Date of Award


Degree Type


Degree Name

Master of Arts (MA)



Committee Chair

Felicia Roberts

Committee Member 1

Margie Berns

Committee Member 2

Elaine Francis


Previous research has demonstrated price-based lexical variation in American online restaurant reviews and advertising (Freedman and Jurafsky, 2011; Jurafsky, Chahneau, Routledge & Smith 2014). According to the theories of Bourdieu (1986, 1987), and Douglas and Isherwood (2002), this type of variation may be the result of different social classes using differing cultural capital to evaluate their consumer experiences. As reviewers’ lexical choices reflect the criteria they apply to that good, the reviews of more expensive restaurants were expected to use significantly different lexicons than the reviews of less expensive restaurants. In comparing lexical, grammatical and etymological choices across four price rating levels in Turkish online restaurant reviews from Yelp, this study extends the research done previously in American corpora to see if price-based lexical variation occurs as expected in a new language and culture. The results showed multiple examples of price-based variation in the Turkish reviews, and these variations created lexical and etymological patterns that were unique to the Turkish corpora. When reviewing inexpensive restaurants, Turkish reviewers stressed health, homey authenticity and price. As prices increased, the reviewers used different terms for wait staff. The grammatical analysis showed no significant differences, but the etymological analysis showed a growing preference for words with Germanic and Romance roots as the price increased. In conclusion, price-based lexical variation was measurably present in the reviews, though the patterns differed widely from the ones in the American corpora. The measurable lexical differences supported the hypothesis that review criteria vary by the perceived class ranking of the good. Finally, as socioeconomic class is often a neglected variable in sociolinguistics and corpus studies, this study was also valuable in demonstrating a method for identifying class-based register differences in corpus studies.