The Likert scale is the most commonly used scale in survey research. Typically when analyzing Likert-scale data, a total score is computed by summing respondent's responses and a higher score is assumed to represent a large amount of ability or attribute being measured. This assumption is not always true. For example, a regular swimmer may “strongly disagree” with the statement “You like to participate in physical activity regularly, such as walking and running” even if the survey's intent is to measure physical activeness. Fortunately, the Generalized Graded Unfolding Model (GGUM; Robert, Donoghue, & Laughlin, 2000), an item response theory (IRT) model, was recently developed to address these issues. An unfolding model suggests that higher item scores should in all probability, be observed to the extent that the individual and the item are located close to each other on the latent continuum. The purpose of this study is to explore the GGUM model in analyzing health behavior data and compare its performance with the more commonly used IRT model. The data used in this study was from a nutrition and eating behavior survey, in which 479 women (Mean age = 51.82; 61% African American, 37% White & 2% other minority) responded, using the Likert-scale (Strongly disagree = 1 and Strongly agree = 5), to nine items regarding their “changes in healthy eating.” The data was analyzed using both GGUM and rating scale models and were completed using the GGUM2004 software. The model-data fit was evaluated using Infit and Outfit statistics. Both models fit the data well. GGUM model, however, calibrated item and threshold parameters by item (mean thresholds: Category 1 = -4.90, 2 = -2.99, 3 = -5.02, & 4=-5.06), which provided rich information on the items and its categorization while only one set of threshold parameters (Category 1 = -2.73, 2 = -3.36, 3 = -2.02 and 4 =-1.73) were estimated for all items under the rating scale model analysis. The correlation between item difficulty parameters by two models is -.446, but the correlation between respondents' abilities estimates is .81, indicating the GGUM model provided different information regarding both item and respondents. This supports the conclusion that the IRT unfold models should be employed when predicting abilities or attributes based on the distances between a give individual and each item in the measurement scale. Keyword(s): assessment, measurement/evaluation