6 Assessing productive and interactive skills Personal refl ection

 
are more directly 
based on evidence. He described three alternative empirical approaches.
 
The fi rst of these, 
empirically derived, binary-choice, boundary-
defi nition (EBB)
 
scales,
 
were introduced by Upshur and Turner (1995). 
Performances (written scripts or recordings of speaking test tasks) are 
divided by expert judges into two categories: stronger and weaker 
performances. The judges state their reasons for deciding which perfor-
mance should go into which category. The two sets of performances 
are each then further subdivided into ‘stronger’ and ‘weaker’ categories. 
Again, reasons are stated. The justifi cations made for each division are 
used in preparing a series of yes/no questions that can be followed in 
rating. Each choice is simple and concisely worded and narrows down 
the range of scores that can be used to describe the performance (see 
Figure 6.5
). The rater goes through the yes/no questions step by step 
until he or she arrives at a fi nal score.
The second empirical approach involves 
scaling descriptors
. Large 
numbers of descriptions of performance are collected from rating 
scales or other sources. These are broken down into their constituent