Statistical Parametric Mapping

At each voxel, the analysis models measured signal variation using the experimental design matrix. This matrix links observations to conditions or groups, allowing the model to estimate how strongly each factor relates to local brain signals. Repeating the same model across spatial locations produces a brain-wide set of statistical estimates that can be examined for condition-related differences.

The design matrix formally represents the structure of the experiment, including the measurements being compared and the conditions or groups associated with them. Because the voxel-wise model uses this matrix to organize the data, an appropriate design is essential for testing the intended scientific question. Its structure determines which effects can be estimated and compared.

Contrasts specify the particular comparison or effect to test within the fitted model. Rather than treating every modeled component as equally relevant, a contrast can target differences between experimental conditions, groups, or other modeled effects. The resulting statistical map shows where the selected hypothesis is supported, making interpretation dependent on the question encoded by the contrast.

Testing many voxels creates a risk that some locations will appear significant by chance. SPM addresses this problem with multiple-comparison procedures, including random field theory or other correction methods. These approaches adjust the interpretation of statistical evidence across the spatially distributed tests, helping distinguish findings that are more consistent with a meaningful brain-wide pattern from isolated chance results.

A typical workflow begins with spatially distributed fMRI or PET measurements and an experimental design that represents the conditions or groups. A general linear model is then fitted separately at each voxel, followed by testing prespecified contrasts. Multiple-comparison control is applied to the resulting statistics, producing maps that can be interpreted in relation to the original neuroscience question.

The framework is useful when researchers need to localize task-related activity or compare spatially distributed brain signals between conditions. It supports analyses of both fMRI and PET data and can be applied to patient-versus-control comparisons. By connecting experimental design with voxel-level statistical evidence, it helps relate measured neural responses to specific research hypotheses.

Statistical maps can identify locations where neural responses differ across patient and control groups or across experimental conditions. Researchers can also examine relationships between these mapped responses and behavior, provided that relationship is represented in the analysis design. This creates a route from spatial brain measurements to population-level conclusions about neural activity and behavioral variation.