It’s quite interesting how level design has evolved stylistically throughout the past 10 years.
With Doom, level design evolved such that realistic lighting and feats not possible with the technology were feigned. Lots of very small sectors created the effect of light fading out, and transparent textures created bridges that could be walked over and underneath. Some more subtle tricks were used to give geometry a sense of depth beyond the limited technology and evade the ‘stuck in a box’ feeling. Mappers who could pull off tricks like these were given much praise.
Quake was all about the shadows and complete 3D freedom. To master Quake level design, you had to really understand how to use the entirely 3D environment appropriately in response to the gameplay you want. The point-based lighting (not radiosity) was also something that took a while to master - not only was white your only option, but you could only do so much with it given the Quake theme. Castles typically had lots of walltorches, open ceilings, or (more likely) faked radiosity to give areas brightness that the tech hadn’t. Being able to swim around in water was also something that needed very careful consideration.
Quake 2 took Quake a step further with coloured lighting, texture-lights (and with it, sky lighting). The best levels for Quake 2 made subtle and sensible use of each, using coloured lighting enough to bring the environment alive and direct the gameplay, but not enough to be the focus. It was about this time that lighting really became a very important skill, since it still required faking and invisible texture-lights and careful brightness choices. At the same time, environments started getting detail that couldn’t be afforded before.
Unreal, being roughly the same era as Quake 2, was also about colour. However, it also had a focus on large areas. Again, the best levels used these sensibly, but not too much.
Both Q2 and Unreal were important milestones. It were they that created screenshots we could awe at due to their beauty rather than the tech. They started to become works of art, and in this, reflect the artistic flair of the designer. Of course, Quake 1 also had this (Ikka Keranen was particularly good at embossing a very individual style onto his maps), but Q2 made it the norm to think about prettyness alongside gameplay.
With Half-Life, this died off. It was hard to put your own style to a map and also appeal to the theme of the game (‘realism’). Instead, it was drawn towards how you achieved small things, like realising doors, rooms, staircases, handrails and so on. At the same time, level design became a team effort where before, an entire game could (feasibly) have been mapped out by a single person. This has continued just about all games since - maps now expose the character of the game rather than the designer (which is really how it should be).
These days of course, it’s hard to come across a map that isn’t the product of a dozen hands, and in this, maps become a combination of many styles watered down to fit within the confines of the game theme. The only remaining area to demonstrate your own touch is in the gameplay itself.
Of course, some games aren’t like this. Games which feature large fantasy elements and style (think Unreal, Quake rather than Half-Life or Splinter Cell) are fantastic for putting your own style through, simply because you can create something so different to the rest of the game, and it will still seem at home. Those games almost expect the maps to be unique and varied. Other games expect the maps to be realistic and believeable.
As a mapper, I was never good at the former. I always wished I could do the far-out fantasy manic creations of people like Ikka, but the imagination (and texture-making ability) wasn’t there. That’s probably why the advent of HL was such an omen to people like me, who instead had the task of taking existing ideas and taking them further. It felt more natural, easier and more rewarding - the focus was on creating new gameplay rather than something new to gawp at.