Purpose: It has been reported that the second-order kernel response components of multifocal electroretinograms (mERGs) reflect the electrical activity of the inner retinal layers. In this study, we have investigated whether the amplitudes of the second-order kernel response components correlate with the spatial distribution of human retinal ganglion cells.Methods: Multifocal electroretinograms were recorded using the Veris III system from 5 healthy subjects with different stimulus and recording parameters. The mERGs were analyzed using the Veris Science software programs. The stimuli consisted of densely arranged arrays of 103, 61, 37 or 19 hexagonal elements. Four minutes were required to record one set of mERG responses using 8 sessions, and 8 minutes using 16 sessions. The second-order kernel response components were extracted and analyzed using the Veris Science program.Results: The signal-to-noise ratio of the first-order kernel response components was improved considerably by the summation of the nine reproducible responses from the same subject but the second-order kernel response components were not. The summation of the nine reproducible responses was insufficient to identify an array of the second-order kernel response components. Both the first- and second-order kernel response components were larger when fewer hexagonal elements were used. There was no significant difference in the individual responses between the 4-minute and the 8-minute recordings. A response density analysis revealed a weak correlation between the amplitude distribution of the second-order kernel response components and the spatial distribution of human retinal ganglion cells.Conclusions: The distribution of the amplitudes of the second-order kernel response components of the mERGs elicited from normal subjects did not correlate with the distribution of human ganglion cells. This suggests that the theory that second-order kernel response components arise from the activity of retinal ganglion cells should be reconsidered.