On Mon, 3 Feb 1997, Eliezer Yudkowsky wrote:
> So, with those newfangled PET scanners and other close-to-real-time
> neuroimaging tools, they finally get a chance to see what happens when
> people *retrieve* LTM. Guess what lights up?
>
> Not the hippocampus.
> Not the cerebral cortex.
> The cerebellum.
It didn't light up much on the scans we were shown. But I know why: this
was a symposium on dyslexia, and everyone was interested in semantic
memory. While there is definitely some arguments about excatly what is
involved in LTM (or even if there is any clear distinction between STM and
LTM) the "consensus" seems to be like this:
Hippocampus and maybe parts of thalamus do the encoding of STM into LTM
for *declarative memory* (i.e. the kind of memory you can tell me about,
like semantic ("Kyzyl is the capital of Tuva"), episodic ("At 13.00
yesterday I was watching a movie") or metamemory ("Yes, I know most of the
periodic system")). Nondeclarative memory is handled by other systems,
procedural memory is most likely the basal ganglia and cerebellum. It
should be noted that the memories seem to be encoded in cortex, especially
inferotemporal cortex (perceptual memory) and the prefrontal cortex.
I have seen several scans suggesting that when you are encoding a
declarative memory the left prefrontal cortex shines up (not the
hippocampus, which is surprising), while during retrieval the right
prefrontal cortex lit up. There are several other parts of the brain that
light up, including the cerebellum, but they seem to be rather transient
and variable. Most people I have talked with agree that the cerebellum is
involved in a lot of things beside movement control; in fact the lateral
cerebellum seems to be quite involved in verbalization (it lit up in some
scans of people reading or listening).
Of course, one should be a bit critical of PET and fMRI scans - they just
show that something happens, not what. After all, it could be inhibiting
activity or just the "shadow" of signals sent from (say) hippocampus. But
they are suggestive.
> And if anybody can even *begin* to explain this, I haven't heard about
> it.
Just check the literature. Things like this tend to produce amazing
amounts of models...
> My own guess is that the cerebellum does constraint propagation or
> some other kind of constraint-to-visualization assembly, and that is
> what motor skills, LTM retrieval, and (another guess) spatiotemporal
> visualization have in common.
Interesting. I personally think it does error correction, a bit like a
general system for keeping neural signals on track if some kind of error
can be defined. We should also remember that some of its nuclei are
involved in motivation.
-----------------------------------------------------------------------
Anders Sandberg Towards Ascension!
nv91-asa@nada.kth.se http://www.nada.kth.se/~nv91-asa/main.html
GCS/M/S/O d++ -p+ c++++ !l u+ e++ m++ s+/+ n--- h+/* f+ g+ w++ t+ r+ !y