My main intent is on (some aspects of) how to think about prepositions. Finding exactly what preposition is correct in any given situation is a much harder (and less interesting; and much, much lengthier) task, I do not claim to have the mastery needed, and I will not make the attempt. (Also see an excursion.)
For now, I will deal only with English. However, chances are that much carries over to other languages heavy in prepositions, after corresponding mental adjustments.
For convenience, I will often speak of “word”, even where something more generic (e.g. “phrase”) or something more specific (e.g. “noun”) might have made sense. Contrast e.g. “father of X” with “adopted father of X”: In one case, “of” modifies a single word (“father”); in the other, in my main interpretation, two (“adopted father”). The role of “of” is the same, however. (Alternatively, “adopted” could be seen as modifying “father of X”, instead of just “father”, in the second case. By sticking with “word” and an understanding that “word” is a potential oversimplification, I dodge such issues.)
A good way of thinking about prepositions is as something that provides “slots” where modifying information can be attached to a word. In principle, each existing preposition, then, provides a potential slot (but with the reservation that not all prepositions and their associated slots necessarily make sense with the word at hand). Consider “the nail was driven into the wall, by the carpenter, with the hammer, on Monday, in the garage, during an eclipse”. (Where not every comma is necessary. I use them in part for emphasis of prepositions. Note, however, that the presence or absence of a comma can alter meanings. For instance, “by the carpenter with the hammer” might best be seen as “with the hammer” modifying “the carpenter”—and this would hold even more strongly for e.g. “by the carpenter with the blue jacket”.)
I originally used “a [sic!] hammer” and “a [sic!] blue jacket” above, but switched to “the” for consistency with a later discussion, where I wish to have carpenter and hammer on equal footing.
However, the original formulations seem more natural above, with a slight difference in meaning, and the reader should feel free to mentally replace “the” with “a” (more generally, to swap definite and indefinite articles in any position).
Note, in particular, how each and every prepositional slot can be removed (and how others can be added) without doing grammatical harm to the sentence. (But too many prepositions can be harmful for e.g. stylistic reasons.) Removing “into the wall”, admittedly, results in something awkward, but the problem is more of a semantic than grammatical nature. (As can be seen by a modified sentence that uses, say, “the car was driven” as the starting point and makes other modifications to match the new context.)
A complication is that several logically distinct slots can be associated with the same preposition, as with “he walked in the forest, the rain, and the morning” or “he walked in the forest, in the rain, and in the morning”. Unless deliberate humor is attempted (first case) or some type of stylistic effect sought (second case), such collisions should be removed through a rewrite (e.g. “it rained as he took a morning walk in the forest”).
However, prepositions are not the only way to add slots. For instance, “with the hammer” could be removed in favour of “using the hammer”. (This might even be the preferable formulation, as it is slightly more specific.) In other contexts, cases might serve to introduce slots (and some languages do so far more often than English), as with “I [nominative] gave him [dative] a present [accusative]”.
For precision and easier comparisons, I use more “international” case terminology than is usual in English texts. Whether this is formally correct can be debated, but here it seems better. Generally, the modern English take on grammar is a jumble of confusing and misleading terminology (as with e.g. the “possessive” case), failures to distinguish what should be distinguished (as with the idiocy of grouping all words ending with “-ing” under the label “ing words”, while ignoring their underlying nature), and whatnot.
As an aside, some languages, e.g. German, can combine prepositions and cases for slots, as with “er ging in dem [dative] Museum und in das [accusative] Café”—“he walked in [within, while inside, whatnot] the museum and into the café [that was present inside the museum]”. (The example illustrates dual use of “in”, if in a contorted manner.)
What preposition to use can often depend on perspective, context, and exact intent, as with “on [the island of] Cuba” vs. “in [the country of] Cuba”—even when they are both true at the same time (they need not be) they give a different perspective, the one might be more relevant in one context and the other in another context, and which matches the intents of the speaker can vary accordingly.
Questions about correct use of prepositions (e.g. “Is it ‘wait on’ or ‘wait for’?”, cf. below) are often impossible to answer for a similar reason. Several alternatives, corresponding to different slots, are often grammatically correct, but have different implications. In order to answer the question, the actual intended meaning can be of central importance.
