Play Therapy

Abstract

Play therapy is a form of psychotherapy in which play, the natural language of childhood, becomes the medium of treatment, letting children express and work through difficulties they cannot yet put into words. This article traces the approach from Klein's and Anna Freud's use of play in child analysis, through Axline's non-directive method and its roots in Rogers's client-centered theory, to Landreth's child-centered play therapy and Schaefer's catalogue of the therapeutic powers of play. It sets out the evidence base, anchored in Bratton and colleagues' meta-analysis of a moderate-to-large pooled effect, the moderators that raise that effect, and the dosage relationship that governs it, and it surveys applications in trauma, schools, and early intervention. Three demonstrations model the effect-size distribution, the pooling of moderated effects, and the session dose-response.

Keywords: play therapy, child-centered play therapy, therapeutic powers of play, effect size, meta-analysis

Ask a distressed six-year-old to describe what is wrong and the request is likely to fail, not because the child feels nothing but because the verbal, reflective account an adult therapy assumes is beyond them. Play therapy begins from the observation that children do their thinking, remembering, and working-through in play, and that a therapy built around a doll house, a sandtray, and a set of figures meets the child in the medium they already command. It is the systematic use of play's therapeutic powers to help children prevent or resolve psychosocial difficulties and achieve optimal growth (Bratton & Ray, 2000). The approach is old, contested, and, on the best current evidence, effective across a moderate-to-large range of outcomes (Bratton et al., 2005).

What Play Therapy Is

Play therapy is a form of psychotherapy, classified by MeSH under the psychotherapies, in which a trained therapist uses the child's play as the primary means of communication and change. Its premise is developmental: symbolic and pretend play is the arena in which young children represent experience, rehearse roles, and regulate feeling, so a treatment conducted in play speaks to the child in a language they have already mastered, where direct verbal enquiry would not. What a child cannot say about a frightening event they can often stage with figures in a sandtray, gaining distance from the material precisely because it is enacted at one remove (Drewes & Schaefer, 2015). This reliance on symbolic representation links play therapy to the broader development of theory of mind and social cognition in the years it typically treats.

The field is not one method but a family, arranged along a continuum from non-directive to directive. At the non-directive pole the therapist follows the child's lead, trusting the child to set the content and pace and providing a permissive, accepting relationship within firm limits; at the directive pole the therapist structures activities toward specific goals, drawing on behavioural principles of reinforcement and skills practice. The best-studied and most widely disseminated form is child-centered play therapy, the non-directive tradition running from Axline to Landreth, and it is the approach most of the outcome literature has evaluated (Ray et al., 2001).

Figure 1

The Directive Continuum of Play Therapy

Historical Origins

Play entered the clinic through psychoanalysis. Klein treated the child's spontaneous play as the equivalent of the adult's free association, reading the arrangement of toys as a direct route to unconscious phantasy, while Anna Freud used play more cautiously, as a way to build the therapeutic alliance before interpretation. Their disagreement over how far play was itself analytic material framed the first generation of child treatment and established that play could be the medium, not merely the setting, of therapy (Bratton & Ray, 2000).

The decisive turn toward the modern research tradition was Axline's. Applying Rogers's client-centered theory to children, she proposed that the same conditions held to drive change in adults, a genuine, accepting, non-directive relationship, would free a child's own drive toward growth, without the analyst's interpretation. Her eight principles, warmth, acceptance, permissiveness, reflection of feeling, respect for the child's capacity to solve their own problems, following rather than directing, patience with the therapeutic process, and only the limits necessary to anchor it in reality, became the operational core of non-directive play therapy. Landreth later systematized this tradition as child-centered play therapy, codifying its rationale, its training, and its characteristic stance that it is the relationship, not the technique, that heals, and it is his formulation that most contemporary trials evaluate (Ray et al., 2013).

The Child-Centered Approach

Child-centered play therapy operationalizes Axline's principles into a small set of therapist responses. The therapist uses tracking, describing aloud what the child does without directing it; reflects content and feeling, naming the emotion the play appears to carry; returns responsibility to the child, letting them make choices and solve problems rather than supplying solutions; and sets limits, stating the few firm boundaries, on safety and on the frame of the session, that make the permissiveness safe. The stance rests on an assumption of an inherent growth tendency: given a relationship of consistent acceptance, the child will move of their own accord toward more organized, regulated, and adaptive functioning (Wilson & Ray, 2018).

