The 100 point scale and the letter grade systems have been proxies for student performance since the late 19th century. As students began transferring between institutions, the need to standardize the grading system grew. However, this proved difficult due to a lack of interrater reliability: several studies found that the 100 point scale was highly unreliable and suggested the use of fewer gradations (e.g., the five-letter scale).

In the 1900s, as intelligence tests (e.g., IQ) became popular, some assumed that student performance in the classroom should follow a normal (or bell) curve. As a result, distributing grades according to a normal curve became popular as a “solution” to the subjective nature of grading since ranking students based on performance could be done reliably. However, the rationale for this approach was based on questionable assumptions, which was clear from the inception of curved grading: "Immediately, even some proponents of curved grading recognized problems with comparing levels of aptitude in the population with levels of classroom achievement among a population of students. For a variety of reasons, a given classroom might not include a representative sample from the general population...." (Schinske & Tanner, 2014, p. 162). In addition, many education scholars objected to this approach, referred to as norm-referenced grading, because it dissociates grades from content knowledge and learning. For example, Benjamin Bloom  pointed out that in grading on a curve “it matters not that the failures of one year performed at about the same level as the C students of another year. Nor does it matter that the A students of one school do about as well as the F students of another school" (Bloom, 1968, p. 20).

Interestingly, for almost 100 years, the University of Michigan measured achievement on a simple absolute scale of “passed,” “not passed,” or “conditioned.” “Faculty insisted that this system was the best; that their students flourished amidst this kind of mastery grading, not needing any explicit competition or ranking to motivate them” (McKay). However, institutional considerations – such as increasing enrollments and the need to identify students for admittance to Phi Beta Kappa – led to the adoption of letter grades. For example, the Colleges of Engineering and LSA adopted a letter grade system in 1907 and 1912, respectively.

References:

• McKay, T. Analytics at Michigan–A Brief History", the third video in the MichiganX/EdX course Practical Learning Analytics by Tim McKay The clip is between minute markers 1:18 and 5:30. 
• Schinske, J., & Tanner, K. (2014). Teaching more by grading less (or differently). CBE–Life Sciences Education. 13, 159-166.

Today, the meaning of grades remains unclear, and grading systems remain controversial. As Clark and Talbert point out, “...despite nearly a century of growing objections, traditional grading practices persist, more or less unchanged since 1897. They have remained, not because they have stood the test of time and are the logical endpoints of centuries of mindful refinement, but primarily because they are useful— efficient, scalable, and interoperable—particularly so for the large bureau­cracies that educational systems have become. But in order to attain that usefulness, grades have lost what pedagogical meaning they once had” (2023, p. 16).

Grades affect students in many different ways, from eligibility for scholarship and financial aid awards, to entrance to graduate or professional schools and gaining employment. Some researchers argue that grades can be psychologically harmful to students and discourage them to persist to a degree. So it is worth re-examining our systems and asking:

• How might grades (or what systems might) better represent the knowledge and skills students have learned?
• What systems might provide useful feedback for students?
• How might a grading approach motivate students to learn intrinsically (because they see the value of learning) rather than extrinsically (because they need an A)?