Summary
Social skills training programmes teach eye contact, conversation initiation, turn-taking, reading facial expressions, small talk—usually through modelling, role play, and reinforcement. The assumption is that autistic people lack social skills.
This assumption is radically incomplete. SST produces modest short-term behavioural changes but limited lasting benefit for social satisfaction or quality of life. Most programmes don’t teach social skills; they teach masking.
What the evidence shows
Modest short-term effects
A 2025 meta-analysis found moderate effects (d = 0.53) across 19 studies and 741 children. A 2024 meta-analysis found skill transfer better in younger individuals, face-to-face contexts, and with cognitive rather than behavioural interventions.
These effects describe what happens in structured settings during and immediately after training. The question of whether skills translate to genuine social connection in daily life is where evidence weakens.
Limited transfer and durability
Trained behaviours often don’t generalise to natural settings, don’t persist, and don’t produce improved social satisfaction or reduced loneliness.
An autistic teenager may learn eye contact in therapy but find it exhausting in the classroom. A child may learn conversation initiation but find scripted openings artificial. The behaviour is learned; the social experience doesn’t improve.
What is actually being taught
Most SST programmes teach neurotypical social conventions: eye contact (often aversive for autistic people), “appropriate” facial expressions, small talk (a bonding ritual many autistic people find pointless), reading body language and tone (unreliable sensory channels for autistic people), turn-taking conventions.
These are not universal social skills but neurotypical ones—cultural technologies designed by and for a specific neurotype. Teaching them to autistic people is like teaching a left-handed person to write right-handed: the behaviour can be learned at a cost, but it doesn’t change the underlying hand dominance.
The masking connection
SST that teaches neurotypical performance without acknowledging it as performance contributes to masking—suppression of autistic behaviour to appear neurotypical. The evidence is clear: masking is associated with anxiety, depression, burnout, and suicidality. Programmes teaching masking without naming it, acknowledging the cost, or offering alternatives cause harm.
Not all communication support is harmful. The distinction: teaching communication skills someone wants differs from training neurotypical performance as default. An autistic person who wants to learn small talk for a job interview or family gathering is choosing their goal; the cost is acknowledged; alternatives exist. Requiring an autistic child to maintain eye contact and initiate small talk without informed consent, without acknowledging it as performance, is normalisation.
What would be better
An alternative approach: support autistic social preferences by creating spaces respecting autistic communication norms—smaller groups, quieter environments, explicit rules, interest-based connection, written options. The diffusion chain studies show autistic people communicate effectively when conditions match their processing style.
Train neurotypical people. If communication difficulty is bidirectional, so should be the solution. Teaching neurotypical teachers, peers, and colleagues to understand autistic communication—flat affect is not disinterest, gaze aversion is not rudeness, directness is not aggression—does as much good as training autistic people to perform neurotypicality.
Modify the social environment (group size, noise level, communication mode, sensory conditions) to change outcomes without requiring the autistic person to change. Facilitate social connection between autistic people through interest groups, online communities, and neurodivergent-normed spaces, addressing loneliness more effectively than teaching neurotypical small talk.
Open questions
Is there a version of social skills training that is genuinely helpful — that teaches skills autistic people want, acknowledges the cost of performance, and respects autistic social preferences? Some programmes are moving in this direction. Whether they succeed depends on whether they start from the person’s goals rather than from neurotypical norms.
Key sources
- Meta-analysis of social problem-solving interventions (2025): d = 0.53 across 19 studies
- Nature meta-analysis on social training transferability (2024)
- Crompton et al. (2020, 2025). Diffusion chain studies
- Milton, D. (2012). The double empathy problem
- Howard, P.L. & Sedgewick, F. (2021). Communication mode preferences