We study how job attitudes, such as job satisfaction, job insecurity, and organizational commitment, are transmitted among between individuals through email communication. We combine longitudinal psychometric surveys with the content of millions of email messages exchanged between organizational members to train machine learning models to identify different linguistic signatures that can predict job attitudes. Initially expecting that messages personally written by each subject (i.e., sent messages) would be most predictive of the employees' responses on the surveys, we instead find that for many subjects the messages they receive (i.e., received messages) provided a better model of job attitudes. We use our models to then impute the net effect of each dyad in the network and classify individuals as either net senders ("influencers) of job attitudes to others or net sinks ("absorbers") of job attitudes in the network. We then analyze other sources of available data to develop a theory of individual differences to which explains why individuals are either attitudinal senders or sinks.