In my June brain column for Discover, I wrote about the bizarre idea that there are single neurons in your head that can respond to individual people. The so-called “grandmother cell” started out 40 years ago as a thought experiment riffing on Philip Roth’s novel Portnoy’s Complaint. By the 1970s, most neuroscientists considered it more of a joke than a valid concept, but in the years since it hasn’t quite gone away.
In my column, I described the work of the work of Rodrigo Quian Quiroga of the University of Leicester:
For the past eight years, he and his colleagues have been studying epilepsy patients who have had electrodes implanted in a region of their brains called the medial temporal lobe, as part of a study to identify the source of their seizures. Quian Quiroga showed the subjects 100 pictures. The pictures included photos and drawings of celebrities as well as landmarks and various familiar objects. The patients had to press one button if a picture was of a human face and another if it was not.
In their first such study, Quian Quiroga and his team were able to observe the individual activity of 993 neurons. They found that 132 of them responded to at least one picture. And of those responding neurons, 51 fired in response to only a single person or thing. One neuron responded only to Halle Berry, for example.
Amazingly, the “Halle Berry” neuron responded to any picture of her, including one in which she was dressed as the masked Catwoman. Even the name Halle Berry triggered that neuron, which was silent at the sight of other actresses or their names.
Quian Quiroga does not, however, believe these neurons are grandmother cells, at least as they were initially conceived. He suspect that a very sparse network of neurons–perhaps hundreds out of the billions in our heads–can develop this kind of response to an individual. Quian Quiroga just happen to stick his electrodes near single neurons that belonged to these networks.
Which brings us to Quian Quiroga’s latest paper, published in Current Biology. He analyzed the signals from 750 electrodes implanted in seven patients as they looked at pictures of some celebrities like Oprah Winfrey. Quian Quiroga found neurons that responded strongly to the sight of these individuals–and they also responded strongly to their written names and even the sounds of their names.
Quiroga and his colleagues also ran the same test using themselves rather than the celebrities to probe for neurons. They discovered neurons that responded strongly only to individual researchers, too–and once more, the same neurons responded to the sight and sound of their name. Bear in mind–the patients had only met the scientists a day or two earlier. So these neurons had developed their grandmother-ish response in a very short time.
These results offer some clues to how these sparse networks are arranged. Some of the neurons probably get signals from other regions of the brain that recognize faces. Others tap into auditory networks, and others language centers. Yet, remarkably, the information from these far-flung parts of the brain get funneled into tiny sets of neurons that can then encode concepts of people.
They may not be the Grandmother Cells of legend, but in their own way, they’re very cool.
Reference: Quian Quiroga et al., Explicit Encoding of Multimodal Percepts by Single Neurons in the Human
Brain, Current Biology (2009), doi:10.1016/j.cub.2009.06.060
July 23rd, 2009 at 1:03 pm
I think the Jodie Foster and Helen Hunt neurons in my brain are entangled because I get them confused constantly.
July 23rd, 2009 at 1:51 pm
It would be cool to see what kind of results would be found for “self”.
July 23rd, 2009 at 5:09 pm
I definitely have a Halle Berry neuron
July 23rd, 2009 at 5:31 pm
Agreed, the cells are very cool. What I think the finding represents — similar to the idea of “sparse networks” — is that neurons can represent remarkably high-level perceptual attributes.
So, if the activation of the sparse network of Halle Berry cells is not the same as recognizing Halle Berry what is the true brain correlate of recognizing Halle Berry? Or, turning the question on its head, if neuroscientists were able to selectively stimulate any set of neurons, what stimulus set would lead to the perception of Halle Berry? and what would the relationship be between that subset and the set of Halle Berry cells?
July 23rd, 2009 at 6:39 pm
I’m a software engineer. Something in this whole set of data is screaming “hash table” at me.
July 23rd, 2009 at 8:48 pm
Yet another result pointing to that brains, or rather the neocortex somehow self-organize (they have seen that too) into symbolic thinking. One neat thing compared to other forms of learning is that there are little or no errors from over-training, a symbol most likely remains robust in such cases and that is what is observed.
Another neat thing I can think of is that this may be directly or indirectly connected with the development of language.
July 27th, 2009 at 10:14 pm
One interesting thing would be to stimulate, using very weak intracortical microstimulation, near an identified “Halle Berry” neuron and see if that caused the patient to “think” of or in some other way perceive Halle Berry. Only then would it be legitimate to claim that the neuron was part of a neural circuit encoding multimodal percepts of Halle Berry.
Then notion of neurons encoding multimodal percepts appears to be closely related to the notion of neurons “representing” things or persons. I suspect that the notion of individual neurons “representing” things is nothing more that a kind of reflex genuflexion to a discredited, nihilistic philosophy (materialistic reductionism). Perhaps, as suggested by Bennett and Hacker in their fine book, The Philosophical Foundations of Neuroscience, the very concept of neural representations should be abandoned. When a neuron responds to picture or thing, all we really know is that the neuron responds in the presence of that picture or thing. It is possible that the activity of that neuron (and the other neurons it is connected with) is necessary for perceptual awareness of that picture or thing. After all, every individual thing we can perceive as unique will be associated with a somewhat unique pattern of activity in the nervous system. And, if injury or disease disrupts the involved neural circuits, perceptual awareness will be lost. However, to leap to the conclusion that a particular neuron’s activity is sufficient for such perceptual awareness, i.e., that it somehow “represents” or “encodes” the picture or thing in some sort of magic inner perceptual space or theater in the cerebral cortex, is to adopt a particular philosophical stance. Indeed, the whole idea of inner subjective awareness opposed to an outer objective world is just another variety of mind-body dualism. Carl, your fine book, Soul Made Flesh, describes how mind-body dualism transformed into a sort of brain-world dualism.