This is a demanding discipline precisely because it withholds what an adult instinctively offers, advice, praise, and correction, in favour of accurate reflection and restraint. Limit setting is the point at which the model is most often misunderstood: the non-directive therapist is permissive about feeling and expression but not about behaviour that is unsafe or that breaks the frame, and the skill lies in acknowledging the feeling while holding the limit. The consistency of that relationship, held session after session, is what the approach treats as curative, and it is the ingredient that Landreth's dictum places above any specific technique (Ray et al., 2001).

The Therapeutic Powers of Play

Rather than posit one mechanism, Schaefer's framework catalogues the therapeutic powers of play, the specific change agents through which play produces benefit (Drewes & Schaefer, 2015). Several recur across theoretical orientations. Play affords emotional expression and catharsis, a route to discharge and label feeling that direct conversation does not open. It provides symbolic distancing: enacting a threatening experience through figures at one remove lets a child approach material that would overwhelm them if addressed directly, a graded exposure the child titrates themselves. It supports mastery and rehearsal, as the child replays events to convert a passive experience of helplessness into active control of the scenario. And it works through the therapeutic relationship, the accepting bond that provides the security from which exploration becomes possible.

These powers map onto recognizable cognitive and emotional processes. Symbolic play recruits the same representational capacities that underlie perspective-taking and mentalizing, so that staging a conflict between figures is also practice at modelling other minds, connecting play therapy to the development of mentalization. Repeated enactment builds the self-regulatory control that later supports sustained attention and impulse management, which is why gains in play therapy so often appear as reductions in disruptive, dysregulated behaviour (Wilson & Ray, 2018). The framework's value is that it refuses a single explanation, treating play as a bundle of distinct mechanisms whose relevant members differ by child and by problem.

Table 1. Four recurring therapeutic powers of play, drawn from Schaefer's framework (Drewes & Schaefer, 2015).

The Evidence Base

The empirical case for play therapy rests on a sequence of meta-analyses. The landmark synthesis by Bratton and colleagues pooled ninety-three controlled outcome studies and reported a large overall effect, on the order of d = 0.80, meaning the average treated child moved almost eight-tenths of a standard deviation beyond the average untreated child across a range of behavioural, social, and self-concept outcomes (Bratton et al., 2005). An earlier, independent meta-analysis by LeBlanc and Ritchie had already found a comparable moderate-to-large effect, so the headline conclusion did not rest on a single team (LeBlanc & Ritchie, 2001). Narrowing to the dominant modality, Lin and Bratton's meta-analysis of child-centered play therapy specifically confirmed a statistically significant effect of similar magnitude, and controlled trials since have added randomized evidence on impairment and on specific presenting problems (Lin & Bratton, 2015; Ray et al., 2013).

An effect size of that magnitude is easy to over-read, so it repays a concrete reading. A d of 0.80 does not mean every treated child improves, nor that treated and untreated groups no longer overlap; it describes a shift in the whole distribution. The first demonstration makes the shift visible, letting the reader separate two outcome distributions by an adjustable d and read off the fraction of treated children who exceed the untreated average, the probability that a randomly chosen treated child outscores a randomly chosen untreated one, and the still-substantial overlap that remains.

This distributional reading also disciplines the critics' objection. Early sceptics argued that the play-therapy literature was methodologically weak, its effects inflated by small, uncontrolled studies; Ray and colleagues answered by showing that the effect survived when the analysis was restricted to more rigorous designs, though they granted the field's need for stronger trials (Ray et al., 2001). The randomized trials that followed, on anxiety and on functional impairment, were part of that response (Stulmaker & Ray, 2015; Ray et al., 2013).

Moderators of the Effect

A pooled effect averages over heterogeneous studies, and the moderator analyses are where the practically useful findings sit. Two moderators stand out in Bratton and colleagues' synthesis. The first is theoretical approach: humanistic, non-directive treatments produced a larger effect than non-humanistic, more directive ones, part of why the child-centered tradition dominates practice (Bratton et al., 2005). The second, and stronger, is who delivers the treatment: when a parent or caregiver was trained to conduct the play sessions, in the filial or Child-Parent Relationship Therapy model, the effect was larger still, a result later corroborated in a meta-analysis focused on behavioural outcomes (Parker et al., 2021b). The filial model itself originates with Guerney, whose rationale was to train parents as the primary agents of therapeutic change with their own children rather than to treat the child directly, and the caregiver-involvement moderator is, in effect, that half-century-old proposal vindicated (Guerney, 1964). The second demonstration builds an inverse-variance pool over a synthetic study set constructed to reproduce these moderators, so that toggling subgroups moves the pooled estimate in the direction the literature reports.

