The spread of solution after a standardized paravertebral injection was studied to determine the precision and predictability of paravertebral spread. The spread of 5 ml of a solution of radiological contrast medium (sodium iothalamate) and local anesthetic mixture after 45 (34 thoracic, 11 lumbar) paravertebral injections was studied in 31 patients by radiography and computed tomography and correlated with the clinical effects. Spread confined to the paravertebral area occurred after only eight (18%) injections. Spread was epidural after 31 (70%) injections and exclusively so in 14 (31%) injections. Mean sensory loss was greater after epidural spread, but a wide range of sensation loss was observed with all patterns of spread. Intrapleural spread occurred after three injections, as did spread into the psoas muscle. In addition, measurements were made of 114 paravertebral spaces in 20 patients by means of computed tomography. Dimensional factors identified as possibly leading to complications of a paravertebral injection included narrow width of the thoracic transverse processes (mean, 3.18 cm; range, 2.1-4.2 cm) and the wide range in paravertebral dimensions. The distance from bony landmarks to pleura frequently fell outside the limits recommended by many standard texts. We conclude that the spread of a small volume of solution after paravertebral injection is imprecise and unpredictable. Neurolytic and diagnostic paravertebral injections performed without the aid of radiological imaging and contrast media should be regarded as hazardous and interpreted with extreme caution.