Good Grades Don’t Tell the Whole Story

Eric Mazur had good reason to think his Harvard physics students understood the material. His courses were full and his end-of-semester feedback was excellent. “For a long while,” he has said, “I thought I was doing a really, really good job.” Then he gave them the Force Concept Inventory, a test of whether students grasp the ideas underneath the formulas, and they did poorly. Some asked whether they should answer according to what he had taught or how they actually thought about the problem. “That’s when it started to dawn on me,” Mazur recalls, “that something was really amiss.”

He rebuilt his course around short explanations, conceptual questions and student discussion, a method now known as peer instruction. The remedy came from inside a school, and it involved more teaching of the right kind, not less.

None of this means grades are hollow. A grade measures something real, and the details matter.

Researchers at the University of Chicago Consortium on School Research followed 55,084 Chicago Public Schools graduates who enrolled in four-year colleges between 2006 and 2009. They found that high school GPA was about five times stronger than ACT score at predicting who would graduate. GPA predicted consistently across schools, while the ACT didn’t. The authors’ explanation is that a GPA reflects effort sustained across a semester and across many kinds of classes, which a single test can’t.

So a strong transcript means something. It means a student can show up and do well at what a school asks over a long stretch. If your teenager has that, you should be pleased, and nothing below is a reason to be less so.

The employer who stopped asking

One counterexample is hard to dismiss. Laszlo Bock, then Google’s head of people operations, told the New York Times in 2013 that Google had stopped asking most applicants for transcripts and test scores, except people a few years out of school, because the company had found they predicted nothing useful. After two or three years on the job, he said, how someone had performed in school had no relation to how they performed at Google.

One company’s internal analysis doesn’t overturn a study of 55,000 students, and there’s a statistical reason to read Bock carefully. Google was measuring only the people it had hired, a group already screened in part on academic records. Within a group filtered on grades, grades will predict less, the way height says far less about scoring among NBA players than it does among the general population. The broader hiring market has moved in Bock’s direction anyway. In the National Association of Colleges and Employers’ annual survey, the share of employers who said they screen applicants by GPA fell from about three in four in 2019 to under half in recent years. Grades are a reliable measure of how a person handles school, and the measure says less as the work moves further from school. That’s a point about the limits of any measure, and it says nothing against the teachers who produced the grades.

Where the hint lives

A laboratory result gives the same lesson a sharper edge. In a classic set of experiments, Mary Gick and Keith Holyoak gave people a puzzle: how to destroy a tumor with radiation without destroying the healthy tissue around it. Few solved it on their own. Many who had just read a story from a completely different domain, about a general who splits his army so the columns converge on a fortress, didn’t use it either, even though the story contains the solution. When they were told the story might be useful, most did. The knowledge was already in the room. What was hard was noticing that it applied.

Now consider an ordinary school assignment. It arrives inside a unit with a title. The question is posed, the method was demonstrated on Tuesday, the rubric names what counts, and the materials are provided. That structure isn’t a flaw. A teacher has to isolate something to teach it, and isolation is how most of us learned anything. But it means the assignment supplies the hint. A grade then records how well a student did with the hint, which is a real accomplishment. What it leaves untested is a second skill: recognizing which of the things you know applies when nobody has told you which chapter you’re in.

The missing middle is teaching

The tempting response is “let them build things.” That misses what Mazur’s students and Gick and Holyoak’s participants needed, which was better-arranged explanation, practice and feedback. Capability comes from knowing the concepts well enough to see them in unfamiliar clothes, and that takes explanation, worked examples, practice with corrections, and then attempts with the guidance gradually withdrawn. Letting someone loose without any of it mostly teaches them what frustration feels like.

Imagine a sixteen-year-old named Jonas who wants to understand why bridges are built from triangles. First he’s taught: why a rectangle of sticks racks sideways and a triangle can’t change shape without changing the length of a side, how forces split into tension and compression, and how to predict which members of a simple truss are pulled and which are pushed. He works several examples with an adult checking his reasoning. Then he builds a small truss from craft sticks and loads it until it breaks. It fails, but not where his calculations said. The members hold. A glued joint lets go. His model assumed ideal joints, and his bridge has real ones. He has to learn something that wasn’t in the lesson, redesign the joints with overlapping gussets, and test again. Then he writes down what he thinks happened and why the second version held more.

That sequence has a teaching phase, an attempt, a response from reality, a revision and an explanation. A strong grade on a unit test about trusses could coexist with none of it.

Does it travel?

A good result on the bridge doesn’t answer the harder question. Could he use the idea somewhere that looks different, without being told the analogy? Suppose someone then asks him why a three-legged stool never rocks on an uneven floor while a four-legged chair often does. “Triangles are rigid” is the tempting answer, and it doesn’t explain the stool. The stool is steady because any three points define a plane, so all three feet can always touch the floor; with four, one can be left hanging. Both problems involve triangles, but the rigidity he learned does no work here. Noticing that’s the transfer skill: knowing when an idea applies, and when a familiar-looking case calls for a different one. One good bridge shows a start and proves little about broad competence.

Where schools already do this well

Many teachers already teach this way, and many do it well. Good science courses run labs, good English teachers assign open-ended writing, and a teenager on a robotics team or in a serious ensemble gets plenty of unscripted problems. Nothing here says schools never ask students to apply what they know, or that grades are meaningless. The narrower question is about one young person. In the areas that matter to them, have they had enough chances to use what they know when the setup stops looking like an assignment? A kid who spends hours in a robotics shop may not need more of this. A kid whose week is entirely assigned work might.

Finding out at home

“What grade did you get?” is a fine question. It just leaves most of the picture untouched. Ask your teenager to explain something from this week to someone who has never heard of it, and listen to the reasoning more than the answer. What happened when they tried to use the idea? Where might it work somewhere else, and where would it break? None of this needs to feel like a quiz, and the reasoning will tell you more than a score does. A teenager who can say where an idea stops working probably understands it.

At Yonder we try to build projects the way the truss example runs, with teaching before the attempt and a chance to use the idea somewhere new afterward.

The grade remains good evidence of how a student performs when the setup is provided. A family can start finding out tonight what happens when it isn’t.