Postdoctoral fellowship opportunity at UCSD Center for Research in Language

For any of you neurolinguists out there, the University of California’s Center for Research in Language has a postdoctoral fellowship opening in its training program entitled “Language, Communication and the Brain” for the 2012-13 academic year. This training program emphasizes new technologies and new theoretical frameworks in cognitive science and neuroscience. They are looking for scientists interested in the mental and neural mechanisms that underlie language learning, language use and language disorders.

The application deadline is April 23^rd. Since it is an NIH-funded position, the recipient must be a US citizen or permanent resident.

Detailed information and application instructions can be found at the following website:

http://crl.ucsd.edu/training/fellowships/postdocannounce.php

 

The Center for Research on Language is a great community to be a part of and is well connected with UCSD’s more general world-class neuroscience and linguistics communities.  It’s a great opportunity.

Emily Dickinson on Brains

Everyone should have at least one joke memorized for emergency situations.  One poem is probably worth burning into your brain as well for the occasional apropos moment.  For you neurogeeks out there, perhaps this could do the trick:

The Brain — is wider than the Sky –
For — put them side by side –
The one the other will contain
With ease — and You — beside –

The Brain is deeper than the sea –
For — hold them — Blue to Blue –
The one the other will absorb –
As Sponges — Buckets — do –

The Brain is just the weight of God –
For — Heft them — Pound for Pound –
And they will differ — if they do –
As Syllable from Sound –

-Emily Dickinson (1830-1886)

John Baldessari’s painting "Brain/Cloud"

Video lectures/tutorials on fMRI/MRI methods

I started working with MRI/fMRI data for this first time this summer and haven’t had the opportunity yet to take an official training course in MRI/fMRI methods. Thankfully, at least a couple of groups have made videos of such workshops freely available online. The most comprehensive set of videos I know of is from the UCLA Advanced Neuroimaging Summer Program:
http://www.brainmapping.org/NITP/Summer2011.php

introduced to me by Miklos Argyelan at the Feinstein Institute for Medical Research.  The videos cover everything from MR physics and basic experimental design to more contemporary techniques such as network analysis, genetics, and pattern classification.

Recently, the folks at the Wellcome Trust Centre for Neuroimaging have placed online a series of videos as well:
http://www.fil.ion.ucl.ac.uk/spm/course/video/

The videos are from a course for using their SPM MATLAB-based software for neuroimaging, but the topics (e.g., general linear models, dealing with a large number of statistical tests) should be informative regardless of what software package you use.

A tip of my hat to the speakers and everyone else responsible for making those videos and placing them online. It takes a good amount of work to make that happen but it’s a great resource for the cognitive neuroscience community. I hope such resources are developed further (more exercises please) and that other groups follow their lead.

Why I am not a naturalist

Philosopher Timothy Williamson published in NY Times an essay titled What is Naturalism? I find it unsatisfactory for many reasons. I want to share here my thoughts about the method of science in this context.

Williamson defines naturalism as a position which maintains that “there is only the natural world, and the best way to find out about it is by the scientific method.”  This statement is meaningless unless “natural world” and “the scientific method” are defined clearly. He characterizes the scientific method as:

For naturalists, although natural sciences like physics and biology differ from each other in specific ways, at a sufficiently abstract level they all count as using a single general method. It involves formulating theoretical hypotheses and testing their predictions against systematic observation and controlled experiment. This is called the hypothetico-deductive method.

If that is what he means by the scientific method, I am not a naturalist. I don’t think the method described above is sufficient for studying nature. What’s described is so vague and sketchy that it hardly counts as a method. According to this description, this is what a naturalist says: “The scientific method is great. It’s the best way to find out things about the nature world. How do you do it? You first come up with a hypothesis, and then you test it! If you do this a lot, you’ll find out a great deal about nature.” Duh! Obviously something is left out. The problem is this: exactly how do we formulate a hypothesis? The space of possible hypotheses is vast. Without a strategy, it’s very unlikely that we’ll ever come up with any hypothesis that has any chance of being right. If all the hypotheses we formulate are bad, this method will not work because the predictions will fail all the time, and we’ll never get anywhere.

