A little bit of a few things.

Wow, more than a month since my last post. I’ve been a bit braindead as I try to finish up my thesis. I’m going to be continuing in my current lab next year, only I’ll be doing my PhD. So that’s pretty exciting.

I’ve seen a couple interesting talks in the past while, and figured I’d post on a couple of them, briefly.

Imaging and intelligence: The presenter is a big g theorist, which given that this appears to be the current consensus isn’t such a big deal, but see a list of discussion links here and an excellent critique of the origins of g-factor theory. I need to look up my notes, but essentially the presenter’s group used various tasks and controls to find activation via fMRI for what they claim to be g in various brain regions, including that magical region, the IFG.

I went to a workshop/seminar on rhythm in music and speech. It was neat, and there was some discussion of comparing musical compositions to speech. The main problem I saw was that a lot of the studies seemed to compare compositions to spontaneous speech, or novel reading aloud. The two aren’t really comparable. A better comparison would be a composition with a composed and practised speech, or improvisational jazz with spontaneous speech. There was a talk after, where they discussed the use of tabla drumming as a speech-like code. The cooler part was a discussion of language-specific differences in supposedly universal aspects of acoustic perception. Apparently, when humans hear a continuous stream of tones which proceeds long-short-long-short-long-short-long-short… we order it short-long, rather than long-short. And when we hear a continuous high-low-high-low-high-low… we parse it as high-low rather than low-high. And this is thought to reflect inherent biases of auditory processing. But the speaker’s group found that this is true in very young Japanese and Canadian infants. But by the time phonemic pruning occurs, infants begin to acquire language-specific biases, and so Japanese infants will hear long-short and sometimes low-high.

Then there was a cool talk on emotion and speech. This one got me thinking, since it dealt with the minimum amount of an utterance you need to hear before you can accurately identify the emotion being expressed, assuming the words themselves are neutral, or if the sentence uses pseudowords. But I’ve been thinking that interjections like “Ugh!” or “Argh!” bypass this to some extent, and may also be culture- or language-specific. It’s part of a broader interest I’ve been developing in socially stereotyped behaviours, which posses a kind of social exemplar. We can all picture, and even imitate, a stereotypical sneeze, and I think this cultural idea of a ‘sneeze’ actually shapes how people sneeze. The same for laughter, stubbing our toes, wiping our eyes when crying… I think this idea of stereotyped behaviours is important, and could be studied in the context of verbal or gesture+speech communication as an efficient communication code or cipher. I need to bone up on my ethology.

I went to a comps presentation on universal grammar and connectionist accounts of language transfer. The speaker pointed out that neither camp makes sufficiently different predictions here for either to be falsified. I’m still sort of amazed that there are still UG people around, but I guess the theory has some explanatory power.

Monday there was a talk on detecting white matter activation in fMRI. I’ll explain sometime why this is generally treated as improbable, but essentially while there’s a good explanation for why we see changes in blood oxygenation levels coupled with grey matter activity, there’s no real explanation for what it would mean to see the same changes associated with white matter.

Tuesday I went to CRIUGM to see a talk on machine learning applications of multivariate pattern analysis in resting-state fMRI (where the participant does nothing except be scanned) and real-time fMRI. I need to go through my notes and do something more thorough, but it was pretty exciting, and showed some ways that we might eventually be able to combine fMRI with real-time conversations, and note relevant activations with specific parts of the discourse.

Alright. That’s enough for now. Things are coming together, so hopefully I’ll get back to posting more regularly.

Thesis, in brief.

So it’s been a while since I went on about science per se, and longer since I went on about my own field of speech neuroscience.

Part of this has been my working on my thesis. I’ve been sort of buried in data and discussion and haven’t had an easy time breaking it down in my head, into a simpler form. And my thesis is looking more like a journal article than a thesis, so I’ve got plenty of work ahead of me.

For those of you who don’t know what I’m doing, here goes. This is at once more general, but much more descriptive, than my previous update.

I’ve got two imaging studies to work with, both older studies that were finished before I joined my current lab. In one, participants were put in an fMRI scanner while they viewed a picture, a word, or heard a word spoken. In each case it was an object, like “bread”. Half the time they just perceived the word passively. The other half of the time, they spoke the word aloud after perceiving it.

