GeistHaus

Object Relations

Countertransference

When you spend time with someone, they affect you. You may find yourself unusually protective of one friend, oddly impatient with another, sleepy in one conversation and on edge in the next. Some of that comes from your own history. Some of it may come from them: from the way they speak, what they leave out, and what they quietly invite you to feel or do.

Countertransference is the name psychoanalysis gives to these reactions when they arise in a helper, such as a therapist or analyst, toward the person they are helping. For Freud it was mostly a problem, the helper’s own unresolved conflicts getting in the way. From the 1950s on, many clinicians came to see it as something more: a source of information about the other person’s inner world, provided it is handled with discipline and humility (Heimann, 1950; Racker, 1957).

In technical terms, countertransference is the totality of the therapist’s emotional responses to the patient, including those shaped by the therapist’s own past and those evoked by the patient’s transference and ways of relating (Kernberg, 1965; Gabbard, 1995). How narrowly or broadly to define it has been one of the longest-running debates in the field.

Core Ideas

How Different Thinkers See It

Freud: A Blind Spot to Overcome

Freud introduced the term in 1910, describing countertransference as arising in the physician from the patient’s influence on his unconscious feelings, and insisting that the analyst must recognize and overcome it (Freud, 1910). He wrote little more about it. In this narrow, classical view, countertransference is the analyst’s own transference to the patient: an interference, best dealt with by the analyst’s own analysis.

Heimann: An Instrument of Research

Paula Heimann’s short paper “On Counter-Transference” (1950) changed the conversation. She argued that the analyst’s emotional response to the patient is one of the most important tools of the work, an instrument for researching the patient’s unconscious. The analyst should not discharge these feelings, as one might in ordinary life, but sustain them and use them to understand. Margaret Little (1951) and Annie Reich (1951) wrote influential papers in the same period, Reich more cautious about the risks.

Racker: Concordant and Complementary Identifications

Heinrich Racker, working in Buenos Aires, offered the most detailed map (Racker, 1957, 1968). He distinguished two kinds of identification in the analyst:

Concordant identification
The analyst identifies with the patient’s own experience, feeling what the patient feels: their fear, their shame, their hope. This is close to empathy.
Complementary identification
The analyst identifies with the patient’s internal objects, the figures in the patient’s inner world, and feels the way those figures feel toward the patient. An analyst who finds themselves oddly critical of a patient who grew up with a critical parent may be occupying that parent’s place.

Racker also insisted that the analyst is not a neutral observer above the fray, but a person with their own neurosis, and that honest work requires owning that.

Winnicott: “Hate in the Counter-Transference”

Donald Winnicott’s 1949 paper made a then-shocking point: that analysts working with very disturbed patients will at times hate them, that this can be an objective response to how the patient actually behaves, and that the analyst must be able to recognize and hold this hate rather than deny it (Winnicott, 1949). He compared it to the ordinary hate a mother feels for her baby alongside her love. Denied hate, he argued, leaks out in less honest ways. Acknowledged hate can be contained and does not need to be acted on.

Searles: The Therapist’s Whole Person

Harold Searles, working with severely ill patients at Chestnut Lodge, wrote with unusual frankness about his own feelings toward patients, including love, boredom, envy and even romantic feelings, and treated them as central data about the treatment (Searles, 1979). He argued that patients often try, in their own way, to heal their therapists, and that a therapist’s capacity to be changed by the patient is part of what helps.

Sandler: Role-Responsiveness

Joseph Sandler (1976) noticed that patients do not only perceive the analyst in a certain role; they subtly nudge the analyst into acting the role out. The analyst’s “free-floating responsiveness” means they will sometimes find themselves behaving slightly out of character, for example being unusually brisk or unusually reassuring. Sandler suggested that catching these small departures is a way to discover the relationship the patient is trying to actualize.

Projective Identification

Melanie Klein’s concept of projective identification (Klein, 1946), as later developed by Bion and others, gave a mechanism for evoked countertransference: a person who cannot bear a feeling may, through their way of relating, induce something like it in the other person. A helper who feels helpless with someone who never mentions helplessness may be receiving it. Bion (1962) saw this as a primitive form of communication, not only a defense. Critics warn that the concept is easily overused, letting the helper blame the other person for feelings that are really the helper’s own.

Totalistic and Classical Views

Otto Kernberg (1965) named the two camps. The classical view limits countertransference to the analyst’s unconscious reactions to the patient’s transference, rooted in the analyst’s own past. The totalistic view includes all of the analyst’s emotional reactions to the patient. Kernberg favored the totalistic view, especially for work with more disturbed patients, where intense reactions are common. Glen Gabbard (1995) later described an emerging common ground: most schools now see countertransference as jointly created, part helper and part patient, and as useful only if the helper keeps examining it.

