OBJECTIVE Results of a previous report showed that, if administered by intraventricular injection to access tissue normally protected by the blood-brain barrier, gadopentetate dimeglumine produced acute excitation, persistent ataxia, and widespread brain lesions in rats at 5-micromol/g brain but not at 3.8-micromol/g brain. The present study using gadodiamide was undertaken to see whether the effects were agent-specific. METHODS Rats, surgically prepared with a lateral ventricular cannula, were administered a slow injection at 2 microL/min of gadodiamide into the lateral ventricle, and behavioral and neuropathologic changes were noted. RESULTS Both gadodiamide and gadopentetate dimeglumine produced focal and generalized myoclonus over several hours. Gadodiamide did not produce the medium-term tremor or persistent ataxia seen after treatment with gadopentetate dimeglumine. Neuropathologic changes developed over 1 to 3 days and took three distinct forms: vacuolated thalamic lesions closely resembling those produced by gadopentetate dimeglumine; small but similar vacuolated symmetrical caudate lesions not produced by gadopentetate dimeglumine; and severe swelling and astrocytic hypertrophy and hyperplasia in the cerebellar vermis, again not produced by gadopentetate dimeglumine. Unlike gadopentetate dimeglumine, gadodiamide produced no spinal cord lesions. The cerebellar changes were seen at 1.25-micromol/g brain and above, behavioral changes at 2.5-micromol/g brain and above, and thalamic and caudate lesions at 10-micromol/g brain, the maximal dose used. Markedly reducing the rate of injecting the same volume over 28 hours prevented the acute excitation but did not reduce the severity of the morphologic effects. CONCLUSIONS The acute excitatory effects of high intraventricular doses of gadopentetate dimeglumine and gadodiamide are similar and appear to be attributable to local action at the infusion site, but differences exist between the two agents in the character and topography of the distant morphologic changes. The cerebellum was the brain area most sensitive to gadodiamide in this experimental model. It is unlikely that gadodiamide would gain access to the brain at these tissue doses when used intravenously for conventional clinical imaging, but our experimental model suggested that it had some unexpectedly specific neuropathologic potential.