In the second study, participants were in the scanner while they watched a video of someone speaking a sentence. In some cases they saw a still image of the person’s face while the audio track played; in some cases the video track played silently; and in some cases they saw both together. Also, in some trials they had to decide if the speaker was angry or happy, and in others if the speaker was asking a question or saying a statement.

There’s a couple places I’m going with this. Firstly, I’m looking at how the brain activation during the listening condition in experiment #1 and all three modalities (auditory, visual, audiovisual) in experiment #2. Here, I’m trying to see if and how much perception activity overlaps with production activity, and whether it stays constant when you go from words to complete sentences. This is motivated partly by that motor theory of speech perception I talked about a while ago.

After that, I’m going to look at whether frontal or motor regions in the brain are more active for auditory, visual, or audiovisual perception. Another group of researchers has the idea that when you see someone talking, the visual information is sent on to motor regions involved in speech production. The brain then matches the gestures you’ve seen and the sounds you’d expect to come from those gestures if you’d made them yourself to the auditory speech information coming in your ears. This is known as hypothesis-and-test. And it predicts that speech that contains visual information should activate frontal and motor regions more than speech you just hear. So I’m also looking at that.

Lastly, if neither of those are confirmed, I’m going to look at activation for each modality in experiment #2 and try to explain how the activity we see is actually heavily dependent on both the task at hand and on the modality being used. A lot of motor theories suggest that activity during speech perception should be very general – that is, it shouldn’t matter much what sort of listening you’re doing, or what you’re thinking about doing while you’re listening. They often seem to say that there are certain regions which are always involved in speech perception, and part of my thesis will suggest that this is not, in fact, the case. Our lab generally thinks that activity is very dependent on the specifics of what the participant is being asked to do.

So far, that looks to be true.

Endogenous cortical rhythms.

So I read an interesting paper today, from a couple years back.

There’s this idea, called Asymmetric Sampling in Time (AST), which posits that auditory input is processed differently by the left and right hemispheres. The left hemisphere, it is thought, processes auditory input at a high sampling rate, over 25-50ms windows. This is long enough to perceive, say, a single phoneme like /p/ or a short syllable like /da/. The right hemisphere is supposed to process auditory input over longer time windows, of around 150-300ms. This allows it to better process pitch variation, like you get in music or verbal prosody.

It would explain a few things, like the general tendency of the left hemisphere to be more active during language processing and the right hemisphere to be more active during music processing. It would explain why the left hemisphere seems more sensitive to temporal (timing) information, and the right hemisphere seems more sensitive to spectral (frequency) information. Note I said general – the right hemisphere is definitely involved in speech and language, and the left certainly contributes to music perception. The point is that it’s not balanced – the contributions are asymmetric.

Now, as I’ve said, the idea is appealing because it has some explanatory power. It makes sense of a few things that are confusing, explains some findings, and is in general pretty nifty. But it’s not very tested. The originators of the idea (to the best of my knowledge) have written a couple papers on it, but it hasn’t been until the last several years that people have actively investigated the implications of the idea rather than suggested how it might explain existing findings.

Which leads me to the paper my prof sent me today. Essentially this is what they did. They stuck some people in an MRI, but also put electrodes on their scalp. This is pretty critical – as I’ve mentioned before (I think), MRI is very good for localization, but you can only detect changes over a matter of seconds, not milliseconds, which is much too slow for AST. Electrodes can detect very fast changes in brain activity, but because of the skull and some issues with physics the signal gets smeared out and you can’t really tell where the changes are occurring. Combining EEG and MRI is basically the best of both worlds, but I hear it’s a huge pain to process the data. So kudos right there for grinding through the data processing.

The nifty part is they didn’t do anything in there. They weren’t so much interested in looking at a specific task. Your brain is active all the time, to varying degrees, even if you aren’t doing anything in particular. That’s the ‘endogenous’ in the post title – the brain activity they were looking at (on the surface of the brain, so ‘cortical’) wasn’t provoked by any particular stimulus, it was just the normal sort of activity.

So they did that for about 40 minutes. This feels sort of like a cooking show – more evidence that ‘The Magic School Bus’ was a good guide to science.

