ESL placement testing is commonly conducted at the beginning of students' studies to determine which level of study would be most appropriate. However, serious mismatches can occur between ESL placement results and the actual progress made by students between levels. The present study was conducted at the University of Hawaii to develop a placement test that would match the curriculum for ESL reading. The usual placement procedures were conducted (N= 194), using a well‐established instrument to test reading comprehension. The 61 students who placed into our reading course were then retested using the same instrument at the end of the 15‐week term. The results were analyzed using classical norm‐referenced and criterion‐referenced item analyses statistics. When the results of the norm‐referenced analyses (item facility and discrimination indexes) were combined with those of the criterion‐referenced analyses (particularly the difference index), serious mismatches between the testing procedures and our ESL program were revealed. The combined information from both sets of analyses was used to revise the placement test and to improve the match between the test and our specific read