1. Intracellular recording, antidromic activation, and horseradish peroxidase (HRP) injection techniques were employed to characterize the receptive-field properties and morphology of the superior collicular (SC) neurons in the hamster that projected to the lateral posterior nucleus (LP) or the dorsal lateral geniculate body (LGNd). 2. Twenty-three tecto-LP and 21 tecto-LGNd cells were successfully characterized, filled with HRP, and recovered. Additional physiological information was obtained from four tecto-LP and five tecto-LGNd neurons in which HRP injections did not completely label the cell, but did provide information as to the laminar location of the soma. Recovered neurons were classified as wide-field or narrow-field vertical cells, marginal cells, stellate cells, or horizontal cells on the basis of their soma-dendritic morphology. They were categorized as stationary responsive (SR), movement sensitive (MV), or directionally selective (DS) on the basis of their physiological responses (3, 37). 3. The somas of the recovered tecto-LP cells were located, with two exceptions, in, or near, the borders of the stratum opticum (SO). Tecto-LGNd neurons, with two exceptions, had their cell bodies in the upper one-half of the stratum griseum superficiale (SGS). Fifty-two percent of the recovered tecto-LP cells were wide-field vertical cells, 22% were narrow-field vertical cells, 13% were stellate cells, 9% were horizontal cells, and 4% could not be classified according to the scheme that we employed. Twenty-four percent of the recovered tecto-LGNd cells were marginal cells, 24% were stellate cells, 38% were narrow-field vertical cells, 5% were horizontal cells, 5% were wide-field vertical cells, and 5% could not be classified. The difference between the distributions of morphological cell types that contributed to the tecto-LGNd and tecto-LP pathways was statistically significant (chi 2 = 15.8, P less than 0.01). 4. Sixty-seven percent of the tecto-LP cells had MV receptive fields, 11% were DS, 7% had SR fields, and 15% were unresponsive. The distribution of receptive-field types for tecto-LGNd cells was somewhat different: 54% had SR fields, 15% were MV, 19% were DS, 4% were somatosensory, 4% were unresponsive, and 4% were incompletely classified. These differences between tecto-LP and tecto-LGNd cells were statistically significant (chi 2 = 18.4, P less than 0.001). The strongest correlation between morphology and receptive-field type was observed for the wide-field vertical cells that projected to LP. All but one of these had MV receptive fields.(ABSTRACT TRUNCATED AT 400 WORDS)