Free, open, online homework help sites appear to be extremely popular and exist for many school subjects. Students can anonymously post problems at their convenience and receive responses from forum members. This mode of tutoring may be especially critical for school subjects such as calculus that are intrinsically challenging and have high attrition rates. However, educational research has focused on tutoring sessions that instruct students on a pre-determined set of material or topics, and there has been no systematic research on these dynamic, free, open, online tutoring communities. In order to distinguish the student-initiated e-help episodes from traditional tutoring sessions, we refer to them as "tutorettes." Each tutorette was assigned a participation code that contained information on the number of contributions by each participant, the sequence of contributions, and the number of different participants. Student problem solving activity, defined by mathematical contributions and efforts, was measured for initial postings and for subsequent contributions. Finally, each tutorette was examined for evidence of mathematical errors and these were classified according to type: pre-calculus, operational, and conceptual. A tutorette on the limit concept is provided to demonstrate how mathematical queries are resolved in an SOH e-help community. Participation and problem solving attempts provided evidence of active student learning. Instead of simply using the tutors to do their homework, many students made initial attempts at solutions, queried tutor responses, and applied the help they received to make progress on solving problems. This behaviour appeared to be influenced by the actions of the tutor: Providing solution sketches accompanied by asking direct questions encouraged dialogue, whereas providing quasi-complete worked solutions seemed to have the opposite effect. In contrast to classroom instruction, students in this e-help community appeared comfortable in presenting incorrect work and tutors were open and forthright in their commentaries, evaluations, and explanations. In addition, tutors modulated their responses according to the type of error. Pre-calculus errors and operational (calculus) errors were not accorded the same depth of explanation as conceptual misunderstandings.