Currently the test adds a Field to the test Document containing a serializable Reader:

SerializationTestHelper.SerializableStringReader reader = new SerializationTestHelper.SerializableStringReader();
field = new Field( "ReaderField", reader, Field.TermVector.WITH_OFFSETS );
doc.add( field );

Since this reader is defined in another test helper it creates a dependency to this helper. Changing it (for example renaming) will break the de-serialization of old data files. The coupling to the hlper should be avoided. Also it is questionable whether it makes practical sense to have a serializable reader. We should write tests instead verifying what happens if existing JVM provided Reader classed get added. I guess in this case exceptions are thrown.