When I moved to another job, they talked about how, in theory, code reviews was the number one most effective way to increase code quality. But they had given up on it, because they felt it came down to a discussion of where to put the dots and commas (or, rather, semi-colons and parentheses).
Quite aside from the issue of code reviews, I had come to realize that I spent much more time reading my code than I spent writing it. Every debugging session is spent reading code over and over. Every time you have to add a feature or change some functionality you have to read the code, and re-read it to avoid breaking stuff. Don't tell me tests will do it for you. Now don't get me wrong, tests are great and I strongly advocate test-first coding, it's a great way to achieve focus and clarity of thought. But when a test fails, you're thrown into debugging mode, which means reading code.
So I concluded it was worth spending a little extra time typing longer variable names, and taking the time to find descriptive names. It was worth spending a little more time breaking down those long methods and simplifying those complex structures. Whenever I was reading code that made me stop and think, I would usually refactor it to be clearer (although the term refactoring hadn't been invented yet). I would also change existing code to make a new feature fit in better, in a more readable and more logical way.
In "The Pragmatic Programmer" the distinction is made between "programming by coincidence" and "programming by intention". We all have to do it occasionally, a little trial-and-error programming, because we're not quite sure how things work. That's "programming by coincidence". Before you check your code in, you want to make sure you understand what each statement does and that all unnecessary statements are removed. Then you've transformed the code from coincidental to intentional.
But that's not enough. Your very functional and intentional code is going to lose you valuable time unless you also transform it to readable code, which clearly displays your intent.
A much-touted wisdom is that you should document and comment your code. Fair enough, that works, but it has many weaknesses. Your energy is far better spent making the code explain itself. Comments often lie, but the code is always pure truth, so prove it in code. Only comment on "why" you are doing something, and that only if you cannot make it evident in the code.
I like the following sound-bite from "Refactoring": "Any fool can write code that a machine can understand, it takes a good programmer to write code that a human can understand."
Test-first "anything" is efficient and focused because it sets up the criteria for success and the means to measure it up front. So what's the best way to test if your code is readable? Get another person to read it, i.e. a code review.
I'm very grateful to those who review my code carefully and pick on every detail, it makes the code better and it helps assert that my thinking was clear. That gratitude gives me the energy to return the favour by reviewing their code equally mercilessly.
You will sometimes, but rarely, find bugs by just reading code (only because everybody has a brain-fart now and then). But the real value of the reviews is in the "dot and comma" discussions and especially in picking good names. In addition to making sure that the code is easy to read, it will sometimes bring a real little nasty bug to the surface.
An example: An index into an array of values is stored into a variable called "value". When the reviewer makes you change the name to "valueIndex" instead, some parts of your code may start to look weird (the bug was exposed).
Clarity of code really is clarity of thought.