Great Behavioral Tasks can Validate Mouse Models of Autism

Ooh Ooh Ooh! They Finally Did It!

I spent a fair portion of my day yesterday on twitter discussing a new model for autism and treatment options in a mouse model. I was haranguing anyone that would listen about the fact that the paper in question overinterpreted their tasks, and that they chose the wrong paradigms. These conversations are storified Here thanks to @avinashtn.

In this post I want to discuss an example of when behavioral testing on a mouse model of human disease is done entirely correct. In this case, the right task was chosen for mice, and was developed in very close collaboration with the clinician whose work in humans was being modeled by the mice. This approach, called endophenotyping, actually studies non-core features of autism in a mouse model, but rather models cognitive processing in human autism using a mouse model.

The manuscripts discussed in this post are “Cognitive Abilities on Transitive Inference Using a Novel Touchscreen Technology for Mice” Jill Silverman in Jackie Crawley’s lab at the UC Davis MIND Institute, which is a behavioral mouse model designed to replicate “Transitive Inference in Adults with Autism Spectrum Disorders.” By Marjorie Solomon and colleagues, also at the UC Davis MIND Institute. The articles can be found on Pubmed. I am going to focus most of my discussion on the Silverman and colleagues study because it is their study that I find near revolutionary in the field of behavioral genetics associated with autism.

Disclaimer: I did one year of postdoctoral research the UC Davis MIND Institute and am currently a collaborator with Marjorie Solomon. I have also had many discussions with Jackie Crawley concerning the task I am going to describe. These collaborations in no way have biased my analysis. In fact, I was always highly critical of the whole project, which I guess makes my excitement all the more surprising.

What is Being Studied and Why?

The original study from Marjorie Solomon set out to evaluate transitive inference in a population with autism to test specific rule learning and generalization. They hypothesized that learning associations would be fine, and that there would be some deficits for logical inference. They found that participants with autism learned very quickly, faster then the control group, and showed relatively normal transitive inference. However, there did seem to be a deficit when the participants were asked to compare the end conditions.

The study by Jill Silverman and colleagues set out to replicate these findings an a mouse model proposed to show autistic-like behavior, the BTBR mouse. They found that the mouse showed comparable patterns of performance to the autistic participants in the human study. They took this to suggest their mouse does in fact demonstrate autistic-like features.

I find these effects intriguing because I flat do not understand and cannot make any sense of the pattern of behaviors in the autism groups relative to the control populations. However, I do not have to. They are what they are and Solomon and colleagues did not set put to prove a grand theory, but rather to evaluate performance on a behavioral task. The fact that the work with human participants was replicated in the proposed mouse model (but not I a comparison mouse model), suggests the effects are real-albeit hard to understand at present. Further experimentation, possibly using the mouse models may shed light on the observed effects.

An Innovative Approach

As background, transitive inference is a paradigm in which a person or animal has to learn a series of associations. Normally these are A rewarded when presented with B (A>B), B>C, C>D, and D>E. Once these associations are learned, tests of novel configurations of stimuli are performed, with the critical test being B vs D in which B is rewarded. The assumption is that a logical transfer of the relationship is required to understand A>B>C>D>E. A vs E should be easy since A was always rewarded and E was never rewarded.

Transitive inference in mice is not a new paradigm, in fact it has been tested over and over and over in the 1989s and 1990s as a model of hippocampus function in rats, and more recently has been performed in mice. However, as these studies involved rodents, the experiments have all been based on odors, whereas the experiments in humans all involved visual stimuli.

What Silverman and colleagues did was to explicitly model a visual transitive inference paradigm from humans into mice. A quick note of the author lists also provide insight to their success, Marjorie Solomon, who designed and performed the human task, was involved at every step in the mouse task design.

Silverman and colleagues used a Bussey Touchscreen system to administer the paradigm and display the stimuli, allowing the mice to physically tap their choice in each trial and to have large, standardized, easy to discriminate visual objects for the associations. Importantly, using this visual presentation and a touchscreen was as close to a direct methodological replication of a human task I have ever seen in mice. Not only were the same cognitive domains being tested, they were being tested the exact same was across species.

What really makes this outstanding is that, if one looks at the learning and performance curves from the tasks, they are very similar! suggesting that the task was, in fact, valid. Silverman and colleagues’ hypothesis that the BTBR mouse would show similar performance at each phase as autistic participants was confirmed, especially since a C57BL/6J mouse control did not show similar deficits. Additionally, the performance of the mice were quite consistent-suggesting this is a task that can be widely adopted.

Take Home Message

The take home message here is that a great behavioral task is needed to really determine if a mouse model of a human condition really is valid. Prior to this paper, I was a nonbeliever in the BTBR as a model of autistic-like features, mostly because I attack the paradigms that were used in making this determination. In fact, I have been something of an outspoken critic of Jackie Crawley’s lab in general for not modeling the tasks used to evaluate human autism well enough.

No more! With this post I am officially eating my words! Even though I do not know how to interpret the effects seen in either of these studies, the BTBR mouse and human participants with autism appear to match each other in performance-and that is more than enough evidence for me that the BTBR is, at least, an adequate model for autism-like behaviors.


BTBR mice purported to model autism appear to actually do so, at least so far as performance on a transitive inference paradigm in concerned. When behavioral tasks for mice like this are designed to explicitly model the human paradigms, especially when developed inclose collaboration with scientists that study the population in question, everyone wins. Models can be validated. Outcome measures can be generated. Data are valid. It is a beautiful thing. Let’s do more of this…please.


I would love to hear your thoughts on this!

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s