diff --git a/Doc/library/csv.rst b/Doc/library/csv.rst
index 1eb3fe0d4903539..64065a75917af83 100644
--- a/Doc/library/csv.rst
+++ b/Doc/library/csv.rst
@@ -314,6 +314,11 @@ The :mod:`!csv` module defines the following classes:
       is given, it is interpreted as a string containing possible valid
       delimiter characters.
 
+      If several delimiters fit the sample equally well ---
+      for example if both ``','`` and ``';'`` split every row consistently ---
+      the delimiters ``','``, ``'\t'``, ``';'``, ``' '`` and ``':'``
+      are preferred, in this order,
+      no matter how many times each of them occurs.
 
    .. method:: has_header(sample)
 
diff --git a/Lib/csv.py b/Lib/csv.py
index 6efa889ed84677e..92bb9b55a4d3eba 100644
--- a/Lib/csv.py
+++ b/Lib/csv.py
@@ -243,6 +243,10 @@ def __init__(self):
     def sniff(self, sample, delimiters=None):
         """
         Returns a dialect (or None) corresponding to the sample
+
+        If several delimiters fit the sample equally well, the
+        delimiters listed in the preferred attribute are preferred, in
+        that order, no matter how many times each of them occurs.
         """
 
         quotechar, doublequote, delimiter, skipinitialspace = \