Relationships between arousal, LC activity, and behavior.

A) The hypothesized Yerkes-Dodson (inverted-U) relationship between arousal (abscissa, which can be assessed in terms of LC firing rates or non-luminance-mediated changes in pupil diameter) and task performance (right ordinate), with peak performance at moderate levels of arousal. This relationship is thought to reflect specific patterns of LC activation (left ordinate), including task-evoked responses that peak at moderate levels of LC baseline (“phasic mode”) and decline as baseline increases (“tonic mode”). B) Baseline measurements of pupil diameter (top, blue) and the activity of an isolated LC neuron (bottom, red) from a single session. Both were estimated using residuals after accounting for slow (linear) drifts across a recording session (solid lines are linear regression fits). The top inset shows an example pupil response for one trial (green circle) aligned to beep onset. The bottom inset shows an example peristimulus time histogram aligned to beep onset. Shaded regions correspond to the time windows used to quantify baseline (gray) and evoked (magenta) responses.

Comparisons of baseline and evoked relationships between LC single-unit activity and pupil diameter.

The first four columns show four example units/sessions (in each panel, solid and dashed lines are linear regression fits and 95% confidence intervals). The last column summarizes results across all units/sessions from the two monkeys (darker points indicate units with H0: partial correlation=0, p < 0.05; horizontal bars indicate medians across units, thick bars indicate H0: median=0, p < 0.05). First row: Baseline LC firing rate (FR) versus baseline pupil. Second row: Evoked LC versus evoked pupil. Third row: Evoked pupil versus baseline pupil. Fourth row: Evoked LC versus baseline LC. Fifth row: Baseline LC versus evoked pupil. Sixth row: Evoked LC versus baseline pupil.

Comparisons of raw baseline and evoked relationships between LC single-unit activity and pupil diameter.

Summary of correlation coefficients across all units/sessions from the two monkeys (darker points indicate units with H0: partial correlation=0, p < 0.05; horizontal bars indicate medians across units, thick bars indicate H0: median=0, p < 0.05), plotted as in Figure 2 (last column). Baseline values did not account for potential drifts over time. First column: Baseline LC firing rate (FR) versus baseline pupil. Second column: Evoked LC versus evoked pupil. Third column: Evoked pupil versus baseline pupil. Fourth column: Evoked LC versus baseline LC. Fifth column: Baseline LC versus evoked pupil. Sixth column: Evoked LC versus baseline pupil.

Log-likelihood ratio comparisons of linear and quadratic fits to baseline and evoked relationships.

Distribution of log-likelihood ratios comparing the linear and quadratic fits for (from left to right): baseline and evoked pupil responses, baseline and evoked LC activity, evoked pupil responses and baseline LC activity, and evoked LC activity and baseline pupil.

Across-session comparisons of baseline and evoked relationships between neuronal activity and pupillary responses.

Measures of LC spiking activity and pupil diameter (z-scored separately for each session/unit) were pooled across trials, sessions, and units for each monkey (rows). First column: Baseline LC firing rate (FR) versus baseline pupil (including non-beep trials). Second column: Evoked LC versus evoked pupil. Third column: Evoked pupil versus baseline pupil. Fourth column: Evoked LC versus baseline LC. Fifth column: Baseline LC versus evoked pupil. Sixth column: Evoked LC versus baseline pupil. Points (data from individual sessions/units) and lines (linear regression fits) are colored based on whether the within-session relationship between baseline pupil diameter and LC activity was significantly positive (green), negative (orange), or was based on all data (grey; see Fig. 2).