diff --git a/pandas/core/arrays/numpy_.py b/pandas/core/arrays/numpy_.py
index 47517782e2bbf..791ff44303e96 100644
--- a/pandas/core/arrays/numpy_.py
+++ b/pandas/core/arrays/numpy_.py
@@ -222,7 +222,7 @@ def __getitem__(self, item):
             item = item._ndarray
 
         result = self._ndarray[item]
-        if not lib.is_scalar(result):
+        if not lib.is_scalar(item):
             result = type(self)(result)
         return result
 
diff --git a/pandas/tests/extension/numpy_/__init__.py b/pandas/tests/extension/numpy_/__init__.py
new file mode 100644
index 0000000000000..e69de29bb2d1d
diff --git a/pandas/tests/extension/numpy_/conftest.py b/pandas/tests/extension/numpy_/conftest.py
new file mode 100644
index 0000000000000..daa93571c2957
--- /dev/null
+++ b/pandas/tests/extension/numpy_/conftest.py
@@ -0,0 +1,38 @@
+import numpy as np
+import pytest
+
+from pandas.core.arrays.numpy_ import PandasArray
+
+
+@pytest.fixture
+def allow_in_pandas(monkeypatch):
+    """
+    A monkeypatch to tell pandas to let us in.
+
+    By default, passing a PandasArray to an index / series / frame
+    constructor will unbox that PandasArray to an ndarray, and treat
+    it as a non-EA column. We don't want people using EAs without
+    reason.
+
+    The mechanism for this is a check against ABCPandasArray
+    in each constructor.
+
+    But, for testing, we need to allow them in pandas. So we patch
+    the _typ of PandasArray, so that we evade the ABCPandasArray
+    check.
+    """
+    with monkeypatch.context() as m:
+        m.setattr(PandasArray, '_typ', 'extension')
+        yield
+
+
+@pytest.fixture
+def na_value():
+    return np.nan
+
+
+@pytest.fixture
+def na_cmp():
+    def cmp(a, b):
+        return np.isnan(a) and np.isnan(b)
+    return cmp
diff --git a/pandas/tests/extension/test_numpy.py b/pandas/tests/extension/numpy_/test_numpy.py
similarity index 84%
rename from pandas/tests/extension/test_numpy.py
rename to pandas/tests/extension/numpy_/test_numpy.py
index 7ca6882c7441b..4c93d5ee0b9d7 100644
--- a/pandas/tests/extension/test_numpy.py
+++ b/pandas/tests/extension/numpy_/test_numpy.py
@@ -6,7 +6,7 @@
 from pandas.core.arrays.numpy_ import PandasArray, PandasDtype
 import pandas.util.testing as tm
 
-from . import base
+from .. import base
 
 
 @pytest.fixture
@@ -14,28 +14,6 @@ def dtype():
     return PandasDtype(np.dtype('float'))
 
 
-@pytest.fixture
-def allow_in_pandas(monkeypatch):
-    """
-    A monkeypatch to tells pandas to let us in.
-
-    By default, passing a PandasArray to an index / series / frame
-    constructor will unbox that PandasArray to an ndarray, and treat
-    it as a non-EA column. We don't want people using EAs without
-    reason.
-
-    The mechanism for this is a check against ABCPandasArray
-    in each constructor.
-
-    But, for testing, we need to allow them in pandas. So we patch
-    the _typ of PandasArray, so that we evade the ABCPandasArray
-    check.
-    """
-    with monkeypatch.context() as m:
-        m.setattr(PandasArray, '_typ', 'extension')
-        yield
-
-
 @pytest.fixture
 def data(allow_in_pandas, dtype):
     return PandasArray(np.arange(1, 101, dtype=dtype._dtype))
@@ -46,18 +24,6 @@ def data_missing(allow_in_pandas):
     return PandasArray(np.array([np.nan, 1.0]))
 
