Clinical trials have come to be regarded as the gold standard for treatment evaluation. However, many doctors and their patients experience difficulties when discussing trials, leading to poor accrual to trials and questionable quality of informed consent. We have previously developed a typology for ethical communication about Phase II and III clinical trials within four domains: (a) shared decision making, (b) sequencing information, (c) type and clarity of information, and (d) disclosure/coercion. The aim of this study was to compare current clinical practice when seeking informed consent with this typology. Fifty-nine consultations in which 10 participating oncologists sought informed consent were audiotaped. Verbatim transcripts were analysed using a coding system to (a) identify the presence or absence of aspects of the four domains and (b) rate the quality of aspects of two domains: (i) shared decision-making and (ii) type and clarity of information. Oncologists rarely addressed aspects of shared decision-making, other than offering to delay a treatment decision (78%). Moreover, many of these discussions scored poorly with respect to ideal content. The oncologists were rarely consistent with the sequence of information provision. A general rationale for randomising was only described in 46% of consultations. In almost one third of the consultations (28.8%) doctors made implicit statements favouring one option over another, either standard or clinical trial treatment. Doctors complied with some but not other aspects of a standard procedure for discussing clinical trials. This reflects the difficulty inherent in seeking ethical informed consent and the need for communication skills training for oncologists.