Proofreading is the reading of a galley proof or an electronic copy of a publication to detect and correct production errors of text or art. Proofreaders are expected to be consistently accurate by default because they occupy the last stage of typographic production before publication.
A proof is a typeset version of copy or a manuscript page. They often contain typos introduced through human error. Traditionally, a proofreader looks at an increment of text on the copy and then compares it to the corresponding typeset increment, and then marks any errors (sometimes called line edits) using standard proofreaders' marks. Thus, unlike copy editing, proofreading's defining procedure is to work directly with two sets of information at the same time. Proofs are then returned to the typesetter or graphic artist for correction. Correction-cycle proofs will typically have one descriptive term, such as bounce, bump, or revise unique to the department or organization and used for clarity to the strict exclusion of any other. It is a common practice for all such corrections, no matter how slight, to be sent again to a proofreader to be checked and initialed, thus establishing the principle of higher responsibility for proofreaders as compared to their typesetters or artists.
Copy holding or copy reading employs two readers per proof. The first reads the text aloud literally as it appears, usually at a comparatively fast but uniform rate. The second reader follows along and marks any pertinent differences between what is read and what was typeset. This method is appropriate for large quantities of boilerplate text where it is assumed that the number of errors will be comparatively small.
Experienced copy holders employ various codes and verbal short-cuts that accompany their reading. The spoken word digits, for example, means that the numbers about to be read aren't words spelled out; and in a hole can mean that the upcoming segment of text is within parenthesis. Bang means an exclamation point. A thump made with a finger on the table represents the initial cap, comma, period, or similar obvious attribute being read simultaneously. Thus the line of text: (He said the address was 1234 Central Blvd., and to hurry!) would be read aloud as: “in a hole [thump] he said the address was digits 1 2 3 4 [thump] central [thump] buluhvuhd [thump] comma and to hurry bang”. Mutual understanding is the only guiding principle, so codes evolve as opportunity permits. In the above example, two thumps after buluhvuhd might be acceptable to proofreaders familiar with the text.
Double reading. A single proofreader checks a proof in the traditional manner, but then passes it on to a second reader who repeats the process. Both initial the proof. Note that with both copy holding and double reading, responsibility for a given proof is necessarily shared by two individuals.
Scanning, used to check a proof without reading it word for word, has become common with computerisation of typesetting and the popularisation of word processing. Many publishers have their own proprietary typesetting systems, while their customers use commercial programs such as Word. Before the data in a Word file can be published, it must be converted into a format used by the publisher. The end product is usually called a conversion. If a customer has already proofread the contents of a file before submitting it to a publisher, there will be no reason for another proofreader to re-read it from copy (although this additional service may be requested and paid for). Instead, the publisher is held responsible only for formatting errors, such as typeface, page width, and alignment of columns in tables; and production errors such as text inadvertently deleted. To simplify matters further, a given conversion will usually be assigned a specific template. Given typesetters of sufficient skill, experienced proofreaders familiar with their typesetters' work can accurately scan their pages without reading the text for errors that neither they nor their typesetters are responsible for.