neurodudes

We’ve heard in the past about neurogenesis in adults, but as far as we understand, this only happens in limited locations throughout the brain. However, what if those new neurons migrate to different places?

New evidence in mice suggests that after being born, new neurons can travel along the flow of spinal fluid to end up in the olfactory bulb.

If there is migration to other locations in the brain, the ramifications for computational models of brain systems are significant.

–Stephen

A Resilient, Low-Frequency, Small-World Human Brain Functional Network with Highly Connected Association Cortical Hubs (Achard et al., 2006)

A study on network properties of the whole brain (functional connectivity data from fMRI)… interesting to see this type of work published in J. Neurosci. Building on previous fMRI/whole brain connectivity studies, the authors use a set of wavelet basis functions to estimate the correlations between different anatomical regions.

Also includes some analyses on resiliency of the system (via a metric like “largest connected cluster”) to random and targeted attack (ie. node deletion). It would be neat if they also did some analysis of common stroke damage. I would think that a stroke probably doesn’t qualify as a “targeted attack”, in the traditional sense, but, due to the predefined structure of the major circulatory structures (eg. circle of Willis), there are likely regions that are near the most commonly blocked arteries, etc. Perhaps someone with some medical qualifications could weigh in here?

There is also a nice discussion of why the human brain does not appear to be a scale-free network: That nodes do not seem to follow the “rich-get-richer” rule of preferential attachment. Evolutionarily recent structures like prefrontal seem to be among the hubs of the system and older structures like limbic regions do not dominate. Here’s a picture of the connectivity map from the paper:

Full abstract after the jump.
(more…)

By the way, this nytimes article got me thinking about mirror neurons. Mirror neurons are neurons that respond both when you do a certain thing, but also when you see others doing the same thing.

The article implies that mirror neurons could be important for imitation. I wonder if they could be a system to let your brain literally try to imitate the patterns in another person’s brain? If (and given the possibility of different representations in different brains, I think this is a big “if”) imitative learning turns out to be literally one person’s brain patterns matching a teacher’s brain patterns, then an imitative learning mechanism would be a mechanism for thought “synchronization”.

If you have a set of neurons in your brain that are mirror neurons for task A, and Katy has a set of mirror neurons for task A, and Katy does task A with you watching, then maybe your sets of mirror neurons will fire in roughly the same patterns at the same time. So the mirror neurons would be somewhat like a telephone line between your brain; nearby neurons could cause Katy’s neurons to change their pattern, and presuming that this causes her to change her outward behavior, your mirror neurons would absorb the changed pattern. So neurons next to your mirror neurons could receive a signal from neurons next to Katy’s mirror neurons.

I’d like to note that I don’t know much about mirror neurons, and I certainly don’t know if mirror neurons have the properties necessary to be actually capable of this. Do they influence behavior? Do sets of mirror neurons actually end up matching the patterns of firing of groups of another person’s mirror neurons, or do they just become more likely to fire on an individual level? It seems pretty unlikely that they’d do this.. but it’s neat to think about.

Anna Fisher and Vladimir Sloutsky have an interesting paper called “When Induction Meets Memory: Evidence for Gradual Transition From Similarity-Based to Category-Based Induction”. They are trying to get at the question of whether children around the age of six have category-based induction or similarity-based induction.

What is category-based induction and similarity-based induction?
If I tell you that object A has property P and object B does not, and then I show you objects X, Y, and Z and ask which ones probably have property P, how do you make the decision?

Do you rely on which categories the objects fall into? (”object A is a mammal and object B is a fish; X and Y are mammals, and Z is a fish, so I’ll guess that X and Y have property P, and Z does not”)

Or do you rely on similarity? (”object A is brown and object B is yellow; X and Z are brown and Y is yellow; so I’ll guess that X and Z have property P, and Y does not”)

Humans apparently tend to prefer category-based induction, at least when they are familiar with the categories. What about children around age 6? This study argues that they prefer similarity-based induction.

