Every time I try to buy oats, I run into trouble. I find myself frustratedly scanning the shelves in the baking section, looking next to the whole wheat flour and the white flour and the corn flour and the various types of sugar. Pretty soon I'm at the chocolate chips and there still aren't any oats. I'm left wondering if my grocery store has just stopped selling oats, or if I should be kicking myself for being really unobservant.
Eventually, I give up and continue with my shopping, only to come across oats a few minutes later, in the cereal aisle.
I'm not sure why my brain can't seem to remember that oats are, at least according to grocery store designers, considered "cereal" and not "baking ingredients." And of course, there's no doubt that a store with the opposite layout would confuse some other group of shoppers. The problem is, some foods really belong in multiple grocery aisles at the same time.
The oats problem also gives us an insight about language. If a language was as rigidly structured as the layout of a grocery store, as various thinkers over the centuries have advocated, things would actually become more confusing.
The logic generally goes something like this: Wouldn't language be so much more efficient if the individual parts of a word actually had predictable and organized meanings? There are many rational appeals to such a system. For one thing, it's easily learned, a matter of remembering a relatively short list of basic concepts that can then be combined infinitely. For example, if you know that pa means "early" and lo means "food," then clearly palo means "breakfast." Furthermore, if you need to talk about something new, you could just combine a few existing syllables in a new way, and everyone would still know what you were talking about. So if ni is "late," then clearly nilo is "dinner."
Once we have the world divided up into basic categories, also known as a taxonomy, all we need to do is give each of them an associated syllable, and we'll have the most efficient, logical language the world has ever seen. Everyone shall learn it in a few hours! Universal world comprehension will be ours at last!
Unfortunately, it's not quite that easy. Sure, several people have tried to create universal taxonomic languages: some relatively well-known examples are Loglan, Lojban, and Ithkuil, which famously developed a cult following in Russia, while less familiar ones include A Common Writing, Real Character, Ars Signorum, and Ro. (For a more detailed examination of taxonomic language, check out In the land of invented languages by The Week's Arika Okrent.) All of them try to break the world down into basic semantic concepts, more basic than any natural language, and then re-build the words for more complex concepts by combining the pieces.
But it turns out that carving the world up into basic concepts isn't quite as easy as it looks. Some combinations of meanings can have multiple interpretations: for example, palo as early+food could mean "breakfast," but it could also mean "hors d'oeuvres." And some things, like oats, belong plausibly to multiple categories. (Let's not even get into the fact that my grocery store puts the ball-shaped mozzarella in the deli section with the fancy meats but the brick-shaped mozzarella in the dairy section with the yogurt.)
While taxonomies have found some application in things like the Dewey Decimal system and the categories in Roget's Thesaurus, they require considerable knowledge of the conventions of the system or a lot of cross-referencing in order to be useful. For actually speaking a language, fuzzy categories and ambiguity end up being more efficient.