Trying to answer such questions by following the examples of others, comparing counts from Internet searches, whatnot, is very dangerous—there is no guarantee that the uses found actually reflect the same intent. (Or that others actually have it right, even when the intent does match.) Yes, it is “Waiting for Godot”—but the waiter is still supposed to wait on his customers.
Often, such complications come with a great risk of mistakes even for natives—and the more so for us non-natives. Consider “talking to” (a good fit for someone holding a speech, lecturing a disobedient child, or using an ostensible conversation as an excuse for a monologue) and “talking with” (an actual conversation or other mutual exchange). A waiter might well wait for a customer (to order, to pay, to whatnot), but, by and large, he waits on his customers; customers, on the other hand, depressingly often find themselves waiting for a waiter and only wait on their respective waiter under truly exceptional circumstances.
A particularly interesting case involves agency, where I (but not necessarily others) try to differ between the performer of an action and the instrument/tool/whatnot of the action. To return to nails and hammers, many would say that “the nail was driven into the wall by the hammer”, if no further slots (or “competing” slots) are filled. I take the view that the hammer is not the agent, just a tool, and that “with the hammer” (or some other non-agent formulation) is preferable. The carpenter, on the other hand, is an agent and we then have “by the carpenter”.
This distinction is a comparatively new development in my own use and older texts have will typically reflect older habits. It might also be that I do not pay enough attention in the now and deviate from my own recommendations.
Also note that a somewhat immediate agency is intended in this case. If someone were to program a robot to use a hammer to drive in a nail, I would see the robot as the immediate agent, not the programmer (“by a robot, with a hammer”). However, in a less immediate context, this can change, e.g. in that “the code was written by the programmer, with a computer, for the robot, to drive a nail into the wall”.
To some approximation, the hammer can be viewed as a quasi- or metaphorical agent, should no true agent is present. If so, “by the hammer” might well be acceptable. However, we then have an inconsistency in use with phrases like “by the hammer”, “by the carpenter”, and “by the carpenter, with the hammer”. Keeping “with the hammer” in all cases removes this inconsistency. A further benefit is that we can always switch to “by the hammer”, with an increase in precision, whenever the hammer does have agency. (Unlikely in real life, but very possible in e.g. a fairy tale—and language must be open to all its uses.)
Other cases of legitimate non-agent “by” exist, as with “the carpenter drove in the nail by using the hammer”. However, while there is an agent-and-tool aspect involved, this “by” has a different nature. Ditto e.g. “live/die by the sword” and “the runner won by a second”. Care must be taken when comparing superficially similar cases.
This leads to the issue of ambiguity and differing intents. In “the carpenter drove in the nail by using the hammer”, the “by” might seem superfluous. Why not leave it out? In this case, we probably can, because the positioning and overall context makes a (non-trivially) different interpretation unlikely. However, there is some risk of ambiguity, and the more so in slightly different constellations. Consider “the carpenter drove in the nail [X] using his smartphone”, where “[X]” can signify an empty string, “by”, or “while”. Here, we would likely take the empty string to be approximately equivalent to “while”, and assume that the carpenter did not actually use the smartphone as an improvised hammer; however, we cannot say this with absolute certainty (he might, e.g., have been involved in a product demonstration for a “rugged” smartphone). With a hammer in a sentence like that, we would almost certainly have expected “by”, but here, too, some counterexample might be found. In more ambiguous cases, we might be left guessing, and the more so with a different order. Consider “the carpenter, using his smartphone, talked with his wife”—in the early days of mobile phones, it would have been a near given that he talked with her over the phone; these days, chances are that he was rudely doing something else, e.g. browsing the Web or playing a game, while simultaneously talking directly with his wife.
An interesting case arose in another text: I found myself speaking of replacing a poor worker by someone better. I promptly changed the “by” to “with”, because there is a potentially drastic change in meaning with “by”. Consider the phrase, “X was replaced with Y by Z” (where it is understood that X, Y, and Z are all humans). Clearly, “by Y” and “with Y” are very different, even in a shorter phrase (“X was replaced [by/with] Y”). In the specific case at hand, the claims “replaced [by/with] someone better” might both be true, but the role of the someone is different and the implication of the “better” is altered in accordance. (With “by”, the “someone better” might be a manager; with “with”, another worker. In the latter case, the “better” is both relevant and likely to be performance related; in the former, it is of no obvious relevance and might, even, be taken to imply that Y was one of X’s “betters”.)
