The results of a pilot study comparing linguistic behavior of speakers in telephone-only and multimodal communicative environments, are reported. The subjects, native speakers of Japanese, conducted goal-oriented conversations in each environment. Linguistic differences (disfluency rates, and use of intention types, syntactic structures, and deictic expressions) and paralinguistic differences (ease of use, and utilization of media) are discussed. Suggestions are made for incorporating these findings into the design of multimodal spoken language interpretation systems.