How do JavaScript prototypes work?

In my previous post, I showed how “inheritance” works in JavaScript, by reimplementing the forms,, and = baz in plain JavaScript functions. In this post, I show JavaScript’s conventions for creating new “classes” and new objects.

If I just get one thing across in this post, it should be this: there is not just one “prototype” concept in JavaScript! There are two completely different things called “prototype” in JavaScript. If you confuse the two, you’re in for a world of pain. Let’s see why.

A JavaScript value has two slots:

  1. An ownProperties slot. This is a map. A key is a string or symbol, and a value is a reference to another JavaScript value.
  2. A parent slot. (This is what I’m calling it, to avoid confusion!) The parent slot contains a reference to another JavaScript value. The parent slot is not a “property” in the JavaScript sense. Only the values null and undefined do not have a parent slot.

I’ve deliberately not used the term “prototype” in my definition. This is because JavaScript maddeningly sets out to confuse you, and uses the term “prototype” in different contexts to refer to one of two completely different things:

  1. An own property with the string key "prototype".
  2. The parent slot.

Here are some of the common confusing notations:

Casual commentators will say things like “x’s prototype”, but this is ambiguous: it could refer to an own property of x with the string key "prototype", or it could refer to the parent of x. I will avoid this ambiguity by saying either “the parent of x”, or “the "prototype" property of x”.

The JavaScript runtime enforces that all values are linked, by their parent slots, into a big tree. The value null is the root of the tree, with no parent. The basic point of these “parent” slots forming a big tree is to emulate the “class hierarchy” from object-oriented design, but without having a first-class notion of “classes” in the language. An object just “inherits” from another object!

To create new “classes”, JavaScript (ab)uses the function Foo() { ... } form. To create new objects, JavaScript has a new operator. Here are the fundamental JavaScript syntax forms that I’ll explain in this post.

We can re-write the new operator in plain JavaScript. In the spec, this function is called Construct. To be honest, the spec has a lot of guff, but in essence it does this:

// You should be able to replace `new Foo(x,y)` with `Construct(Foo, x, y)`
function Construct(constructor, argsList) {
  const obj = Object.create(constructor.prototype);
  const ret =, ...argsList);
  return ret instanceof Object ? ret : obj;

Here we see that new internally uses Object.create, which is really the only way that JavaScript objects get created. The new object’s parent is the constructor’s .prototype property. JavaScript’s decision to look for the parent under a .prototype property has caused much needless confusion. It should instead be called something like .methods.

When I first wrote the above Construct function, I discarded the return value of You don’t usually see a constructor that returns a value, and I assumed that it would just be ignored. I was wrong!! A constructor can return a value, and if it returns an object, this will be used in place of the initial obj (which will probably be quickly garbage-collected). This means you can write things like:

const loggers = new Map();
function Logger(filename) {
  if (loggers.has(filename)) {
    return loggers.get(filename);
  else {
    loggers.set(filename, this);
    this.filename = filename;

const logger1 = new Logger("/var/log/myapp");
const logger2 = new Logger("/var/log/myapp");

// logger1 === logger2! They are the same object!

Now let’s look at how to define a “class” in JavaScript. I put “class” in quotes, because really a “class” is just anything that would be accepted as the constructor argument of Construct, above.

When we define a function in JavaScript with the function(){} notation, we should imagine that JavaScript inserts a bunch of instructions after it, like this:

function Dog(a,b,c) { this.a = a; console.log(b, c); }


const methods = Object.create(Object.prototype);  // New, empty methods for the user to fill out
Object.defineProperty(Dog, 'prototype',
  { value: methods, writable: true });  // Store methods under Dog.prototype, for `new` to find
Object.setPrototypeOf(Dog, Function.prototype);   // Confusing, but provides things like Dog.apply(...)
Object.defineProperty(methods, 'constructor', { value: Dog });  // Unimportant; lets us do `new dog.constructor(4,5,6)`

Note that every time you write function(){}, it creates a constructor, with a new methods object, and an implicit this parameter. This is very often not your intention. To avoid this, you should use the arrow function syntax, which is conceptually simpler. An arrow function does not have a .prototype, and it does not have an implicit this parameter.

Note the methods object is set to inherit from the standard Object methods, stored in in Object.prototype (which should be called Object.methods). Effectively, it defaults to Dog extends Object. If you want a different class hierarchy, you need to manually fix what JavaScript created for you! For example, if you want the hierarchy Dog -> Animal -> Object, you can patch it up with Object.setPrototypeOf:

function Animal() {} = function() { console.log(`Running with ${this.numLegs()} legs`); };

function Dog() {}
Dog.prototype.numLegs = function() { return 4; };

Object.setPrototypeOf(Dog.prototype, Animal.prototype);

const lassie = new Dog();;  // Running with 4 legs

You’ll see other approaches to patching up a class hierarchy. Some others write things like Dog.prototype = new Animal(), but this is just gross. Others, who are less bad, might write Dog.prototype = { ... }, overwriting the prototype property entirely, but this is also ugly, because it overwrites the constructor property too. But having said that, is the constructor property useful? It lets us write things like:

let foo1 = new Dog(1,2,3);
let foo2 = new foo1.constructor(4,5,6);

IMO, the .constructor feature is not useful. I never see this in real-world JavaScript. It’s probably a mis-feature.

Note that when we assign new methods, like = ..., we have to do it with the function() {} form so that the function gets an implicit this parameter. Unfortunately, this also makes all of our methods constructors too! It’s really a JavaScript design mistake that we can’t bind this without also creating a new, unnecessary prototype.

You might wonder how to call super(...) in this world of prototypes. There is no elegant answer. You basically have to know which function is the super function, and explicitly .call() it. One way is like this:

function Animal(prefix = "> ") {
  this.prefix = prefix;
Animal.prototype.speak = function(words) {
  console.log(this.prefix + words);

function Dog() {
  // calling super("DOG: ") from constructor, "DOG: ");

Dog.prototype.speak = function(words) {
  // calling super('WOOF' + words) from method, 'WOOF ' + words);

Object.setPrototypeOf(Dog.prototype, Animal.prototype);

const lassie = new Dog();


Modern JavaScript has a class notation, which basically de-sugars into this prototype-based stuff. I’ll do a future post on how exactly it works.

Tagged #programming, #javascript.
👋 I'm Jim, a full-stack product engineer. Want to build an amazing product and a profitable business? Read more about me or Get in touch!

More by Jim

This page copyright James Fisher 2020. Content is not associated with my employer. Found an error? Edit this page.