So how do scientists come up with scientific theories? My opinion is that we don’t really know. Perhaps there isn’t a thing called the scientific method. The way I see it, what we have is a collection of tricks that we use to attack the research problems that we have at hand and we improvise when necessary. Furthermore, I also believe that this bag of tricks is share by plumbers, mathematicians, engineers, programmers, game designers, movie directors, novelists, politicians, chefs… etc. This would disqualify me as a naturalist because I doubt that science has a fixed method. My intuition would have to be researched by sociologists and cognitive anthropologists and I don’t insist on it. However I want to emphasize that scientists’ bag of tricks does includes mathematics. This is a point I think is worthy of noting because Williamson claims that the existence of mathematics posts a problem for naturalism. His argument, as far as I can tell, is this: How can the scientific method be the best method? Isn’t math equally good or even better? I don’t see how this can be a problem if we count math as part of the scientific method. Williamson says this move is problematic because if we count math as a science, “the description of scientific method just given is wrong”. No. It does not follow. His description of scientific method is not wrong. It is just sketchy.

I believe what Williamson tries to say is that science is justified by empirical results whereas math is justified by deduction (ie. axiomatic proofs). Since these two modes of justification are so different, they don’t mix together. I disagree with it because it leaves out a big chunk of theoretical sciences*. A lot of work in theoretical sciences is deductive. I want to bring up one particular example that I am more familiar with. I am sure that physicists can give much better examples but I want to show that even in biology and psychology, two branches of science that are only loosely integrated with math, there are deductive studies. In fact, there are a lot.

Given some knowledge about the structure of the eye and the retina, it is possible to use geometry, optics, and sampling theorems (from signal processing) to derive a theoretical limit of optical performance of human vision. The procedure is purely deductive. The result is also justified deductively. In other words, we know that the result is right because the derivation is flawless, not because of experimental verification. The theoretical limit cannot be verified experimentally because the conditions assumed in the calculation are idealistic and they are never realized by any real biological systems. However, this piece of information is important (it is actually the foundation of any vision research) because experimental biologists and psychologists can compare the theoretical values to experimental results and use the difference to make inferences about the mechanisms that contribute to the deviation. The theoretical limits are also used as baselines when we compare across different subjects or different species. This does not fit into the hypothetico-deductive method that Williamson described but it is done in psychology and biology all the time. It’s so commonplace that the practice is not reserved for theorists anymore. Even experimental scientists know enough math to do the calculations these days.

I want to contrast this type of research to the kind of work that is made famous by Einstein’s theory of relativity. Special relativity was derived almost completely deductively but it was (at least partially) justified experimentally. It was not justified purely deductively because it was proposed as a theory of physics. It is supposed to make precise predictions about nature and therefore the predictions have to be verified. On the other hand, the theoretical limit of human vision is not a theory of vision. It makes no prediction by itself (except that the performance of any biological system cannot exceed the limit). In fact it does not by itself say anything directly about nature. What it does is that it creates an artificial world. It creates a world that is simple enough that we can understand, so that we can evaluate how much of the complexity in nature is capture by something simple. I am tempted to call this practice not a science but an evaluation of science. I am even tempted to call it philosophy because it is what some philosophers claim that they do. But it is much simpler if we just call it science. In fact I suspect that it supplies the missing component of the hypothesis-testing idea because it explains how hypotheses are formulated and how theories are advanced. We start with a very simple model of the world. We compare it against nature and check what is not explained by the model, and then we try to add more complexity to the model. The model does not always make predictions. Instead it is used to evaluate what isn’t sufficiently explained.

This view of science is continuous with math. Math is essentially artificial worlds that we created. Some mathematical worlds were created to approximate nature (euclidian geometry, for example), whereas some were created to be very different from nature. Scientists like their mathematical worlds to be close enough to nature whereas mathematicians like to venture a lot farther. There is no real difference between the two.

* It also leaves out a big part of math. Mathematicians do proofs but they very seldom do axiomatic proofs. A lot of math is not axiomized. But this is a different topic.

