Want to increase generalization? Try matrix training

Let’s talk about matrix training. It’s a language intervention that’s been around for a while (e.g., Goldstein, 1983), and there’s quite a bit of evidence to demonstrate that it can be used as a framework to teach vocabulary and functional language skills. But, probably more importantly, kids can learn new and untrained language targets by participating in matrix training. Tell me more, right?

Well, the way that it works is that language targets (for instance, adjectives and nouns) are written on the vertical and horizontal axes of a table—aka: matrix. The targets can be either all unknown words or a combination of known and unknown depending on the client. Then, those targets are used to create various combinations. So, if you’re focusing on increasing your client’s understanding of adjectives and nouns, your matrix could look like this:

December graphics (1).png

Matrix training is based on the idea of recombinative generalization (Goldstein, 1983)—which basically means that if we teach a word combination {brown + bear} and then expose the child to one of the same words again but in a new combination {brown + bird}, the child can generalize their understanding or use of untrained, related targets {brown + horse}.

So, the matrix essentially serves as a visual guide that can be used to identify the sequence for instruction by making it easy for you to see which treatment targets are related. For instance, if you choose targets that are in the yellow diagonal cells (brown bear, red bird, blue horse, yellow duck, green fish) as well as the teal cells just above them (brown bird, red horse, blue duck, yellow fish), many of the same adjectives and nouns will be repeated in your intervention. By doing this, your client can then generalize their understanding to related, untrained targets (such as, brown horse, brown duck, red duck, brown fish, red fish, blue fish).

In this study, matrix training was used as a framework for intervention with three children between the ages of 22 and 35 months with severe language delays and/or ASD. The training focused on teaching the toddlers simple, one-step directions (e.g., shake) that included animals (e.g., dog). Three individual matrices were created based on each child’s knowledge of the target actions and animals before the training began. The intervention targeted six (or, 30%) of the one step action–object instructions with each child, and included components of discrete trial training (antecedent, responses, and consequences) as well as verbal and physical prompts and reinforcement. After training, the other 70% of the action-object combinations that weren’t directly taught were probed, and each of the three children demonstrated some level of generalization to these unknown targets (and, one of the toddlers demonstrated understanding of all of the untrained action-object combos!). 

Talk about more getting a little more bang for your buck, am I right?!

Although this study only included 3 participants, the findings suggest that matrix training is a simple framework to implement that has the potential to cut down on time spent on teaching language targets in intervention. This is particularly exciting news for our youngest clients who demonstrate the most significant language delays.

Curiel, E.S.L., D. M., Sainato, D. M., & Goldstein, H. (2018). Matrix training for toddlers with Autism Spectrum Disorder and other language delays. Journal of Early Intervention40(3), 268–284.