Testing the myth of the encoding–retrieval match