Please refer to the paper if you’d like a more detailed explanation, as well as a discussion of the history of the question and citiations to other experiments supporting similarity-based or category-based induction in children.

The details of the study
In the control condition, subjects were given pictures of animals and told to study them for a subsequent recognition test. In the experimental condition, subjects were given a picture of a cat, and told that it has “beta cells inside its body”. Then they were given pictures of other animals, some of which were also cats, and asked which animals had beta cells inside their bodies.

In both conditions, subjects were later shown some of the same pictures and some different ones, and asked which pictures they had seen before (a recognition test).

Adults (and 11-year olds) in the experimental condition were worse than adults (and 11-year olds) in the control condition at the recognition test. But kids of age 7 did about the same in both conditions.

The experimenters argue that this is because the adults in the experimental condition were doing category-based induction and the 7-year-olds were doing similarity-based induction.

More details

They support this assertion by redoing the experiments with some twists. Most notably they redo it with 7-year olds, but only after training them to do category-based induction (see page 590 of the paper for how they did this). These 7-years performed similar to adults on the recognition-memory test. This supports the assertion that the difference between the 7-year olds and the adults was the use of category-based induction.

They also do the opposite experiment, where they create a situation in which the adults don’t know what the categories are and so are forced to use similarity-based induction. In this case, the adult performance on the recognition test becomes similar to the children’s.

Discussion
This paper is methodologically similar to the pigeons study, although they are about different things. In both studies, we first identify a cognitive task that someone else can do better than human adults (someone else being children or pigeons). This in itself is interesting, because it implies that there are tradeoffs in “intelligence”, and that the greater performance of human adults on some tasks comes at the price of worse performance on others (see Chris Chatham’s weblog entry on this study for more on this).

It also allows us to do interesting experiments. In both studies, the researchers wanted to identify the cause of the bad performance of human adults. They did this by training the group who used to be better than human adults to be just as bad as them (or almost just as bad). This strongly suggests that whatever they taught the others was the cause of the bad performance in the human adults. It also shows that what they taught the others was something that they weren’t doing before.

So, in the case of the pigeons, the researchers conclude that prior experience with matching tasks was the cause of the base rate neglect in the human subjects. In the induction study, the researchers conclude that children around age 6 don’t normally prefer category-based induction.

There’s an interesting series of studies from Edmund Fantino, Inna Kanevsky, Shawn Charlton, J Hartl, A Goodie, and D Case about heuristics in human decision-making. They find a task at which pigeons are actually better than humans, because they don’t have the same biases. Then they managed to train the pigeons to show the same biases as the humans. The studies are about base-rate neglect.

What is base rate neglect? (skip down if you already know)
To explain base-rate neglect, let’s say that in a certain sort of situation, one of two mutually exclusive events occurs, and generally event A happens with X% probability, event B with Y% probability (those probabilities are the “base rates”). Your job is to determine which event happened in some particular situation. There is also evidence left behind which helps you decide which event happened (for instance, event A may cause E to happen with a 5% chance, but event B has a 20% chance of causing E; so if you see evidence E, that pushes you in the direction of thinking that event B happened).

The logically optimal thing to do in these sorts of situations is to crunch the numbers using Bayes’ rule; there’s an equation which, when given all of the relevant probabilities and evidence, will tell you whether event A or B is more likely to have occurred.

But what humans often actually do is underestimate the effect of the base rates, and rely too much on the immediate evidence. To see why this is stupid, imagine that you’re coming home at the end of a cloudy day — you think it may have rained but you were inside at work all day and you’re not sure. When you get home you find that the window is open and your carpet is all wet next to the window. Your roommate accidentally leaves the window open about 50% of the time, so you get mad at him for leaving it open and letting the rain in. But he says, “Wait a minute man, yeah, okay, maybe I left the window open, I really don’t remember, and maybe it rained, I didn’t notice, but before you get all mad at me, consider this: we both agree there’s about a 50% chance of it raining on a day like this. So even if i HAD left the window open, there’s only a 50% chance that it would cause the carpet to get wet. But let’s say that wet, slimey aliens came through the window to look at our house during the day. If that had happened, there would be a 100% chance of causing the carpet to get wet. So, the evidence (the wet carpet) really supports the slimey aliens hypothesis better than your I-left-the-window-open hypothesis. So it probably wasn’t my fault”.

