Both documents and queries use linguistic encoding of their content. This s
uggests that linguistic text analysis methods improve effectivity and user
friendliness, A retrieval match on the basis of linguistic content represen
tation should lead to better retrieval systems than a match on the basis of
words or even characters. However it is not at all clear to which extend l
inguistic text analysis methods can improve retrieval systems. Evaluations
of retrieval systems with linguistic components lead to different, some tim
es contradictory results, even though the necessary computational linguisti
cs has made enormous progress, We discuss the question where these unintuit
ve results came from. First the state of the art of linguistics in IR is su
mmarized and then results of various evaluations are discussed and compared
.