Happy new year! I've wanted to blog this for a while, but sickness intervened. May all your new years be free of bacterial warfare. At any rate, here's my small contribution to the burgeoning Ngram literature:
Link for a bigger picture.)
The Google corpus seems to be pretty sparse before 1800, so I would not take the big spike of "tyranny" of around 1760 as evidence of much. But I'm curious about the slow decline of "tyranny" and the slow increase of "dictatorship" as a catch-all term for the pathologies of political regimes.
"Tyranny" is the older, Greek term. Originally a more or less neutral designation for a "usurper" (as opposed to a legitimate heir in a dynasty) it was later transformed into the term for the worst form of government in Plato and Aristotle, and the sense stuck. Tyranny, however, was never precisely characterized by any institutional features; though there was a loose association between tyranny and "lawless" or "arbitrary" monarchy, ultimately the tyrant was simply the unjust ruler. Thus all regimes can become tyrannical; the distinction between tyrannical and non-tyrannical government is moral rather than institutional.
By contrast, "dictatorship" is a Roman term that is far more directly tied to a particular set of institutions. (For a quick and useful potted history of the term, see Jennifer Gandhi's "Political Institutions under Dictatorship"). The dictator was originally a magistrate chosen by the Senate for a limited time (six months) and formally empowered to act extralegally in situations of crisis. The term acquired a bad connotation after Sulla and later Caesar abused the office in various ways, but it still retained an association with a particular institutional context: the dictator is a sole ruler, typically acting extralegally and commanding substantial force, and so on. It is not used to refer to a distinct political regime by any of the early modern political thinkers I'm aware of, even those like Montesquieu who are interested in classification matters (Montesquieu prefers despotism to tyranny as well), but it is revived by Marx in a sort of paradoxical turn of phrase: the "Dictatorship of the Proletariat." (I imagine the paradox was intentional: in the Roman context, the dictator had often been the instrument of the ruling class to put down revolts from below). What distinguishes the dictator from other rulers is not the justice or injustice of his actions, but the fact that he can "dictate" - impose a command on others.
With Marx we also see a return to the more "neutral" original sense of dictatorship; and though the term still carries a sting - to call someone a dictator is typically to imply something bad about them - it typically needs to be qualified or intensified with some adjective ("brutal" or "totalitarian" dictator, for example). In political science, in fact, some people use "dictatorship" to mean simply non-democracy. (This is not to say that they ignore variation within non-democracy; they use dictatorship as an umbrella term that encompasses everything from Mexico under the PRI to the Soviet Union under Stalin, but they do pay careful attention to some forms of institutional variation within this vast array of regimes).
But why does the more "institutional" term - in fact, a range of such terms, from autocracy and totalitarianism to authoritarianism - seem to displace the more "moral" terms (like tyranny, and to a lesser extent despotism, which also seems to have been popular in the 19th century) as labels for political pathology? Part of this must be the rise of democracy as the recognized "good regime" - even if, in actual practice, democracies often disappoint (but they disappoint less!). If the good regime is institutionally identifiable, then the bad regime might also be institutionally identifiable. Here's another Ngram:
Around 1900, the term "democracy" rises enormously in popularity, while the usage of older terms for different kinds of political regime decrease significantly in English. (Here's the Ngram for the Spanish corpus: some interesting differences, similar broad pattern). It's like the distinctions between non-democratic political regimes "flatten," as Norberto Bobbio argued in his "Democracy and Dictatorship" (though I think he got this from Hans Kelsen): the key dimension of difference we see today among political regimes is whether authority is imposed from above ("dictatorially") or emerges from below ("democratically"). (In social science practice, the key dimension tends to be whether executive recruitment happens through genuinely competitive elections, but it is not clear that this always corresponds well to the contrast between "imposed" authority and "consent" just mentioned).
I also wonder whether this sort of phenomenon is a problem: do the Ngrams tell us anything about the potential loss of conceptual distinction, or merely about words? The conceptual distinctions need not go away: if political scientists start using "dictatorship" as an umbrella term for non-democracy, this does not mean they ignore all variation among non-democracies. (In fact, regime classifications have proliferated in recent years). But if the terms themselves incorporate important conceptual distinctions, their decline suggests a loss of conceptual variety. And then we might ignore, for example, the moral dimensions of variation among political regimes to focus on institutional variations that do not have sufficient moral relevance. But my thoughts on this point are still too muddled, so best to stop here.