Background: Interventions for young children with autism are increasingly delivered through parents. This thesis investigated baseline and process variables associating with quality of parent-therapist alliance in a parent-mediated intervention for autism, including variables relating to parents' causal beliefs and perspectives. Participants: 77 parents and 6 therapists from a RCT of a parent-mediated intervention for autism (PACT). Method: A sequential exploratory mixed methods approach, with an intermediary instrument development phase. Baseline variables, investigated for the full sample of 77 cases, comprised demographic variables, parental dichotomous causal belief variables, and a therapist average fidelity variable. Thematic analysis of intervention session transcripts informed the development of the Parental Perspectives Coding Scheme (PPCS), a video-based scheme for rating the quality of parent 'Expression' and therapist 'Integration' of parental perspectives during intervention. 5 overarching themes were initially identified, these were collapsed into three items for the PPCS; Interpretation of the Child (IOC), Parent Actions and Strategies (PAS) and Parental Self Disclosures (PSD). Item inter-rater reliabilities were satisfactory to good. Parent-therapist dialogue, for a sub-sample of 20 cases, was coded using the PPCS to create process variables for the Expression and Integration of parental perspectives. Initial analyses identified baseline and process variables with significant univariate associations with alliance; these were included in separate multivariate models of parent-rated alliance and therapist-rated alliance. Results: Parent-rated and therapist-rated alliance did not correlate. PPCS Expression and Integration scores were higher in the high parent-rated alliance group but the difference was non-significant. Parents who cited MMR as a possible cause of their child's autism rated the alliance significantly lower than those who did not. Parents with no post-16 qualifications rated the alliance significantly higher than those with higher qualifications. Each factor contributed independently to a multiple regression model, together explaining 18.3% of variance in parent-rated alliance. Therapist-rated alliance significantly correlated positively with therapist fidelity and with PPCS variables for parent Expression and therapist Integration; together these explained 58.8% of variance in therapist-rated alliance. Conclusions: Therapists should be aware that parents may rate the alliance differently from themselves and that different factors associate with their ratings. Parents' causal beliefs and level of education may influence their ratings of alliance in specific interventions.