Despite an intensive attention on extracting product features from textual reviews, there is limited understanding of the interplay between numerical ratings and textual reviews, and how rating system design moderates the interplay. In this study, we use data from two leading restaurant review websites with different rating systems to analyze the impact of rating system design on opinion sharing. We find that multidimensional (MD) ratings do not substitute textual reviews. Consumers tend to share longer opinions in textual reviews in a more objective way using the MD system. Results from randomized experiments corroborate that MD ratings do not substitute text reviews. Consumersâ€™ decisions to read reviews are not affected by the rating system. MD system primes consumers to generate a more comprehensive numerical overall rating of all dimensions as well as more comprehensive textual reviews. In addition, consumers are also found to use more neutral words in their textual reviews.