An Introduction to Applied Linguistics

Figure 15.2 The conceptualization of inference in ability testing
One is the ‘ability’ approach, which defines the construct as an unobservable 
trait that is not tied to any particular context of language use. Ability testing 
is based on the point of view that performance on a test is the result of some 
underlying capacities, which are also responsible for performance in non-test 
settings, as illustrated in Figure 15.2. Constructs such as grammatical ability reflect 
an ability perspective toward construct definition because they refer to something 
which would affect performance across many different contexts from talking on 
the phone to a plumber to writing a letter of application for a job at home. Other 
ability constructs would be reading comprehension and vocabulary knowledge.
Performance on a test
Performance in non-
test settings
Inference
Figure 15.3 The conceptualization of inference in performance testing
Another is the ‘performance’ approach, as illustrated in Figure 15.3, which 
aims to make inferences more ‘directly’ from test performance to performance 
outside the test setting. ‘A defining characteristic [of a performance test] is that 
actual performances of relevant tasks are required of candidates, rather than more 
abstract demonstration of knowledge’ such as that required by tests of ability 
(McNamara, 1996: 6). Tests used to measure writing and speaking are often 
referred to as ‘performance tests’ because examinees are asked to use language 
with a simulated purpose, and an inference is made about their probable success 
in speaking or writing in other similar situations.
An ability test is sometimes referred to as ‘indirect’ because of the abstract 
relationship between what examinees do on the test and their potential future 
performance. Performance tests, in contrast, are sometimes called ‘direct’ because 
of the relative directness of the inference. However, this dichotomy is misleading 

---

251
Assessment
because as Bachman (1990) pointed out, all test performance bears an indirect 
relationship to what is being assessed. Test users are always interested not in test 
performance and test scores themselves, but in what test scores mean, that is, 
the inferences that can be drawn from them and what decisions they can make 
with the scores. Moreover, in many testing situations, applied linguists need 
to define the constructs they measure as hybrid ability-performance constructs 
which include an ability as well as a set of contexts in which the ability would be 
relevant (Chapelle, 1998; Bachman, 2007).
Specific/General Purpose Constructs
A second important distinction drawn among various types of construct definitions 
is their degree of specificity, as illustrated by the continuum in Figure 15.4. On 
the left end is the type of construct underlying a specific purpose test for which:
... content and methods are derived from an analysis of a specific purpose target language 
use situation, so that test tasks and content are authentically representative of tasks in 
the target situation. Such a test allows [test users] to make inferences about a test taker’s 
capacity to use language in a specific purpose domain.
(Douglas, 2000: 19)
Specific purpose
General purpose