To characterize the time course of microcirculatory alterations and their relation to outcome in patients with septic shock. Prospective, observational study. Thirty-one-bed, medico-surgical intensive care unit in a university hospital. Forty-nine patients with septic shock. The sublingual microcirculation was investigated with an orthogonal polarization spectral imaging device on the day of onset of septic shock (baseline) and each day until resolution of shock. Five sequences of 20 secs each were recorded and analyzed off-line by a semiquantitative method. Data were analyzed with nonparametric tests and presented as median (25th-75th percentiles). Three patients died after the resolution of shock from unrelated causes and were excluded. Of the other 46 patients, 26 survived and 20 died: 13 due to unresolving shock and seven due to persistent multiple organ failure after resolution of shock. At the onset of shock, survivors and nonsurvivors had similar vascular density (5.6 [4.7-7.0] vs. 6.2 [5.4-7.0]/mL; p = nonsignificant) and percentage of perfused small vessels (65.0 [53.1-68.9] vs. 58.4 [47.5-69.1]%; p = nonsignificant). Small vessel perfusion improved over time in survivors (analysis of variance, p <.05 between survivors and nonsurvivors) but not in nonsurvivors. Despite similar hemodynamic and oxygenation profiles and use of vasopressors at the end of shock, patients dying after the resolution of shock in multiple organ failure had a lower percentage of perfused small vessels than survivors (57.4 [46.6-64.9] vs. 79.3 [67.2-83.2]%; p =.02). Microcirculatory alterations improve rapidly in septic shock survivors but not in patients dying with multiple organ failure, regardless of whether shock has resolved.