diff --git a/pandas/core/dtypes/cast.py b/pandas/core/dtypes/cast.py
index 75434fcc2b40d..fc0a5375b5d75 100644
--- a/pandas/core/dtypes/cast.py
+++ b/pandas/core/dtypes/cast.py
@@ -7,7 +7,7 @@
 
 from pandas._libs import tslib, lib
 from pandas._libs.tslib import iNaT
-from pandas.compat import string_types, text_type, PY3
+from pandas.compat import string_types, text_type, PY3, _default_fill_value
 from .common import (_ensure_object, is_bool, is_integer, is_float,
                      is_complex, is_datetimetz, is_categorical_dtype,
                      is_datetimelike,
@@ -255,6 +255,8 @@ def changeit():
 
 
 def maybe_promote(dtype, fill_value=np.nan):
+    if fill_value is _default_fill_value:
+        fill_value = np.nan
 
     # if we passed an array here, determine the fill value by dtype
     if isinstance(fill_value, np.ndarray):
diff --git a/pandas/core/generic.py b/pandas/core/generic.py
index ba6cfddbc452b..ca53ad44a733a 100644
--- a/pandas/core/generic.py
+++ b/pandas/core/generic.py
@@ -7263,6 +7263,9 @@ def _align_frame(self, other, join='outer', axis=None, level=None,
         clidx, cridx = None, None
 
         is_series = isinstance(self, ABCSeries)
+        if fill_value is _default_fill_value:
+            # XXX: per-column?
+            fill_value = np.nan
 
         if axis is None or axis == 0:
             if not self.index.equals(other.index):
diff --git a/pandas/core/internals.py b/pandas/core/internals.py
index 59813c2a24bfd..0534308cb4857 100644
--- a/pandas/core/internals.py
+++ b/pandas/core/internals.py
@@ -4417,9 +4417,15 @@ def reindex_indexer(self, new_axis, indexer, axis, fill_value=None,
         else:
             if fill_value is None:
                 fill_value = _default_fill_value
-            new_blocks = [blk.take_nd(indexer, axis=axis, fill_tuple=(
-                fill_value if fill_value is not _default_fill_value else blk.fill_value,))
-                for blk in self.blocks]
+
+            new_blocks = []
+            for blk in self.blocks:
+                if fill_value is not _default_fill_value:
+                    fill_tuple = (fill_value,)
+                else:
+                    fill_tuple = (blk.fill_value,)
+            new_blocks = [blk.take_nd(indexer, axis=axis, fill_tuple=fill_tuple)
+                          for blk in self.blocks]
 
         new_axes = list(self.axes)
         new_axes[axis] = new_axis
diff --git a/pandas/core/series.py b/pandas/core/series.py
index 886567500ce71..348a77fd8d9cc 100644
--- a/pandas/core/series.py
+++ b/pandas/core/series.py
@@ -3217,7 +3217,7 @@ def _reindex_indexer(self, new_index, indexer, copy):
             return self
 
         from pandas.core.dtypes.missing import na_value_for_dtype
-        fill_value = na_value_for_dtype(self.dtype)
+        fill_value = na_value_for_dtype(self.dtype, compat=False)
         new_values = algorithms.take(self._values, indexer,
                                      fill_value=fill_value)
         return self._constructor(new_values, index=new_index)
diff --git a/pandas/core/sparse/frame.py b/pandas/core/sparse/frame.py
index 2cefbea722098..b282450efb037 100644
--- a/pandas/core/sparse/frame.py
+++ b/pandas/core/sparse/frame.py
@@ -6,7 +6,7 @@
 # pylint: disable=E1101,E1103,W0231,E0202
 
 import warnings
-from pandas.compat import lmap
+from pandas.compat import lmap, _default_fill_value
 from pandas import compat
 import numpy as np
 
@@ -690,7 +690,7 @@ def _reindex_columns(self, columns, method, copy, level, fill_value=None,
         if level is not None:
             raise TypeError('Reindex by level not supported for sparse')
 
-        if notna(fill_value):
+        if not (isna(fill_value) or fill_value is _default_fill_value):
             raise NotImplementedError("'fill_value' argument is not supported")
 
         if limit: