Purpose: Teacher ratings were used to compare children with developmental language disorders (DLD) and their typically developing peers on 2 subtypes of social withdrawal (shyness and unsociability). Measurement invariance analysis was utilized to determine if teachers rated the 2 groups using the same underlying construct for each of the rating scale items that have been designed to assess withdrawn behavior. Method: The Teacher Behavior Rating Scale (TBRS; C. H. Hart & Robinson, 1996) was administered to the teachers of 173 children with DLD and 182 typically developing children (age range: 5;0-12;11 years;months) to compare 2 subtypes of withdrawal, shyness and unsociability. Measurement invariance analysis was used to establish an appropriate basis for comparing the latent group means and other structural parameters, and partial invariance models were used to compare the groups. Results: For the TBRS, shyness and unsociability were measured by 4 and 5 items, respectively. The measurement invariance analysis indicated that classroom teachers approached 1 item on each of these scales using a different underlying construct when rating the 2 groups of children. Taking this into account, the groups were compared. Teachers rated children with DLD as significantly more withdrawn on both shyness and unsociability in comparison with typically developing children. Age and gender were not significant factors. Conclusions: When conducting assessments, it should not be assumed that teachers or other raters approach even commonly used rating scale items using the same underlying perception for typically developing children and children with disorders. However, the analysis of invariant items on the TBRS reconfirmed that children with DLD are at risk for social withdrawal as operationalized by assessments of both shyness and unsociability.