Best Model to Predict Vocabulary Learning Gains: <scp>ILH</scp> , <scp>ILH</scp> Plus, or <scp>TFA</scp> ?

ABSTRACT The field of vocabulary instruction has witnessed the emergence of three models that operationalize the construct of elaborate processing and thus help predict the effectiveness of vocabulary learning activities: Involvement Load Hypothesis (ILH), Involvement Load Hypothesis Plus (ILH Plus), and Technique Feature Analysis (TFA). However, no research has compared their predictive power for vocabulary gains. The present study filled this gap. To this end, it employed a between‐group pretest–posttest–delayed‐posttest quasi‐experiment to examine the learning outcome out of ten nonwords through a gap‐filling ( n = 63), a sentence‐writing ( n = 66), and a composition‐writing task ( n = 67). A meaning‐recall test was conducted right after the experiment and 2 weeks later to measure vocabulary uptake and retention. A control group ( n = 21) was also included in this experiment to gauge test‐taking effects. All three activities were found to yield vocabulary gains: Composition‐writing &gt; Sentence‐writing &gt; Gap‐filling. The gain size was well‐predicted by ILH Plus, but not by ILH or TFA. Time‐on‐task could explain both vocabulary uptake and retention, but prior L2 proficiency could only do so for vocabulary retention. These findings provide useful implications for both instruction and research practice.