Technical Translation: Usability Strategies for Translating Technical Documentation

Bottleneck Theories 
Bottleneck theories fall under the categories of early selection and late se-
lection models and they generally revolve around some variation on the 
notion of filters. Indeed, the idea of filters is a key element of both early 
and late selection theories. If we cast our mind back to the idea of the 
cocktail party effect, we will recall that we can filter out stimuli and focus 
on one particular stimulus. 
Early Selection Filter Models 
In early selection filter models, we work on the assumption that only one 
source of information can be processed (Ellis & Hunt 1993:52). Logically, 
this means that unattended information is filtered out before cognitive 
processing takes place, i.e. before the information reaches STM. We can 
see, therefore, that early selection takes place early on in the information-
processing chain of events. 
Perhaps the most well known early selection filter model is 
Broadbent’s 
Switch 
(Ellis & Hunt 1993:52ff ). Broadbent (1958) proposed that our atten-
tion is determined by a filter and a detector located between sensory mem-
ory and STM (Gavin 1998:35). Using the idea of a switch means that we 
process information from one source or channel only in an “all or nothing” 
manner (Ellis & Hunt 
ibid.
). Essentially, if one stimulus is being processed, 
all other stimuli are effectively blocked out. But if we are blocking all of the 
remaining sources of information, how do we remain sufficiently aware of 
our surroundings to be able to operate the switch and shift our attention? 
How do we decide how, when and where to focus our attention? Accord-
ing to Broadbent, the unattended information is subjected to a pre-attentive 
125 

---

Understanding Users 
analysis (i.e. it is analysed before we become aware or conscious of it) 
which examines the physical nature of the incoming information. From our 
discussion of sensory memory earlier we should recall that information is 
stored here in a detailed and unprocessed form. This means that any other 
form of analysis of the information would be impossible before the infor-
mation is passed on to STM. Information which is selected on the basis of 
physical characteristics is then passed along the low capacity channel and 
into STM for processing. The remaining information is left in sensory 
memory where it decays and disappears after 0.2-0.5 seconds. 
Unfortunately, the notion of an “all or nothing” switch does not explain 
the cocktail party effect. If we are concentrating on one particular conver-
sation to the exclusion of all other sensory input, how can we detect our 
name being spoken and change the focus of our attention? The audio input 
arising from our name being spoken is not processed cognitively and as 
such, the physical representation of the sound in sensory memory has no 
meaning for us. This problem is also highlighted by Gavin (1998:36) and 
Ellis & Hunt (1993:54-56) in their discussions of experiments carried out 
by Cherry (1953) and Treisman (1960). Treisman discovered during ex-
periments involving dichotic listening and shadowing
1
that subjects were 
able to report back on some of the content of the unattended information 
and that the unattended information could even affect the performance of 
the attended, shadowing task (Ellis & Hunt 1993:55). It is obvious from this 
that the unattended information is subject to at least some limited form of 
cognitive processing. This presents obvious problems for the application of 
the basic switch model.