A major impact of email spamming is the ubiquitous deployment of filtering mechanisms removing or tagging emails they classify to be spam. Presumably, spam filters classify email messages into "genuine" and "spam". However, a closer look at the technology reveals significant differences between definitions of spam used in spam discourse (unsolicited commercial email or bulk email) and those operationalized by spam filters. Discourse defines spam almost always in a user-referential way: “unsolicited” refers to the message recipient and his or her attitude towards receiving the message. Definitions of spam operationalized by spam filters, however, focus on representational characteristics of messages. Accordingly, messages may be classified as "solicited" or "unsolicited" regardless of the recipient's attitude towards receiving those messages. In this paper we discuss reasons for the gap between definition and operationalization and look at concepts originating from traditional information filtering/retrieval in order to conceptualize spam filtering in a systematic, theoretically grounded way.