This study investigates the deconvolution of impulse responses in event-related BOLD fMRI to improve temporal resolution and quantify neuronal activity. The BOLD response in sensorimotor and auditory cortices was measured during finger tapping with metronome pacing tones. The response was found to be nonlinear, with amplitude varying slowly for short stimuli. A modified version of Buxton's balloon model predicted this behavior. Wiener deconvolution was used to deblur responses from concatenated short episodes of finger tapping at different rates (1–4 Hz). The deconvolved response at each rate agreed well with separate scans at each rate, indicating that linear deconvolution is effective when stimuli are separated by at least 4 seconds. The deconvolution filter must be measured for each subject using a short-stimulus paradigm. The study concludes that deconvolution can reduce hemodynamically induced temporal blurring and has potential applications in quantitating responses in event-related fMRI. The study also examined the linearity of the BOLD response in sensorimotor and auditory cortices. For short stimuli, the response was nonlinear, with undershoots and amplitude variations. However, deconvolution could recover responses when stimuli were separated by at least 4 seconds. The impulse response function was measured for each subject and used to deconvolve responses from multiple episodes. The results showed that deconvolution could separate individual episodes even when their responses overlapped, provided the episodes were sufficiently spaced. The study used a 1.5-T scanner with high-performance gradients and receiver coils. Data were acquired using a 2D single-shot spiral gradient-recalled sequence. The block trial data were analyzed using cross-correlation with a sine wave. Repeated-trial data were processed to remove trends and averaged over time frames. The results showed that the BOLD response in sensorimotor and auditory cortices was delayed and blurred by the hemodynamic system. The response had a long negative undershoot, which could distort measured time series. The study found that the BOLD response in sensorimotor and auditory cortices was nonlinear, with amplitude varying with stimulus duration. However, deconvolution could recover responses when stimuli were separated by at least 4 seconds. The study also showed that the BOLD response in auditory cortex was more linear for short stimuli, but highly nonlinear for longer stimuli. The results suggest that deconvolution can improve the temporal resolution of event-related fMRI by reducing the blurring caused by the hemodynamic system. The study concludes that deconvolution is effective in diminishing hemodynamically induced temporal blurring and has potential applications in quantitating responses in event-related fMRI.This study investigates the deconvolution of impulse responses in event-related BOLD fMRI to improve temporal resolution and quantify neuronal activity. The BOLD response in sensorimotor and auditory cortices was measured during finger tapping with metronome pacing tones. The response was found to be nonlinear, with amplitude varying slowly for short stimuli. A modified version of Buxton's balloon model predicted this behavior. Wiener deconvolution was used to deblur responses from concatenated short episodes of finger tapping at different rates (1–4 Hz). The deconvolved response at each rate agreed well with separate scans at each rate, indicating that linear deconvolution is effective when stimuli are separated by at least 4 seconds. The deconvolution filter must be measured for each subject using a short-stimulus paradigm. The study concludes that deconvolution can reduce hemodynamically induced temporal blurring and has potential applications in quantitating responses in event-related fMRI. The study also examined the linearity of the BOLD response in sensorimotor and auditory cortices. For short stimuli, the response was nonlinear, with undershoots and amplitude variations. However, deconvolution could recover responses when stimuli were separated by at least 4 seconds. The impulse response function was measured for each subject and used to deconvolve responses from multiple episodes. The results showed that deconvolution could separate individual episodes even when their responses overlapped, provided the episodes were sufficiently spaced. The study used a 1.5-T scanner with high-performance gradients and receiver coils. Data were acquired using a 2D single-shot spiral gradient-recalled sequence. The block trial data were analyzed using cross-correlation with a sine wave. Repeated-trial data were processed to remove trends and averaged over time frames. The results showed that the BOLD response in sensorimotor and auditory cortices was delayed and blurred by the hemodynamic system. The response had a long negative undershoot, which could distort measured time series. The study found that the BOLD response in sensorimotor and auditory cortices was nonlinear, with amplitude varying with stimulus duration. However, deconvolution could recover responses when stimuli were separated by at least 4 seconds. The study also showed that the BOLD response in auditory cortex was more linear for short stimuli, but highly nonlinear for longer stimuli. The results suggest that deconvolution can improve the temporal resolution of event-related fMRI by reducing the blurring caused by the hemodynamic system. The study concludes that deconvolution is effective in diminishing hemodynamically induced temporal blurring and has potential applications in quantitating responses in event-related fMRI.