@@ -223,8 +223,12 @@ def read_parquet(self, path, columns=None, metadata=None, schema=None,
         table : pyarrow.Table
         """
         from pyarrow.parquet import ParquetDataset
-        dataset = ParquetDataset(path, schema=schema, metadata=metadata,
-                                 filesystem=self)
+
+        if metadata is not None:
+            raise ValueError(
+                "Keyword 'metadata' is not supported with the Dataset API")
+
+        dataset = ParquetDataset(path, schema=schema, filesystem=self)
         return dataset.read(columns=columns, use_threads=use_threads,
                             use_pandas_metadata=use_pandas_metadata)