Technical Translation: Usability Strategies for Translating Technical Documentation

Types of Closed Questions 
There is a range of ways in which closed questions can be structured in or-
der to elicit a particular form of data. These range from simple checklists to 
more complex multi-point scales and semantic differential scales. 
190
•
•
•
•
•

---

Survey Methods
Checklists 
are the simplest form of closed question and they ask for basic 
responses to specific questions. This type of closed question is ideal for fac-
tual information such as which software packages respondents have used 
etc. 
Multi-point 
or 
scalar rated questions 
ask respondents to rate a specific 
statement between two polar opposites. This approach is suitable for deter-
mining user opinions. 
Likert scales 
are similar to multi-point scales but in this case, respondents 
are asked to indicate the extent to which they agree or disagree with a 
statement. According to Kirakowski (2000:8), it is necessary to prove “that 
each item of the questionnaire has a similar psychological ‘weight’ in the 
respondent’s mind”. In order to prove the reliability of such scales, some 
form of psychometric evaluation is necessary (
ibid.
). 
Ranked order questions 
dispense with scales and ask respondents to 
number, in order of preference, a series of options. This approach is best 
used with a limited number of options, otherwise respondents may give ar-
bitrary answers (Preece 1994:633).
Semantic differential questions 
are similar to Likert scales but they ask re-
spondents to rate an interface on a scale between two diametrically opposed 
adjectives, e.g. clear-confusing or interesting-boring (Preece 1994:632). 
Multiple-choice questions offer a range of explicit responses and respon-
dents are asked to select either one or more of these options.
A concern with questions that involve some form of scale is the granular-
ity or number of rating points provided to the respondent. This relates to 
the level of detail an evaluator wants to achieve in the collated data. If a 
broad general idea of preferences or opinions is required, a simple three-
point scale is adequate. However, if a more detailed breakdown of prefer-
ences and opinions is required, a greater number of rating points can be 
used. However, simply adding more and more points on the scale can 
prove counter-productive. If we use a ten-point scale, it is likely that some 
respondents may find it difficult to differentiate between any two adjacent 
points on the scale with the result that they may arbitrarily pick points (Dix 
1998:433). 
On a related note, Kirakowski (2000:10) raises the question of whether 
to use an odd or even number of rating points on the scale. The reasoning 
behind this is that with odd-numbered scales where the central point corre-
sponds to neutral opinions or undecided users, respondents may “go on 
auto-pilot” and select neutral points without giving any real thought to 
their choice. As a way of preventing this, an even numbers of option can be 
191

---

Assessing Usability 
used to “force” respondents “to go one way or another” (
ibid.
). This is, ac-
cording to Kirakowski, unhelpful in that it does not cater for respondents 
who genuinely have no preference or strong opinions. It is also common 
for respondents to randomly pick alternate options from the two middle 
options.