What’s wrong with your roommate’s argument? Base rate neglect. Sure, slimey aliens would be more certain to get the carpet wet than leaving the window open. But the prior probability of slimey aliens coming in on any particular day is much, much lower than the prior probability of your roomate leaving the window open by accident. The huge difference in the priors outweighs the fact that the aliens would be more certain to cause the evidence that you see.

Any adult human can see what’s wrong with the slimey aliens excuse, although they probably wouldn’t phrase it in terms of “base rate neglect”. But here’s a similar situation that humans tend to get wrong.

1% of women at age forty who participate in routine screening have breast cancer. 80% of women with breast cancer will get positive mammographies. 9.6% of women without breast cancer will also get positive mammographies. A woman in this age group had a positive mammography in a routine screening. What is the probability that she actually has breast cancer?

I won’t give away the answer here in case you want to work it out yourself. Please see this link for the correct answer, as well as a detailed explanation of how to get it. If you get it wrong, don’t feel bad; so do 85% of doctors (heh heh, I bet now you feel even worse, but for a different reason).

OK so what’s the experiment with the pigeons?
On each trial, first they showed the pigeons a picture of either a vertical or a horizontal line (they call this the “sample”). Then they gave the pigeons two choices; peck at a button with a picture of a vertical line, or peck at a button with a picture of a horizontal line. If they picked the vertical line, there was a 75% chance of getting a reward; if they picked the horizontal line, there was a 25% chance of getting a reward. The sample was a red herring; it didn’t give you any information about the right answer. The optimal strategy is always to pick the vertical line.

This is the task at which pigeons are better than humans. Pigeons do pretty close to the optimal thing. Humans choose the choice that matches the sample about 55% of the time (Goodie and Fantino (1995), but that’s not online; this article summarizes on pages 18-19).

The researchers hypothesize (on page 19 of the previous link) that this is because human have been trained over years and years to do tasks in which matching is important. In other words, for this task, humans are “overeducated”:

…from early childhood humans learn to match like colors and shapes at home, at play, and at school
(e.g., in playing with blocks and puzzles and in reading picture books with their parents).
Pigeons, on the other hand, have not experienced a rich history of matching.

They test this hypothesis by training the pigeons to do a matching task, and then they have them redo the original task (creating “overeducated” pigeons; Fantino has a funny cartoon picture in his talk but I couldn’t find it online). And indeed, they report that these pigeons do about the same thing that humans do, choosing the choice that matches the same about half the time.

So, they have succeeded in educating the pigeons until they become as stupid as us.

So, Edge has a new question for 2006 for its All-Stars of Academia to answer: What is your dangerous idea? (Suggested to Edge by Steven Pinker, who perhaps got the idea from a colloquium series at his old haunting grounds.)

Offhand, one might expect a broad range of perceived dangerous ideas, varying by research interests and such. What’s surprising is that many of the luminaries think that the “most dangerous idea” is this particular, same idea: As neuroscience progresses, popular realization that the “astonishing hypothesis” — that mind is brain — will create a potentially cataclysmic upheaval of society as we know and have profound (negative) moral implications as people claim less responsibility for their actions.

Of course, this just isn’t true. But, would you believe that
Paul Bloom,
VS Ramachandran,
John Horgan,
Andy Clark,
Marc Hauser,
Clay Shirky,
Eric Kandel,
John Allen Paulos,
and, in a more genetic context, Jerry Coyne and Craig Venter
are all very worried about this issue? (And I didn’t even read 50% of the Edge dangerous ideas… there might be even more… ) Is this really the most dangerous idea out there to all of these talented thinkers?

I feel strongly that science and morality have always been separate domains and that any worry that, by “debunking” the mind, we automatically become immoral machines is just ridiculous. Through this scientific knowledge, we might gain some humility, maybe better see our close relatedness to nonhuman primates and place in nature, etc., but we’re not going to flip out and become crazed zombies. This just isn’t going to happen.

