1. Extracellular recordings were made from 81 primate spinothalamic (STT) neurons in the L7-S1 segments of the spinal cord. The majority of the sample was recorded from within laminae IV-V. 2. The temporal features of the responses to noxious thermal stimulation of glabrous and hairy skin were studied in an attempt to determine whether natural groupings of STT neurons could be identified on the basis of response time course alone and whether these groups were skin type dependent. The relationship between these groups and those based on static response features was also explored in an attempt to define more fully their potential functional roles. 3. In most STT neurons, the thermally evoked responses typically appeared to have two response components, particularly at stimulus temperatures above 49°C. The first response phase typically peaked within 1-12 s of stimulus onset and then adapted. The second phase slowly rose to a maximum, typically 15-30 s following stimulus onset. 4. The existence of natural groupings of STT neurons based upon the characteristics of these two response components was assessed with a k-means cluster analysis. On the basis of the onset and early peak latencies, two well-defined short and long latency neuronal clusters were found in the responses evoked from both glabrous and hairy skin; these were referred to as the S(P1) and L(P1) classes, respectively. The glabrous and hairy skin S(P1) classes did not differ significantly in either onset or early peak latency for stimuli of 47-55°C. However, the hairy skin L(P1) class had significantly shorter onset latencies than the glabrous skin L(P1) class for stimuli of 49-53°C, as well as shorter peak latencies for stimuli of 49 and 51°C. The S(P1) class constituted 62% of the hairy skin subset, whereas the L(P1) class constituted 57% of the glabrous skin subset. 5. A cluster analysis of the late-peak latencies also revealed two subgroups. In the responses evoked from both glabrous and hairy skin, the longer latency classes (L(P2)) constituted more than 80% of the samples. With one exception, no dependence upon the type of skin that was stimulated was found in the latencies of either the L(P2) class or the shorter latency S(P2) class. 6. Prior conditioning of the skin with a 30-s thermal pulse of 51-55°C led to a suppression of the early response phase and an enhancement of the late phase in nearly all cases examined (n = 11). This pattern was independent of skin type. 7. On the basis of thermal threshold, response profile, and the effects of conditioning, it is hypothesized that the S(P1) class receives a strong input from C polymodal nociceptors, whereas the glabrous and hairy skin L(P1) classes are presumed to receive input from two subclasses of Aδ polymodal nociceptor. Similar considerations suggest that the S(P2) and L(P2) classes receive input from the type I Aδ polymodal nociceptors. 8. A hierarchial cluster analysis that incorporated the mechanical and thermal clustering results from the preceding paper was performed to determine if a higher-order clustering was present. Because of the poor responsiveness of the A(mnr) class, the data set was limited to the B(mnr), C(mnr), and D(mnr) classes. The results of this analysis suggest that the B(mnr) and L(P1) classes share common members as do the C(mnr), D(mnr), and S(P1) classes. These results suggest that the subset of the STT cell population most concerned with coding cutaneous stimuli is composed of essentially two classes: a class suited to coding potentially damaging cutaneous stimuli and another class suited to coding frankly damaging cutaneous events.