Ondertussen kennen we via John Hattie het gegeven van meta-analyses in onderwijs. Maar vaak komt al snel de commentaar dat bij het vergelijken van alle bestaande onderzoeken, bepaalde nuances verloren gaan. Dat klopt als je enkel naar het eindcijfer kijkt.

Er is nu een nieuwe meta-analyse gepubliceerd door Jens Dietrichson, Martin Bøg, Trine Filges en Anne-Marie Klint Jørgensen richt zich op een welbepaalde nuance: namelijk ‘wat werkt’ er voor kinderen uit gezinnen met een lage SES.

Deze grafieken tonen de resultaten:

Hierbij merken de onderzoekers wel enkele zaken op:

• geen enkele aanpak zal de kloof tussen kinderen uit gezinnen met lage SES en hoge SES volledig dichten, maar de beter werkende aanpakken verkleinen wel die kloof.
• Voor sommige aanpakken, net onder de top 4 is het mindere effect mogelijk te danken aan het beperkt aantal studies.

Zelf wil ik ook nog 2 opvallende zaken meegeven:

• Merk vooral op hoe extrinsieke beloningssystemen voor zowel leerlingen als leerkrachten geen enkel effect lijken te hebben.
• In Figure 5 wordt ook duidelijk waarom 1 effect size cijfer niet zo veel zegt: het gaat over een interval, waarbij sommige van de aanpakken gaan tussen geen en veel effect (bijvoorbeeld small-group instruction). Om te weten wanneer welke effectgrootte speelt, kijk je best naar de oorspronkelijke studies. I know, het is dus niet zo simpel.

Tot slot, Larry Ferlazzo omschreef de 4 beste aanpakken kort, zodat je ook weet waarover het gaat:

1. TUTORING

Tutoring interventions were activities where students got supplemental pedagogical support from an instructor, either one-to-one or in a small group (five students or fewer). Tutors could be volunteers, paid non-teachers, or professional teachers. The interventions included in the tutoring category were often highly structured programs (e.g., manual based) implemented over a limited time period, typically 12 to 20 weeks.

2. FEEDBACK & PROGRESS MONITORING

This category included interventions that added a specific feedback or progress monitoring component, where teachers or students received detailed information about the students’ development. The objective was often to customize instruction to the individual student’s needs. Note that tutoring and cooperative learning are also likely to contain increased feedback, but because such feedback is embedded in the regular set up of these programs, these interventions are not coded in this category. Interventions had to add an extra component of feedback or progress monitoring to be coded here.

3. SMALL GROUP INSTRUCTION

Interventions in this category included instruction where students are placed in groups smaller than regular class sizes. These interventions differed from those in which learning in small groups are built in, such as cooperative learning and tutoring. There was no cooperative learning element explicitly included in the interventions coded in this category, and the groups were larger than what normally counts as tutoring (here defined as more than five students per group).

4. COOPERATIVE LEARNING 

Cooperative learning, or peer-assisted learning, referred to interventions where students work together in pairs or small groups in a systematic and structured manner. Examples included students acting as pedagogical instructors for each other, as when more able students help less able students.

Abstract van het onderzoek:

Socioeconomic status is a major predictor of educational achievement. This systematic review and meta-analysis seeks to identify effective academic interventions for elementary and middle school students with low socioeconomic status. Included studies have used a treatment-control group design, were performed in OECD and EU countries, and measured achievement by standardized tests in mathematics or reading. The analysis included 101 studies performed during 2000 to 2014, 76% of which were randomized controlled trials. The effect sizes (ES) of many interventions indicate that it is possible to substantially improve educational achievement for the target group. Intervention components such as tutoring (ES = 0.36), feedback and progress monitoring (ES = 0.32), and cooperative learning (ES = 0.22) have average ES that are educationally important, statistically significant, and robust. There is also substantial variation in effect sizes, within and between components, which cannot be fully explained by observable study characteristics.