July 28th, 2009 at 10:43 am
Response to S. David Stoney:
I’m a neuroscientist who, for a living, records from single cells and describes response properties. We who do this work recognize that the findings are just correlations and do not show cause and effect. Nonetheless, as the tool is remarkably powerful. The (very difficult) experiment you describe, stimulating the Halle Berry cells, is a great concept for an experiment, but probably impossible to do. The tone of your post sounds critical of the project of recording individual cells. Yet this type of experiment has led to transformation of our understanding of visual processing and, in the area that I work (hippocampus), a revolution in concepts about central function.
I don’t know any neuroscientist who believes in the simple model of grandmother cells. As I understand it, the grandmother cell concept has always been a straw man. It separates those who think that neurons are significant, high-level operators from others who think that significant neural codes are widely distributed. So it is is remarkable and important that single neurons can have remarkably specific and high-level correlates. High-level neurons that have been found in various domains. They have also proven to be powerful tools in analyzing signal processing in specialized neuronal networks.
Concerning the philosophical problem you have with perceptual dualism. A version of the experiment you suggested has been done, and was done a long time ago. The famous neurosurgeon, Wilder Penfield, pioneered techniques of localized electrical stimulation of awake patients during brain surgery. What he found was that stimulating regions of the brain led to percepts that were frequently indistinguishable from actual perceptions. Stimulating the hand region of somatosensory cortex could lead to a perception that something is stroking the hand, stimulating auditory cortex could lead to the patient hearing a familiar song, and stimulating deep in the temporal lobe could elicit memories, or memory-like percepts. Strong support for dualism.
July 28th, 2009 at 2:07 pm
Correction to my last post. I shouldn’t have said that Penfield’s experiments are “strong support for dualism”.
The Penfield studies are support for the neuroscientific model that the world is represented (re-presented, modeled) in the brain. As I understand dualism, it suggests a non-material aspect of mind. The Penfield studies don’t address this.
July 30th, 2009 at 3:22 pm
HiHo JohnK,
Thank you for your thoughtful reply. However, I don’t think Penfield stimulated SI or SII and caused a feeling of the hand being stroked. I thought that he stimulated temporal lobe with surface electrodes and large (ma) currents. That rather complete memories/replications of past episodes of experience were elicited is a remarkable and important finding. It is not clear to me, however, that the finding provides “strong support” for an internal neural representation of the world. The truth is that only rarely does electrical stimulation of the nervous system result in fully formed, coherent percepts. Stimulation of SI, for example, does not produce a sense of touch or stroking, but rather a tingling or shock-like feeling in the appropriate body part. Similarly, stimulation in primary, secondary, or tertiary visual cortex usually does not, as far as I know, produce perception of line segments, colors, faces, or objects moving in the visual field, but rather phosphenes (light flashes). Similarly, stimulation of Broca’s or Wernicke’s area does not produce speech, but rather disrupts speech. So, while I know that activation of a unique neural circuit is necessary for every single unique percept I have, I remain unconvinced that the existence of internal neural representations of the external world has been clearly established.
Maybe, as suggested by Alfred North Whitehead, perception is the cognition of prehensive unification of subject and object. Such could, I suppose, be possible in a quantum universe with nonlocality, while not being possible in a Newtonian universe comprised of insentient matter/energy and empty space. If so, then – as Chris Koch notes in his review of Tononi’s work (http://www.scientificamerican.com/article.cfm?id+a-theory-of-consciousness) – consciousness may be universal. I have made an abortive stab at developing a process philosophical approach to neuroscience at http://home.earthlink.net/~icedneuron/, especially sections VII and VIII. Please ignore my sadly unsuccessful attempt to bring attention to the earth’s ongoing climate cycle by emphasizing the next glacial phase. Note also that the website is badly out of date…
Has anyone tried microstimulation near “place cells” in the hippocampus? If one could do that, with the effective extent of the stimulating current restricted to the cells near the one being recording from, what do you predict would be the effect on the animal’s behavior?
As for being critical of recording single cell receptive fields, yes, I admit that I’m dubious about its utility. But who knows? That project may keep people off the streets who might otherwise be dangerous <;-)
All the best – David
August 4th, 2009 at 9:01 pm
S. David Stoney,
thanks for the qualifications on brain stimulation work. I have on old video where Penfield is interviewing a patient where the recall the surgery. As described, stimulating S1 produces a kind of sensation on the hand. Stimulating another place (not specified) caused the patient to say that, during surgery, she heard a song and thought that others in the OR ought to be able to hear it also. This is hardly representative, but interesting.
I don’t think anyone has tried to stimulate place cells. A major reason that the prospect is bad is that hippocampal place cells are not topographically organized. Local stimulation would likely stimulate dozens or hundreds of place cells, with no relationship to each other. Neocortex is generally topographically organized, so local stimulation may stimulate cells with related properties. Not so in the hippocampus. The Halle Berry cells are, I think, in entorhinal cortex. There is on known topography there, so the there is little likelihood that local stimulation will stimulate related neurons.
The link in your text seems dead. I don’t really understand your ideas about consciousness. This is a difficult and fascinating subject. My sense of consciousness studies that they are diverse and disconnected; like the blindmen describing an elephant. Each is feeling a different part of the animal and not finding any common territory.