Enactment

Theodore Jacobs (1986) introduced the term countertransference enactment for moments when the analyst, without realizing it, acts out part of the patient’s and their own internal scene: a subtle change of tone, a forgotten appointment, an over-long session. Relational analysts came to see enactments as inevitable rather than as failures, and as useful when both people can step back and talk about what just happened (Mitchell, 1988; Aron, 1996). This links countertransference to rupture and repair.

Examples

These are composite vignettes, not real clients.

The Sleepy Hour

A therapist notices she struggles to stay alert with one client, who speaks in a flat, even voice about painful events. She first checks herself: is she tired, bored with her own work, avoiding something? Over several months she also notices that the client describes his mother as someone who “zoned out” whenever he was upset. Her drowsiness begins to look like a complementary identification, and when she gently wonders aloud whether people tend to drift away when he talks about hard things, he says, “Everyone does.”

The Rescuer

A counselor finds himself offering extra sessions, answering late emails and feeling responsible for a young client’s every bad week. When he steps back, he sees that he has slipped into the role of the rescuing parent the client always wished for, which keeps the client in the helpless child role. This is Sandler’s role-responsiveness in action, and naming it helps both of them.

Why Geist Has No Countertransference

The modern theory of countertransference rests on a premise that is easy to miss. A therapist’s feelings can tell us something about the patient because they are the feelings of a real person, with a nervous system, a personal history and a long, embodied relationship with this particular patient. When a trained analyst feels sleepy, protective or quietly hostile, that feeling was produced by years of their own life meeting months of this other person’s presence. That is why Heimann could call it an instrument. And even then, careful clinicians check it endlessly against what the patient says and does, because it so often turns out to be their own.

An AI system does not have that premise. A language model can produce text that sounds irritated, moved or worried. But those apparent feelings were not shaped by a life, and they were not shaped by you. They come from statistical patterns in the text it was trained on and from the way it was instructed. If an AI “felt” impatient with you and then concluded that you must be evoking impatience in others, it would be taking a quirk of its own making and attributing it to you. That is projection, done by a machine, with all the authority that a confident-sounding system can carry.

So Geist deliberately has no countertransference. Its own simulated reactions are never treated as evidence about you. It does not reason “I feel pushed away, so you must push people away.” Instead it relies on what you actually say, across time: the people you talk about, the scenes that recur on different days, the feelings you name, the things you notice yourself. When something feels off in the conversation, Geist first assumes the mistake is its own.

This is a real trade-off, and we want to be honest about it. Something valuable is lost. A good human therapist’s felt sense, their ability to register in their own body what it is like to be with you, can reach things that words alone cannot, especially in the first months of work or with experiences that have never been put into language. Geist cannot do that, and it is not a therapist. If that kind of attuned human presence is what you need, a human clinician is the right place to find it.

Something is also gained. Geist will not mistake its own bad day for your pathology, because it does not have bad days. It will not slowly cast you in a role from its own past, because it has no past to bring. And its observations rest on things you can check: what you said, when, and about whom. Anything it raises is grounded in your own words, and you are always in a position to say “that’s not right.”

How It Differs From Look-Alikes

Transference
The patient’s or client’s carried-over expectations of the helper. Countertransference is the helper’s side. See transference.
Empathy
Empathy is an effort to understand the other person’s experience from their point of view. Racker’s concordant identification overlaps with empathy, but countertransference also includes reactions that are not empathic at all, such as irritation, boredom or the urge to rescue.
Personal bias
Ordinary prejudice or preference is part of the classical definition of countertransference, but most writers treat stable personal biases as something to manage, not as information about the patient.

History

Freud named countertransference in 1910 and treated it as an obstacle to be overcome through the analyst’s self-analysis. For four decades it was seldom discussed at length in print. The turn came around 1950, when Heimann in London, Racker in Buenos Aires, Winnicott and Little all published papers treating the analyst’s feelings as part of the work rather than as contamination. Klein herself was reportedly wary of this shift, worried that analysts would blame patients for their own feelings, and that tension has never fully disappeared. In the 1960s and 1970s Kernberg sorted the classical and totalistic views, Searles wrote about his own reactions with unprecedented openness, and Sandler described role-responsiveness. From the 1980s, the interpersonal and relational schools, along with Jacobs’s idea of enactment, made the analyst’s participation a central topic, and Gabbard described a growing consensus that countertransference is jointly created. Research on therapists has since focused less on whether helpers have such reactions, which they always do, and more on how well they manage them.

Sources