 
-@pytest.fixture
-def na_value():
-    return np.nan
-
-
-@pytest.fixture
-def na_cmp():
-    def cmp(a, b):
-        return np.isnan(a) and np.isnan(b)
-    return cmp
-
-
 @pytest.fixture
 def data_for_sorting(allow_in_pandas):
     """Length-3 array with a known sort order.
diff --git a/pandas/tests/extension/numpy_/test_numpy_nested.py b/pandas/tests/extension/numpy_/test_numpy_nested.py
new file mode 100644
index 0000000000000..cf9b34dd08798
--- /dev/null
+++ b/pandas/tests/extension/numpy_/test_numpy_nested.py
@@ -0,0 +1,286 @@
+"""
+Tests for PandasArray with nested data. Users typically won't create
+these objects via `pd.array`, but they can show up through `.array`
+on a Series with nested data.
+
+We partition these tests into their own file, as many of the base
+tests fail, as they aren't appropriate for nested data. It is easier
+to have a seperate file with its own data generating fixtures, than
+trying to skip based upon the value of a fixture.
+"""
+import pytest
+
+import pandas as pd
+from pandas.core.arrays.numpy_ import PandasArray, PandasDtype
+
+from .. import base
+
+# For NumPy <1.16, np.array([np.nan, (1,)]) raises
+# ValueError: setting an array element with a sequence.
+np = pytest.importorskip('numpy', minversion='1.16.0')
+
+
+@pytest.fixture
+def dtype():
+    return PandasDtype(np.dtype('object'))
+
+
+@pytest.fixture
+def data(allow_in_pandas, dtype):
+    return pd.Series([(i,) for i in range(100)]).array
+
+
+@pytest.fixture
+def data_missing(allow_in_pandas):
+    return PandasArray(np.array([np.nan, (1,)]))
+
+
+@pytest.fixture
+def data_for_sorting(allow_in_pandas):
+    """Length-3 array with a known sort order.
+
+    This should be three items [B, C, A] with
+    A < B < C
+    """
+    # Use an empty tuple for first element, then remove,
+    # to disable np.array's shape inference.
+    return PandasArray(
+        np.array([(), (2,), (3,), (1,)])[1:]
+    )
+
+
+@pytest.fixture
+def data_missing_for_sorting(allow_in_pandas):
+    """Length-3 array with a known sort order.
+
+    This should be three items [B, NA, A] with
+    A < B and NA missing.
+    """
+    return PandasArray(
+        np.array([(1,), np.nan, (0,)])
+    )
+
+
+@pytest.fixture
+def data_for_grouping(allow_in_pandas):
+    """Data for factorization, grouping, and unique tests.
+
+    Expected to be like [B, B, NA, NA, A, A, B, C]
+
+    Where A < B < C and NA is missing
+    """
+    a, b, c = (1,), (2,), (3,)
+    return PandasArray(np.array(
+        [b, b, np.nan, np.nan, a, a, b, c]
+    ))
+
+
+skip_nested = pytest.mark.skip(reason="Skipping for nested PandasArray")
+
+
+class BaseNumPyTests(object):
+    pass
+
+
+class TestCasting(BaseNumPyTests, base.BaseCastingTests):
+
+    @skip_nested
+    def test_astype_str(self, data):
+        pass
+
+
+class TestConstructors(BaseNumPyTests, base.BaseConstructorsTests):
+    @pytest.mark.skip(reason="We don't register our dtype")
+    # We don't want to register. This test should probably be split in two.
+    def test_from_dtype(self, data):
+        pass
+
+    @skip_nested
+    def test_array_from_scalars(self, data):
+        pass
+
+
+class TestDtype(BaseNumPyTests, base.BaseDtypeTests):
+
+    @pytest.mark.skip(reason="Incorrect expected.")
+    # we unsurprisingly clash with a NumPy name.
+    def test_check_dtype(self, data):
+        pass
+
+
+class TestGetitem(BaseNumPyTests, base.BaseGetitemTests):
+
+    @skip_nested
+    def test_getitem_scalar(self, data):
+        pass
+
+    @skip_nested
+    def test_take_series(self, data):
+        pass
+
+
+class TestGroupby(BaseNumPyTests, base.BaseGroupbyTests):
+    @skip_nested
+    def test_groupby_extension_apply(self, data_for_grouping, op):
+        pass
+
+
+class TestInterface(BaseNumPyTests, base.BaseInterfaceTests):
+    @skip_nested
+    def test_array_interface(self, data):
+        # NumPy array shape inference
+        pass
+
+
+class TestMethods(BaseNumPyTests, base.BaseMethodsTests):
