We examined the reliability of three traditional P50 auditory event-re
lated potential indices under paired-click conditions: (a) the conditi
oning response (C), (b) the testing response (T), and (c) the testing
to conditioning suppression ratio (T/C). Three alternative indices, (a
) the (C - T) difference, (b) the (C - T)/(C+ T) adjusted difference,
and (c) the (T - T') residualized difference, where T' is the regressi
on of Ton C, were also studied. The N100 wave was used as a generaliza
bility check. Although C and T amplitudes were reliably measured by tr
aditional means, the T/C suppression ratio was not. For psychometric r
easons that are described, the reliability of the suppression ratio is
undermined principally by the correlation between C and T. The C - T
difference score is a promising alternative to the unreliable T/C supp
ression ratio. Theoretical consequences of changed metrics are discuss
ed.