Technical Translation: Usability Strategies for Translating Technical Documentation

Numbers of Participants 
A critical element determining the success of any experiment in terms of 
validity and accuracy is the numbers of participants involved. Quite simply, 
there have to be enough participants in the test to ensure that the results are 
truly representative of the real users as a whole and that the results are not 
just the idiosyncratic behaviour of participants (Dumas & Redish 
1993:128). Nevertheless, it would be impractical and indeed extremely dif-
ficult to use vast numbers of participants. Obviously some form of com-
promise is needed. 
Nielsen (2001) maintains that for the purposes of usability testing in in-
dustry, 3-5 participants are generally sufficient in order to gain an insight 
into the usability of a product. Nielsen makes the point that after the fifth 
participant, most of the problems have already been discovered and any 
subsequent participants will only repeat and confirm what is already known. 
This is echoed by Dumas & Redish (1993:128) who say that “after you’ve 
seen several people make the same mistake, you don’t need to see it a 10
th
or 20
th
or 50
th
time”. As we are comparing two versions of an interface (i.e. 
one user guide with iconic linkage and one without) we will, of course, 
need twice the number of participants. This fits in with claims by Dumas & 
Redish that usability tests generally involve 6-12 participants divided among 
a number of subgroups. A problem with this, however, is that using a small 
number of participants renders it very difficult to perform statistical analyses 
to establish the statistical reliability of the data collected. Many of the more 
commonly used statistical tests such as t-tests etc. require much larger num-
bers. There is, however, a way around this using the Wilcoxon-Mann-
Whitney exact test. This test will be discussed in more detail later.