+
+    @pytest.mark.skip(reason="TODO: remove?")
+    def test_value_counts(self, all_data, dropna):
+        pass
+
+    @pytest.mark.skip(reason="Incorrect expected")
+    # We have a bool dtype, so the result is an ExtensionArray
+    # but expected is not
+    def test_combine_le(self, data_repeated):
+        super(TestMethods, self).test_combine_le(data_repeated)
+
+    @skip_nested
+    def test_combine_add(self, data_repeated):
+        # Not numeric
+        pass
+
+    @skip_nested
+    def test_shift_fill_value(self, data):
+        # np.array shape inference. Shift implementation fails.
+        super().test_shift_fill_value(data)
+
+    @skip_nested
+    def test_unique(self, data, box, method):
+        # Fails creating expected
+        pass
+
+    @skip_nested
+    def test_fillna_copy_frame(self, data_missing):
+        # The "scalar" for this array isn't a scalar.
+        pass
+
+    @skip_nested
+    def test_fillna_copy_series(self, data_missing):
+        # The "scalar" for this array isn't a scalar.
+        pass
+
+    @skip_nested
+    def test_hash_pandas_object_works(self, data, as_frame):
+        # ndarray of tuples not hashable
+        pass
+
+    @skip_nested
+    def test_searchsorted(self, data_for_sorting, as_series):
+        # Test setup fails.
+        pass
+
+    @skip_nested
+    def test_where_series(self, data, na_value, as_frame):
+        # Test setup fails.
+        pass
+
+    @skip_nested
+    def test_repeat(self, data, repeats, as_series, use_numpy):
+        # Fails creating expected
+        pass
+
+
+class TestPrinting(BaseNumPyTests, base.BasePrintingTests):
+    pass
+
+
+class TestMissing(BaseNumPyTests, base.BaseMissingTests):
+
+    @skip_nested
+    def test_fillna_scalar(self, data_missing):
+        # Non-scalar "scalar" values.
+        pass
+
+    @skip_nested
+    def test_fillna_series_method(self, data_missing, method):
+        # Non-scalar "scalar" values.
+        pass
+
+    @skip_nested
+    def test_fillna_series(self, data_missing):
+        # Non-scalar "scalar" values.
+        pass
+
+    @skip_nested
+    def test_fillna_frame(self, data_missing):
+        # Non-scalar "scalar" values.
+        pass
+
+
+class TestReshaping(BaseNumPyTests, base.BaseReshapingTests):
+
+    @pytest.mark.skip("Incorrect parent test")
+    # not actually a mixed concat, since we concat int and int.
+    def test_concat_mixed_dtypes(self, data):
+        super(TestReshaping, self).test_concat_mixed_dtypes(data)
+
+    @skip_nested
+    def test_merge(self, data, na_value):
+        # Fails creating expected
+        pass
+
+    @skip_nested
+    def test_merge_on_extension_array(self, data):
+        # Fails creating expected
+        pass
+
+    @skip_nested
+    def test_merge_on_extension_array_duplicates(self, data):
+        # Fails creating expected
+        pass
+
+
+class TestSetitem(BaseNumPyTests, base.BaseSetitemTests):
+
+    @skip_nested
+    def test_setitem_scalar_series(self, data, box_in_series):
+        pass
+
+    @skip_nested
+    def test_setitem_sequence(self, data, box_in_series):
+        pass
+
+    @skip_nested
+    def test_setitem_sequence_mismatched_length_raises(self, data, as_array):
+        pass
+
+    @skip_nested
+    def test_setitem_sequence_broadcasts(self, data, box_in_series):
+        pass
+
+    @skip_nested
+    def test_setitem_loc_scalar_mixed(self, data):
+        pass
+
+    @skip_nested
+    def test_setitem_loc_scalar_multiple_homogoneous(self, data):
+        pass
+
+    @skip_nested
+    def test_setitem_iloc_scalar_mixed(self, data):
+        pass
+
+    @skip_nested
+    def test_setitem_iloc_scalar_multiple_homogoneous(self, data):
+        pass
+
+    @skip_nested
+    def test_setitem_mask_broadcast(self, data, setter):
+        pass
+
+    @skip_nested
+    def test_setitem_scalar_key_sequence_raise(self, data):
+        pass
+
+
+# Skip Arithmetics, NumericReduce, BooleanReduce, Parsing