In natural environment, higher frequency components of sounds typically have relatively lower intensity compared with lower frequency components. This is also true for animal vocalizations as well as human speeches. Integration of spectrum information is essential for sound cognition, but it is not known how the lower level of higher frequency components is compensated. Previously, frequency-dependent change of response latency has been shown in the primary auditory cortex (A1). Here we tested the hypothesis that frequency-dependent change in response latency is a mechanism for spectrum compensation. To this end, we first quantified the frequency-latency relationship in the guinea pig A1, by recording responses evoked by 20 - 80 dB SPL of narrowband noises centering at 1 kHz or 16 kHz (± 0.25 octave) from the A1 with in vivo optical imaging technique. At 60 dB SPL, 16-kHz-noise evoked responses ~4 ms earlier than 1-kHz-noise at the same intensity. 1-kHz-noise at 70 dB SPL and 16-kHz-noise at 40 dB SPL evoked responses at the same latency in spatially separated positions. To examine whether a 4-ms latency difference can affect cortical integration, we measured responses evoked by 1-kHz-noise and/or 16-kHz-noise in the A1, presented either simultaneously or with one sound systematically delayed at 4-ms steps. Suppressive effect was observed over all locations in A1 for all delays. At the best-frequency position for 16 kHz, the suppressive effect was significantly stronger when 16-kHz-sound was delayed for 4 ms than other cases. This result suggests that the shorter latency of the response to 16-kHz-noise may function to make the response resistant to suppression by lower frequency components, thereby compensating the lower level of high frequency components.