I just finished reading the question asked by Bobnix, in which RegDwight referred to another question with an interesting answer by Kosmonaut. Kosmonaut refers to the great number of pictograms (Kanji or Hanzi) available in Japanese and Chinese, and mentions that the task of memorizing our weirdo spellings pales in comparison to learning vocabulary in one of those languages.
That got me to thinking. When I first started studying Japanese, I first learned the two written versions of the syllabary, hiragana and katakana. And when faced with the formidable task of memorizing thousands of characters and their various readings, I wondered why, given the phonetic language, Japanese still stuck with all those originally Chinese characters. Were they just masochists?
But I dug in, and as I learned more and more kanji a strange thing happened. I realized it was actually easier to read the language with the kanji than without them, because so many Japanese words sound alike (or at least their parts do) and to render them in hiragana would force me to slow down and try to figure out which ほう (hou) they meant: 保, 俸, 倣, 剖, 報, 方, 法 or any of the others. Learning the more complicated writing method actually let me read faster, and to understand words almost pre-apprehensively. By that I mean something a little like looking at the hands of an analog clock and understanding the time without relating it to a numerical equivalent.
Now for English. We have sound-alike words like to, two, and too (or even tu, if you count Shakespeare's imagining of Julius Caesar's dying line). If we went to a strict phonetic spelling system, all those would be spelled the same. I think there are cases where such a thing would actually slow us down. And it may be that the more difficult and idiosyncratic the spelling is, the more likely we are (as Kosmonaut said) to remember it. Further, having remembered it may mean we are more likely to recognize it more easily. Or something like that.
This is just a supposition on my part. It has plausibility and feels right to me, but that doesn't mean it is right. I'd be interested if anyone knows of any information or research done on either side of this argument.