## 7.3 Words and Eras

For most of the book, I have worked from a dataset that started with the JSTOR word list, then made two major edits. First, I cut out words that don’t look like they are part of the content of the papers I’m focussing on. Second, I cut out words that only appear 1 to 3 times in a paper. For this section, mostly still excluding the non-contentful words - though I’ll note one point below when I consider them - but I’m restoring the words that appear 1-3 times.

On the other hand, I’m restricting attention to the 5000 most common words in the data set. In practice, that means restricting attention to words that appear about 2200 times or more. That is, the word has to appear about once ever 15 articles. That doesn’t mean it has to appear in $$\frac{1}{15}$$ of the articles; it could appear very often in a few articles. But it does have to get those 2200 appearances somewhere. The reason for this restriction will soon becme clear.

To start, here are the most common words from each era - remembering that we’ve filtered out a lot of ‘stop words’.

Table 7.8: Most common words in each era
1876-1945 1946-1965 1966-1981 1982-1998 1999-2013
sense sense theory theory theory
nature theory case case case
experience fact true true true
fact true sense might might
knowledge philosophy might argument argument
time case fact moral first
mind use first sense example
philosophy moral time first account
theory first argument truth just
world time moral fact truth

That doesn’t tell us a lot. It is a bit interesting that ‘theory’ and especially ‘case’ are much more prevalent after World War II than before it, but that’s about it.

We learn more by comparing the frequency of each word in an era to that word’s frequency across the whole data set. So rather than ask how many times a word appears in an era, we might ask what percentage of the word’s appearances are in that era. That gives us a sense of the characteristic words of a given era.

But as it stands, that doesn’t work either. A lot of words have 100% of their appearances in one or other era. For instance, we don’t see any occurrences of ‘elga’, ‘kolodny’, ‘knobe’, ‘weatherson’, ‘rayo’, ‘obama’ or ‘greaves’ until era 5. So they’d all be tied for being the most characteristic words of the 1999-2013 era, since 100% of their appearances in this era. And while some of those words are somewhat important to the era, I don’t think that’s quite what we’re looking for.

So what I decided to do, following common practice, is to restrict the study to the 5000 most common words. This excludes all the words I just listed. (I’d have to extend it a fair bit further to get them; ‘elga’ is at position 10468, ‘greaves’ at 16583, and the others in between.) And then we can look at which words from that 5000 have the highest percentage of their occurrences in a given era.

But it turns out even that isn’t quite what we want, though it’s not clearly not what we want. If you do what I just said, most of the words that show up are between the 4000th and 5000th most common words. It’s just much easier for rarer words, especially names, to appear at one particular time. So I decided to show you a whole bunch of tables.

For every one of the following tables, I restricted attention to the top n thousand words in the data set, and then asked of those words, which have the highest percentage of their occurrences in different eras. I think looking at the tables for each value of n from 1 to 5 is useful. There will be some repetition; sometimes a common word has a distinctive distribution. But you learn a new thing each time you increase n. I cut it off at n = 5, but you could keep going beyond that. (Though if you did, you’d find a few more latex words, and journal names, and OCR errors, that in retrospect I might have wanted to delete from the dataset.) So here’s what we get restricting attention to the 1000 most common words.

Table 7.9: Most distinctive words in each era (among 1000 most common words)
1876-1945 1946-1965 1966-1981 1982-1998 1999-2013
consciousness statements jones intentional epistemic
reality statement legal rationality models
unity analytic quine rawls normative
soul ethical predicates frege population
feeling philosopher criteria realist intuitions
sensation art wants strategy worlds
ultimate men logically beliefs modal
whole phenomenological wittgenstein desires strategy
absolute descriptive rawls probabilities agents
quality aesthetic behaviour realism david

Is ‘jones’ at the top of 1966-1981 because of Sellars, or Gettier, or Frankfurt? I think the answer is, all of the above! I think ‘frege’ appears prominently in 1982-1998 because of “Frege cases”, not because of a particular upsurge in attention to Frege’s own writings. Both ‘quine’ and ‘rawls’ are a bit later than their most famous writings, which makes sense. And note ‘intuitions’ turning up as a distinctive word in the 21st century literature. It’s interesting, I think, that it isn’t used as much in the era intuitions were allegedly dominating philosophy as in the era when metaphilosophy became such a big deal.

We can look at the graphs of how frequently these words appeared over time to get a sense of what it means for them to be the distinctive words of an era. I’ll just graph the top 5 for each era, because otherwise the graphs get too cluttered.

You can confirm that the words in question really do peak in the era in question.

