Identifying What Makes Science Vocabulary Easy or Hard to Learn

• In Event: Examining Vocabulary Practices Through the Lenses of Linguistic Frameworks and Analytic Tools

Abstract

OBJECTIVES: The aim of this study was to learn which features of words influence the difficulty of a set of science words. Specifically, we asked, within the domain of science, what word characteristics predict word difficulty before instruction and influence the likelihood that the words would be learned in the course of instruction
THEORETICAL FRAMEWORK: Knowledge about what makes particular words easier or harder to learn in content-area instruction and how that knowledge might impact the selection and instruction of words are less well developed than understandings about general characteristics of effective vocabulary instruction. As Flanigan and Greenwood (2007) note, all words are too often treated equally in content-area lessons—irrespective of the characteristics of the words or differences in their roles in a given text or topic.
METHODS & DATA SOURCES: This study used a large corpus of data on vocabulary from efficacy studies of instructional units of integrated literacy and inquiry-based science. The data in this study comes from assessments of 2,718 students before and after instruction across three grades (Gr. 2: 535; Gr. 3: 676; Gr. 4: 1507). Item types varied from matching items to providing a heading for a list of items. Each of the 95 words in the assessments was assigned a value of dispersion, polysemy, length, frequency of the word in written English, frequency of a word’s morphological family members, concreteness/abstractness, and part of speech.
RESULTS: We first looked at individual predictors and then built models using the individually significant predictors of pretest scores. Polysemy and frequency were significant predictors at all three grades. Length was a significant predictor at grades 2 and 3 but not at grade 4. Stepwise forward and backward regression was used to examine relationships among and variance explained by the individually significant predictors (i.e., polysemy, frequency, and length) and the outcome variable (pretest score). When significant features of polysemy, length, and frequency at grades 2 and 3 were entered into a single regression model, the overall model accounted for 35% and 39% of variance in item p-value. At grade 4, polysemy and frequency accounted for 23% of the variance in item p-value.
To investigate the effects of instruction on vocabulary learning, we used the significant predictors of pretest score to predict growth. Polysemy was a significant predictor of growth at grades 2 and 3, and frequency predicted growth at grades 3 and 4. Length did not predict growth score at any grade level.
SCHOLARLY SIGNIFICANCE: Our results suggest that rare and multiple-meaning words yielded the lowest performance before instruction and predicted growth in word knowledge after science-literacy instruction. Given that many words in science have both a specialized scientific meaning and a more common everyday meaning (e.g., property, energy), it may be useful to target such words for additional instruction. This research also calls into question the practice of adopting a uniform instructional routine for all science vocabulary words. Knowing more about how words are learned by students can help in selecting and instructing words in more informed ways.

Authors

• Gina Cervetti, University of Michigan

• Elfrieda H. Hiebert, TextProject

• P. David Pearson, University of California - Berkeley