diff --git a/doc/developers/utilities.rst b/doc/developers/utilities.rst
index 5f4364bcc1fddd880cd8392cae9639dcd32da0c3..e1232657dcce7511f5207a0e79bf36a32ed34ef2 100644
--- a/doc/developers/utilities.rst
+++ b/doc/developers/utilities.rst
@@ -192,8 +192,6 @@ ARPACK
   method.  A limited version of ``svds`` is available in earlier
   scipy versions.
 
-- :func:`fixes.arpack_eigsh` [TODO: remove this from spectral_clustering
-  in favor of the above back-port]
 
 Benchmarking
 ~~~~~~~~~~~~
diff --git a/sklearn/cluster/spectral.py b/sklearn/cluster/spectral.py
index f0acb80b647ff17821b0ff05a008e91268266b11..bb4590c86f078d8acd3f62c4c49c1e8d3244a9e1 100644
--- a/sklearn/cluster/spectral.py
+++ b/sklearn/cluster/spectral.py
@@ -61,7 +61,7 @@ def spectral_embedding(adjacency, n_components=8, mode=None,
     """
 
     from scipy import sparse
-    from ..utils.fixes import arpack_eigsh
+    from ..utils.arpack import eigsh
     from scipy.sparse.linalg import lobpcg
     try:
         from pyamg import smoothed_aggregation_solver
@@ -102,8 +102,8 @@ def spectral_embedding(adjacency, n_components=8, mode=None,
                 # csr has the fastest matvec and is thus best suited to
                 # arpack
                 laplacian = laplacian.tocsr()
-        lambdas, diffusion_map = arpack_eigsh(-laplacian, k=n_components,
-                                              which='LA')
+        lambdas, diffusion_map = eigsh(-laplacian, k=n_components,
+                                        which='LA')
         embedding = diffusion_map.T[::-1] * dd
     elif mode == 'amg':
         # Use AMG to get a preconditioner and speed up the eigenvalue
diff --git a/sklearn/utils/fixes.py b/sklearn/utils/fixes.py
index 54c1c5a2aa8958eebca4021f974b3569a9e180b9..3f831933f1b7cb6d5f6283ecb6757e48642d081c 100644
--- a/sklearn/utils/fixes.py
+++ b/sklearn/utils/fixes.py
@@ -130,19 +130,6 @@ def qr_economic(A, **kwargs):
         return scipy.linalg.qr(A, econ=True, **kwargs)
 
 
-def arpack_eigsh(A, **kwargs):
-    """Compat function for sparse symmetric eigen vectors decomposition
-
-    Scipy 0.9 renamed eigen_symmetric to eigsh in
-    scipy.sparse.linalg.eigen.arpack
-    """
-    from scipy.sparse.linalg.eigen import arpack
-    if hasattr(arpack, 'eigsh'):
-        return arpack.eigsh(A, **kwargs)
-    else:
-        return arpack.eigen_symmetric(A, **kwargs)
-
-
 def savemat(file_name, mdict, oned_as="column", **kwargs):
     """MATLAB-format output routine that is compatible with SciPy 0.7's.