As an alternative channel to traditional store shopping, Internet shopping has been steadily increasing for the last decade. In view of the expectation of continued growth in e-shopping and its critical impacts on people’s shopping behavior, transportation and even environmental concerns, understanding why people choose one channel over the others and to what degree the chosen channel is adopted becomes an intriguing topic for retail marketing and transportation planning professionals. The contribution of this study is to use latent class modeling (LCM) to explore the effects of channel-specific perceptions, along with other variables, on purchase channel intention. Using data on book purchases collected from an Internet-based survey of two university towns in Northern California, we develop a latent class model with two segments (final N=373).  Age turns out to be the only observed determinant of class membership, and in the intention model, the mostly-younger segment is more cost-sensitive and the mostly-older segment appears to be more convenience-sensitive.  The results clearly demonstrate the effects on purchase intention of channel-specific perceptions, purchase experience, context and socio­demographics. Comparing the LCM to the unsegmented model and to models deterministically segmented on age indicates that the LCM is substantially better, from both interpretability and statistical perspectives.