Proofreading

Proofreading is a profession. Unfortunately, proofreading studies have not looked at the complex issue of grammatical errors. Instead, they have focused on the detection of spelling errors.

There are two types of spelling errors. In nonword errors, the misspelling creates a string that is not a real word. So if you are trying to spell the and spell teh, this is a nonword error. Proofreaders often catch 90% of all nonword spelling errors.

For word errors, the misspelling produces a text string that is a real word. If you are trying to spell read and type reed, this is a word error. Humans only catch about 75% of these errors.

Spelling checkers will catch most nonword errors, but they will leave most word errors in the text. In other words, they leave precisely the errors that people have a hard time finding.

Source	Event	Word or Nonword Error	Detection Rate
Cohen, G. [1980]	In proofreading, finding an incorrect word in text	Both	76%-94%
Daneman & Stainton [1993]	Proofreading nonword errors 20 minutes after writing. Own / unfamiliar other's	Nonword Own/Other 20 minutes	81% / 89%
Daneman & Stainton [1993]	Proofreading nonword errors 2 weeks after writing. Own / unfamiliar other's	Nonword Own/Other 2 weeks	88% / 91%
Daneman & Stainton [1993]	Proofreading word errors 20 minutes after writing. Own / unfamiliar other's	Word Own/Other 20 minutes	59% / 76%
Daneman & Stainton [1993]	Proofreading word errors 2 weeks after writing. Own / unfamiliar other's	Word Own/Other 2 weeks	68% / 83%
Dittmer	Proofreading of material with seeded errors. Had correct text for comparison. Pretest excluded subjects with <90% discovery rate. N=9. Alphanumeric text, not prose. Correction rate if shown simultaneously on screen / if shown side by side.	Unknown Shown Simultaneously/ Shown side by side	95.3% /89.3%
Healey [1980]	Proofreading nonword errors in prose by word length. 2 / 3/ 4/ 5/ 6-10. Length 3 may be an artifact.	Nonword 2 characters 3 4 5 6-10	92% 99% 94% 77% 72%
Levy & Begin [1984]	Proofreading prose for spelling errors. Nonword errors found in easy prose if familiar by prior reading, unfamiliar. Based on Levy [1983] but without limited proofreading time.	Nonword Simple Familiar/ Unfamiliar	83%/86%
Levy & Begin [1984]	Proofreading prose for spelling errors. Nonword errors not found in difficult prose if familiar by prior reading, unfamiliar. Based on Levy [1983] but without limited proofreading time.	Nonword Complex Familiar/ Unfamiliar	83%-88%
Levy [1983]	Proofreading prose for spelling errors. Nonword errors found in difficult prose if familiar by prior reading, unfamiliar. Limited proofreading time.	Nonword Time-Constrained	61%-74%
Levy, Di Persio, & Hollingshead [1992]	Proofreading prose for spelling errors. Nonword errors found.	Nonword	86%-87%
Levy, Di Persio, & Hollingshead [1992]	Proofreading prose for spelling errors. Word errors found.	Word	69%
Levy, Di Persio, & Hollingshead [1992]	Proofreading prose for spelling errors. Word errors not found in first reading of a passage / fourth reading. Same passage each time. New errors introduced. Two trials.	Word First Reading Fourth Reading (Different Seeded Errors)	69% - 79% / 69% - 73%
Riefer [1991]	Proofreading for nonword misspellings. Simple material / Difficult material / Difficult material, 2 passes	Nonword Simple Material	84%-86%
		Nonword Difficult	47%-58%
		Nonword Difficult, 2 passes	56%-68%
Riefer [1991]	Proofreading. Nonword / Word errors	Nonword	84%-86%
Riefer [1991]	Proofreading. Nonword / Word errors	Word	42%-62%
Spirrison & Gordy [1993]	Proofreading 385-word passage in 3 minutes. 49 seeded errors. Health effects of vitamin D. 73 Males / 120 Females. Undergraduate students.	Unknown	76%
Wallace [1991]	Proofreading for nonword spelling errors in Time magazine passages. Vivid imagers / Poor imagers.	Nonword	93.5%
Wallace [1991]		Nonword	91.2%