BACKGROUND & AIMS This study ran within the framework of the Thales Aphasia Project that investigated the efficacy of elaborated semantic feature analysis (ESFA). We evaluated the treatment integrity (TI) of ESFA, i.e., the degree to which therapists implemented treatment as intended by the treatment protocol, in two different formats: individual and group therapy. METHODS & PROCEDURES Based on the ESFA manual, observation of therapy videos and TI literature, we developed two ESFA integrity checklists, for individual and group therapy, and used them to rate 15 videos of therapy sessions, delivered by three speech-language therapists (SLTs). Thirteen people with aphasia (PwA) were involved in this study. Reliability of the checklists was checked using Kappa statistics. Each session's TI was calculated. Differences in TI scores between the two therapy approaches were calculated using independent sample t-tests. Treating SLTs' views on what facilitates TI were also explored through a survey. OUTCOMES & RESULTS Inter- and intra-rater reliability were excellent (.75 ≤ κ ≤ 1.00) for all but one video (κ = .63). Overall, a high TI level (91.4%) was achieved. Although both approaches' TI was high, TI for individual therapy sessions was significantly higher than for group sessions (94.6% and 86.7% respectively), t(13) = 2.68, p = .019. SLTs found training, use of the treatment manual, supervision and peer support useful in implementing ESFA therapy accurately. CONCLUSIONS & IMPLICATIONS ESFA therapy as delivered in Thales is well described and therapists can implement it as intended. The high TI scores found enhance the internal validity of the main research project and facilitate its replication. The need for more emphasis on the methodological quality of TI studies is discussed.