The December 17th Economist contains an article entitled "In Search of the World's Hardest Language". Such things usually make me groan, but this one is actually pretty good. At the level of detail one can reasonably expect in such a context, the facts seem to be correct, the range of languages considered is broader than usual, and it recognizes that there are multiple factors involved. There are, however, a few points worth making about this article, as well as inferior examples of the genre.
One is that it doesn't really define what is meant by "difficulty". This could mean how hard a fluent speaker has to work to speak or understand the language, how hard it is for a child to acquire the language as his first language, or how hard it is for an adult to acquire as a second language. The context suggests that the last is what is intended. Even there, do we mean how hard it is to acquire near-native fluency, or how hard it is to acquire the rudiments, or some intermediate level of functional ability? These may turn out to be different.
Another question is whether difficulty is a property independent of the learner. It is, I think, the received view, that how difficult a language is to learn depends in part on what languages the learner already knows. Navajo is by all accounts quite difficult for English speakers, but probably not all that difficult for speakers of Apache. Koreans seem to find Japanese grammar much easier to master than English speakers do, presumably because the syntax of the two languages is similar.
It is also quite possible that different aspects of languages are more or less difficult for different learners even if one controls for what languages they start with. Some people may be good at, say, learning complex morphology but not very good at learning an exotic sound system. Others may be good at the latter but not at the former.
A point that frequently arises is the idea that languages that pack a lot of information into words are difficult. Is it really self-evident that it is harder to pack deal with complex words than with multi-word phrases that convey the same information? If a language puts a lot into a word but does so in a transparent way, so that words are easy to parse, interpret, and construct, why should this be difficult? It may well be that the perception of difficulty here merely reflects unfamiliarity, which is likely true of quite a few of the features often cited as leading to difficulty.
Perhaps the largest problem with this entire genre is that factors that might lead to difficulty are enumerated but no empirical evidence as to which languages are actually difficult to learn is ever adduced. There is very little research on this question, so one can't criticize authors for failing to cite it, but in the absence of such research discussion of the factors that lead to difficulty is mostly mere speculation.
The article also seems a bit confused regarding about theoretical debates in linguistics, conflating the question of the extent to which universals particular to language exist and the related question of how and to what extent languages differ from one another, with the question of whether language influences thought. These are two quite different questions. For one thing, the examples cited by proponents of the Sapir-Whorf hypothesis are almost entirely lexical, while Chomsky's proposals regarding universals are almost entirely grammatical. It would be perfectly possible, in principle, for there to be a world in which all languages had exactly the same grammar, differing only in their sound systems and lexica, where the differences in lexicon influenced thought in accordance with the Sapir-Whorf hypothesis.
In any case, in my estimation support for the Sapir-Whorf hypothesis is even weaker than the article suggests. It mentions that critics consider the examples put forward "trivial", but not that even such "trivial" effects are disputed and, I think, usually wrong. Our colleague Lila Gleitman is a real expert at this: she can debunk most Sapir-Whorf claims within seconds.
The example given in the article epitomizes the problem nicely. It points up a correlation between language (exclusive reference to compass directions vs. use body-oriented directions) and behaviour (constant tracking of compass direction vs. the lack thereof), but doesn't even attempt to establish the direction of causation. Surely it is perfectly plausible that what is at work here is a cultural focus on external vs. individual orientation which has linguistic consequences. If you don't care about individual orientation you don't have words for it. The Sapir-Whorf claim would be that it is the lack of words for left and right that causes people to rely entirely on compass directions. Even if you consider this plausible, the facts cited in the article do nothing to favour this direction over the reverse.
Whereas it is hard to say much about which language is the most difficult to learn, the difficulty of learning writing systems is better understood. English does indeed have rather a bad writing system, but it pales in comparison to that of Japanese, particularly in its pre-WWII version, which feature more and more complex Chinese characters and historicizing kana spellings. And although we have little direct information about the difficulty of learning it, the vaguely similar writing system of Hittite must be another contender for the title of most difficult writing system.