On memory (part 2).

Read part 1 here.

Note: not the actual wedding I attended.
Note: not the actual wedding I attended.

Midway through spring, we found ourselves in Chicago for a wedding.  K was asked to be best man, and since N was (is) still breastfeeding, she and I had to tag along.  I’m not a big fan of weddings, but I did sneak in a lovely conversation with the groom’s younger brother.

He, the younger brother, is an artist (music, film, theater) who recently started working on a degree in journalism.  During our chat he mentioned that he’d become interested in memory after a traumatic head injury — he found it fascinating / alarming that there is a gap in his own mental records for the swath of time before and after that incident.  I didn’t ask, but it’s possible this incident inspired his foray into journalism; I can imagine written records taking on a newfound importance for someone subject to an involuntary lesson in the fragility of biological memory.

Part of why I don’t care much for weddings is the perpetual bustle — it’s hard to finish a good conversation because there’s always something else about to happen, somewhere else you’re supposed to be standing.  Next time, perhaps, if we remember, we can finish talking about this.

One thing he did have time to say, though, was: “Memories are physical things, right?  They exist in your brain.  I was wondering whether someone could analyze a brain and learn what memories are there.”

I rattled off only a partial answer before he was whisked away for photographs.  My answer, as it happens, was incorrect as to the current state-of-the-art for mind reading — I’d recently seen this press release titled “Scientists crack piece of neural code for learning, memory.”  After reading it, I had the impression that the group was able to visually inspect a region of the brain and know which of two memories had been encoded, either if you hear a high-pitched sound, move right or a similar memory instructing a rat to move left instead.

3461234232_19b63c79d7_oWhich seemed believable to me; yes, that’s a very hard problem, but if the group was investigating only a single, simple type of memory, one that was always stored in the same part of the brain, and was attempting to differentiate between only two choices… well, yes, it still seemed difficult, but at least they’d know where to look.  The impression I have is that data recovery from crashed computer hard drives is difficult primarily because any given piece of information might be stored in a variety of places, and that the major calamity isn’t usually that your documents, pictures, etc. are lost, but that the pointers, the addresses for where each document, picture, etc. is stored, are lost.

And computers are devices we designed!  Attempting to recover data from a brain, a system in which we don’t really know what the data would look like even if we knew where it was, sounds many orders of magnitude more difficult.

Still, I’d read that press release and thought one tiny piece of the puzzle had been solved.  That post-mortem visual inspection could unveil one particular binary known-to-be-present memory.

I was wrong.

Before typing this post, I downloaded the actual paper (Xiong et al.’s “Selective corticostriatal plasticity during acquisition of an auditory discrimination task,” although I should warn you that it isn’t open access) and attempted to puzzle through what they’d done.  I’m still shaky on the details, so my apologies if I make any mistakes in this ensuing description.

A flat slice of some poor rat’s brain showing the auditory cortex.

They modified a small population of brain cells to respond to light.  After shining light on these cells, they recorded the responses in another part of the brain — the cells in your brain look something like squids, with a head region that collects signals and tentacles that reach out to send those signals somewhere else, and as best we know memories are encoded through the pattern of linkages between those squids… er, cells — as a measure of the link between those brain regions.

After training a rat to move left when it heard a high-pitched sound, they expected the cells in a “we respond to high pitches” brain region to have stronger connections to / more & more powerful tentacles reaching into the “we move left” brain region.  And, yup, while they were training each rat they took occasional breaks to shine light onto the modified cells, and they did observe stronger signals in the “move left” brain region.

They could also distinguish between a brain slice from a dead rat that had been trained to associate “high-pitch / move left” from one that had been trained to move right with the same strategy: shine light on the pitch recognition region, look at signal in the movement region.

My mistake, then, was thinking they could assess synaptic strength visually — I thought they were inspecting brain slices under a microscope to determine what memory was there.  Their measure for signal strength was “local field potential,” though, which sums the contributions of many cells… so it seems possible that they could have arduously traced out the tentacles from each cell in their pitch region to see how many were reaching into each movement region.  Which would obviously be a huge pain — the brain is messy, and those tentacles are very small — but it seems feasible.

And, sure, I hadn’t realized they were using rats whose neurons were modified via viral infection… although it seems like visual assessment could be done using unmodified cells.  With the caveat being, of course, that making a 3D model showing all the outcroppings of every cell in even a small region of the brain would be incredibly difficult.

Anyway, this seems to be the current state of the art for learning what memories were stored in the brain of a deceased rat.  For humans, we can do even less.  If you happen to have a few heinous memories tucked away, don’t worry; if you die soon, those memories will die with you.

Of course, these fields are advancing all the time.  If you take too long to die, all bets are off.


GolgiStainedPyramidalCellp.s. Synaptic strength isn’t just a measure of the number of tentacles… to elaborate on our metaphor, those tentacles also froth forth with little bubbles of neurotransmitters, and the quantity of those waiting bubbles can change, the density of receptors on the next cell in line can change, etc., and all those changes (and more!) seem to play a role in memory formation.  A perfect 3D map of where each neuron’s axons reach to, where each neuron’s dendrites are grasping from, wouldn’t be enough.

And even if we could obtain all that, a full list of every synaptic strength, we’d still have to puzzle out what all the information means.  How do those connections result in an image of your grandmother’s basement?  A narrative of your youth?  No one knows.  So your horrible secrets will probably be safe even if you die a very long time from now.  Unless you forget to burn your diary.  Then future sleuths could simply bypass your encrypted brain.  And wouldn’t you feel silly!