Small volume pneumothorax can be challenging to diagnose in horses. The current standard method for diagnosis is standing thoracic radiography. We hypothesized that thoracic ultrasonography would be more sensitive. Objectives of this prospective, experimental study were to describe a thoracic ultrasound method for detection of small volume pneumothorax in horses and to compare results of radiography and ultrasound in a sample of horses with induced small volume pneumothorax. Six mature healthy horses were recruited for this study. For each horse, five 50 ml air boluses were sequentially introduced via a teat cannula into the pleural space. Lateral thoracic radiographs and standardized ultrasound (2D and M-mode) examinations of both hemithoraces were performed following administration of each 50 ml air bolus. Radiographs and ultrasound images/videos were analyzed for detection of pneumothorax by four independent investigators who were unaware of treatment status. Sensitivity, specificity, positive predictive values, negative predictive values, and agreement among investigators (Kappa test, κ) were calculated for radiography, 2D and M-mode ultrasound. Comparisons were made using a chi-squared exact test with significance set at P < 0.05. Two-dimensional (84%) and M-mode (80%) ultrasound were more sensitive than radiography (48%) for pneumothorax detection (P = 0.02 and P = 0.04, respectively). Specificity and positive predictive values were similar for all three imaging modalities (P = 1). Agreement between investigators for pneumothorax detection was excellent for 2D ultrasound (κ = 1), very good for M-mode ultrasound (κ = 0.87), and good for radiography (κ = 0.79). Findings from this experimental study supported the use of thoracic ultrasonography as a diagnostic method for detecting pneumothorax in horses.