In this paper we focus on cohesive social groups that communicate and establish relationships by mobile phone. Through a methodology which identifies cohesive groups and extracts their temporal motifs, we show how the members of social groups interact by means of calls and text messages. Our analysis rests on an anonymized mobile phone dataset, which is based on Call Detail Records (CDRs). This dataset integrates the usual voice call data with the text messages sent by one million mobile subscribers in the metropolitan area of Milan over the span of 67 days. The findings of our study concern both the structural characterization of cohesive groups and the temporal patterns emerging from the interactions among their members. Structurally, cohesive groups are small and people comprise them in ways similar to other social networks or instant messaging services. Temporally, we observe that communication patterns between pairs of group members are predominant, especially for text message communications, where the nature of the medium tends to lead toward “blocking” interactions. Finally, if members participate in more complex communication patterns, text messages make the diffusion of common information easier.