The study of speech production has been based heavily on spontaneous s
lips of the tongue. Being observational data, these slips ave usually
jotted down immediately after being heard (''on-line'' data), but ther
e is no possibility of reexamining the correctness or completeness of
this record Tape-based collections (''off-line'' data), which allow th
ese shortcomings to be overcome, are rare. Hence, there is good reason
to question the quality of much of the data. This study reports on th
e quality of slip data. The theoretical concepts of reliability and va
lidity originating in the theory of psychological tests are applied to
slip data with emphasis on the method of collection by comparing publ
ished on-line and off-line corpora. The results of the statistical ana
lysis show that on-line data are neither reliable nor valid.