Generalization is the capacity to apply what is learned in one situation to another. Problems in doing so have a wide array of consequences for the design of learning spaces and teaching people with autism in general.

A (somewhat extreme) example of this disability are children who have learned arithmetic through pictures, i.e.: 2 tigers plus 3 tigers make 5 tigers. One can understand this without grasping that the same goes for glasses of lemonade. One fails to see the general principle of 2 plus 3.
Temple Grandin, one of the most famous people with autism, relates how she stores all individual cats and all church steeples she has ever seen in her (photographic) memory, and that she has no concept of a ‘generalized cat’ or church steeple.

This inability has to do with difficulties in dealing with abstraction. Some people on the spectrum tend only to see the concrete and have difficulty (or are altogether unable) to conceive of a general principle. This matter is related to ‘overselectivity’ which is virtually the same as weak Central coherence in which details are so prominent that the whole isn’t perceived or only with great difficulty. The more one is inclined to focus on small differences between situations, the harder generalization becomes since generalization requires a focus on similarities. (See Brown & Bepko, 2012, for a review.)

What has been learned tends to stay “glued” to the concrete situation in which it was experienced. For many people on the spectrum, the ‘learned’ event never becomes assimilated as an abstract experience.  This ‘assimilation of abstract experience’ makes it possible for skills to be generally applied in an infinite variety of circumstances. If this ability is limited or lacking ‘real world’ trainings are preferable to simulated training environments. Not everyone on the spectrum, however, is capable of ‘real world’ engagement; for them, striving for simulated circumstances which resemble the reality in which the skills are needed is the next best option.

Dilemmas in treatment and architecture are discussed elsewhere. Among them is the crucial question behind autism-friendly design: how far could or should one go in adapting the physical environment? Where people with considerable generalization problems are concerned, they contribute to the argument that adaptations should be kept to a minimum.

*Temple Grandin

Grandin, Temple, Thinking in Pictures. And Other Reports from My Life with Autism, New York, Vintage Books, (1995), second edition, January 2006. PDF
Brown & Bepko
Brown, S.M., J.M. Bebko, ‘Generalization, overselectivity, and discrimination in the autism phenotype: A review.’ In: Research in Autism Spectrum Disorders 6 (2012) 733–740.