diff --git a/Point_set_processing_3/doc_tex/Point_set_processing_3_ref/pca_estimate_normals.tex b/Point_set_processing_3/doc_tex/Point_set_processing_3_ref/pca_estimate_normals.tex
index 3edb3e31b96..5216ada48de 100644
--- a/Point_set_processing_3/doc_tex/Point_set_processing_3_ref/pca_estimate_normals.tex
+++ b/Point_set_processing_3/doc_tex/Point_set_processing_3_ref/pca_estimate_normals.tex
@@ -22,7 +22,7 @@
 \ccc{CGAL::pca_estimate_normals()} estimates normal directions at all points of an input point set by linear least squares fitting of a plane over their $k$ nearest neighbors. The result is an unoriented normal for each input point.
 
 The output of the normal estimation algorithm highly depends on the \ccc{k} parameter.
-The number of neighbors controls the size of the point subset considered for plane fitting at each input point. As this parameter is application-specific we do not provide any default value. For noise-free point sets this value can be set to a small number, e.g., 15. Larger values (e.g., 30 or more) lead to smoother normal fields and are more time consuming. We thus recommend using them only for noisy data sets. 
+The number of neighbors controls the size of the point subset considered for plane fitting at each input point. As this parameter is application-specific we do not provide any default value. For noise-free point sets this value can be set to a small number, e.g., 18. Larger values (e.g., 30 or more) lead to smoother normal fields and are more time consuming. We thus recommend using them only for noisy data sets. 
 
 \ccInclude{CGAL/pca_estimate_normals.h}