A central construct from Lyle Bachman and Adrian Palmer's Language Testing in Practice (Oxford University Press, 1996). The TLU domain is the set of real-world language-use situations to which a test is intended to generalise. Test tasks are valid to the extent that their characteristics correspond to the characteristics of language use in that domain.
The construct exists to discipline test development. Without an explicit TLU domain, "reading comprehension at B2" is a label rather than a target; with one, "reading a 700-word university-prospectus excerpt to extract specified factual information under timed conditions" becomes something a passage writer and an item writer can converge on without each importing private assumptions.
Bachman and Palmer (1996) operationalise the TLU domain through five categories of task characteristics. A test task is judged authentic to the extent that its profile across these categories matches the profile of TLU tasks in the domain.
The framework asks the developer to describe each of these for both the TLU tasks and the test tasks, then to inspect the gap. Where the test characteristics drift from the TLU characteristics, the test loses authenticity and risks construct-irrelevant variance.
The TLU lens is what turns construct definition into something a developer can act on. It is the upstream input that should sit on top of every test specification: the spec describes what the test does, the TLU description tells you why those choices are right.
For AI-assisted item generation the TLU domain is the missing input most pipelines lack. Without it, a generator drifts toward whatever passages and questions are statistically frequent in its training distribution: a domain of its own, but rarely the one the test is meant to represent. A formal TLU specification fed alongside the test specification is what closes that gap and lets the generator hit the right target.
Bachman and Palmer also use the TLU framing to redefine authenticity itself, splitting it into two complementary qualities. Situational authenticity is the correspondence between test-task characteristics and TLU-task characteristics, the alignment described above. Interactional authenticity is the degree to which the test task engages the candidate's language ability, topical knowledge, and affective resources in ways characteristic of real language use. A task can be situationally authentic yet interactionally hollow, or vice versa; the framework asks for both.