From two roads to infinitely many
Back on the real line, when you said "the limit of f(x) as x approaches a", you secretly checked just two things: the value f wants from the left and the value it wants from the right. If those two agreed, the limit existed. That is the whole story for one dimension. The moment we step into the complex plane, that comfortable two-sided picture collapses, because a point z_0 can now be approached along a straight line from the east, a spiral curving in from the south, or a zigzag that never settles on a direction at all.
So the definition of a complex limit must be tougher than its real cousin. We say f(z) tends to the value L as z approaches z_0 if we can force f(z) to lie inside any tiny disk we name around L, simply by keeping z inside a small enough disk around z_0. Notice the geometry: "close" is measured by the modulus, so |z - z_0| small and |f(z) - L| small are statements about distances between points in two planes. No left and right anywhere — just nearness in every direction at once.
Every road must lead to the same place
The most useful way to picture the definition is this: the limit L exists only if, no matter which path you take toward z_0, f drags you to the same destination. This is the direction-independence of the limit, and it is exactly where the plane gets its teeth. If approaching along the real axis gives one answer and approaching along the imaginary axis gives another, the limit simply does not exist — there is no single L that all roads agree on.
The classic cautionary tale is the conjugation map f(z) = z-bar, examined near z_0 = 0 through the quotient z-bar / z. Slide in along the real axis, where z = x is real, and z-bar = x, so the quotient is x / x = 1. Now slide in along the imaginary axis, where z = i y, so z-bar = -i y and the quotient is (-i y)/(i y) = -1. Two roads, two different answers — so this quotient has no limit at 0. That single example is the seed of why z-bar fails to be differentiable in the complex sense, a theme the next rung makes precise.
approach along real axis (z = x): z-bar / z = x / x = +1 approach along imag axis (z = i y): z-bar / z = (-i y)/(i y) = -1 1 != -1 => limit at 0 does NOT exist
Splitting the limit into real and imaginary halves
There is a friendly bridge back to the calculus you already know. Writing a function as f = u + i v splits it into two real-valued functions of two real variables, u(x, y) and v(x, y). It turns out that f(z) tends to L = A + i B exactly when u tends to A and v tends to B as the point (x, y) approaches (x_0, y_0). One complex limit is precisely two ordinary two-variable limits stacked together.
This is genuinely reassuring, but read the small print. A two-variable real limit is itself the hard kind that already demands agreement along every path through (x_0, y_0) — the same directional fussiness, just dressed in real coordinates. So splitting into u and v does not make the difficulty disappear; it relocates it into territory your multivariable calculus has already prepared you for. The complex picture and the (x, y) picture are two views of one and the same demand.
A second handle on the same idea works through sequences in the plane: f has limit L at z_0 if and only if f(z_n) tends to L for every sequence z_n approaching z_0 (with z_n never equal to z_0). The sequence view is often the cleanest way to disprove a limit — find two sequences crawling toward z_0 whose images drift to different points, and you are done.
Continuity: when the limit just equals the value
Once limits are in hand, continuity is the easy, expected next step. A function f is continuous at z_0 when the limit of f(z) as z approaches z_0 exists and equals the value f(z_0) — no surprise jumps, no holes, no path-dependence. Geometrically: nudge the input point a little and the output point in the target plane moves only a little. The mapping pictures you built in the earlier guides — translations, rotations, the squaring map, inversion — are all continuous wherever they are defined, which is why their images of regions came out as smooth, connected shapes.
And the u-plus-iv bridge carries over cleanly: f is continuous at z_0 if and only if both real fields are. This is the continuity of u and v, and it lets you reuse every continuity fact from multivariable calculus — sums, products, quotients (where the denominator is nonzero), and compositions of continuous maps are continuous. So z^2, being polynomial in x and y after expansion, is continuous everywhere, while 1/z is continuous everywhere except the origin where it is not even defined.
Why this single idea casts such a long shadow
Here is the quiet payoff that makes this guide worth lingering on. The complex derivative is defined as the limit of the difference quotient (f(z) - f(z_0)) / (z - z_0) as z approaches z_0 — and that quotient is exactly the z-bar/z kind of object we just dissected. For the derivative to exist, this ratio must approach ONE number no matter which direction z comes in from. The whole drama of complex analysis is hiding in that one word: direction.
- Pick the direction along the real axis: the difference quotient must approach the derivative's value.
- Now pick the direction along the imaginary axis: the SAME quotient must approach the SAME value.
- Force those two results to be equal, and out fall constraints linking the partial derivatives of u and v — the seeds of the Cauchy-Riemann equations.
This is why direction-independence is not a dry technicality but the engine of everything ahead. Demanding agreement along merely two directions already pins down a pair of equations; demanding it along ALL directions is so restrictive that a single complex derivative forces a function to be infinitely differentiable and to equal its own power series — a rigidity with absolutely no analogue in real calculus, where a function can be differentiable once and never again. Do not let anyone tell you the complex derivative is "just the same limit" as the real one. It is the same limit asked of a far harder, fully two-dimensional question, and that is the whole point.