Coloured dots indicate positions of probe tips as determined by post-experiment probe length measurements and histology (see Methods). Each colour represents probes from one rat.

a, Current measured at each potential during successive scans with electrical MFB stimulation (red line) in an anaesthetized rat.b, Current–voltage plot from the 5 scans during the 0.5-s period after MFB stimulation. The large increase in current around 0.6 V (dotted line, peak) corresponds to the dopamine redox potential measuredin vitro.c, Time course of current at the dopamine redox potential around electrical MFB stimulation (red line).d–f, Plots, as ina–c, illustrating dopamine response in the same rat, now awake, to room lights being turned on.g, Current changes averaged over all trials (n = 4,418 trials) in which ramping occurred (see Methods) during T-maze running.h, Average current–voltage plot for all identified ramping trials for the time period (−3 to −2 s) indicated by the brackets in the colour plot ing.i, Average current changes induced by tonic MFB stimulation in anaesthetized rats (n = 3).j, Average current–voltage plot from the bracketed time range (10 to 11 s) ini following the onset of stimulation.

a, b, Dopamine concentration recorded in the VMS (a,n = 9 rats) and in the DLS (b,n = 8 rats). Data were first averaged across trials to yield session average traces for each probe in each session. These traces were averaged within rat to obtain one average trace per rat, which were then averaged across rats. These plots differ from those inFig. 1e, f, which considered session averages for each probe to be an independent measure. Shading represents s.e.m across rats.c, d, Distribution of average peak dopamine values for all recordings in VMS (c) and DLS (d). Each colour corresponds to an average peak dopamine concentration measured by a single probe in different sessions.e, f, Proportion of trial averaged dopamine recordings in the VMS (e, out of 300) and DLS (f, out of 262) that displayed a positive (blue) or negative (grey) ramping response during maze running, and an unclassified dopamine profile (red).g, h, Average dopamine concentration in the VMS (g) and DLS (h) for the positive ramping traces (top), the negative ramping traces (middle), and the unclassified traces (bottom). Shading represents s.e.m.

a, Dopamine concentration in a representative trial that included both a phasic response to warning click and a sustained ramping response to goal-reaching.b, c, Average normalized dopamine traces from VMS (b) and DLS (c) probes from all trials (n = 890 and 640, respectively) that showed identified transients after warning click (see Methods). In the VMS, note the sharp increase in dopamine around warning click superimposed on the ramping response that followed the phasic click response. Shading represents s.e.m. calculated across trials.d, Model for dopamine release profiles in the T-maze task. Sharper transient responses are present at the start of maze running (red) and after goal-reaching (cyan). These responses can be superimposed on and modulated independently of the slower ramping signal related to goal proximity (dark blue).

a, Average peak dopamine to unexpected chocolate milk delivery outside the task is positively correlated with peak ramping dopamine measured from the same probes during preceding behavioural training in the maze (n = 146 sessions; Pearson’sR = 0.45,P < 0.0001).b, Average peak dopamine concentration induced by unexpected free reward outside the task (blue) and peak amplitude of dopamine ramping during maze performance just before reward (red; pairedt-test,P < 0.001).

a, Dopamine release modelled as a summation of four weighted transients in response to fixed maze events on short trials (purple) and long trials (orange). The overlap of the transients is reduced on the long trials, resulting in a lower peak dopamine level at the end of the maze run. Thick lines indicate overall average dopamine, and thin lines indicate the averages of each of the 4 transients across 100 simulation runs.b, Relative predicted peak dopamine levels on short and long trials calculated as a linear decay function of trial duration for the simulated model shown ina (black), for the spatial proximity model (light blue), and for the actual experimental data (dark blue).c, An alternative multi-transient model in which the transients (3, inset) are heavily weighted towards the goal location, are highly variable in their time of occurrence, and display a long decay time-course. In this model, the difference between short and long trials is within the noise range of the data. The average of individual transients (inset) across multiple simulations is a smeared version of the single transients that is weighted towards the goal location.d, Data plotted as inb, for the alternative multi-transient model.e, f, Average run speed (e) and acceleration (f) during short (purple) and long (orange) trials, as shown inFig. 2, for all animals and sessions.

a–c, Video tracker traces (a), relative proximity to reward (b), and dopamine concentration (c) measured during a single trial in which a rat paused near the choice point.d–f, Video tracker traces (d), relative proximity to reward (e) and dopamine concentration (f) measured during another trial from a different rat.

a–c, Average peak dopamine levels for M-maze sessions for three individual rats (M31, M36, and M47) in the left (blue) and right (red) end-arms. Blue and red shading indicates sessions in which left and right arm contained the larger reward, respectively. Error bars indicate s.e.m.d, e, Average peak dopamine levels, as ina–c, for value bias T-maze sessions for two rats (M36 and M47).f, g, Average peak dopamine levels, as ina–c, for value bias S-maze sessions for 2 rats (M47 and M35).h–j, Average normalized dopamine levels measured in the high reward arm (light green) and low reward arm (dark green) as rats performed the M-maze (h), T-maze (i) and S-maze (j) tasks.k, l, Dopamine concentration relative to left (k) and right (l) goal-reaching during the first session following a reversal of reward values (session 17 of M47, indicated by an asterisk inc).

a, Average normalized dopamine concentrations measured from VMS probes in one rat (n = 5 sessions) performing the S-maze task as inFig. 3j. Light green line indicates runs to the higher reward goal, and dark green lines to the lower reward goal. Shading indicates s.e.m. Red vertical lines indicate turns.b, Traces, as ina, for the second rat trained on the S-maze task (n = 4 sessions).

a, b, Distribution of selectivity indices, as inFig. 4d, for all probes implanted in the left (a,n = 5 rats) and right (b,n = 3 rats) hemispheres. Note the bias in both groups of the selectivity preference towards negative selectivity indices (right bias, red) relative to the shuffled data (blue).c–e, Biases in average run time (c), percentage of correct responses (d), and arm choices (e) across training blocks. Negative values indicate biases towards the right end-arm.f, Raw average dopamine selectivity indices across training blocks. Note emergence of right bias with training.g, Correlation coefficients (Pearson’sR) computed for each training block between arm choice selectivity indices and dopamine selectivity indices. Error bars indicate confidence limits of the correlations.h, Normalized peak magnitudes of dopamine signals averaged in a 0.5-s window before goal-reaching in sessions with significant (Mann–Whitney U-test,P < 0.05) pre-goal increases. Data are averaged across rats for each training block (for which, left to right,n = 48, 101, 113 and 179 trial averaged recordings, respectively).i, Average (± s.e.m.) dopamine concentration from ramping dopamine sessions in which percentage of correct trials fell above (red,n = 179) or below (blue,n = 92) the learning criterion for T-maze task acquisition (72.5% correct, chi-square test,P < 0.05).j, Average (± s.e.m.) peak dopamine levels from the sessions plotted ini, showing no significant difference between pre- and post-learning periods (t-test,P = 0.44).

This file contains a Supplementary Discussion and additional references. (PDF 265 kb)