Gnosanopsia – what?

In a previous entry about Riddoch syndrome, I mentioned that I am fascinated by a phenomenon which Zeki calls gnosanopsia - awareness without discrimination. When presented with visual stimuli, patients with gnosanopsia claim that they can see something, but they are unable to describe the visual features associated with that something, such as its color or the direction of motion,

Many questions immediately come to mind. Can we take this claim at face value? Maybe the patients are mistaken. Maybe they merely have a vague sense of something happened, but choose to describe this non-visual experience in visual terms. For example, it is entirely possible that what the patients experience is sensing the automatic motor responses of their eyes triggered by visual stimulation. Can we scientifically differentiate this scenario from gnosanopsia, supposedly a real visual experience but without sufficient content to support discrimination?

This is a case where the phenomenology of vision becomes complicated. It will take cognitive scientists and philosophers decades to disentangle the conceptual complexity. In this entry I merely want to point out that this complexity is not something only found in neurological disorders. In fact our everyday visual experience is complicated.

As an example, let’s consider the phenomenon of crowding. The picture below is from a recent review paper  (Whitney & Levi, 2011):

First, look at the lower left corner of the image. What do you see? A little boy in a green shirt. Now fixate on the bullseye symbol in the middle of the image, and then try to convince yourself that you can still recognize that little boy. It’s not that easy. You know something green is there, but it’s so fuzzy that recognition becomes impossible. It really is a very odd feeling. What is this odd feeling, this confusing sense of green things jumbled together? Is this a real visual experience or is it a vague non-visual experience? It’s very hard to say. I’d like to say it’s vague but not so vague that I can’t say it’s green. I’d like to call the experience visual but there seems to be something more visceral to it.

But maybe this is not so strange, after all. It is very well-established that visual acuity degrades with retinal eccentricity. Maybe we cannot recognize the little boy simple because we don’t have enough visual acuity to resolve the fine details about the little boy. But crowding is not just an effect of reduced acuity. When the little boy is not crowded (for example, the boy on the lower-right corner), recognition is a lot easier. As discussed in Whitney & Levi (2011), the mechanism of crowding is very complicated and we know very little about it.

I started to think about crowding in the context of gnosanopsia and blindsight because I wanted to differentiate real vision (also called phenomenological vision, visual qualia, real visual experience and so on) from merely non-visual, vague feelings. But then I realized that the two are not separate categories of things. Real vision is very often associated with vague feelings. Depth perception is another example. What is the the experience of depth? Is depth, like color, a category of qualia? It’s not very easy to say. Try the following exercise: look at whatever object that is directly in front of you, take a long and careful look, and then close one eye. Can you tell the difference? Was your visual experience changed by closing one eye? On one hand I want to say yes because depth perception is definitely lost after the closure of one eye, but the difference is so subtle that most people can’t tell the difference, at least immediately. Is the percept of depth a visual experience, or is it a non-visual, vague[1] feeling?

[1] Note that vague really isn’t the right word because depth judgement is very precise! It’s not vague at all. We’ll discuss this in a future post.

Reference: Whitney, D. & Levi, D.M. (2011) Visual crowding: a fundamental limit on conscious percept ion and object recognition. Trends in Cognitive Science, 15, 160-168.

An increase in neural activity can lead to a decrease in blood oxygenation/flow

Functional magnetic resonance imaging (fMRI) is one of the most popular measures of human brain function. It is based upon changes in blood oxygenation, called the blood oxygenation level dependent activity (BOLD) response. The BOLD response is commonly interpreted as a correlate of neural activity, based on the premise that greater neural activity leads to increased blood flow/oxygenation (increases in blood flow lead to increases in oxygenation — Fox and Raichle, 1985). Although there is some evidence to justify this premise (e.g., Logothetis, 2002), an experiment from Anna Devor and colleagues (2008) shows that this is not necessarily the case. Using anesthetized rats, Devor et al. measured the electrical, hemodynamic, and metabolic response of forepaw stimulation in primary sensory cortex. They found that forepaw stimulation increased electrical activity and energy consumption in both hemispheres. However, while blood flow/oxygenation increased in the hemisphere contralateral to the paw that was stimulated, blood flow/oxygenation showed an overall decrease in the ipsilateral hemisphere (see figure). Although Devor et al. were using spectral and laser spectral imaging to measure blood flow and oxygenation and not fMRI, fMRI experiments in humans have shown that contralateral and ipsilateral hand stimulation also lead to an increased and decreased BOLD response, respectively (Hlushchuk & Hari, 2006).

