This paper offers two contributions to the literature on taxpayer perceptions of the tax system. Our first contribution is that we develop a model that generates predictions over how taxpayers update their beliefs about their marginal tax rates when they experience an unexpected change in their tax liability. Our model is an extension of a standard full information, rational agent model that allows for imperfect knowledge of the income tax schedule. In the model, a household is subject to a linear tax schedule that changes from year to year. The household perceives innovations to the tax schedule with noise due to information gathering and processing costs. When the household is uncertain about the exact tax schedule it faces, it updates its beliefs using signals about the tax schedule generated by interactions with the tax system.