The y-axis measures the frequency of the words among the words in the JSTOR data. So that excludes 1 and 2 letter words, and whatever stop words JSTOR has excluded (like ‘the’, ‘and’, and the like), but includes things like bibliographic information and latex code. It probably overstates the actual frequency of the words by something like 25 to 50 percent. So if it says that a word appears 1 time in 400, its real frequency is, as far as I can tell, more like 1 time in 500 to 600.

Note that the distinctive words of the middle eras are much less frequent than the distinctive words of the early eras or even (to a lesser extent) the later eras. The word ‘consciousness’ seems to have appeared, on average, about once a page in the early years! No word is this prevalent in the later years.

Let’s expand the data set and look at the 2000 most common words.

Table 7.10: Most distinctive words in each era (among 2000 most common words)
1876-1945 1946-1965 1966-1981 1982-1998 1999-2013
organic statements popper davidson epistemic
consciousness usage strawson kripke testimony
reality statement jones women phenomenal
unity factual legal preferences multiple
existent analytic utilitarianism intentional options
soul synthetic punishment van option
feeling ethical quine rationality counterfactual
eternal signs entailment rawls vagueness

Both ‘marx’ and ‘davidson’ turn up one era later than I would have guessed. And I would have thought ‘strawson’ was either an era earlier or an era (or two) later; earlier for the work on descriptions, later for the work on responsibility. So those are a bit interesting. Testimony really was a big topic in the early 21st century. And note ‘worry’ turning up. Recent philosophy has a very distinctive lexicon, which we’ll see more and more of.

Here are the graphs for the first five words in each column.

Table 7.11: Most distinctive words in each era (among 3000 most common words)
1876-1945 1946-1965 1966-1981 1982-1998 1999-2013
psychical ryle marx nuclear williamson
esthetic ayer hare parfit credence
apprehension statements hempel fodor luck
spiritual usage utilitarian computational epistemically
volition statement strawson twin epistemic
organic philosophic chisholm nozick target
impulse western goodman evans representational
intellect peirce geach putnam worry

Apart from in the earliest era, we’re starting to see the majority of the list here be names of famous (male) philosophers. And we get a pretty good sense of when they were being most commonly discussed. The graphs show this in slightly more detail. (The fourth graph is a little busted because one of one year when ‘nuclear’ went nuclear.)

The pattern stays the same as we expand to 4000 words.

Table 7.12: Most distinctive words in each era (among 4000 most common words)
1876-1945 1946-1965 1966-1981 1982-1998 1999-2013
bosanquet ryle marx nuclear williamson
psychical ayer grue parfit scanlon
stout historian hare dummett credence
instinct dewey austin dennett scenario
bergson poem hempel burge bob
esthetic philosophies hart fodor luck
apprehended verifiable popper dworkin arguably
presentations sartre utilitarian computational doxastic
apprehension civilization strawson consequentialist global

I’m a bit surprised to see ‘bob’ here; I’m not sure if this is Stalnaker, or Brandom, or who is being referred to so informally. The graphs don’t show a great deal that isn’t visible on the previous set, so let’s skip over them about move the limit up to the 5000 most common words.

Table 7.13: Most distinctive words in each era (among 5000 most common words)
1876-1945 1946-1965 1966-1981 1982-1998 1999-2013
bosanquet emotive illocutionary deterrence hawthorne
schiller stevenson marx laudan chalmers
psychical ryle hintikka nuclear contextualism
spencer ayer grue churchland williamson
muscular historian capitalist rorty normativity
stout dewey hare parfit woodward
instincts santayana lakatos dummett scenarios
instinct poem alienation dennett scanlon
bergson philosophies austin burge credence
antithesis verifiable hempel fodor egalitarianism

And finally we get that not all the names are of men. Korsgaard is here, and the model doesn’t discriminate between the Churchlands, so at least part of the reason for ‘churchland’ in 1982-1998 is Patricia Churchland.

Apart from the names, the words in the final era are all fairly much as expected. I’ll come back in the last chapter to ‘credence’, because I don’t think everyone realises how new a term it is. And ‘egalitarianism’ was used so much in 1982-1998 that I’m very surprised it can turn up here.

I have no idea why ‘muscular’ is such a common term in the first era. I suspect I wouldn’t be happy to find out.

Here are the graphs for the five most distinctive words in each of the eras.

Here part of the story, as you can tell from the y-axes, is ever-increasing diversity. Figures who feel like they dominate the current age, like Williamson, Hawthorne and Chalmers, are discussed much less than figures like Ayer or Ryle were a couple of generations ago.