Background. The construct validity and reliability of the short form of the Wolf Motor Function Test (S-WMFT) in people with suhacute stroke and chronic stroke (S-WMFT subacute stroke and chronic stroke versions) have not been investigated. Objective. The purpose of this study was to investigate the dimensionality, item difficulty hierarchy, differential item functioning (DIF), and reliability of the S-WMFT subacute stroke and chronic stroke versions in people with mild to moderate upper-extremity (UE) dysfunction. Design. This was a secondary study in which data collected from randomized controlled tnals were used. Methods. Data were collected at baseline from 97 people with chronic stroke (>12 months after stroke) and 75 people with subacute stroke (3-9 months after stroke) at 3 medical centers in Taiwan. Test structure, hierarchical properties, DIF, and reliability were assessed with Rasch analysis. Results. The test structure for both versions was unidimensional. No DIF relevant to sex, age, or stroke location (hemispheric laterality) was detected. The tasks of moving a hand to a box and moving a hand to a table in the S-WMFT for subacute stroke showed a significantly high correlation. The reliability coefficients for both versions were approximately .90. Limitations. The findings were limited to people with stroke and mild to moderate impairment of UE function. Conclusions. The S-WMFT subacute stroke and chronic stroke versions are useful tools for assessing UE function in different subgroups of people with stroke and show evidence of construct validity and reliability. A high correlation between the tasks of moving a hand to a box and moving a hand to a table in the S-WMFT for subacute stroke suggests that the removal of 1 of these 2 items is warranted. Copyright © 2012 American Physical Therapy Association.