That caregiver involvement is the strongest lever has a plausible reading: the trained parent extends the accepting, limit-setting relationship from the fifty-minute session into the child's daily environment, and the mechanism the model treats as curative, a consistent relationship, is precisely what a parent can supply continuously. The moderator finding thus doubles as a theory test, and it favours the relational account over a technique-centred one.

Dosage

If the relationship is the active ingredient, its effect should depend on how much of it a child receives, and the dosage literature bears this out with a curvilinear, not merely monotonic, shape. LeBlanc and Ritchie found that outcome improved with the number of sessions up to an optimum in the region of thirty to thirty-five sessions, beyond which additional sessions added little (LeBlanc & Ritchie, 2001). This matters clinically because much routine practice, and many trials, run far shorter courses, which may under-dose the treatment and understate its effect. The third demonstration renders the reported shape as a schematic curve of effect size against session count, letting the reader move a dosage marker along it to see benefit rise toward the optimum and then plateau.

The dosage curve should be read as an illustration of a reported relationship, not a per-session guarantee; individual children vary, and the optimum is an average over studies. But the qualitative lesson is robust and easily missed: a null result from a six-session trial is weak evidence against an approach whose modelled benefit is still climbing at that point.

Applications

Play therapy has been carried from the clinic into the settings where children are most reachable. In schools, a meta-analysis of child-centered play therapy delivered by counsellors found significant benefits on the academic and behavioural outcomes schools care about, supporting its use as a school-based intervention rather than only a clinic treatment (Ray et al., 2015). In early intervention, a randomized trial embedded in Head Start found that child-centered play therapy reduced disruptive behaviour in preschoolers relative to control, extending the evidence to a young, high-need, economically disadvantaged population (Bratton et al., 2013). Trials targeting specific presentations have added anxiety, where a controlled study found child-centered play therapy reduced anxiety in young children, to the list of outcomes with randomized support (Stulmaker & Ray, 2015).

Trauma is the application under most active development. A systematic review of child-centered play therapy and trauma mapped a promising but still-maturing literature, noting that the non-directive, child-led, symbolically distanced quality of the approach is well suited to children who cannot safely recount what happened to them, while cautioning that the controlled evidence specific to trauma remains thinner than the general outcome base (Parker et al., 2021a). The most direct recent test is a randomized controlled trial by Ray and colleagues examining child-centered play therapy for children with a history of adverse childhood experiences, which added rigorous evidence in exactly the area the review flagged as underserved (Ray et al., 2022).

Worked Example

An effect size is only as useful as one's ability to translate it into what happens to children, and the first demonstration reports the same quantities the arithmetic below produces. Take the pooled effect of play therapy as d = 0.80, and model the treated and untreated outcome distributions as two normal curves of equal spread whose means are eight-tenths of a standard deviation apart. Three standard readings follow directly.

The first is Cohen's U3, the fraction of the treated distribution that lies above the mean of the untreated distribution, which is the normal cumulative probability at d itself. Evaluating the standard normal at 0.80 gives 0.788, so about 78.8 percent of treated children, nearly four in five, end up above the average untreated child. The second is the probability of superiority, the chance that a randomly drawn treated child outscores a randomly drawn untreated child, which for two normal distributions is the normal cumulative probability at d divided by the square root of two. Here that is 0.80 divided by 1.414, or 0.566, and the normal probability at 0.566 is 0.714, so a treated child beats an untreated child about 71.4 percent of the time. The third is the overlap: twice the normal probability at minus d over two, that is twice the probability at minus 0.40, which is 0.689, meaning the two distributions still share roughly 69 percent of their area. The three numbers together give the honest picture of a d = 0.80 treatment. It shifts most children upward and would win a head-to-head comparison seven times in ten, yet it leaves the groups substantially overlapping. A large average effect is not a categorical cure, and stating it in these terms guards against both dismissing the benefit and overselling it.

Discussion

The status of play therapy is that of an intervention with a solid aggregate evidence base and real methodological work still to do. The consistent finding, replicated across independent meta-analyses, is a moderate-to-large pooled effect that does not collapse when weaker studies are removed, together with two robust moderators, humanistic approach and caregiver involvement, that tell practitioners how to obtain the larger effects (Bratton et al., 2005; LeBlanc & Ritchie, 2001; Lin & Bratton, 2015). The moderator on who delivers the treatment is theoretically as well as practically important, because it favours the model's own claim that a consistent accepting relationship, not a proprietary technique, is what drives change (Parker et al., 2021b).

