Use RANDWORDS to make a unique Yes-No test for any frequency scheme or set of levels

SUMMARY: (1) Get 20 random words for a level; (2) add 10 plausible non-words; (3) alphabetize to integrate and scramble the list; and (4) paste into a column of a Word or PDF doc (Click to see PDF or download MS-Word)   

Yes-No Tests are simple to make, quick to administer, and surprisingly accurate and useful.

In a Yes-No Test, learners are presented with mixed sets of real words and 'plausible non-words' and asked to tick or circle each word if they know it.

The non-words are a control on extreme guessing or failing to take the test seriously. Read Meara and Buxton for the rationale.

The test is scored +1 for every real word circled, and -2 for every non-word circled. There are twice as many words as non-words, so that a test taker who simply circles every item scores zero. A typical question set for a k-level is 20 real words and 10 non-words. (Fewer are unreliable, more are time-consuming).

RANDWORDS makes it easy to get a set of random real words for a given level/framework and then insert 50% as many PNWs (from a set devised by Paul Meara who has written widely on the Y-N test concept).

The suggested method is to generate 25 real head words (uninflected forms) for each relevant k-level, select 20 of these intuitively, add in 10 non-words of roughly the same word length, scramble (done for you here on output pages by lightweight Javascript apps), and integrate into your test.

Which k-levels should you test? Only piloting with your learners will tell you this, but for beginners, 1k, 2k, 3k are a good guess; intermediates 2k, 3k, 4k, and advanced 4k, 5k, 6k (assuming a three-levels test in a three-columns arrangement).

This format makes scoring simple, so that a percentage can be calculated at the bottom of each column. For example, in a given k-level, 15 of the 20 are correctly circled for 15 points; two non-words are circled, for a loss of four points; so the score for that level is (15-4=) 11 out of 20, or 55% (i.e., the learner is not strong at that level - strong means above 85%).

A typical set of scores for 1-2-3k would be 90%-70%-50%. These scores could be used as a diagnostic, to suggest that this learner could stand to focus on the second and third k-level, and might not do well reading a text with large proportions of k-3 items or beyond. This score could also be extrapolated to a probabilistic vocabulary size, in this case 900+700+600=2,200 word families from the first 3,000.

A more complex option is to fully randomize the levels across columns.

NOTE ALSO THAT Lextutor's Reaction Time (Instrument Builder) can build an computer version of the Yes-No test.