diff --git a/reciprocalspaceship/dataset.py b/reciprocalspaceship/dataset.py
index 4b0afcf7..8aeb539e 100644
--- a/reciprocalspaceship/dataset.py
+++ b/reciprocalspaceship/dataset.py
@@ -110,28 +110,33 @@ def set_index(self, keys, **kwargs):
         return super().set_index(keys, **kwargs)
 
     def reset_index(self, **kwargs):
+
+        # GH#6: Handle level argument to reset_index
+        columns = kwargs.get("level")
+        if columns is None:
+            columns = list(self._cache_index_dtypes.keys())
+        drop = kwargs.get("drop", False)
+        
+        def _handle_cached_dtypes(dataset, columns, drop):
+            """Use _cache_index_dtypes to restore dtypes"""
+            if drop:
+                for key in columns:
+                    dataset._cache_index_dtypes.pop(key)
+            else:
+                for key in columns:
+                    dtype = dataset._cache_index_dtypes.pop(key)
+                    dataset[key] = dataset[key].astype(dtype)
+            return dataset
         
         if kwargs.get("inplace", False):
             super().reset_index(**kwargs)
-
-            # Cast all values to cached dtypes
-            if not kwargs.get("drop", False):
-                for key in self._cache_index_dtypes.keys():
-                    dtype = self._cache_index_dtypes[key]
-                    self[key] = self[key].astype(dtype)
-                self._cache_index_dtypes = {}
+            _handle_cached_dtypes(self, columns, drop)
             return
-        
         else:
-            newdf = super().reset_index(**kwargs)
-
-            # Cast all values to cached dtypes
-            if not kwargs.get("drop", False):
-                for key in newdf._cache_index_dtypes.keys():
-                    dtype = newdf._cache_index_dtypes[key]
-                    newdf[key] = newdf[key].astype(dtype)
-                newdf._cache_index_dtypes = {}
-            return newdf
+            dataset = super().reset_index(**kwargs)
+            dataset._cache_index_dtypes = dataset._cache_index_dtypes.copy()
+            dataset = _handle_cached_dtypes(dataset, columns, drop)
+            return dataset
 
     @classmethod
     def from_gemmi(cls, gemmiMtz):