Neurons were grown on plastic surfaces that were untreated, or treated with polylysine, laminin, or L1 and their growth cones were detached from their culture surface by applying known forces with calibrated glass needles. This detachment force was taken as a measure of the force of adhesion of the growth cone. We find that on all surfaces, lamellipodial growth cones require significantly greater detachment force than filopodial growth cones, but this differences is, in general, due to the greater area of lamellipodial growth cones compared to filopodial growth cones. That is, the stress (force/unit area) required for detachment was similar for growth cones of lamellipodial and filopodial morphology on all surfaces, with the exception of lamellipodial growth cones on L1-treated surfaces, which had a significantly lower stress of detachment than on other surfaces. Surprisingly, the forces required for detachment (760-3,340 mudynes) were three to 15 times greater than the typical resting axonal tension, the force exerted by advancing growth cones, or the forces of retraction previously measured by essentially the same method. Nor did we observe significant differences in detachment force among growth cones of similar morphology on different culture surfaces, with the exception of lamellipodial growth cones on L1-treated surfaces. These data argue against the differential adhesion mechanism for growth cone guidance preferences in culture. Our micromanipulations revealed that the most mechanically resistant regions of growth cone attachment were confined to quite small regions typically located at the ends of filopodia and lamellipodia. Detached growth cones remained connected to the substratum at these regions by highly elastic retraction fibers. The closeness of contact of growth cones to the substratum as revealed by interference reflection microscopy (IRM) did not correlate with our mechanical measurements of adhesion, suggesting that IRM cannot be used as a reliable estimator of growth cone adhesion.