Does anybody else think that this just isn’t a truly dangerous idea (although certainly an “astonishing” one, in the Crick sense)? Or am I wrong here?

Samples of academic worrying after the jump.
(more…)

In his great blog, Chris Chatham mentions two interesting phenomena that seem to be evidence for similarities in the way that vision and audition are processed:

“Lateral suppression can be observed both within the auditory and the visual systems as well. In the case of vision, this takes the form of mach bands, in which adjacent bars of increasing luminance show contrast effects at their edges. In the case of audition, the masking threshold is much lower for those sounds that coincide with the cutoff frequencies of a non-simultaneous noise mask; in other words, it becomes easier to hear the masked tone when it exists near the edges of the mask because of increased perception of contrast. In both cases, there are contrast effects that appear at the “edges” (whether auditory or visual) of stimuli.

There are also similar figure vs. ground phenomena in both visual and auditory stimuli. In vision, Rubin’s illusion shows how a bistable stimulus can appear to have one of two possible figures (face or vase) superimposed on one of two possible backgrounds (white or black). In audition, a similar effect occurs in pulsation threshold masking, when one auditory stimulus appears to be superimposed over another. Even though neither is truly a “background” sound, since both are pulsating, one is perceived as occurring continuously “behind” the other. Both these scenarios exhibit pattern completion or, to use the gestalt phrase, good continuation.”

Maurer, D., & Mondloch, C. Neonatal synesthesia: A re-evaluation. In L. Robertson & N. Sagiv (Eds.), Attention on Synesthesia: Cognition, Development and Neuroscience, Oxford University Press, 2004. Pp. 193-213.

This article postulates that babies experience synaesthesia.

I’m not convinced of that hypothesis because (although I only skimmed the article), I couldn’t find any evidence of something that infants and synaesthetes do that non-synaesthetes do not do. But it still reviews some interesting facts.

There are apparently a number of tasks that demonstrate, to quote the article, “paradoxical evidence of U-shaped development of cross-modal perception: Babies demonstrated successful linking of information across sensory modalities near birth, failed at similar tasks later in infancy, and then appeared to gradually learn cross-modal links in the second half of the first year of life.”

The article also reviews evidence for mysterious, presumably innate cross-modal correspondences in normal adults. For example, high frequency sounds go together with lighter colors. Angular shapes go with aggression, strongness, and loudness. Brighter light goes with loudness.

news from the future:

Eugene M. Izhikevich. Polychronization: Computation with Spikes. Neural Computation, Vol. 18, No. 2. (February 2006), pp. 245-282.

This readable, highly recommended paper is about a concept that Izhikevich calls “polychronization”. When a bunch of neurons tend to fire at the same time, you say that group is “synchronized”. But what do you call it when a group of neurons tends to participate in a consistent spatiotemporal firing pattern? Izhikevich would say that such a group of neurons is “polychronized”. Note that synchronization is just a special case of polychronization.

For example, maybe you notice that you often observe the sequence {neuron A fires, then 10ms later neuron B fires, then 14 ms later neuron C fires} in a network. These neurons are not synchronized (they are firing at different times), and the timing of the start of the pattern may or may not be timelocked to anything else (stimulus onset, gamma rhythm, etc), but what is important is that they are firing with a consistent pattern of timing within themselves.

(One complication is that the pattern may not repeat exactly; in the above example, suppose neuron B fails to fire 20% of the time. Just as we say that 3 neurons are mostly synchronized even if one of the neurons fails to fire with the others every single time, we will say that a group is polychronized even if the pattern is inexact [1])

The paper proposes a mechanism which could cause these patterns, and also a mechanism for other neurons to detect them. Both mechanisms are based on axonal conduction delay. As a warm-up, consider how we could get synchronization. Suppose that neurons A, B, and C all synapse onto both neuron D and neuron E, and that the axonal conduction delay is constant (say, 5ms). Now suppose that neurons A, B, and C all fire at the same time. If the synaptic weights are sufficently strong, this will cause D and E to fire at the same time. So, we have synchronization between D and E.

