For individuals with developmental disabilities, teaching methods that maximize time and resources are critically important. Matrix training is a teaching method that has been used to maximize teaching time for children with Autism Spectrum Disorder (ASD) by producing recombinative generalization. Recombinative generalization allows individuals to produce and understand novel stimuli, from previously trained stimuli components. This includes color-shape combinations, noun-preposition combinations, and noun-verb combinations in individuals who are vocal, and those with speech-generating devices. However, the percentage of recombinative generalization individuals produce through matrix training ranges anywhere from 0% - 94%. With such a wide range, it is difficult to pinpoint what prerequisites or procedural modifications are necessary for recombinative generalization to emerge. The primary purpose of this paper was to compare the effectiveness and efficiency of two different training procedures: simultaneous training and combination training based on the procedures by Bergmann et al. (2022) for three participants, between the ages of 5-6. Participants were taught arbitrary stimuli using combination training, in which both components are taught at the same time as one stimulus, and simultaneous training, which uses a chained- trial procedure, for two matrices. Results for all three participants were similar to Bergmann et al. (2022), in which simultaneous training produced more recombinative generalization in less training time. After training, a preference assessment was conducted with each participant to determine if they preferred the training condition that was more efficient/effective. Results of the assessment showed that participants did not prefer one condition over the other, as results were mixed. Implications, limitations, and considerations for future research are discussed as well.

A common goal of instructional techniques is to teach skills effectively and efficiently. Matrix training techniques are both effective and efficient as they allow for the emergence of untrained responding to novel stimulus arrangements, a phenomenon known as recombinative generalization. However, it is unclear which type of matrix arrangement best promotes recombinative generalization. The current study compared two common matrix training approaches, an overlapping (OV) design and a non-overlapping (NOV) design, with respect to arranging relations targeted for training. We conducted a replication evaluation of a Wilshire and Toussaint study, and taught two typically-developing preschoolers compound object-action labels in Spanish and used either an OV or NOV matrix training design. Results from both studies demonstrated the participant trained with an OV design produced recombinative generalization and participants trained with a NOV design produced significantly low levels of emergence or none at all. These results suggest that an OV matrix design facilitates recombinative generalization more effectively than a NOV design. Implications for instructional arrangements are discussed.

In recent years, Generative Language Models have exploded into the mainstream with household names like BERT and ChatGPT, proving that text generation could have the potential to solve a variety of tasks. As the number of students enrolled into programming classes has increased significantly, providing adequate feedback for everyone has become a pressing logistical issue. In this work, we evaluate the ability of near state-of-the-art Generative Language Models to provide said feedback on an automated basis. Our results show that the latest publicly available model GPT-3.5 has a significant aptitude for finding errors in code while the older GPT-3 is noticeably more uneven in its analysis. It is our hope that future, potentially fine-tuned models could help fill the role of providing early feedback for beginners, thus significantly alleviating the pressure put upon instructors.