There’s a cool thing you can do with EEG. Rather than looking at the up-and-down wave patterns recorded from each electrode, you can look at the frequency of the waves, and see which frequency dominates at each electrode, or at certain times. Now, a time window of 20-50ms is approximately covered by activity in the 20-50 Hz frequency range, and a window of 100-300ms by around 3-6 Hz. One hertz, by the way, is one cycle per second. So you can see there’s about 20-50 cycles per second to process something that occurs over 20-50ms, and about 3-6 cycles per second for things that occur over a couple hundred milliseconds.

Essentially, what they then did was combine the EEG ‘power’ or ‘spectral’ data with the fMRI location data, so they could see where in the brain the different frequency patterns dominated.

Now, their findings aren’t quite as supportive of AST as I think the paper tries to make them out to be, but the findings support a nuanced version of the same idea. Slower, low frequency rhythms dominate in primary auditory regions on the right, and faster high frequency rhythms dominate on the left. Now, since I’m interested in motor systems and speech/audition, I found some of their other findings particularly interesting.

Firstly, there’s definitely some bilateralness in frequency dominance. There are slow areas on the left, and fast areas on the right. But motor regions and inferior frontal regions in both hemispheres appear to be dominated by fast rhythms, according to their figures. And for further niftiness, it looks as though the motor regions more involved with the mouth are dominated by slow rhythms, and those involved with the tongue are more associated with fast rhythms. This makes some surface sense, since tongue movements are typically more extreme and faster than jaw movements, and slow rhythms correspond to full syllables. There’s a suggestion that syllable length fits with jaw movement rhythms, so again, the findings make some sense.

Now, the findings might be slightly more ambiguous than I’ve been discussing them, but science is always a bit fuzzy. It was a neat paper, and it’s given me some food for thinking on my own ideas.

Giraud AL, Kleinschmidt A, Poeppel D, Lund TE, Frackowiak RS, Laufs H. Endogenous cortical rhythms determine cerebral specialization for speech perception and production. Neuron. 2007 Dec 20;56(6):1127-34.

Proposal, learning.

So, between working feverishly on my thesis proposal and my exam, I’ve been a bit pre-occupied this week.

The proposal’s come along well. I think I’ll like it more when I’ve fleshed it out into a full thesis. There’s some sections that could be laid out in more detail, and I skipped over some good papers in order to make my main points in a decent amount of space.

Essentially I’m taking a conjunction of three speech production tasks, done at the word level; extracting that conjunction as a region-of-interest; and then using that functional ROI to look at three speech perception tasks, which involved viewing either auditory, audiovisual, or purely visual recordings of someone speaking a sentence. They’re mostly abstract, non-action sentences, though I still need to double-check them for that. My hypothesis right now is there will be motor activity during perception of higher-level speech, in the same regions as during production of simple, single nouns. If there is, I’m predicting there will be more activity during the audiovisual and visual tasks than the purely auditory ones.

I’ve been pretty influenced by two papers in particular, Skipper et al 2007 and Tremblay & Small 2010. When it comes time to my thesis though, I’m going to broaden my background and will probably end up referring to the dorsal-ventral path theory put forth in Hickok & Poeppel 2004.

In related news, I spent much of Thursday and all of Friday teaching myself how to run a structural vector autoregression connectivity analysis using 1dSVAR, a program written in R for use with AFNI. No one else in my lab knows how to use it, and there’s not much of a manual for it, so I’m pretty pleased with myself. I still need to figure out how to interpret my output, and to double-check my input, but I’m on my way to being able to do another type of fMRI analysis, which is pretty cool.

On walking back.

So I was in the pub with some lab mates, and we were talking mirror neurons. I’ve cooled a bit on them since I first read me some Pulvermuller, Rizzolatti, Arbib, etc, but I still like the idea of them. Part of the problem is since people like the idea SO MUCH, they’ve invaded things from neuroscience, to speech, to dodgey autism treatments, to group psychotherapy, to aesthetics.

The point is, while there is evidence of observation-execution paired neurons in macaques, claims beyond that should be made pretty cautiously. Mirror neurons are used in speech research because of an initial study that showed some neurons activated while listening to the sound of the action which execution also activated them. It provided an interesting mechanism for the motor involvement in speech perception, and revived motor/gestural theories of speech perception.

