Shood I ettle on ty-pin speed or accuracy?

Thir examples uise a simple definition o final-character accuracy. Actual studies micht uise minimum string distance tae deal wi inserted, left-out an substituted characters, sae the examples shouldnae be taen as a reproduction o ilka site's accuracy measure. The measurement literature explains why scorin maun specify the intended an transcribed strings. Soukoreff et al., 2003 Soukoreff et al., 2001

Errors left an errors corrected are different

A body that maks mony mistaks an fixes them aw can finish wi accurate text, but spend a hantle o time correctin it. Somebody that leaves mistaks micht seem faster while producin less uissable ootput. Final error rate, correction effort an entry rate sae answer different questions. Soukoreff et al., 2003

The desktop dataset reports baith uncorrected errors an measures related tae correction. Faster participants generally haed fewer errors, but the study wis observational. It disnae shaw that juist forcin ilka body tae type faster will mak them mair accurate, or that slowin doon by a set ammount causes a specified improvement. Dhakal et al., 2018

Why there is nae ae ideal percentage for aw

The cost o an error depends on the task an whit happens efter it. A transcription benchmark measures supplied text; writin or editin involves ither decisions an aw. Foundational reviews o ty-pin an text entry distinguish thir parts o the task an explain why rates fae ae setting dinnae transfer directly tae aw wark. Salthouse, 1986 MacKenzie and Soukoreff, 2002

A laboratory study fand muckle differences atween ty-pin ordinary text an random strings, even amang experienced participants. Familiarity wi the passage is pairt o whit a score measures. A personal best on a simple word list disnae necessarily shaw the same performance on prose wi plenty punctuation. Feit et al., 2016

Device help changes whit the score means

Autocorrect micht replace text without a ane-for-ane sequence o typed characters; prediction can insert a hale wird. The mobile study discusses the limits this brings tae keystroke-level error measures. Comparin assisted phone entry wi unassisted desktop entry as gin their accuracy percentages war identical can hide meaningful differences. Palin et al., 2019

Whit's a sensible wey tae practise?

Uise the same task an record the rate alang wi the relevant error measure. Compare groups o similar unfamiliar passages insteid o pickin ae run. This is a recommendation aboot measurement, no a claimed best dose o trainin. The evidence on practice history amang everyday typists backs task-specific proficiency, but disnae set a universal speed-versus-accuracy trainin ratio. Keith et al., 2007 MacKenzie and Soukoreff, 2002

Gin a faster run leaves mair errors, describe that as a trade-off insteid o an improvement wi nae qualification. Gin final accuracy stays the same while correction effort faas, that's uissfu information an aw. The point o uisin mair nor ae measure is tae mak the chynges clear, insteid o squashin ilka ootcome intae ae flattering nummer. Soukoreff et al., 2003 Soukoreff et al., 2001

Improvement should be described at the level that wis tested an aw. Better ty-pin is a uissfu skill; the cited keyboard studies dinnae mak WPM or accuracy a measure o intelligence, attention disorder or general cognitive enhancement. Their conclusions concern text entry in defined conditions. Dhakal et al., 2018 Feit et al., 2016