People who study creativity and innovation talk a lot about the value of “recombination” — bringing existing ideas, practices, processes, or technologies together in new ways or applying them in fresh contexts or markets. It’s a model that has led to many popular consumer products, such as leak-proof water bottles that borrow nozzles from shampoo dispensers, and home cholesterol testers that incorporate the inject/eject mechanism from CD players.

Over the last three decades, research has shown that the people most likely to innovate via recombination talk with groups of people who don’t talk to each other. In the language of social networks, these people span diverse clusters, which gives them a “vision advantage,” as the sociologist Ronald Burt is fond of putting it. Their position as “network brokers” allows them to see things that others can’t.

So, if you want to innovate, you should become a network broker, right? Unfortunately, it’s not that simple. Changing your network is hard, because it’s not just about who you talk to. It’s about who your friends and coworkers communicate with — and you can’t dictate that, even if you forge new relationships.

But not to worry. Our recent research using survey data on communication networks among software engineers at a large mobile services company shows that people who are firmly ensconced in a single, densely connected community (“non-brokers,” for short) can be as innovative as network brokers if they follow one simple, counterintuitive strategy: narrow their focus.

Though brokers benefit from allocating their attention in a balanced way across all their contacts, scouting for recombination opportunities across communities, non-brokers fare poorly when they take a similar tack. We found that they can be up to four times less likely to innovate when they scan the horizon of their more-constrained networks than when they selectively attend to a few communication partners who can help them see anomalies and inconsistencies among local ideas within their domain.

It doesn’t seem to matter much who non-brokers pay attention to as long as they focus intensely on a few key contacts who talk to each other frequently. When they do that for a sustained period, they begin to hear differences in the way those people present information, approach problems, and think through solutions on the same topics.

Becoming aware of these differences that most others don’t notice — because they aren’t paying close enough attention — can help non-brokers develop novel ideas that build cumulatively on knowledge. We call this way of innovating “interrogation,” since it’s a bit like the work that police detectives do when they question several people, compare their stories, and look for inconsistencies that will lead to breakthroughs.

This is essentially the approach that engineers at SpaceX took to reduce the cost of building its Falcon 1 Rocket. They did not explore ideas, practices, or technologies that they could import from other rocket builders or from other domains. Instead, the team of engineers charged with reducing the build cost focused its attention on other engineers within the company who had differing ideas about how to design components with lower material costs. By comparing those solutions, looking for commonalities or differences across them, and triangulating a new design based on these comparisons, they were able to radically alter the way the rocket was built. As a result, they produced the Falcon 1 for $7 million at a time when the next cheapest rocket on the market was $65 million.

In our study, non-brokers who focused their attention on a small number of contacts were judged by senior managers to be as innovative as brokers. And brokers who gave equal attention to all their communication partners were much more likely to be rated by senior managers as innovative than brokers who focused their attention on a selected few contacts.

So for the purpose of innovation, neither group should change its approach to networking — but both should pay attention to whom they pay attention.