The article presents a framework and empirical investigation to demonstrate the role of task difficulty in the effectiveness of collective intelligence. The research contends that collective intelligence, a form of community engagement to address problem solving tasks, can be superior to individual judgment and choice, but only when the addressed tasks are in a range of appropriate difficulty, which we label the 'collective range'. Outside of that difficulty range, collectives will perform about as poorly as individuals for high difficulty tasks, or only marginally better than individuals for low difficulty tasks. An empirical investigation with subjects randomly recruited online supports our conjecture. Our findings qualify prior research on the strength of collective intelligence in general and offer preliminary insights into the mechanisms that enable individuals and collectives to arrive at good solutions. Within the framework of digital ecosystems, the paper argues that collective intelligence has more survival strength than individual intelligence, with highest sustainability for tasks of medium difficulty © 2013 IEEE.