The open problems are the ones the field's own advocates name. The primary-study base, though large, contains many small trials, and the randomized evidence, while growing, is still modest relative to the confidence with which the approach is practiced; the response to early critics was to add rigorous trials rather than to dismiss the objection, and that program is unfinished (Ray et al., 2001; Ray et al., 2013). Dosage is routinely under-attended, so that short trials may understate what an adequately dosed course would show (LeBlanc & Ritchie, 2001). And the trauma literature, where the approach is most intuitively apt, is precisely where the controlled evidence is thinnest (Parker et al., 2021a). What is not in serious doubt, running from Axline's first application of client-centered theory to children through to the current randomized trials, is that a therapy conducted in the child's own medium can produce measurable, and sometimes substantial, benefit.

Current Directions

The most active current work is on trauma and adversity. The systematic review of child-centered play therapy and trauma set an explicit agenda, identifying the approach's fit for children who cannot verbally process what happened to them while documenting how few rigorous, trauma-specific trials exist (Parker et al., 2021a). The randomized trial of child-centered play therapy for children with adverse childhood experiences is a direct response, testing the approach in a population defined by cumulative early adversity and adding the kind of controlled evidence the review called for (Ray et al., 2022). In parallel, the meta-analytic literature has continued to refine the moderator picture, with a recent behavioural-outcome meta-analysis reinforcing that caregiver-involved and humanistic delivery yield the larger effects (Parker et al., 2021b). Two priorities recur across this work: larger randomized trials to match the confidence of practice with the strength of evidence, and closer attention to dosage, so that the treatment tested is the treatment the dosage curve predicts will work.

Common Misconceptions

Play therapy is just letting a child play; it is not real therapy.

The play is the medium, not the treatment in itself. A trained therapist tracks, reflects feeling, returns responsibility, and sets limits within a structured relationship, and the approach carries a moderate-to-large pooled effect across controlled studies (Bratton et al., 2005).

Non-directive means the therapist never sets any limits.

Non-directive refers to following the child's lead in content and feeling, not to an absence of boundaries. The child-centered therapist holds firm limits on safety and on the frame of the session, and acknowledging the feeling while maintaining the limit is a core skill of the method (Ray et al., 2001).

A large effect size means play therapy works for essentially every child.

An effect of d = 0.80 shifts the whole outcome distribution, but the treated and untreated groups still overlap by roughly seventy percent. It is a substantial average benefit, not a categorical cure, and individual response varies (Bratton et al., 2005).

Glossary

Child-centered play therapy.

The non-directive tradition running from Axline to Landreth, in which an accepting, permissive therapeutic relationship within firm limits is held to free the child's own growth tendency; the most researched form of play therapy.

Child-Parent Relationship Therapy.

A filial model in which parents are trained to conduct child-centered play sessions with their own children; caregiver-delivered treatment is the strongest moderator of the play-therapy effect.

Cohen's d.

A standardized effect size expressing the difference between two group means in units of their pooled standard deviation; the pooled play-therapy effect is near d = 0.80.

Directive play therapy.

Approaches at the structured pole of the continuum, in which the therapist arranges play activities toward specific goals, often drawing on behavioural principles, in contrast to the non-directive tradition.

Filial therapy.

The general practice of training parents or caregivers to deliver play therapy to their own children, extending the therapeutic relationship into daily life; Child-Parent Relationship Therapy is its manualized form.

Limit setting.

The child-centered therapist's statement of the few firm boundaries, on safety and on the session frame, that make permissiveness about feeling and expression safe; a defining skill of the non-directive method.

Meta-analysis.

A statistical synthesis that pools effect sizes across studies into a single weighted estimate; the principal form of evidence for play therapy's overall efficacy and its moderators.

Non-directive play therapy.

The stance, originating with Axline, in which the therapist follows the child's lead in content and pace rather than directing the play toward chosen goals; the basis of the child-centered approach.

Probability of superiority.

The chance that a randomly chosen treated case outscores a randomly chosen untreated case; for two equal-variance normal distributions it is the normal probability at d divided by the square root of two, about 71 percent when d = 0.80.

Reflection of feeling.

The child-centered response of naming aloud the emotion the child's play appears to carry; one of Axline's principles and a core operational skill of the non-directive method.

Sandtray.

A tray of sand with miniature figures in which a child stages scenes; a characteristic medium of play therapy that lets a child enact experience at a symbolic remove.

Symbolic distancing.

The therapeutic power by which enacting threatening material through figures at one remove lets a child approach content that would overwhelm them if addressed directly, as a self-titrated graded exposure.

Therapeutic powers of play.

