Do social stories work?

They are on a 2020 evidence list, and a 2010 meta-analysis still found the average result low to questionable. That is the short answer. Clearing a study-count bar is not the same as a change you can expect for one child.

What one looks like, with a tiny example, is in what a social story is. One narrow situation already written out is the trick-or-treating story. Another is Christmas. We checked the papers below on November 2, 2026.

What the evidence lists actually counted

The National Clearinghouse on Autism Evidence and Practice reviewed autism intervention research published from 1990 through 2017. With the earlier review, that was 972 articles, ages birth to 22, sorted into 28 evidence-based practice categories. Social Narratives is one category. Carol Gray's Social Stories™ is also named on its own: the summary says that manualized method now has sufficient evidence to be classed as evidence-based, inside the Social Narratives group (NCAEP executive summary; Hume et al., 2021).

A practice qualifies, in that summary, with group studies from separate teams, or with five high-quality single-case studies from three teams and at least 20 participants, or with one group study plus three single-case studies. That is a count of acceptable studies. It is not an average effect.

Hume and colleagues say the 2021 paper is a systematic review, not a meta-analysis, so it did not measure how large the effects were. Studies that found no effect were left out, and only peer-reviewed journal articles were used (Hume et al., 2021).

The National Standards Project, Phase 2, reached a similar "worth trying" place with a different label. Story-based Intervention is one of 14 Established interventions for people under 22. The report says those 14 have been thoroughly researched and have sufficient evidence for the project to state that they are effective. Social Stories™ is the best-known example in that group. The project counted 21 articles in phase 1 and 15 in phase 2, with participants ages 3 to 14. The outcomes they list include communication and learning readiness, interpersonal skills, self-regulation, and fewer problem behaviors (NSP2, 2015).

Those lists answer whether there is enough research to try one. The meta-analysis below is about the size of the average result.

What the 2010 meta-analysis measured

Kokina and Kern pooled single-subject research on this intervention with autistic students (Kokina & Kern, 2010). The abstract says:

• Overall effectiveness was low to questionable.
• Results looked better when the goal was reducing inappropriate behavior than when the goal was teaching social skills.
• Results looked better in general education settings, and when the child was their own intervention agent, meaning the child helped carry it out rather than only hearing it.
• Age, diagnosis, skill level, format, length, and comprehension checks "also was explored." The abstract does not say those factors improved outcomes.

Aiming at one inappropriate behavior is the better use in that review. Aiming at a new friendship skill is the weaker use. Many of the underlying studies follow a few students. The average can shift as larger studies arrive.

Free every Tuesday

A little support for the week ahead.

Real parenting moments, an expert’s honest take, and one small idea to try. In your inbox every Tuesday.

Your email address

Join free

Get Social Stories every Tuesday. Unsubscribe whenever you like.

How to tell if one is helping

Neither review scores your child. This home check is a parent habit, not an extra finding from Kokina and Kern.

Pick one action you can see: leaving the chair, covering ears, or hitting. Write how often it happens, for a few days, in the same kind of moment, before anyone reads anything. Read on calm days. For about two weeks, write the number again in that same moment. Helping means the number moved the way you hoped. One easy afternoon is not a trend.

Let your child hold the pages, point at a photo, or speak a known line. That matches the 2010 note about the child as their own intervention agent.

If the number stays flat, narrow the behavior or use photos of the real place. If you are worried during those days, call the pediatrician or the therapist who knows your child.

When to try something else

• Unsafe behavior, or fear blocking school or medical care. Call the pediatrician or therapist now. Reading is not the plan for that.
• About two weeks of notes and no movement, and you are worried. Call, and ask what else belongs in the plan.
• The hope is a broad friendship skill. Kokina and Kern found that use weaker. One recess moment is closer to the better use than a general "make friends" goal.
• The hard part is a noise, a smell, or a touch that reading will not remove. Changing the visit, or asking the adult there to skip that part, may matter more.
• The job is a speech sound. A printable sound sheet drills one sound at home. That is a different task, and it is not an evaluation.

Clinicians may still fold this into a larger plan. These papers do not replace that plan.

Another one can come by email. Subscribe to Social Stories.

FAQ

Does "evidence-based" mean it will work for my child?

No. In the 2020 clearinghouse review, the label means enough good studies, from more than one team, showed positive effects for some outcomes (Hume et al., 2021). Kokina and Kern still found low to questionable overall effects in the single-subject studies they pooled. Try one behavior. The label is not a personal guarantee.

Did the 2021 review measure how big the change was?

No. Hume and colleagues call their paper a systematic review, not a meta-analysis, so effect size was not examined. Studies that found no effect were left out. The 2010 meta-analysis is the one that speaks to the average, and that average was low to questionable.

Which goals looked better in Kokina and Kern?

Reducing an inappropriate behavior looked better than teaching social skills. General education looked better than other settings. Results also looked better when the child helped carry it out, instead of only listening. Age, diagnosis, and skill level were explored. The abstract does not say those factors improved outcomes.

When should I call a pediatrician?

If behavior is unsafe, if fear is blocking care or school, or if about two weeks of notes on one behavior has changed nothing and you are worried. It does not replace that conversation. NSP2's count covered ages 3 to 14. The clearinghouse review covered practices from birth to 22. Neither range predicts one child. Everyday uses, such as a move, are not what the 2010 meta-analysis measured.