The temporal aspect of PE is usually studied by keylogging or screen recording, and the output measures are either translated words per hour or the total amount of time needed to translate a specific text. In both cases, the time of PE can be compared to HT (or from scratch translation). From an economic point of view, simply saving time would be beneficial. However, in recent years, the time devoted to PE came to be seen as an indicator of cognitive effort, too, since shorter processing time can be associated with the completion of an easier cognitive task.

In general, the results of studies on the time needed for post-editing are mixed, but in most cases, they tend to support its temporal efficiency (e.g., Robert, 2013; Guerberof, 2009, 2010; Plitt & Masselot, 2010; Läubli et al., 2013; Aranberri et al., 2014; Carl et al., 2015; Daems et al., 2017; Arjani & Jamshidiha, 2024; Lee & Choi, 2023). However, some studies failed to find differences in the speed of PE and HT (Garcia, 2010; Carl et al., 2011), and many others stressed that some variables of the translation setting had a huge impact on PE time. In more detail, PE was more efficient when used with certain types of texts, usually, but not always with technical texts (Aranberri et al., 2014) (Jia et al., 2019). Similarly, Wang and Jalalian Daghigh (2023) found that delivery time was significantly longer for expressive texts and operative texts in the HT condition than in the PE condition. However, no significant differences were found for informative texts (Wang & Jalalian Daghigh, 2023). Results in other studies showed large variations depending on what language pairs were involved (Zhechev, 2014; Terribile, 2024), and individual translators seemed to save time with PE to varying degrees (Guerberof, 2014; Koehn & Germann, 2014; Escartin & Arcedillo, 2015). Also, Läubli and his colleagues (2013) found that PE did reduce translation time in a realistic work environment, but considerably less than in experimental settings.

As mentioned above, post-editing time is often seen as an indicator of cognitive effort (Koponen et al., 2012). Nevertheless, time spent on post-editing can be influenced by factors other than cognition; in consequence, several other techniques are used to assess cognitive effort in PE. Such techniques include keylogging (pauses, re-writing segments several times), and eye-tracking (duration of fixations, saccade lengths and regressions). Most of these studies suggest that PE requires less cognitive effort than HT (e.g., Koglin, 2015; Carl et al., 2015; Schaeffer et al., 2016; Daems et al., 2017; Jia et al., 2019; Wang & Jalalian Daghigh, 2023). Studying cognitive efforts in PE is one of the most popular research fields, many studies, however do not focus on comparing HT and PE, but on other factors that may interact with cognitive effort in PE (e.g identifying types of texts or segments of texts whose post-editing is more demanding – O’Brien, 2004; Doherty and O’Brien, 2014, Popovic et al., 2014; or the predictive power of average MT error rate – Daems et al., 2015). Wang and Jalalian (2023) studied the effect of text type on cognitive effort and found that PE was less demanding than HT regardless of text types.

Some studies (e.g., Gaspari et al., 2014, Carl et al., 2015) compared translators’ perception of effort and objective measures of errors (i.e., data from keylogging or eye-tracking). An interesting finding is that translators often perceive PE to be time-consuming and inefficient, but objective data suggest the opposite. This contradiction can be related to negative attitudes that translators have towards PE.

A third aspect of productivity is technical effort, which is most often expressed in edit distance metrics (computing the number of edits between the MT output and the edited version; this is often referred to as technical effort). By definition, these metrics cannot be used to compare HT and PE, but provide useful information on what types of errors need the largest number of edits or what characterises texts or sentences that must be extensively edited. Technical effort can also be studied in relation to temporal effort (how much time is needed to edit certain segments/errors) and cognitive effort (identifying errors and planning corrections) (e.g., Koponen, 2012, 2016). Koponen suggested that temporal effort is related to both technical and cognitive effort, with cognitive effort being more dominant. Interestingly, she found that technical effort is not always in line with cognitive effort, meaning that low cognitive effort (finding mistakes fast and planning corrections quickly) is not always accompanied by low technical effort (introducing the changes fast), and the other way round. In a more recent study, Cumbreño and Aranberri found only weak correlations between the three types of efforts, which led the authors to the conclusion that just one measure of the three may not give a full picture of efforts needed in a particular task (Cumbreño & Aranberri, 2021).