of background context and formal training. I’m not trying to be a snobby educated person or anything, but those images will show different things to a layman vs. someone who has even the least bit of knowledge about brain imaging, such that the latter would never think to describe it as “lit up”. And I don’t mean to rip into this one person in particular, it happens time and time again, and every time I see something like this I just experience a cycle of anger, disgust, reluctant acceptance, and fear that someone will “learn” something from this article. Again, I don’t necessarily believe that it’s one person refusing to do their job properly, but that they lack the proper training to interpret neuroscience studies.

This problem is better highlighted in an article that’s not such a blatant offender. The Guardian article, for example, was great. It did not stretch the interpretation of the study, and generally described the results in line with how the authors did in the paper itself. But after a close reading, it is apparent that it is only that: an accurate recount of the findings in the study that skip details whenever possible, and with little actual understand of the topic at hand. For example, “functional connectivity” means a very different thing to a fMRI scientist, because it does not imply actual connectivity, as I outlined above. But for a lay person exposed with the basics of the brain, connectivity in the context of neuroscience probably means physical connections, i.e. synapses. In other words, I wouldn’t put it past someone to interpret this finding as “LSD increases brain connections.” Also, posting the figure without a detailed explanation of what the colors and blobs mean is a dangerous thing to do, because then it allows any one person’s literal interpretation of the image with the apparent validity of “science”, such that you get shit like “LSD lights up the brain”. This is not a hypothetical situation, I know because I also came to my own interpretation of the image after reading the news article (maybe I’m just not critical?), which was promptly corrected after reading the actual paper.

So why is this a problem? Well, on a smaller scale, it’s not. Worst case scenario is that some people are slightly misinformed about the effect of LSD, and you might say that the people who will get it wrong don’t matter because it’s just a light bathroom reading for them, and the people who will get it right will get it right anyway because they don’t base their perception on secondary reporting in news articles (though let’s be honest, who’s going to read the actual paper?) This assumes a two-tier population in terms of scientific understanding, which leads us to think about who these news report of science are targeting - and I think this gets at the actual problem. The reality is, there is a spectrum of lay people in terms of their engagement with science, and depending on what field you work in, even scientists get relocated on this spectrum. By that I mean, for example, I will read neuroscience papers for their findings and implications but I get most of my understanding of new breakthroughs in physics through the news app on my phone, and not the primary sources themselves, because I don’t have the understanding required to evaluate the study in its raw form.

I think to accurately describe this situation, we need a model with three groups of people. The first group is the “scientists”, people that care tremendously about the details of a particular reporting and will make informed choices based on what they read. Most people that fall within this group are probably actual scientists and graduate students, but lay people who are particularly interested in a line of research for whatever reason also fall in here. The second group is what I would call one-and-doners: people who read these things purely as a past-time and does not take any of this into their actual lives. Classic example: your grandma. These first two groups are not interesting, because they are the people who will get it right anyway and the people for whom it really doesn’t matter.

The third group is what I would call “armed and dangerous”, and given the dramatic increase in our ability to access information via the internet, this is a growing group of concern. Basically, “armed and dangerous” means that you are educated enough to understand the information, perhaps because you have an education in sciences or a related field, and you WILL make life decisions based on it. Not only do these people have the power to change their own lives, they have the power to influence people around them and even society as a whole via voting and lobbying and protesting. For these people, they will trust the news articles for what they are, perhaps because they don’t have the ability or desire to read the actual papers (like me with physics). In this case, it’s absolutely crucial to convey the information in the most accurate way possible, and “increased connectivity” vs. “increased functional connectivity” vs. “the brain lit up” all have very different outcomes for these people. In the context of LSD, it probably doesn’t really matter, because it’s quite a niche thing in people’s daily lives. But if we look at other health-related research that can potentially alter someone’s lifestyle and behavior, such as marijuana research, nutrition research, and, oh god, vaccine research, conveying information accurately becomes very, very important. It’s the same thing as politics, for democracy to work, your voting base should be as informed as possible. While this is proving to be more and more untrue as this clusterfuck unfolds in the U.S., I would say it’s even more important for the voting base to be informed accurately regarding science.

This, then, becomes an engineering problem for science journalism: how do you write for the lay-but-educated population, for whom actionable outcomes are a very real possibility, while still keeping the content engaging so information is disseminated and the company gets paid? Well, for one, maybe it helps to have semi-experts that can outline the important assumptions and interpretations of a particular study. Grad students and neuroscientists all know, for example, that any fMRI research comes with the invisible disclaimer in the first line that says “we don’t actually know what we’re measuring but it should be fairly representative of actual brain activity”. But regular people that have other things to do in their lives don’t know that. And if, somehow, in a weird parallel universe, a particular health policy hinges on whether LSD makes actual brain connections or “functional connections”, then it’d be damn important for people to know what that actually means.