diff --git a/CHANGELOG.md b/CHANGELOG.md
index d07c664..f08b29f 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -1,3 +1,4 @@
+7/13/18: Fix usage of randoms in correlation functions (issue #85)
 7/11/18: Update code to python3
 3/17/16: Update documentation to Sphinx standard and add documentation build files (issue #17)
 2/17/16: Changes to correlation function plots & documentation (issue #77)
diff --git a/stile/sys_tests.py b/stile/sys_tests.py
index 276dbb9..bffe1ff 100644
--- a/stile/sys_tests.py
+++ b/stile/sys_tests.py
@@ -364,6 +364,16 @@ def makeCatalog(self, data, config=None, use_as_k=None, use_chip_coords=False):
         catalog_kwargs['config'] = config
         return treecorr.Catalog(**catalog_kwargs)
 
+    def hasObjects(self, cat):
+        if hasattr(cat, '__len__'):
+            try:
+                return len(cat)>0
+            except TypeError:
+                return len(numpy.atleast_1d(cat))>0
+        elif hasattr(cat, 'nobj'):
+            return cat.nobj>0
+        return False
+
     def getCF(self, correlation_function_type, data, data2=None,
                     random=None, random2=None, use_as_k=None, use_chip_coords=False,
                     config=None, **kwargs):
@@ -429,11 +439,12 @@ def getCF(self, correlation_function_type, data, data2=None,
 
         # First, pull out the TreeCorr-relevant parameters from the stile_args dict, and add
         # anything passed as a kwarg to that dict.
-        if (random and len(random)) or (random2 and len(random2)):
-            treecorr_kwargs[correlation_function_type+'_statistic'] = \
-                treecorr_kwargs.get(correlation_function_type+'_statistic', 'compensated')
         treecorr_kwargs = stile.treecorr_utils.PickTreeCorrKeys(config)
         treecorr_kwargs.update(stile.treecorr_utils.PickTreeCorrKeys(kwargs))
+        if ((self.hasObjects(random) or self.hasObjects(random2))
+            and correlation_function_type in ['nn', 'ng', 'nk']):
+            treecorr_kwargs[correlation_function_type+'_statistic'] = \
+                treecorr_kwargs.get(correlation_function_type+'_statistic', 'compensated')
         treecorr.config.check_config(treecorr_kwargs, corr2_valid_params)
 
         if data is None:
@@ -500,11 +511,11 @@ def getCF(self, correlation_function_type, data, data2=None,
                 func_dr = None
         elif correlation_function_type == 'nn':
             func_random = treecorr_func_dict[correlation_function_type](treecorr_kwargs)
-            if len(random2):
+            if self.hasObjects(random2):
                 func_random.process(random, random2)
             else:
                 func_random.process(random)
-            if not len(data2):
+            if not self.hasObjects(data2):
                 func_rr = treecorr_func_dict['nn'](treecorr_kwargs)
                 func_rr.process(data, random)
                 if treecorr_kwargs.get(['nn_statistic'],
@@ -519,7 +530,7 @@ def getCF(self, correlation_function_type, data, data2=None,
             else:
                 func_rr = treecorr_func_dict['nn'](treecorr_kwargs)
                 func_rr.process(random, random2)
-                if treecorr_kwargs.get(['nn_statistic'],
+                if treecorr_kwargs.get('nn_statistic',
                    self.compensateDefault(data, data2, random, random2, both=True)
                    ) == 'compensated':
                     func_dr = treecorr_func_dict['nn'](treecorr_kwargs)
@@ -555,10 +566,10 @@ def compensateDefault(self, data, data2, random, random2, both=False):
         indicates that both data sets if present must have randoms; the default, False, means only
         the first data set must have an associated random.
         """
-        if not random or (random and not len(random)):  # No random
+        if not self.hasObjects(random):  # No random
             return 'simple'
-        elif both and data2 and len(data2):  # Second data set exists and must have a random
-            if random2 and len(random2):
+        elif both and self.hasObjects(data2):  # Second data set exists and must have a random
+            if self.hasObjects(random2):
                 return 'compensated'
             else:
                 return 'simple'
diff --git a/tests/test_correlation_functions.py b/tests/test_correlation_functions.py
index 90d191f..7206ff6 100755
--- a/tests/test_correlation_functions.py
+++ b/tests/test_correlation_functions.py
@@ -124,6 +124,21 @@ def test_generator(self):
         self.assertEqual(type(stile.sys_tests.BaseCorrelationFunctionSysTest()), 
                          type(stile.CorrelationFunctionSysTest()))
         
+    def test_randoms(self):
+        """ Run some correlation functions with random catalogs to make sure that also works. """
+        stile_args = {'ra_units': 'degrees', 'dec_units': 'degrees', 'min_sep': 0.05, 'max_sep': 1,
+                      'sep_units': 'degrees', 'nbins': 20}
+        lens_data = stile.ReadASCIITable('../examples/example_lens_catalog.dat',
+                    fields={'id': 0, 'ra': 1, 'dec': 2, 'z': 3, 'g1': 4, 'g2': 5})
+        source_data = stile.ReadASCIITable('../examples/example_source_catalog.dat',
+                    fields={'id': 0, 'ra': 1, 'dec': 2, 'z': 3, 'g1': 4, 'g2': 5})
+
+        object_list = ['GalaxyShear', 'BrightStarShear', 'StarXGalaxyDensity',  'StarXGalaxyShear', 
+                       'StarXStarShear', 'GalaxyDensityCorrelation', 'StarDensityCorrelation']
+        for object_type in object_list:
+            object_1 = stile.CorrelationFunctionSysTest(object_type)
+            results = object_1(lens_data, source_data, lens_data, source_data, config=stile_args)
+        
     
 if __name__=='__main__':
     unittest.main()