Which I think was a good thing. But it’s got a bit out of hand. Only one study can claim to be strong evidence for the existence of mirror neurons in humans, and there’s definitely dispute over it. But there’s a lot of evidence to suggest that there are regions of the brain which show activation during perception and production of the same tasks.

So I’m willing to accept, provisionally, that there may well be mirror neurons in humans brains. Even Greg Hickok, definitely not a supporter of the mirror neurons in speech crowd, says he’d be surprised if turned out that there were no mirror neurons in humans. So we’ll assume that, somewhere in the inferior frontal gyrus or the vental premotor cortex, there exists some population of neurons that respond both to observation and execution of the same task or gesture.

Now, there’s two major streams of research involved in mirror neuron theories of speech. These are sometimes easy to confuse, and given that the same groups often pursue both avenues it can be tricky for newbies like myself to untangle.

In one, mirror neurons are thought to provide a basis for action understanding. Because they fire when you see someone reaching out as well as when you reach out, they seem like they could be a mechanism for understanding that you move, and so do other people (I’m not working from the papers here, mostly memory, so bear with me if I’ve muddled it a bit). Part of this is the mirror system hypothesis, which is basically a evolutionary theory of language, which suggests that speech is built on an earlier neural system for gestural communication. Arbib’s the big name here, and to his credit it’s an interesting theory. I have some issues with it, but I won’t go into them now.

So this research has focused a lot of things like action sentences, or action verbs, which are related to hands or feet or mouths, the theory being that an ‘action understanding’ of language would cause those regions in the motor cortex to light up while listening to those words. There’s a lot of studies where people see if reading those words, or hearing them spoken, facilitates activation of those regions of the motor homunculus (using TMS) or causes increased neural activation in those regions (using fMRI). Technical issues aside, it’s difficult to say that this is facilitating understanding or comprehension – the best I’ve seen was a study showing faster response times to action words when the relevant part of the motor cortex was stimulated.

The other, and I think more likely avenue of inquiry, has been looking at potential mirror neuron involvement in phoneme perception, particularly comparing labial sounds like /pa/ to dental sounds like /ta/ or gutterals like /ka/. The evidence here is, to my mind, more compelling, particularly that there is a role for mirror neurons in multi-modal integration of phonemes. The link there goes to one of my favourite papers, and if this has been at all interesting I suggest reading it. Other researchers have found evidence for mirror neuron involvement in masked/noisy speech, or non-native speech, so the idea of motor involvement at least for difficult or demanding speech perception is increasingly uncontroversial.

Really, though, I’m beginning to think that people should be cautious about using the term ‘mirror neuron’. I think most of the time, the best you can say is ‘motor contribution’ or ‘speech production region involvement’ in speech perception. Mirror neurons just seem like too specific a claim to be making, particularly using fMRI, when you see an output region activating during an input task. And they come with a lot of baggage these days, so I’m keeping that in mind as I work on my thesis.

HSDD and problems in neuroscience.

So I had something I’d wanted to post on, and now (frustratingly) cannot remember what it was.

Instead, I’ll link to these posts on a recent study purporting to study neural differences in women with and without low sex drives. The whole issue really strikes at a few major issues in neuroscience. And, unfortunately, it’s been widely covered in the Telegraph as Women with low libidos ‘have different brains’ and the BBC as Libido problems ‘brain not mind’.

1. Any study needs to have a good idea of what exactly it’s studying. There’s some sketchy history over hypoactive sexual desire disorder. It’s not disputed that people might have low libidos and be frustrated over this, but as Dr. Petra Boynton points out, many of the screening tools are too vague to be reliable indicators of what’s being supposedly measured ie, bothersomely low libido. So even the diagnosis is pretty vague.

2. The groups were 19 women ‘with’ HSDD and 7 without. Now, having twice as many people in your clinical group as your control group is pretty sketchy, and you wouldn’t think it’d be hard to get more controls – hell, most of the time the problem is too many people in your control group and too few in the clinical. And in another study done by the authors of this study, looking at themes to include in research-purposed erotic clips, they do not even take into account the sexuality of their participants.

