diff --git a/django/db/models/expressions.py b/django/db/models/expressions.py
index 2b883c53ac8..79add29f368 100644
--- a/django/db/models/expressions.py
+++ b/django/db/models/expressions.py
@@ -5,8 +5,7 @@ from django.conf import settings
 from django.core.exceptions import FieldError
 from django.db.backends import utils as backend_utils
 from django.db.models import fields
-from django.db.models.constants import LOOKUP_SEP
-from django.db.models.query_utils import Q, refs_aggregate
+from django.db.models.query_utils import Q
 from django.utils import six, timezone
 from django.utils.functional import cached_property
 
@@ -306,24 +305,6 @@ class BaseExpression(object):
         c.copied = True
         return c
 
-    def refs_aggregate(self, existing_aggregates):
-        """
-        Does this expression contain a reference to some of the
-        existing aggregates? If so, returns the aggregate and also
-        the lookup parts that *weren't* found. So, if
-            existing_aggregates = {'max_id': Max('id')}
-            self.name = 'max_id'
-            queryset.filter(max_id__range=[10,100])
-        then this method will return Max('id') and those parts of the
-        name that weren't found. In this case `max_id` is found and the range
-        portion is returned as ('range',).
-        """
-        for node in self.get_source_expressions():
-            agg, lookup = node.refs_aggregate(existing_aggregates)
-            if agg:
-                return agg, lookup
-        return False, ()
-
     def get_group_by_cols(self):
         if not self.contains_aggregate:
             return [self]
@@ -482,9 +463,6 @@ class F(Combinable):
     def resolve_expression(self, query=None, allow_joins=True, reuse=None, summarize=False, for_save=False):
         return query.resolve_ref(self.name, allow_joins, reuse, summarize)
 
-    def refs_aggregate(self, existing_aggregates):
-        return refs_aggregate(self.name.split(LOOKUP_SEP), existing_aggregates)
-
     def asc(self):
         return OrderBy(self)
 
diff --git a/django/db/models/query_utils.py b/django/db/models/query_utils.py
index 5d087869257..c0ae9030853 100644
--- a/django/db/models/query_utils.py
+++ b/django/db/models/query_utils.py
@@ -93,25 +93,6 @@ class Q(tree.Node):
         query.promote_joins(joins)
         return clause
 
-    @classmethod
-    def _refs_aggregate(cls, obj, existing_aggregates):
-        if not isinstance(obj, tree.Node):
-            aggregate, aggregate_lookups = refs_aggregate(obj[0].split(LOOKUP_SEP), existing_aggregates)
-            if not aggregate and hasattr(obj[1], 'refs_aggregate'):
-                return obj[1].refs_aggregate(existing_aggregates)
-            return aggregate, aggregate_lookups
-        for c in obj.children:
-            aggregate, aggregate_lookups = cls._refs_aggregate(c, existing_aggregates)
-            if aggregate:
-                return aggregate, aggregate_lookups
-        return False, ()
-
-    def refs_aggregate(self, existing_aggregates):
-        if not existing_aggregates:
-            return False
-
-        return self._refs_aggregate(self, existing_aggregates)
-
 
 class DeferredAttribute(object):
     """
@@ -301,20 +282,6 @@ def deferred_class_factory(model, attrs):
         return type(str(name), (model,), overrides)
 
 
-def refs_aggregate(lookup_parts, aggregates):
-    """
-    A helper method to check if the lookup_parts contains references
-    to the given aggregates set. Because the LOOKUP_SEP is contained in the
-    default annotation names we must check each prefix of the lookup_parts
-    for a match.
-    """
-    for n in range(len(lookup_parts) + 1):
-        level_n_lookup = LOOKUP_SEP.join(lookup_parts[0:n])
-        if level_n_lookup in aggregates and aggregates[level_n_lookup].contains_aggregate:
-            return aggregates[level_n_lookup], lookup_parts[n:]
-    return False, ()
-
-
 def refs_expression(lookup_parts, annotations):
     """
     A helper method to check if the lookup_parts contains references
diff --git a/docs/ref/models/expressions.txt b/docs/ref/models/expressions.txt
index d7c4105c4f6..aada3b448dc 100644
--- a/docs/ref/models/expressions.txt
+++ b/docs/ref/models/expressions.txt
@@ -526,23 +526,6 @@ calling the appropriate methods on the wrapped expression.
         A hook allowing the expression to coerce ``value`` into a more
         appropriate type.
 
-    .. method:: refs_aggregate(existing_aggregates)
-
-        Returns a tuple containing the ``(aggregate, lookup_path)`` of the
-        first aggregate that this expression (or any nested expression)
-        references, or ``(False, ())`` if no aggregate is referenced.
-        For example::
-
-            queryset.filter(num_chairs__gt=F('sum__employees'))
-
-        The ``F()`` expression here references a previous ``Sum()``
-        computation which means that this filter expression should be
-        added to the ``HAVING`` clause rather than the ``WHERE`` clause.
-
-        In the majority of cases, returning the result of ``refs_aggregate``
-        on any nested expression should be appropriate, as the necessary
-        built-in expressions will return the correct values.
-
     .. method:: get_group_by_cols()
 
         Responsible for returning the list of columns references by