This paper presents empirical tests supporting the linear transform model of functional magnetic resonance imaging (fMRI) responses in human primary visual cortex (V1). The model posits that fMRI responses are proportional to the local average neural activity, averaged over time and space. Three key findings support this model: (1) fMRI responses depend separately on stimulus contrast and timing; (2) responses to long stimuli can be predicted from shorter ones; and (3) noise is independent of contrast and temporal period. These results are consistent with the linear model, allowing estimation of the temporal fMRI impulse-response function and contrast-response function of V1. The study used two types of stimuli: periodic and pulse. Periodic stimuli involved slowly moving vertical bars of checkerboard patterns, while pulse stimuli involved brief flickering checkerboard patterns. The fMRI responses were analyzed by measuring the amplitude of the sinusoid best fitting the time course of the response. The results showed that fMRI responses increase with stimulus contrast and are delayed and blurred relative to the stimulus. The time-course data were consistent with time-contrast separability, where responses to different contrasts were scaled copies of each other. The linear transform model was also tested using pulse stimuli, where responses to longer pulses were predicted by summing responses to shorter pulses. However, the predictions were systematically inaccurate, possibly due to neural adaptation. Noise analysis showed that fMRI noise is independent of stimulus contrast and temporal period, supporting the linear model. The study also estimated the temporal fMRI impulse-response function and contrast-response function of V1. The impulse-response function was derived from the best-fit parameters of the model, showing a delay of about 2 seconds after stimulus onset. The contrast-response function showed a nonlinear relationship between stimulus contrast and fMRI response, consistent with the nonlinear response of V1 neurons. The results support the linear transform model of fMRI responses in V1, suggesting that fMRI responses are proportional to the average neural activity. The model allows for the estimation of the temporal impulse-response function and contrast-response function of V1, which are critical for understanding the relationship between neural activity and fMRI responses. The study also highlights the importance of time-contrast separability in fMRI analysis and the need for further testing of the linear model using additional data.This paper presents empirical tests supporting the linear transform model of functional magnetic resonance imaging (fMRI) responses in human primary visual cortex (V1). The model posits that fMRI responses are proportional to the local average neural activity, averaged over time and space. Three key findings support this model: (1) fMRI responses depend separately on stimulus contrast and timing; (2) responses to long stimuli can be predicted from shorter ones; and (3) noise is independent of contrast and temporal period. These results are consistent with the linear model, allowing estimation of the temporal fMRI impulse-response function and contrast-response function of V1. The study used two types of stimuli: periodic and pulse. Periodic stimuli involved slowly moving vertical bars of checkerboard patterns, while pulse stimuli involved brief flickering checkerboard patterns. The fMRI responses were analyzed by measuring the amplitude of the sinusoid best fitting the time course of the response. The results showed that fMRI responses increase with stimulus contrast and are delayed and blurred relative to the stimulus. The time-course data were consistent with time-contrast separability, where responses to different contrasts were scaled copies of each other. The linear transform model was also tested using pulse stimuli, where responses to longer pulses were predicted by summing responses to shorter pulses. However, the predictions were systematically inaccurate, possibly due to neural adaptation. Noise analysis showed that fMRI noise is independent of stimulus contrast and temporal period, supporting the linear model. The study also estimated the temporal fMRI impulse-response function and contrast-response function of V1. The impulse-response function was derived from the best-fit parameters of the model, showing a delay of about 2 seconds after stimulus onset. The contrast-response function showed a nonlinear relationship between stimulus contrast and fMRI response, consistent with the nonlinear response of V1 neurons. The results support the linear transform model of fMRI responses in V1, suggesting that fMRI responses are proportional to the average neural activity. The model allows for the estimation of the temporal impulse-response function and contrast-response function of V1, which are critical for understanding the relationship between neural activity and fMRI responses. The study also highlights the importance of time-contrast separability in fMRI analysis and the need for further testing of the linear model using additional data.