3. This isn’t really peer reviewed. I think there are lots of problems before this even gets to peer review, but it when it was put about in the media it hadn’t even been published, only presented at a conference. Such findings usually reflect ongoing research which has not been finalized or thoroughly reviewed, and as such shouldn’t be considered authoritative.

4. Not that you’d know that by the researchers themselves. Here’s Dr. Diamond, the lead researcher: “Us being able to identify physiological changes, to me provides significant evidence that it is a true disorder as opposed as opposed to a societal construct.”

Now, at pointed out here by Julian Savulescu, a neuroethicist, this is bullshit. The study found different patterns of brain activation. Well bully for them. Say some women complain of low libido, and others don’t. Put them in an MRI, and they will have different patterns of brain activation. Say some women complain of chronic boredom, and others don’t. Put them in an MRI, and they will have different patterns of activation.

Why? Because mental states are the same bloody thing as physical states, and I strongly question the capability of any neuroscientist unable to grasp this. Social constructs create different mental states -> mental states = brain states -> brain states are physical. Bam, there you go.

All this report shows is that there is a difference in mental processing between people in one construct and people in another – or as has been pointed out, people who might be tired, sick, feeling incompatible with their partner, or any other number of things. Perhaps what they found was simply that sexually satisfied people process erotic images differently than people who aren’t.

So big deal. What do the findings -mean-? From the excellent deconstruction by the Neurocritic, quoting the report:

“…women with normal sexual function had greater activation in superior frontal and supramarginal gyri. Women with HSDD exhibited greater activation in the inferior frontal, primary motor, and insular cortices. Additionally, normal women had greater activation in the posterior cingulate cortex while women with HSDD appeared to recruit the midcingulate region.”

And yet there seems to be no discussion about just why those regions have differential activation. Different brain regions have broadly different functions, and any study looking at something like this ought to at least try to explain their findings in some coherent way. Simply saying ‘different patterns of activation between groups X and Y’ is a cop-out.

There’s two final points I’d like to make.

One, go read Ben Goldacre’s take on the issue of ‘neuro-realism’, whereby people feel subjective mental states, like fear or pain, need to be validated as ‘real’ through brain imaging. Of course they’re real – if they weren’t people wouldn’t feel them. There’s no such thing as something that’s ‘just in your head’. If you feel it, it’s real. This isn’t talking about facts – if you ‘feel’ Atlantis was real you’re wrong – but internal states of being. And all neuroimaging can do is localize regions involved in these states, not somehow validate their existence.

Secondly, and I can’t emphasize this enough, is that scientists in general and social scientists in particular need to take responsibility for their research. Anything involving humans needs to be approached carefully. Science rightly has a good deal of authority, and things which are glossed with science (much in the same way as pseudoscience) are often accepted uncritically, particularly when they confirm existing attitudes. I would suggest, given how counter-intuitive scientific findings can be, that any findings which seem to confirm existing social rules or attitudes should be strongly scrutinized. Even if your research isn’t pseudoscience, and is conducted with the best of intentions, you need to look very hard at how you’re asking your question, why you’re asking it, how you will interpret your findings; and in this day and age, how the media will interpret your findings.

After all that, as this nonsense shows us, your findings can still be reported as confirming social attitudes even if they don’t.

T. L. Woodard, N. T. Nowak, S. D. Moffat, M. P. Diamond, M. E. Tancer, R. Balon. Wayne State University School of Medicine, Detroit, MI. CEREBRAL ACTIVATION PATTERNS IN WOMEN WITH HYPOACTIVE SEXUAL DESIRE DISORDER (HSDD) VERSUS WOMEN WITH NORMAL SEXUAL FUNCTION. American Society for Reproductive Medicine’s 66th Annual Meeting.

An exciting life

The discussions I had in my lab today:

Q: Should you align your functional images to your anatomicals, or vice-versa?
A: Anatomicals to functionals, since I’m looking at activity patterns.

Q: Why the hell is everything misaligned in the GUI now?
A: Who cares, once everything’s been aligned to standard space it won’t matter.

Q: Should you warp your images to standard space before or after convolving your hrf with a piece-wise linear spline and polynomial baseline model with autocorrelation matrix?
A: Further argument. Warping your betas afterwards means there’s less interpolation of data before the analysis, but doing it before makes things easier when doing lots of group comparisons. The argument is left unresolved.

