There are unique challenges to estimating causal effects of preschool for students with special needs that have not received attention in the literature. We revisit the Head Start Impact Study (HSIS) to illustrate that when and how special needs is defined has implications for the internal validity of and interpretation of special needs subgroup impact estimates. We find that the treatment group in the HSIS was three percentage points more likely to be classified as special education (SD = 0.11, p < 0.001) at baseline, likely biasing the impact estimates for this subgroup. We also find that the estimated intent-to-treat effects of Head Start on cognitive and socioemotional measures are sensitive to subgroup definition.