A test frame TF is a construction – typically, but not necessarily a sentence – containing a slot for a linguistic item LI to be inserted. Its purpose is to determine whether LI has a certain binary property BP or belongs to a certain class. Insertion of LI into TF converts TF into a complete expression of the language whose acceptability may be assessed. In particular, the resulting construction may be

1. grammatical vs. ungrammatical,
2. semantically unproblematic vs. semantically anomalous,
3. pragmatically appropriate vs. pragmatically inappropriate.

In the first of each of these alternatives, LI passes the test. This is interpreted to signal that LI has, or does not have, BP.

Using TF as a method for assigning LI to a class or ascribing it a vcertain property presupposes the following steps:

1. A binary property BP, or two mutually exclusive classes, of the LI to be tested is formulated.
2. TF is formulated.
3. The function of TF as a criterion for deciding whether LI has BP is formulated; i.e. ‘passes TF’ is mapped onto +BP, and ‘does not pass TF’ is mapped onto -BP (or vice versa).

In the passage from step 1 to step 2, the capability of TF to bring BP in LI out must be deduced; i.e. it must be shown why TF is an appropriate method of determining whether LI has BP. Generally, TF has a crucial property which is syntagmatically compatible only with BP (or its absence). This property of TF may consist in the presence of a lexical item or in a constructional feature. This property of TF, in its turn, has been explained beforehand and can now be taken for granted.

In the clearest cases, the result of putting LI in TF admits a grammaticality judgement. In less clear cases, there is only semantic or pragmatic (in-)compatibility. This means that for one class of LI, the resulting construction conforms to the standard, while for the other class it requires special interpretive effort or special contextual circumstances in order to be passable.

In many cases, in the combination of TF with LI, the crucial property of TF overrides BP; i.e. the result can always be interpreted as acceptable. This is often the case if the crucial property of TF is more explicit than BP. For instance, if temporal adverbs are combined with tenses, the time reference of the result is usually determined by the adverb, overriding the information that the tense might contribute. To that extent, TF does not do what it was supposed to do. Instead it shows that BP is weaker, i.e. less informative, than hypothesized.