Behavior for mammals is built out of multiple muscles acting in a coordinated fashion. Prima facie, there are three principal ways to increase an animal's behavioral repertoire size. The first is to, for each new behavior type, create a set of new muscle types (e.g. triceps, sartorius, etc.) with new functions specifically devoted to the implementation of that behavior type. If this were the case, then although each behavior is built out of many muscles, behavior is not built in a combinatorial fashion out of muscles. The second is similar to the first in that new behavior types are implemented via new muscle types, but, instead, muscles are used in a combinatorial fashion, so that it is the combination of the new muscle type with existing muscle types that makes the new behavior type possible. This is analogous to the addition of new words in a language. The third way behavioral complexity may be scaled up is to increase the complexity of behavioral expressions themselves (rather than increasing the number of muscles types), namely by having more muscles involved in an average behavior. This is analogous to uttering longer sentences in a language. My main task in this paper is to examine which of these ways underlies the increase of behavioral complexity among mammals. Behavioral repertoire sizes from the ethology literature were accumulated for mammals from two dozen species across eight orders, and the number of muscle types was estimated from atlases of anatomy across eight mammalian orders. The manner in which behavioral complexity actually increases among mammals appears to favor the second possibility mentioned above: greater behavioral complexity is achieved primarily by increasing the number of muscle types, and by using muscles in a combinatorial fashion. The theoretical framework I describe allows us to interpret the manner in which the number of muscle types scales with behavioral repertoire size, and I conclude that the number of degrees of freedom in the construction of behavioral expressions is on the order of three, which is probably due to neurobiological limitations in forming behaviors. The ontogeny of behavior in rat is also discussed within this framework. Finally, I show that there is a strong positive relationship between behavioral repertoire size and encephalization among mammals.