I can watch you learning

Human Stereotaxic System. Photo credit: The Mind Project
Human Stereotaxic System. Photo credit: The Mind Project

Recording directly form the healthy living human brain has always been a coveted goal of many neuroscientists, thus bypassing the limitations of non-invasive techniques or animal work. But, understandably, nobody would seek or grant approval for inserting an electrode in the healthy living human brain, on moral and ethical grounds. The next best thing is to insert an electrode into the not so healthy living human brain.

Ison, Quiroga, & Fried (2015) got lucky and gained access to 14 patients with intractable epilepsy that had electrodes implanted in their brain to find where the seizure focus is (for possible surgical resection later on). Using these electrodes, they recorded the activity of single neurons within the medial temporal lobe (MTL, a brain area paramount for learning) while the patients performed some simple association tasks. First, they presented images of places, people, and animals to the patients to see “which (if any) of the recorded neurons responded to a picture” (p. 220). When they got a neuron responding to something, they rushed out, did some data and image processing, and after an hour they started the experiment. Which was showing the patient the picture to which the neuron responded to (e.g. Stimulus 1 = patient’s daughter) overimposed on a background that the neuron did not respond to (e.g. Stimulus 2 = the Eiffel tower). After one single trial (although there was some variability), the patients learned the associations (i.e. Stimulus 3 = daughter in front of Eiffel tower) and this learning was mirrored by how the neuron responded. Namely, the neuron increased its activity by 200% to 400% (counted in spikes per second) when shown the previously un-responded to image alone (i.e. Stimulus 2).

Excerpt from Fig. 5 from Ison, Quiroga, & Fried (2015).
Excerpt from Fig. 5 from Ison, Quiroga, & Fried (2015). “Average normalized neural activity (black squares) and behavioral responses (green circles) to the non-preferred stimulus as a function of trial number. Data were aligned to the learning time (relative trial number 0)”, i.e. when they showed the composite image between Stimulus 1 and Stimulus 2. “Note that the neural activity follows the sudden increase in behavioral learning”.

The authors recorded from over 600 neurons from various MTL regions, out of which 51 responded to a Stimulus 1. From these, only half learned, that is, they increased their activity when Stimulus 2 was shown. For the picky specialist, the cells were both Type 1 and Type 2 neurons, located 6 in the hippocampus, 4 in the entorhinal cortex, 11 in the parahippocampal cortex, and 1 in the amygdala. And the authors controlled for familiarity, attentional demands, and other extraneous variables (with some very fancy and hard to follow stats, I might add).

The paper settles an old psychology dispute. Do we learn an association gradually or at once? In other words, do we learn gradually that A and B occur together, or do we learn that the first time we are shown A and B together and the next trials serve just to refine and consolidate the new knowledge? Ison, Quiroga, & Fried (2015) data show that learning happens at once, in an all-or-none fashion.

Reference: Ison, M. J., Quian Quiroga, R., & Fried, I. (1 July 2015). Rapid Encoding of New Memories by Individual Neurons in the Human Brain. Neuron, 87(1): 220-30. doi: 10.1016/j.neuron.2015.06.016. Article | FREE PDF

By Neuronicus, 4 October 2015