The linear transform model of functional magnetic resonance imaging (fMRI) posits that fMRI responses are proportional to the average neural activity averaged over a period of time. This study reports three empirical tests that support this hypothesis: (1) fMRI responses in human primary visual cortex (V1) depend separately on stimulus timing and contrast; (2) responses to long-duration stimuli can be predicted from responses to shorter duration stimuli; and (3) the noise in fMRI data is independent of stimulus contrast and temporal period. Despite these tests not proving the correctness of the linear transform model, they do not reject it. The authors proceed to estimate the temporal fMRI impulse-response function and the underlying neural contrast-response function of human V1. The results show that fMRI responses are separable functions of stimulus timing and contrast, and that the noise in fMRI data is independent of both stimulus contrast and temporal period. The estimated impulse-response functions and contrast-response functions are consistent with the linear transform model, suggesting that fMRI responses can be predicted by convolving the time course of neural activity with a shift-invariant linear temporal filter. The study also discusses the implications of these findings for understanding neural activity and the limitations of the linear transform model.The linear transform model of functional magnetic resonance imaging (fMRI) posits that fMRI responses are proportional to the average neural activity averaged over a period of time. This study reports three empirical tests that support this hypothesis: (1) fMRI responses in human primary visual cortex (V1) depend separately on stimulus timing and contrast; (2) responses to long-duration stimuli can be predicted from responses to shorter duration stimuli; and (3) the noise in fMRI data is independent of stimulus contrast and temporal period. Despite these tests not proving the correctness of the linear transform model, they do not reject it. The authors proceed to estimate the temporal fMRI impulse-response function and the underlying neural contrast-response function of human V1. The results show that fMRI responses are separable functions of stimulus timing and contrast, and that the noise in fMRI data is independent of both stimulus contrast and temporal period. The estimated impulse-response functions and contrast-response functions are consistent with the linear transform model, suggesting that fMRI responses can be predicted by convolving the time course of neural activity with a shift-invariant linear temporal filter. The study also discusses the implications of these findings for understanding neural activity and the limitations of the linear transform model.