Schaefer's framework identifying the specific change agents through which play produces benefit, including emotional expression, symbolic distancing, mastery, and the therapeutic relationship.

Tracking.

The child-centered response of describing aloud what the child is doing in play, without directing or evaluating it, conveying attentive acceptance and following the child's lead.

U3 (Cohen's U3).

The proportion of the treated distribution lying above the untreated mean, equal to the normal cumulative probability at d; about 79 percent of treated children exceed the average untreated child when d = 0.80.

Key Researchers

Virginia Mae Axline (1911-1988). American psychologist who applied Rogers's client-centered theory to children, founding non-directive play therapy; her eight principles remain the operational core of the child-centered approach, and her case study Dibs in Search of Self carried the method to a wide readership. Wikipedia - Wikidata

Sue C. Bratton. Professor Emerita and Director Emerita of the Center for Play Therapy at the University of North Texas; lead author of the field's landmark meta-analysis and co-developer of Child-Parent Relationship Therapy, she did much to put play therapy on a quantitative evidence footing. Center for Play Therapy - Google Scholar

Peggy L. Ceballos. Professor and department chair in counseling at the University of Texas at San Antonio; a contemporary child-centered play therapy researcher whose work on caregiver-involved treatment and early intervention includes the Head Start trial and the recent adverse-childhood-experiences randomized study. Faculty Page - Google Scholar

Melanie Klein (1882-1960). Austrian-British psychoanalyst who pioneered the play technique in child analysis, treating the child's spontaneous play as the equivalent of adult free association and establishing play as the medium of child therapy. Wikipedia - Wikidata

Garry L. Landreth (1937-2026). Founder of the Center for Play Therapy at the University of North Texas; he systematized the non-directive tradition as child-centered play therapy and, through his text on the art of the relationship, trained generations of practitioners in the stance that it is the relationship, not the technique, that heals. Center for Play Therapy

Dee C. Ray. Regents Emerita Professor of counseling at the University of North Texas; a leading child-centered play therapy outcome researcher whose meta-analyses and randomized trials on impairment, anxiety, school settings, and adverse childhood experiences anchor much of the modern evidence base. Center for Play Therapy - Google Scholar

Charles E. Schaefer (1933-2020). American psychologist known as the father of play therapy and co-founder of the Association for Play Therapy; he catalogued the therapeutic powers of play, reframing the question of how play heals as one of identifying its distinct change agents. Wikipedia - Wikidata

Frequently Asked Questions

What is play therapy? Play therapy is a form of psychotherapy that uses play, the natural medium of childhood expression, as the primary means of communication and change, letting children work through psychosocial difficulties they cannot yet put into words (Bratton & Ray, 2000).

How is play therapy different from ordinary play? The play is the vehicle, not the treatment in itself. A trained therapist tracks the child's play, reflects feeling, returns responsibility, and sets limits within a structured therapeutic relationship, and the approach has measurable outcomes across controlled studies (Ray et al., 2001).

What is child-centered play therapy? It is the non-directive tradition, running from Axline to Landreth and grounded in Rogers's client-centered theory, in which an accepting, permissive relationship within firm limits is held to free the child's own growth tendency; it is the most researched form of play therapy (Ray et al., 2013).

Does play therapy actually work? Independent meta-analyses converge on a moderate-to-large pooled effect, on the order of d = 0.80, that persists when weaker studies are removed, supported by a growing set of randomized trials (Bratton et al., 2005; LeBlanc & Ritchie, 2001).

What makes play therapy more effective? Two moderators stand out: humanistic, non-directive approaches outperform more directive ones, and treatment delivered by a trained parent or caregiver, as in filial or Child-Parent Relationship Therapy, produces the largest effects (Bratton et al., 2005; Parker et al., 2021b).

How many sessions does play therapy take? The dosage literature reports a curvilinear relationship, with outcome improving up to an optimum near thirty to thirty-five sessions and adding little beyond it, which means short courses may under-dose the treatment (LeBlanc & Ritchie, 2001).

Is play therapy used for trauma? Increasingly. Its child-led, symbolically distanced quality suits children who cannot safely recount what happened to them, and while trauma-specific controlled evidence is still maturing, recent randomized work has begun to fill the gap (Parker et al., 2021a; Ray et al., 2022).

What ages is play therapy for? It is designed principally for young children, roughly the preschool and primary years, whose symbolic play outstrips their capacity for verbal, reflective conversation; trials span populations from Head Start preschoolers to school-age children (Bratton et al., 2013; Ray et al., 2015).

References