It is the whole RTS, a simulation and creator of "me" that I can not tell how is done by neurons. If I am to have genuine free will (without a soul postulated) then this computational procedure and not be deterministic as all digital computers and even analogue computers that do not self modify or "short term evolve" are.* I strongly suspect the reason why it is not and yet not random, is that it is highly "self referencing - such as "This sentence is false." is. I.e. not will well defined truth values for its statements.(That example sentence makes a simple declaration that is neither true, nor false.)
While a very successful type of Learning or self-taught, not programed, computer, commonly called a "neural network computer", does exist (See simple illustration of one below) it normally I not "self-referencing"; however, it could be, especially in a network of many trillion such units with the input of many units coming from the output of others. Certainly that is the way the brain is "wired up" and there even the "wiring" (synaptic connections) are constantly changing. So while I don't know the details, I suspect that parietal based "computation" is how "I" am created as information only in the brain, not a physical body, which would be deterministically governed by the laws of nature.
* We know brains learn and "short term evolve" as when you learn the name of person just introduced to.
Output can be more than two possibilities with more "nodes," some of which (or all) can be inter connected to be "winner takes all" or not. I suspect conscious decisions are made with "winner takes all" interconnects. I.e. you decide to buy or not to buy item A. etc. but prior to the conscious level more like weighted decisions /evaluations are made. Not known precisely but the number of "brain nodes" not part of consciousness are probable 100 times more numerous than those that are; However, as just discussed, surely output from conscious related node feed back as inputs to unconscious nodes AND conversely.
A three pounds (1.35kg), human brain is composed of neurons, glial cells, and blood vessels. The number of neurons was estimated in 1988 as 100 billion (1011), interconnected by their 100 trillion (1014) synapses.[5] In 2009, the estimate was lowered to 86 billion neurons, of which 16.3 billion are in the cerebral cortex, and 69 billion in the cerebellum.[ (Ref:
http://en.wikipedia.org/wiki/Human_brain)
So if we neglect the cerebellar neurons, that mainly store learned motor pattern commands, there are ~5E10 neurons most having little to do with the RTS and even fewer to do with creating "me." Lets make a wild guess that 0.2% are. I.e. "I" am the result of the collective interactions of E8 neuronal operations. There are 10 nodes in the above diagram, so there could be 10 million such "connection like machine" creating "me." However, many are more complex than 10 nodes and because of the many postulated "feed backs" making it my postulated "self referencing" system. Only a million such machines make "me."
To understand how powerful that "Me, making" computer array of neurons is, note that ONLY one man-made connection computer with less than 25 nodes, mostly inputs, can do a very good, near perfect, job so recognizing faces or telling hornless goats form sheep etc. and it is not at all with self-referencing connections!