Last week I did a twitter and it got a big response (for me, that is). It was about a recent paper on language that appeared in an economics journal and it lit a fire under other people as well. The paper is called “Do Linguistic Structures Affect Human Capital? The Case of Pronoun Drop” and it’s by Horst Feldmann. I thought that in addition to dunking on that paper on Twitter, I’d spell out some of the fundamental problems with it. Here goes.
Hide your students! Hide your TAs!
Economists are doing linguistics again. 🙄 https://t.co/SBeZA60HFu
— Joe McVeigh is also @EvilJoeMcVeigh@mstdn.social (@EvilJoeMcVeigh) December 4, 2018
Got the Roley on my arm
This is the very first sentence: “Recent research in economics suggests that certain linguistic structures affect both individuals’ behavior and aggregate outcomes.” That would be fine, but the recent research he’s talking about is all seriously flawed and, as Feldmann makes clear, he is not competent enough to recognize the problems with those studies.
Feldmann references Chen 2013, a paper that seemed to show that people are more likely to save for the future if their language does not mark the future tense, and that countries in which these languages are the majority have higher national savings. But Feldmann doesn’t reference Roberts, Winters & Chen 2015, a paper which disproved the claims of Chen 2013. And he doesn’t seem to recognize any of the problems with the way that Chen simplified the linguistic analysis in his study (hint: it’s a problem).
Feldmann’s second paragraph says that his research will draw on the linguistic relativity hypothesis, aka the Sapir-Whorf hypothesis. The only problem with this is that the Sapir and Whorf hypothesis has been thoroughly debunked for generations. Feldmann claims that “the past few decades, several experimental studies in linguistics and psychology have corroborated the hypothesis (for a survey of these studies, see Everett 2013).” Studies over the past few decades have not corroborated the Sapir-Whorf hypothesis, but rather a very narrow and limited version of it, sometimes called the weak version of the Sapir-Whorf hypothesis, and even Everett (2013) is careful with how he talks about these studies and what they mean. But Feldmann presents the Sapir-Whorf hypothesis as a monolithic theory that has been proven, rather than a collection (and oversimplification) of several theories that have been the center of fierce debates in the field of linguistics. Feldmann’s oversimplification of linguistics seems to be a theme.
The third paragraph starts like this:
This paper empirically studies the effect of grammatical rules that permit speakers to drop a personal pronoun when it is used as a subject of a sentence (pronoun drop, for short). Davis and Abdurazokzoda (2016) give a good example of such a rule: “pronoun drop is permitted in Spanish, such that the English sentence ‘I speak’ may be translated either as ‘Yo ablo’ or as ‘Ablo,’ dropping the subject pronoun ‘Yo.’ In contrast, pronoun drop is not permitted in English, as the pronoun ‘I’ is required to make sense of the sentence” (p. 544). (Feldmann 2)
This is bad for several reasons. First, and most obvious, is the part claiming that Yo ablo is Spanish. I haven’t taken a Spanish class since the fifth grade but I know that hablo should have an “h” in front of it. The fact that this got past three authors and multiple editors shows that there’s a willful ignorance in regards to language in these studies – not exactly what you want in research ABOUT LANGUAGE. The second problem has to do with how Feldmann claims that English grammar requires the pronoun “I” in a sentence like I speak to make sense of the sentence. While we can say that “I speak” is a sentence in English, it is highly unusual and unidiomatic. If you speak English, think about if you would ever say (or write) just “I speak”. If someone asked you if you speak English, would you reply “I speak”? No, you wouldn’t. Because that’s not how people speak English. In fact, in the Corpus of Contemporary American English, which includes 520 million words of spoken and written English, there is only one example of “I speak” as a sentence on its own. But look where it comes from:
# A Methodist church. John’s son, Bob, has arranged it. John’s coffin, pale gray anodized metal on a chrome stand. Same color as Muriel’s, years ago. # I step up to the podium. A polished hardwood podium, tightly jointed with chamfered edges. Where each Sunday morning the minister stands. I look out at the small gathering. In the front row, Bob and his wife and sons, and Hilary and Dutch, between them my empty chair. # I speak . # ” Friends. This morning I was awake early as usual, sitting at the table in the dining room – the table most of you probably think of as John’s and Dutch’s and my table. By the window. As often happens, I watched Caspar walk across the yard. He has done this nearly every morning since I came to the Manor. On the table in front of me I had a piece of paper and a pencil. I intended to write what I would
That passage (a piece of fiction from the New England Review) also includes sentences like “A Methodist church.” and “Same color as Muriel’s, years ago.” Not exactly sentences as you would think of them (we can have the debate over what a sentence is later, just hang with me).