Similar confusion can arise in other areas, and is a reason to pay great attention to cases and left out words. Consider “he can throw the javelin farther than [me/I]”: Both are arguably grammatically correct, with the implication/as an abbreviation of “[...] he can throw me” resp. “[...] I can”. These implications, however, are very, very different.
This also shows why a blanket “than me” should not be allowed. Even someone who argues that “than” serves as a preposition cannot avoid the unnecessary ambiguity. (A “prepositionist” might also see further ambiguities, e.g. in that “me” could be some type of distance, e.g. “my” height. Note phrases like “he can throw the javelin farther than 80 meters” and “[...] the club record”, which are poor matches with “[...] me”, should the intended implication be “[...] I can”.)
I might go as far as advice against leaving these words out at all. This removes the risk of confusion both when “abbreviationists” and “prepositionists” might clash and when other words, with a lesser case differentiation, are used. (For instance, “Tom” and “you” do not vary when plugged into “[...] X can” and “[...] he can throw X”.)
Disclaimer: I do not always practice what I preach, because my “feel” for language is largely built on what uses I hear and read, incorrect uses are very common, and my own uses are strongly influenced by that “feel”, even when I “know better”.
The earlier parts of this text already show that picking the right preposition is tricky—and the list of potential prepositions, slots, and words where they can be tagged on with different effects is very long.
However, it is worse than that:
For the below examples, I go by subjective impressions, not actual statistics. Corresponding reservations apply, but are not that important in a bigger picture. (The underlying issue of inconsistent use remains, even should the specific examples be wrong.)
The preferred combination of preposition and implication can differ from one variation of English to another, over time, and even between highly skilled users of the same time, country, whatnot. (Let alone users who are not highly skilled.) My own development (as a non-native speaker) has certainly been hampered by how often I have seen inconsistent uses among the natives.
A good example is “about”, which seems to grow ever more common, often at the cost of “of”. (“I thought about you” over “I thought of you”, e.g.) I have tried to steer my own use towards “of”, as this matches what I seem to encounter more often in older books (and I usually try to err on the side of “too old” use, rather than “too new”), but I seem to naturally drift towards “about”, likely through the influence of modern TV/movies.
Another is whether constructions like “about 5 o’clock” or like “around 5 o’clock” are better, and the choice might be largely a matter of personal taste. Here, too, there might be a drift towards “about” over time, but there might also be a U.S./U.K. or other synchronic difference.
Both “about” and “around” can serve as adverbs and here the lines between the two roles are blurred: A third alternative is “approximately 5 o’clock”, using a pure adverb.
An interesting test is to see what happens with the insertion of a comma: By current rules, “approximately, 5 o’clock” is possible or even preferred, while “[about/around], 5 o’clock” is nonsense. Ditto, re-formulations like “5 o’clock, approximately” vs. “5 o’clock, [about/around]”. In the future, this might change. (And I have not spent enough time on the issue to judge how strong such tests actually are.)
(Such blurring might be a future sub-topic. Also note how “about” in “he hang about” and “he hang about the house” can be viewed as an adverb in one case and a preposition in the other, which is not entirely satisfactory to me.)
The above discussion of a seemingly superfluous preposition is a special case of a mostly harmful tendency to leave out “unnecessary” words. The word “that” is a particularly common victim.
A longer text is in my backlog, but a few words in the interim:
The idea that something unnecessary should be left out has some merit, as with Strunk’s “omit needless words”. However, most adherents of this idea, much unlike Strunk, seem to leave out words that actually bring value, e.g. because they remove ambiguity and reduce the time/effort needed to understand the text on an even syntactical/grammatical/whatnot level. Consider “he whispered, he was tired”: is the intent that “he whispered and he was tired”, “he whispered that he was tired”, or something else yet?
Too eclectic removals can also lead to inconsistent sentences: It is often the case that (a) a particular word could be removed in the specific sentence at hand, but (b) it cannot generally be removed from sentences of a similar shape. Removing it in one case, at best, leads to inconsistencies. at worst to errors, because a sentence is misjudged by the author. Indeed, I have seen examples of sentences that might have been clear at the time of writing but have since become ambiguous, and where that missing word would be needed in the now. (E.g. because some other word in the sentence has taken on new meanings or new grammatical roles.) The removal of “that” is particularly common and problematic, e.g. through creating garden path sentences.
TODO gather examples for that longer text