Individuals with different attachment classifications (Secure, Avoidant and Preoccupied) may experience emotional closeness differently, in their intimate relationships but also as clients in psychotherapy. However, evidence for this assumption almost exclusively comes from research with self-report questionnaires. In this paper, we use observer-rated measures to explore in depth how patients with different attachment classifications experience closeness and distance from the therapist in different phases of therapy. Three patients' and their therapists' narratives about the therapeutic relationship at three time points during therapy were extracted and analyzed with two transcript-based observational measures: The Patient Attachment Coding System (PACS), which classifies patients' attachment according to their discourse behavior, and the therapeutic-Distance Scale-Observer version (TDS-O), which assesses the therapeutic relationship in terms of closeness, distance, autonomy and engagement. Cases were chosen from a larger research project due to their different prototypical attachment classification on the PACS. The narratives were obtained from Relationship Anecdote Paradigm (RAP) interviews in which the patients and their therapists narrated separately about meaningful interactions with each other, at early, middle and late phases of therapy. In addition, we followed patients self-report of the alliance and symptoms (OQ-45). Although all patients reported experiencing discomfort with feeling distant from the therapist the therapeutic distance, the secure patient was able to reflect on his feelings and, in the therapist's recollection, was able to share them with the therapist. This allowed the therapist to harness these feelings for the benefit of the therapy. The avoidant and the preoccupied patients both experienced the therapist as distant, but the avoidant patient prevented closeness by a minimal expression of feelings, and the preoccupied described strong frustration with the therapist in a one-sided manner that prevented collaborative processing and left the therapist confused. It appears that patient discourse is a stable (trait-like) component of attachment, while the therapeutic-distance is a process (state-like) component that may change along therapy. The discourse of insecure patients may hinder therapists' ability to adjust the therapeutic-distance to patients' needs. Therapists' knowledge about the ways patients with different attachment classifications communicate their proximity wishes may improve their attunement.