Q: Should I be taking functional ROIs computed from a conjunction analysis of another study, or paint anatomical ROIs on my individual subjects.
A: Man, that last one sounds like a lot of work. We’ll try the first option for now.

It was actually all pretty exciting. I’m actually getting to know enough to have these discussions.

Speaking and listening – interbrain communication

So I’ve been writing a lot of ‘these things bug me’ posts lately. I figured I’d put up something cool today, and save my latest rant for another post.

One of my profs sent me a cool new article on speech imaging. It’s a study conducted at Princeton, by Stephens et al.

One of the big topics in speech research is communication. I know this seems obvious, but there are a lot of studies devoted to how people perceive speech – how phonemes make words, how words have meaning, how words fit together to make a grammatical sentence; and to production, with how people pick appropriate words, pronounce the phonemes, construct grammatical sentences, and convey meaning through prosody. Already I’ve left a lot out, and made a bunch of assumptions about hotly debated topics.

Now, my lab is big on the motor theory of speech. That’s a topic for another day, but suffice to say we are very interested in how people go from perceiving to producing. Because speech is not static. We hear things and need to be able to respond very quickly, almost at the rate of unconscious action. We only speak as though we were picking words out of a lexicon and parsing them syntactically when we’re speaking an unfamiliar language, have some injury or disorder, or are in a debilitated condition. That last is why I’m holding off on a delicious whisky sour until after I’m done writing this post.

There are a bunch of ideas explaining in different way how we are able to communicate so effectively, seamlessly picking up our side of a conversation in real time, and following along with a spoken narrative even under noisy conditions or confusing accents.

Stephens et al recorded functional MRI from a participant who related an unrehearsed story in English, about an experience during a high school prom. They also recorded a similarly unrehearsed personal story from a native Russian speaker, also while recording functional MRI. This was done a few times to acquaint the participant with the procedure, and then a single story for each speaker was picked.

MRI machines make a lot of noise, so they used a clever technique with two microphones that allowed them to remove the scanner noise from the story in post-processing, without making the speech sound artificial. I expect we’ll be seeing this technique more often, as it’s useful to be able to record verbal responses.

After all that, they played the story to a bunch of listeners, and recorded their brain activity using fMRI while the listeners listened to the stories in both Russian and English.

Now, here’s the cool part. Using a moving window of -4 to -1.5 to 0 to 1.5 to 4 seconds, they compared brain activity in the speaker to brain activity in the listeners during the entire length of the story. Afterwards, the listeners were tested on their recall and comprehension.

There was, excitingly, a large overlap between active regions in the speaker and the listeners. This is cool because a number of theories of speech postulate that the same regions involved in producing speech are also involved in perceiving it. Further, they found that much of this activation in the listeners lagged behind the speaker by a few seconds, implying that there was a causal relationship – that the regions active in the speaker were, via the speech, causing the same regions to become active in the listeners. In this way, a spoken narrative can be thought of as a method for one brain to induce specific activity in another (see my previous post on us being our brains).

But even cooler, was a few regions which in the listeners were active a few seconds ahead of the speakers. There were mostly frontal regions involved in planning and other executive functions. The authors suggest this is due to anticipation on the part of the listeners, a sort of heuristic predictive action to facilitate comprehension. Indeed, those listeners which had more activity in advance of the speaker’s showed improved comprehension and recall after the experiment.

Furthermore, this was all dependent on actual understanding of the story. The only speaker-listener coupling during the Russian story was in the primary auditory regions, implying that while the listener’s brains were being activated by the sounds being heard, there was no further processing specific to the story.

Even more fun than that, when they asked the speaker to relate a new story while being scanned, and compared their brain activity patterns to this listeners from the original story (to see if there is simply a regular spatial-temporal pattern of activity while relating and listening to a narrative), they found very little activity coupling. It turns out that while the generalities of activation are there, actual activity coupling only seems to occur during shared communication.

I really look forward to seeing these results replicated. The actual analysis sounds a little tricky, but the experiment itself is simple and elegant, and if the results bear out it has some interesting implications for the study of coupled speech perception and production.