“At first blush, these claims regarding the relative biological plausibility of localist (grandmother) and distributed coding schemes seem well founded. As reviewed by Gross (2002), few neuroscientists have taken localist representations seriously. Indeed, the term grandmother cell (a term often attributed to Jerry Lettvin; cf. Gross, 2002) is generally used to ridicule the claim that complex and meaningful stimuli are coded by individual cells in the cortex. Finkel (1988) called them “infamous grandmother cells” (p. 787).”
What a shame the term doesn’t derive from Fodor’s granny…
Most definitely. (I see that Bowers cites Page with approval.) It’s even more worth reading (as well as the Bowers) if you find yourself *disliking* grandmother cells, but aren’t sure why. Many people dislike them just because they’ve been made fun of by famous people.
I know I dislike them because in every sensory system I’ve studied empirically there is a distributed code.
Of course the right question to ask is the degree of distributed-ness of the code, not whether it is distributed. There are two poles, from 1 (each thing represented has one neuron representing it) to 0 (each thing represented has every neuron representing it).
The truth depends on the system being studied (e.g.,vertebrate, invertebrate, etc), the modality (vision is not necessarily the same as somatosensory processing (the latter is my specialty), where in the processing heirarchy you are (the retina, IT, etc), the type of stimuli being used (e.g., full-field spatiotemporal white noise, little tiny dots, pictures of people’s faces).
There isn’t one answer to the question.
Eric Thomson
I focused on sensory systems because I know about them, but in motor systems, the number seems closer to 1 than to 0. Georgopolous and all that. But we can quantify the proportion of neurons needed to most accurately control a robotic arm using motor cortex, so motor systems may be an even better place to get an empirical bead on the issue.
Eric Thomson
Err, switch the 0 and 1 there (0 is fully distributed, 1 is crazy grandmother code).
How neural coding is achieved is a polarized debate.
But i guess that those working in Neural Interfaces Technologies could some day make big insights on this, becuase they need to detect and tramsit the neural signal for effector control helping poeple with moor limitations.
But what if grandmother cells, gnostic cells or pontifical cells do not code for specific faces (identity) but the configurational shape of faces?
I do not think that every student all over the world has got a passion of expository essay writing! Nevertheless, people ,which don’t have writing skillfulness have to utilize an assistance of distinguished essays online service and enjoy a success.
This is also worth reading if you like localist (grandmother-style) models:
Page, M. (2000). Connectionist modelling in psychology: A localist manifesto. Behavioral and Brain Sciences, 23 (4), 443-467.
https://www.bbsonline.org/documents/a/00/00/05/05/index.html
“At first blush, these claims regarding the relative biological plausibility of localist (grandmother) and distributed coding schemes seem well founded. As reviewed by Gross (2002), few neuroscientists have taken localist representations seriously. Indeed, the term grandmother cell (a term often attributed to Jerry Lettvin; cf. Gross, 2002) is generally used to ridicule the claim that complex and meaningful stimuli are coded by individual cells in
the cortex. Finkel (1988) called them “infamous grandmother cells” (p. 787).”
What a shame the term doesn’t derive from Fodor’s granny…
Most definitely. (I see that Bowers cites Page with approval.) It’s even more worth reading (as well as the Bowers) if you find yourself *disliking* grandmother cells, but aren’t sure why. Many people dislike them just because they’ve been made fun of by famous people.
I know I dislike them because in every sensory system I’ve studied empirically there is a distributed code.
Of course the right question to ask is the degree of distributed-ness of the code, not whether it is distributed. There are two poles, from 1 (each thing represented has one neuron representing it) to 0 (each thing represented has every neuron representing it).
The truth depends on the system being studied (e.g.,vertebrate, invertebrate, etc), the modality (vision is not necessarily the same as somatosensory processing (the latter is my specialty), where in the processing heirarchy you are (the retina, IT, etc), the type of stimuli being used (e.g., full-field spatiotemporal white noise, little tiny dots, pictures of people’s faces).
There isn’t one answer to the question.
I focused on sensory systems because I know about them, but in motor systems, the number seems closer to 1 than to 0. Georgopolous and all that. But we can quantify the proportion of neurons needed to most accurately control a robotic arm using motor cortex, so motor systems may be an even better place to get an empirical bead on the issue.
Err, switch the 0 and 1 there (0 is fully distributed, 1 is crazy grandmother code).
How neural coding is achieved is a polarized debate.
But i guess that those working in Neural Interfaces Technologies could some day make big insights on this, becuase they need to detect and tramsit the neural signal for effector control helping poeple with moor limitations.
But what if grandmother cells, gnostic cells or pontifical cells do not code for specific faces (identity) but the configurational shape of faces?
I do not think that every student all over the world has got a passion of expository essay writing! Nevertheless, people ,which don’t have writing skillfulness have to utilize an assistance of distinguished essays online service and enjoy a success.