The sensitivity of conventionally collimated SPECT systems is essentially fixed by the number of resolution elements desired. A variety of collimation methods are proposed or in use that distribute sensitivity and resolution differently among the various projection elements, but the effects on image quality have not really been documented. Of the many existing systems, the camera-based systems, which are useful for both head and body imaging, will be most useful in a clinical environment. For brain imaging, it is clear that fan beam collimation offers a large performance advantage over parallel collimation. However, comparison of the various instruments based on specifications alone is difficult if not impossible, and careful 3-dimensional imaging studies are an absolute necessity. In view of the fairly healthy price of dedicated brain instruments, it is interesting to speculate on their role in nuclear imaging. Their cost is from $50,000 to $200,000 less than a multi-headed general purpose system, and they have excellent resolution and sensitivity for brain imaging. They cost from $175,000 to $300,000 more than a single head camera-computer system yet offer 3 to 4 times the resolution-adjusted sensitivity. This suggests that a special purpose instrument would need to be heavily utilized or located (physically or politically) where the body imaging capability of a two or three-headed system would not be used. A good case might be made for locating a dedicated instrument in a neurological ICU in order to avoid transporting critically ill patients out of the unit. The various non-conventional imaging methods show promise for exceeding the sensitivity limit of regular collimators. At present it is speculative as to how much can be gained and whether it will be worth the added complexity and/or cost. Nevertheless, they are very interesting and certainly justify additional research.