Next, suppose that there are longer delays between A,B,C and neuron E than between A,B,C, and D; it still takes 5ms for a spike to get from A,B, or C to D, but that now it takes 10ms for a spike to get from A,B, or C to E. Now if A,B, and C fire at the same time, it causes a pattern: first D fires, then E fires 5 ms later. This is how axonal conduction delays can cause these patterns.

Now, detection. In the previous two examples, D and E acted as synchrony detectors; they “detected” when A, B, and C fired together. Now what if we add 3ms to the time it takes for a spike to get from neuron C to either D or E? Here are all our axonal conduction delays now:

A->D: 5
A->E: 10
B->D: 5
B->E: 10
C->D: 8
C->E: 13

Now how should A,B,C fire in order to excite D and E? Well, C needs to fire 3ms before A and B if we want D and E to receive 3 spikes at once. So, now D and E are detecting not synchrony, but rather polychrony; they are detecting the pattern {C fires; then 3 ms later, A and B fire}.

Furthermore, D and E’s response to their pattern is not a synchronized burst, but rather it is a different firing pattern; {D fires; then 5 ms later, E fires}.

The first contribution of this paper is in providing us with a language to describe this sort of computational system. In the above example, we might notice that following pattern recurring in network activity: {C fires; then 3ms later, A and B fire; then 5ms later, D fires; then 5ms later, E fires}. The neurons involved in this pattern are termed a polychronous group [1] . Note that a single neuron might participate in multiple polychronous groups; for instance, if the above example is part of a larger network, perhaps there is another firing pattern involving neuron F, G, and C.

The second contribution of this paper is providing a simple 1000-neuron model which exhibits this sort of behavior [2]. The model itself is one page of Matlab code, and is based on the spiking neuron model in (Izhikevich, 2003). The model uses STDP plasticity to update synaptic weights as the model runs, and STDP is key to the formation of the polychronous groups [3].

The third contribution is the analysis of this model. The model displays different network states including slow and fast (”delta” and “gamma”) rhythms. The emergence of polychronous groups is robust to changes in some of the model parameters. Polychronous groups appear and disappear and change over time. Application of an external stimulus can cause a “response” consisting of the probabilistic activation of a certain subset of polychronous groups.

The most important result of the paper in the view of the authors is that the number of polychronous groups exceeds the number of neurons in the network (remember, each neuron may be part of multiple groups). This is important because if the real “logic elements” in the computation are the polychronous groups, not the individual neurons, then the memory capacity or computing capacity of the network may be larger than otherwise expected.

The forth contribution is an interesting discussion of various aspects of how the brain might work if its computation is indeed based on polychronous groups.

FOOTNOTES

[1] Actually, technically, Izhikevich defines a polychronous group based whether the group of neurons has the POTENTIAL to fire in such a pattern, based upon its anatomical connectivity. So, technically, “polychronous group” is an anatomical property, not a functional one.

[2] Although I guess Izhikevich has published similar models before so you might say that “contribution” belongs to his previous papers. Whatever.

[3] Which is analyzed further in (Izhikevich, 2004). Although I don’t think Izhikevich has conclusively shown that STDP is the ONLY kind of plasticity that could cause this.

This is from 9 months ago. It’s about a quadriplegic patient with a Cyberkinetics implant (100 electrode array) who can control a cursor decently. Here’s the link. Amazingly, Matthew says he learned to use the interface in only a couple of days (by which I infer he meant “start to basically use it”, not “have precise, skilled control of it”).

I have to point out that in my opinion, the article’s statement that “Neuroscientists can record and roughly translate the neural patterns of monkeys” is very misleading. It’s not that we can record ALL of the neural activity in a monkey, or completely translate all its thoughts; the best I’ve heard of is recording a tiny fraction related to motor planning and translating it into roughly which direction the monkey wants to move a limb.