In Kandel's "Principles of Neural Sciences" in the chapter about the anatomical organization of the brain one reads (p. 323, 4th ed.):

»Although a variety of [relay] neurons are involved at each stage in information processing, these neurons generally fall into two functional classes: principal (or projection) neurons and local interneurons. The axons of principal neurons convey information to the next stage in the system. Interneurons may receive inputs from the same sources as the principal cells, but they contact only local cells involved in the same processing stage [(relay)].«

This means: All neurons of a "relay" (= functional unit) receive external input, but only some give external output. I wonder if there are examples of functional units (nuclei, cortical areas, etc.) with significant numbers of

  • interneurons receiving input only from local cells
  • principal neurons receiving input only from local cells

If there were, there would also be distinguished input neurons, receiving external input but not giving external output.

  • $\begingroup$ I don't think assigning those classes the adjective "real" makes any sense. Additionally, in an ANN, there aren't any interneurons if we think about it like the brain. Otherwise almost the entire CNS is interneurons, except for the motoneurons and sensory neurons themselves. $\endgroup$ – Bryan Krause Feb 7 at 16:26
  • $\begingroup$ @BryanKrause: Of course, the entire CNS is interneurons (except motor, endocrine, and sensory neurons), but I asked for smaller populations of neurons like nuclei and cortical areas. And why shouldn't a layered ANN have interneurons? $\endgroup$ – Hans-Peter Stricker Feb 7 at 16:38
  • $\begingroup$ I think I wasn't very clear. What I meant is that the definition of what an "interneuron" is varies with systems, and ANNs and the brain are not organized very similarly. One could just as much consider a hidden layer of an ANN to be a collection of neurons that project to another layer. It's fine to call them interneurons in an ANN context, I'd just be hesitant to carry that definition over to analogies. $\endgroup$ – Bryan Krause Feb 7 at 17:31
  • $\begingroup$ @BryanKrause: I dropped the ominous adjective "real" because it adds nothing to my question and was responsible for some confusion, as I am afraid. Now the question should be rather clear-cut - or isn't it still? Also, mentioning ANNs didn't add anything essential to my question, so I dropped it, too. $\endgroup$ – Hans-Peter Stricker Feb 7 at 22:54
  • 1
    $\begingroup$ I see. I think that in most cases brains are simply much more complicated than that; there are too many good reasons to keep, for example, an efferent copy of an output signal (making the output neurons...improper). Probably the retina is a relatively complex structure that fits at least some of your requirements, but computationally it is quite boring compared to other brain areas. $\endgroup$ – Bryan Krause Feb 7 at 23:47

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.