4 years ago
Authors used cyclic voltametry to measure dopamine concentrations in realtime (10 readings per second) in freely moving rats in mazes. Dopamine concentration ramps up as rats approach a reward. Instantaneous [DA] correlates better with position-on-track than time-from-start-of-trial. The ramps are more apparent in Ventral-medial striatum (which is considered to be more involved in motivation), than in Dorsal-lateral striatum (which is more strongly linked to motor output). On some trials, [DA] ramped downwards as rats approached food reward. [DA] isn't related to running speed. And the steepness of the ramp is related to the expected size of the upcoming reward (these expectations were constant within training days, not quickly varying - but different parts of the maze won the rats different sizes of reward).
Authors suggest that extended ramps of [DA] line the path to distant goals, providing motivation along the way.
Rats searching for food on mazes - very ethologically relevant. It's easier in this case to imagine the dynamics of dopamine applying to real-world learning; I find the Schultz-like setups a little too contrived, and too likely to be interpreted through the implicit assumptions of the motivated learning framework. Here we have more of a natural observation.
Most of the paper is observational, rather than forcing the hypothesis that is (a little too strongly) suggested in the title: that these DA profiles are in fact a signal. Such a specific thing would be very hard to demonstrate. Lots of value in the focus on description of DA dynamics under different conditions (changing reward sizes, forced/unforced, straight track vs. track with choices).
For all the papers inferring dopamine concentrations from the spiking of a small number of VTA or SN neurons, it's very cool to see the ACTUAL dopamine readout. Comparing this with said spiking will be a really important step in bridging the literature. I hope voltametry brings some clarity to the diversity of spike patterns in neurons. Even within genetic subsets, individual neurons have very different correlates (citation needed); maybe milti-site voltametry can bring some clarity to that.
The paper would benefit a lot from more single-trial analysis. Averaging a signal across trials can result in what looks like a 'ramp', when in fact each individual trial has only a single 'step' at a different time.