Cronbach envisioned the test, which he labelled “multiple true false”, as a useful measure
of precision ofknowledge for technical or content-specific vocabulary. But he also noted
its limitations, including difficulty in testing verbs, abstract nouns, and words referring to
large entities.
Familiarity tasks
The assessment of precision of word knowledge was also examined by Curtis and Glaser
(1983). Their approach involved first presenting items that required only a minimal degree
of familiarity with a word, followed by items that required discrimination among related
concepts. For example, an item that tested minimal familiarity might require selection of
a synonym for desist from among the following: stop, review, consider, debate. A more
discriminating item for “desist” might then present the following choices: pause, halt,
prevent, discontinue.
The notion of assessing both vague and precise recognition of word meanings was used
by Marshelek (1981) to investigate aspects of vocabulary knowledge in high school
seniors. Students’ knowledge was assessed through vague-recognition items in which all
distractors were semantically unrelated to the correct choice; accurate-recognition items,
in which all distractors were semantically related; and a test of ability to provide
definitions for words. This variety of measures allowed examination of the distribution
of words in various states of knowledge.
Marshelek found that many words could be recognised or defined vaguely but not
accurately, or recognised but not defined. More specifically, it was found that often
students could give a correct example of a word’s use but inferred incorrect defining
features and that low-ability students had more words in partial knowledge states than did
high-ability students.
Semantic tasks
A similar finding that completeness or precision of word knowledge differentiates high-
and low-ability individuals resulted from a study by Curtis (1981). In that study
undergraduates took a traditional multiple-choice test and then were interviewed about the
tested words. Curtis’s interviews included measures of semantic range (i.e., whether
82