This leads to another problem with the above claim. Standard Written English may “require” the pronoun to not be dropped (I’m putting “require” in quotes because I’d argue this claim – let’s let it slide for now). But no one speaks Standard Written English – I know it sounds stupid to say this, but here we are. This problem seems simple at first, but it is fundamental. Standard Written English is a written register. It does not follow the same rules as spoken English. If you want to get an idea of what I’m talking about, record a couple of minutes of spontaneous spoken language – your family around the breakfast table, say – and then try to transcribe it. It will look nothing like any written language you’ve ever come across (assuming you’re not a linguist or linguistics major).
For example, consider the following sentences. The verbs missing a pronoun subject have been underlined.
| Source | Example of dropped pronouns |
| COCA:2012:FIC Bk:Sailor |
looking mildly puzzled. “What’s that?” Deiter looked around. Saw nothing notable. Just a wall with a picture of a fox hunt on it |
| COCA:2012:FIC Analog |
straight to the point,” the man said. “I was nearby. Saw your truck pull in, and wanted to talk to you.” He looked |
| COCA:2011:FIC Bk:Serial |
aware of his right hand moving quickly on the lower periphery of her vision. Saw an instantaneous glint of silver. A blade! Something peculiar about it. # |
| COCA:2010:SPOK NBC_Today |
because you’re far too busy to go to theater with me anymore. Saw a brilliant play called “The Pitmen Painters.”
KOTB: Yes. |
| COCA:2009:SPOK NBC_Dateline |
Unidentified Man 10: Oh, really?
Mr-KLEIN: Yeah. Saw this tournament. Looks like it’s going to be a good one? |
| COCA:2009:SPOK NBC_Dateline |
MORRISON: Saw that.
Mr-KLEIN: Heard him. Saw him. Looked him right in the eye. MORRISON: Totally believable? |
| COCA:2009:SPOK NBC_Dateline |
…friend to both Bill and Vickie, lived right across the street from them. Saw the fight that day and what happened next. |
| S73U 1369 | S0525: >> yeah you’re right oh yeah they came out didn’t they? Says your daughter’s asked for a white one we thought we’d |
| S7FU 605 | S0024: are you saying you’ve always thought you had a hamster in your head and you just never told me?
S0144: not always thought it I – I can feel him talks to me S0024: does he? S0144: tells me things S0024: in hamster language? S0144: no he’s got an RP accent |
Those examples are from the Corpus of Contemporary American English and the Spoken British National Corpus (2014). The first three examples are from fiction, i.e. written English. The next four are from interviews on national news shows, so formal-ish spoken language. And the final two are from spontaneous spoken conversation. It took me longer to clean up the formatting than it did to find them. There are many, many more examples.
The thing is, English doesn’t conjugate verbs for person, like other languages do. Or at least, the form of the verb is the same for every pronoun except the 3rd person singular in the present tense:
I talk We talk
You talk You talk
He/she/it talks They talk
It would seem that English speakers are free to drop the pronoun when the implied person is understood, but since the verbs aren’t marked for person, there is a tendency to keep them. It may also be the case that some pronouns get dropped more than others, since I and you are the most obvious implied pronouns when two people are speaking to each other. But consider also how normal it is to drop the pronoun it when it is followed by a contracted is or was:
It is snowing. >> It’s snowing. >> (’s) snowing.
It is packed in here. >> It’s packed in here. >> (’s) packed in here.
Feldmann’s research doesn’t take any of this into account. After what we’ve just seen, would you really divide languages into two categories based on whether they allow pronoun drop? And would you comfortable claiming English is a language that does not allow speakers to drop pronouns? No. But Feldmann would.
But ch’all knew that
Feldmann presents the rules of Standard Written English as being primary, but linguists know that spoken language is language. There is definitely a written language bias in linguistics, because written language is easier to analyze, but no linguist would make claims about an entire language based solely on written data or the rules of its standard written variety. Perhaps more importantly, no linguistics journal would publish a paper making such claims.