Devor and colleagues are not sure why the relationship between blood flow/oxygenation varies depending on the side of stimulation. They speculate that the activity of inhibitory neurons may release signals that promote blood vessel constriction while the activity of excitatory neurons do the opposite. This could explain their results as paw stimulation leads to increased activity of inhibitory neurons in ipsilateral sensory cortex. Another explanation they suggest is that astrocyte activity in cortex ipsilateral to stimulation may lead to reduced blood flow/oxygenation. Regardless of the mechanism, their findings clearly demonstrate that the relationship between neural activity and BOLD is more complicated than a simple correlation with metabolic rate or electrical activity, which complicates trying to relate fMRI data to those of direct measures of brain function like EEG/MEG (e.g., Dale et al., 2000)).

 
Dale AM, Liu AK, Fischl BR, Buckner RL, Belliveau JW, Lewine JD, & Halgren E (2000). Dynamic statistical parametric mapping: combining fMRI and MEG for high-resolution imaging of cortical activity. Neuron, 26 (1), 55-67 PMID: 10798392

Devor A, Hillman EM, Tian P, Waeber C, Teng IC, Ruvinskaya L, Shalinsky MH, Zhu H, Haslinger RH, Narayanan SN, Ulbert I, Dunn AK, Lo EH, Rosen BR, Dale AM, Kleinfeld D, & Boas DA (2008). Stimulus-induced changes in blood flow and 2-deoxyglucose uptake dissociate in ipsilateral somatosensory cortex. The Journal of neuroscience : the official journal of the Society for Neuroscience, 28 (53), 14347-57 PMID: 19118167

Fox PT, & Raichle ME (1985). Stimulus rate determines regional brain blood flow in striate cortex. Annals of neurology, 17 (3), 303-5 PMID: 3873210

Hlushchuk Y, & Hari R (2006). Transient suppression of ipsilateral primary somatosensory cortex during tactile finger stimulation. The Journal of neuroscience : the official journal of the Society for Neuroscience, 26 (21), 5819-24 PMID: 16723540

Nikos K. Logothetis (2002). The neural basis of the blood-oxygen-level-dependent
functional magnetic resonance imaging signal
Philosophical Transactions of the Royal Society B DOI: 10.1098/rstb.2002.1114

Handy software for automatically documenting MATLAB code

I’ve spent a good part of the past two years writing MATLAB code for EEG analysis (some of which is freely available as the ERP Mass Univariate Toolbox) and somewhat recently stumbled across a handy software package for automatically generating documentation for a set of MATLAB functions. The software is called M2HTML (http://www.artefact.tk/software/matlab/m2html/), and, true to its name, it generates a set of interlinked HTML pages from the MATLAB M-files in a branch of directories. Specifically, you get an index page that lists all the M-files and then a page for each M-file that can contain information such as the help header for that function and a list of functions called by that M-file and functions that that M-file calls. You can see an example of this at the Chronux Toolbox’s website. In addition, M2HTML produces a dependency graph like the one below, which illustrates which functions call which other functions.

A dependency graph of M-files produced by M2HTML. Arrows indicate which functions are called by the function that is the source of the arrow. Click on the image to see it full window.

To me, the most helpful aspect of M2HTML is that it makes clear function interdependencies. This is a huge help when I update a function, since I can clearly see what other functions might consequently need to be updated as well. It also helps when trying to make sense of the flow of processing in other people’s MATLAB software packages.

If your personal MATLAB code for a project has grown beyond a dozen or so interdependent functions or if you’re trying to wrap your head around someone else’s rat’s nest of MATLAB functions, I recommend checking it out. It should take just a few minutes to figure out and apply.