| Faculty Comments |
|
|
| Faculty Member |
Comments |
John Kalaska
Université de Montréal, Canada
NEUROSCIENCE
 Hypothesis
 New Finding
 Controversial
|
In a fascinating report, Quiroga et al. present some of the most striking evidence to date for an explicit and invariant sparse-coding mechanism for the mnemonic coding of the identity of people and objects, as opposed to a more broadly distributed mechanism. They describe neurons in memory-related structures of the human brain (including the hippocampus, parahippocampal gyrus, amygdala and entorhinal cortex) that responded highly selectively and invariantly to different visual images of a specific famous person (e.g. a television or sports star), or a famous object (e.g. Tower of Pisa), or sometimes even to viewing the printed name of those people or objects. These provocative results are sure to evoke considerable interest and debate, and raise many new questions to investigate. For instance, although it was not stated explicitly, we can presume that the subjects recognized every image that evoked selective neuronal responses. However, can we also predict that the opposite should never occur, that is, a single cell should never respond selectively to images whose identity was not recognizable by the subject? The authors note another intriguing aspect of their results. If mnemonic coding is so sparse, to paraphrase Humphrey Bogart's famous rhetorical question in "Casablanca", of all the possible visual images of all the people and all the objects in the world, how were they able so often (15% of recorded units) to find a person or object that activated one of the few cells they were recording at any given time from amongst a limited set of test images? The authors speculate that this may indicate that single cells can contribute to the sparse coding of multiple distinct objects or people, which in turn suggests that these high-level abstract representations may be based on a distributed discrete classification process rather than on a continuous representation of the metrics of the images. Another striking feature of the results was that the large majority of the neuronal responses were statistically selective but quite weak, implying that surprisingly few spikes from only a few neurons may be enough to recall the unique identity of one person or object from all the other people and objects in the world. However, could it be that the modest responses reflected the fact that the cells were responding to images of people and places that were largely irrelevant to the everyday life of the subjects? It is unlikely that any of the subjects knew Jennifer Aniston, Mother Theresa or Kobe Bryant personally, or had attended a concert at the Sydney Opera House. It would be very interesting to repeat this experiment using images that are drawn from the everyday experience of the subjects, such as their own friends rather than TV "Friends", their own car, their own home, and of course their own grandmother (how could one ever resist?). If the mnemonic coding of these highly relevant people and objects proves to be equally as sparse as appears to be the case for famous celebrities and landmark architecture, that would be a particularly compelling confirmation for the authors' findings.
Evaluated 1 Jul 2005 |
Kevan A. Martin
Institute of Neuroinformatics ETH/UNIZ, Switzerland
NEUROSCIENCE
 Confirmation
New Finding
|
This paper extends the authors' earlier work in showing that single neurons in the medial temporal lobe (MTL) of awake humans have a remarkable degree of selectivity in their response to pictures of individuals, landmark buildings, animals, and objects. The responses of single or multiple neurons were recorded from chronic electrodes implanted in the MTL in order to localize the focus of epileptic seizures. Since the famous patient HM, who had amnesia following bilateral removal of his MTL, the importance of the MTL for declarative memory has been well established. This new study shows the precision of the neural representations in the MTL. Although most neurons responded to more than one image, some neurons responded only to images of famous people (such as ex-president Bill Clinton, the Beatles or the soap actress Jennifer Aniston) or famous buildings, such as the Sydney Opera House. The data are interpreted as evidence against a highly distributed population code and for a sparse, explicit, invariant representation involving relatively few neurons.
Evaluated 29 Jun 2005 |
Mayank Mehta
Brown University, United States of America
NEUROSCIENCE
New Finding
|
The paper shows that neurons in the hippocampus and related areas in humans respond to abstract visual information in a selective fashion. Using microwires, Quiroga et al. recorded the activity of neurons from several parts of the medial temporal lobe, including the hippocampus, of epileptic patients. The hippocampus is usually associated with learning, memory and spatial representation. The authors show that neurons in the human hippocampus respond robustly to a select subset of images of famous places as well as faces in a sensory-cortical fashion. Further, these neurons seemed to encode information in a highly invariant or abstract fashion in that they responded robustly to the same face when presented in a variety of ways, including caricature and semantic representation. Such selective, yet invariant, representation of abstract information in the human hippocampus raises interesting questions about the nature of neural information processing in the hippocampus.
Evaluated 24 Jun 2005 |
Aina Puce
West Virginia University, United States of America
NEUROSCIENCE
Hypothesis
New Finding
Controversial
|
This fascinating study investigated abstract representation in the brains of 8 patients being investigated for potential seizure surgery. The investigators found that individual neurons in the medial temporal lobe responded selectively to the identity of famous individuals, for example Halle Berry or Jennifer Anniston, and also to other categories of object, including famous buildings, in a series of recording sessions where the responsivity of neurons to multiple visual categories were investigated. These intriguing results challenge current ideas and models of object and face recognition that posit that the output of multiple neurons typically would code the identity of an individual person or specific object.
Evaluated 24 Jun 2005 |
|
|
| Faculty Comments |
|
|
|
How to cite the Faculty of 1000 evaluation(s) for this paper
1) To cite all the evaluations for this article:
Faculty of 1000: evaluations for
Quiroga RQ et al
Nature 2005 Jun 23 435
(7045)
:1102-1107 http://www.f1000biology.com/article/15973409/evaluation
2) To cite an evaluation by a specific Faculty member:
John Kalaska:
Faculty of 1000, 1 Jul 2005 http://www.f1000biology.com/article/15973409/evaluation
Kevan A. Martin:
Faculty of 1000, 29 Jun 2005 http://www.f1000biology.com/article/15973409/evaluation
Mayank Mehta:
Faculty of 1000, 24 Jun 2005 http://www.f1000biology.com/article/15973409/evaluation
Aina Puce:
Faculty of 1000, 24 Jun 2005 http://www.f1000biology.com/article/15973409/evaluation
|
|
|