And the thing is, there isn’t even one variety of Standard Written English. Think about how the writing of newspapers differs from that of fiction. I want to leave that aside for now because, again, written language is not the language that speakers speak. Instead, consider how many different varieties of spoken language there are. Not only are the “national” varieties, such as American English, Canadian English, Irish English, etc., but inside each of these catch-all varieties, there are other varieties. People in Philadelphia do not speak like people in Washington D.C. and neither of those people speak like people in St. Louis – or anywhere else for that matter. Young people do not speak like older people, black people do not speak like white people do not speak like Latinx people. And inside those demographics there are multiple varieties. Rich people do not speak like middle class people do not speak like lower class people. Straight women do not speak like straight men do not speak like gay women do not speak like gay men do not speak like trans people… You get the idea. But Feldmann does not.
Instead, Feldmann (and he’s not alone in this) treats grammatical categories as a yes-no option. We’ve seen that English does not require speakers to include the pronoun to be understood, but Feldmann presents English as if it does. What about other languages? I don’t know, and I’ll bet Feldmann doesn’t either. But he still presents them as either requiring the pronoun or not. Language doesn’t work like that.
If you want to claim that English is a non-pronoun drop language, you’d first have to show that English speakers as a whole tend to not drop pronouns, despite their language allowing them to do so. That would be impossible since we don’t have any reliable corpus of spoken English to represent the entirety of the English speaking world. So you would have to look at some smaller section of English speakers – say, middle-class adults in London. And if you’re doing a study like Feldmann’s, you’d need to make sure that they are all monolingual speakers (I’m sorry, but I can’t even get into the effects of bilingualism here).
As if this weren’t enough, Feldmann doesn’t tell us which languages he marked as pronoun drop and non-pronoun drop. He says that English is an example of a non-pronoun drop language (despite what I’ve just shown you) and that Spanish is an example of a pronoun drop language, but he doesn’t provide any evidence for this beyond some armchair linguistics. And the supplemental material in his article doesn’t tell us which languages were considered pronoun drop and which were considered non-pronoun drop. Kind of a big oversight, don’t you think?
And here’s where I have to remind you that I’ve written 2,000 words about why Feldmann’s paper is wrong and we haven’t made it past his third paragraph. Let’s skip ahead.
The Big Sauce Boss
Olli Silvennoinen points out on Twitter that pro-drop is a notion that comes with a lot of problems. The first is that a language (such as Finnish) isn’t different because it doesn’t do something that English does. That implies that English is the normal language and languages which do things differently – and the speakers who speak those languages – are the “other”. But Feldmann just glosses over the idea of pronoun drop.
The second point raised by Silvennoinen is that linguistics has a history of Anglo- and Euro-centrism. This cannot be overstated. Languages like English, German and French have been seen as the “normal” languages and the most research has been devoted to studying these languages. I’m not the person to point out all of the problems this can cause, but I can say that fields like Semantics have suffered from theories which claimed to describe All Language, but really only worked on English.
Yeah no, he didn’t have to do that
Before I leave this topic (for now), I need to talk about the conclusion. Feldmann says:
According to our regression results, languages that allow to omit pronominal subjects have a negative effect on human capital. Specifically, speakers of such languages have a lower probability of having completed secondary or tertiary education, compared with speakers of languages that do not allow such an omission. Furthermore, countries where the dominant languages permit pronoun drop have lower secondary school enrollment rates. In both cases, the magnitude of the effect is substantial and slightly larger among females.
See what he did there? According to his research, language is the reason that some people – especially “females” – have lower education rates. Feldmann is claiming that the language you speak determines whether you are likely to have been to high school or college. And (surprise!) people who speak Spanish are predetermined to be less educated than those who speak English (the only two languages that Feldmann mentioned in his study). If that sounds bigoted, that’s because it is. But Feldmann ain’t done yet. He also says:
Finally, we find pronoun drop languages to affect schooling directly rather than through contemporary culture. In our view, these languages reflect and perpetuate cultural norms of collectivism that have been formed in the distant past.
This is some linguistic social Darwinism and it should have never been published. Feldmann’s ethnocentric conclusions completely remove all blame from politics and society. Some people are just dumber because of their skin color language. Some people are smarter because their language is better less collectivist. It’s infuriating to even write this nonsense, but Feldmann tries to pass it off like it’s science. It’s not. He started from false notions about language, which means he could really end up with any kind of conclusion. But he found his way to an ethnocentric one. What does that tell you?
If you want to read more, Language Log goes into the problems with the statistical analysis in a post by Bob Kennedy. The LinguaBishes recap the reactions to the paper by linguists and language scholars on Twitter, as well as running down the problems with the paper (in a much more succinct and hipper way than I can do).
That’s all for now. Feldmann and the journal (Kyklos) have a lot of explaining to do.