Univariate selection is a statistical method used to select features by examining each feature individually to determine its strength of relationship with the response variable. It's particularly useful for reducing dimensionality in datasets with a large number of features, by identifying and retaining only those features that have the strongest individual relationships with the outcome.