PURPOSE. TO study cycle-by-cycle recording of small-amplitude flicker-
electroretinogram (ERG) responses and analyze results with robust stat
istical methods to estimate the measurement uncertainty. METHODS. Flic
ker ERGs at 32 Hz were recorded simultaneously from both eyes of patie
nts with retinal degeneration. The ERG was amplified under wide-band (
1-1000 Hz) conditions, digitized at 6144 Hz/eye, and multiplied point
for point (192 points/cycle) by sine and cosine functions within each
1/32-second flash cycle to extract coefficients for six harmonic compo
nents of a discrete Fourier transform in real time. Amplitude windowin
g was not used, and all data were saved for subsequent statistical pro
cessing to identify and remove large-amplitude artifacts discretely an
d to search for quiet recording periods that minimized small-amplitude
noise. RESULTS. Plots of amplitude and phase indicated far outlying n
oise points that were excised from the data. The SD of sequential inte
rvals on a time line of the sine component identified quiet periods th
at minimized small-amplitude noise and improved measurement consistenc
y. The SE of the response mean provided an estimate of measurement unc
ertainty. CONCLUSIONS. The harmonic components of many individual resp
onses are captured quickly (e.g., 500 responses in 15.6 seconds) for p
ost hoc statistical analysis, using mathematical algorithms that are p
recisely reproducible to facilitate comparison of results from all lab
oratories. Graphical time lines of responses allow separation of artif
act transients from gaussian noise for elimination of noisy periods wi
thout disturbing the stored information. Statistical estimates of meas
urement uncertainty are determined on-line to allow immediate feedback
during the recording session. Amplitude-phase plots of the multiple h
armonic components, along with reconstructed analog waveforms, provide
results in a readily assimilated manner for comparison of all testing
sessions.