Are there long-term "literal copies" of visually presented words?

Evidence for the hypothesis that the appearance of visually presented words is stored in "literal copy" form is critically evaluated and shown to be inconclusive. An experiment in which students were required to retain information about zero, one, or two visual properties of words is reported. A strong version of the literal-copy hypothesis predicts that retention of both case and color should be equally good under instructions to remember one of these attributes and instructions to remember both. However, the results were inconsistent with predictions from this hypothesis. A model that assumes that each visual attribute is stored independently in abstract propositional form provided a good quantitative fit to the data.