In training, the odd item was just a different shape (like a rose against a bunch of houses). In testing, single feature changes, like a shift in the location of a single dot, or gestalt properties, like a line made of dots vs a diagonal line made of dots, were used to test the speed and accuracy of finding the odd item. Children with autism categorized as low functioning, typically developing toddlers, and adult tamarins scored equivalently on single feature changes and gestalt changes, making more errors. Adult college students and children with autism who were high functioning favored gestalt principles and responded faster and better-used grouping principles. This indicates another anomaly with tamarins shared with younger humans and autistic humans — the lack of application of gestalt principles reliably.