Hacker News Reader

Comment by jp57 | Hacker News Reader

After I started my PhD program in CS/AI (late 90's), one of the faculty told me that the GRE verbal score was thing on the application most predictive of student success in the program. Of course, most applicants had perfect or near perfect GRE math scores, so there was no predictive power there. But the application also included the GRE subject test, transcripts, and letters of recommendation.

View on HN

I also seem to recall from The Making of the Atomic Bomb that the theoretical physicists of the era were noted as having the highest verbal intelligence of any group of scientists. That struck me since I would not have expected it.

EDIT: I found the quote, from chapter 6:

Psychometricians have closely questioned American scientists of this first modern generation, curious to know what kind of men they were—there were few women among them—and from what backgrounds they emerged. Small liberal arts colleges in the Middle West and on the Pacific coast, one study found, were most productive of scientists then (by contrast, New England in the same period excelled at the manufacture of lawyers).

...

Theoretical physicists averaged the highest verbal IQs among all scientists studied, clustering around 170, almost 20 percent higher than the experimentalists. Theoreticians also averaged the highest spatial IQs, experimentalists ranking second.

I'd think strong verbal skills would make for better collaboration between scientists.

In order to solve what are essentially "word problems" -- i.e. reading a verbal description of a problem and translating it to math -- you have to have very good reading comprehension.

Perhaps survivorship bias explain part of it. Many of those scientists arrived at US fleeing their home countries in WWII. Migrating under those circumstances is tricky, lacking the necessary support network or failing to successfully negotiate an encounter with border patrol may be the difference between reaching a safe place and being dragged to rot in the trenches. Good verbal intelligence is very helpful in this process. Perhaps the same way good verbal intelligence is over-represented in those who escaped the war, the trait of being taciturn was over-represented in all the scientists that got stuck and died in WWII[1].

[1] https://worldscientific.com/worldscibooks/10.1142/q0436#t=ab...

The GRE math bar for excellence was quite low and so a perfect score was only the 92nd percentile or so. It’s not surprising that other metrics correlated better with success

Putnam exam score might be more predictive?

Sure, but also, if you can't read and write well, you're going to struggle in a PhD, because that's really a lot of what you have to do.

Makes sense that verbal ability would line up more with success in CS, especially when math scores are already high across the board. A lot of programming leans on language-type skills: reading and understanding code, navigating docs, naming things clearly, writing maintainable logic etc.

The field probably does itself a disservice by overemphasising math. That framing can push people away who might actually do really well, especially those strong in reasoning, abstraction, or communication. Linked study is a good reminder to rethink how we present programming imo.

It does. The majority of what's popular and gets attention in programming is things like languages and frameworks which do the best smooth talking. The whole point of CS is to focus on the aspects of programming that aren't just vogue trends. With any field or aspect of modern life, language skills are the gateway to privilege and software engineering is no exception. But here it's respect for the math that stops it all from spiraling into a bunch of bs. You can't fool the math people.

Reply on news.ycombinator.com