A level designer is considered the complete package, who can take the rough theme of a game from a document and turn it into an exciting game experience. For me a level that can be remembered long after playing it must be good. I often judge a level a long time after playing it. This might be a little basic but it's the bottom line of a great environment. I then ask myself what makes them so memorable? It could be a clever twist in the story, fun combat situations, perfect pacing? This is the time to explore the level's stories because from them, the significant parts of the level's design will emerge. Regardless of the narrative of the game, each level has the potential to tell many layers of its own background story. Even an empty office has the potential to tell little stories that transform it from a dull set of plain rooms into a real place through artfully placed builder's tools, scrounged furniture, used cups, and discarded rubbish. But the real power of level stories has nothing to do with set dressing, it is the ability to provide you with contextual game play scenarios that the story-based method really shines. I would argue that the power to immerse the player, to absorb his attention completely, is the common attribute of the greatest and most successful games. Gathering and studying reference is critical to creating immersion for the player. It is something that every one involved in the design should be apart of.
Everyone stores simplified constructions of reality in their mind; schemata that codify the critical features of the world around us. We use our schemata to recognize and interpret everything we experience. We also use those same simplified representations of reality to recreate it through art. Because no two people use precisely the same critical features to build their schemata, every person's art has a unique look, filtered through the lens of their uniquely simplified representations of reality. When we are creating worlds in games, immersion is only possible for the player if we can convince the players that the space is authentic (whether stylized or not.) If the critical features on screen don't match up with the critical features of the player's schemata, then he or she will not be fooled by it. So as game makers we must have really precise schemata to convince the widest selection of players. American dumpsters sitting in the back streets of Paris or French road signs on the streets of Chicago might seem acceptable to the developers because they do not mismatch with their very simple schemata of those distant locations, but these contextually inappropriate placements will be laughably inaccurate to people really familiar with those places. Given that games are released worldwide, it is difficult to overestimate the damage to audience immersion and perception done by poorly researched levels for a large percentage of your audience. It's your worldwide reputation on the line.