Perceptual scales are exponential. Light, sound and earthquakes follow this pattern. For sound, for example, 20 db is twice as loud as 10 db, and 30 db is twice as loud as 20 db.

This is why a scale of how much impact an error has on a translation should also be exponential. The ATA has a flowchart for error point decisions that follows a  0, 1, 2, 4, 8 and 16 point scale depending on the effect of the error. For mechanical errors, it works as follows:

0 I, the editor or grader, never knew it was there.
1 Only a typical editor would notice it.
2 A typical reader would notice it.
4 It disrupts my reading. It makes me stumble.


For translation errors, it is approximately as follows.

0 Who knew? It makes no difference.
1 It makes a minor difference.
2 The interference is minimal. I am somewhat disrupted, but not much.
4 The disruption is limited (to a small section). In other words, this error only affects this sentence or paragraph but not much more.
8 It is a serious error, but we can still use the document. People will know what we are talking about.
16 This document is useless.

  • Saying we are in the top five when we are in the bottom two is not OK.
  • The subway runs on Sunday. We can’t say it doesn’t.


The ATA gives an overview of the general expectations of a good translation. At the exam, a candidate can have a maximum of 17 points on each 200-275 words to pass.

This is a reasonable level of quality for certified translators to maintain, once we know what errors we are measuring.