Defending the Yes No Test Nonword Approach with Regression Analysis
Raymond Stubbe · Nagasaki University's Academic Output SITE (Nagasaki University) · 2018
Recent research has questioned the validity of the yes-no (YN) vocabulary test format. It has been found that YN test-takers who do not claim knowledge of any nonwords (zero false alarmers) still overestimated their knowledge of the tested items, thereby questioning the effectiveness of the whole YN nonword approach. This study will demonstrate that the nonword approach should not be abandoned for two reasons. First, the results from the zero false alarmers in one study can be used to create a regression-based scoring formula for adjusting YN test results in a different study to better reflect those test-takers' actual vocabulary knowledge as demonstrated on a translation (L2 to L1) test. Secondly, applying this regression-based formula to the simplest of the established YN scoring formula, h-f, accounted for the increases in over-estimation of word knowledge nonwords were originally designed to indicate. Practically half of the individually predicted scores (n = 455) fell within five percentage points (plus or minus 4.8 words of the 96) of her/his actual translation test score. Fully 85% of these scores were within ten percentage points of matching translation score. This suggests nonwords should be maintained, in conjunction with appropriate scoring formulae. Classroom implications are presented.