To find a limit from a graph, trace the curve toward the x-value from the left and from the right, then check whether both traces converge on the same y-value. The limit is that shared y-value — even when the curve has a hole there, and even when the function's actual value at that x is something else entirely. The condition that trips people up: the limit exists only if the two one-sided traces agree, so a jump, a vertical asymptote, or a sudden break kills it. Everything below builds on that single rule.
What the y-value a graph approaches actually tells you
A limit answers a narrow question: as x gets arbitrarily close to some number a, what number does f(x) get arbitrarily close to? Note what the definition leaves out — f(a) itself. The function can be undefined at a, defined differently at a, or defined perfectly well, and none of it changes the limit. That separation between "where the curve is heading" and "where the dot sits" is the whole game when you read a graph.
Mathematicians write this as the limit of f(x) as x approaches a equals L. The formal version, from calculus textbooks like Stewart's Calculus, uses epsilon and delta: for every tolerance around L, there's a window around a where the function stays inside that tolerance. Visually, you don't need the formalism. You need to watch the pencil line, not the printed point.
Here's the boundary condition in plain terms. If the left side of the graph heads toward y = 3 and the right side heads toward y = 3, the limit is 3. If the left heads to 3 and the right heads to 5, the limit does not exist — mathematicians abbreviate that as DNE. There is no third option where you average them or pick the friendlier one.
One more piece of vocabulary matters for graph reading. A removable discontinuity is a hole — the curve approaches one value from both sides but has a gap there. A jump discontinuity is a vertical break where the two sides land on different y-values. A vertical asymptote is where the curve rockets toward infinity. Each looks different on paper, and each produces a different limit verdict.
Why the traced curve beats the printed dot
Students lose more points to the printed dot than to any other feature of limit graphs. The dot is the function value, f(a). The curve's trajectory is the limit. Instructors draw them apart on purpose, and the gap between them is usually the exam question.
Suppose a graph shows a smooth parabola approaching the point (2, 4), but a solid dot sits at (2, 1). The limit as x approaches 2 is 4. The function value is 1. Both facts are true simultaneously, and neither overrides the other. According to the standard definition used across AP Calculus and university courses, the limit concerns only nearby behavior — the value at the point is irrelevant to it.
The failure case looks like this: a student sees the dot at (2, 1), writes "limit = 1," and moves on. That answer confuses f(2) with the limit. The corrective habit is mechanical — cover the vertical line x = 2 with your finger, then ask what height the curve is squeezing toward on either side of your finger.
Real exam graphs lean on this trap constantly. The College Board's AP Calculus AB framework lists "estimating limits from graphs" as a foundational skill, and released free-response questions routinely show a graph with a displaced point. If the curve and dot disagree, the curve wins for limit purposes every time.
Estimating a two-sided limit in five moves
Reading a limit off a graph takes about thirty seconds once the sequence is automatic. Work in this order and you won't skip the right-hand check that catches most errors.
1: Locate x = a on the horizontal axis and draw an imaginary vertical line there. 2: Trace the curve from the left, moving x toward a, and note the y-value the curve is closing in on. That's the left-hand limit. 3: Trace from the right, same motion, and note the y-value. That's the right-hand limit. 4: Compare the two numbers. If they match, that shared value is the limit. If they differ, the limit does not exist. 5: Check for a hole or displaced dot at x = a and record it separately as the function value — it does not change your limit answer.
If step 2 or step 3 sends the curve off the top or bottom of the grid along a vertical line, the one-sided limit is infinite, and you write that side as approaching positive or negative infinity rather than a finite number. Infinite one-sided behavior means the two-sided limit fails to exist as a real number.
Failure handling matters here. If your left and right traces disagree by a small amount, don't split the difference — the graph is telling you the limit is DNE, and estimating "about 3.5" between 3 and 4 is simply wrong. If the curve has a vertical asymptote, resist writing "infinity" as a limit; write DNE and note the asymptotic behavior separately.
Reading one-sided limits when the two sides disagree
One-sided limits are the raw material of every two-sided answer, so read them with care. The notation lim x→a⁻ means approach from values less than a (the left), and lim x→a⁺ means approach from values greater than a (the right). On a graph, left means literally leftward along the x-axis.
Piecewise functions make this concrete. Consider f(x) = x + 1 for x < 3 and f(x) = 7 − x for x ≥ 3. Approaching 3 from the left, the line climbs toward 4. Approaching from the right, the other line falls toward 4 as well. Both one-sided limits equal 4, so the two-sided limit is 4 even though the graph has a corner there. Corners are fine; breaks are not.
Now shift the second piece to f(x) = 10 − x for x ≥ 3. From the left the curve still heads to 4; from the right it heads to 7. The one-sided limits are 4 and 7, they disagree, and the two-sided limit is DNE. The graph shows a visible jump of 3 units at x = 3, and that vertical gap is the visual signature of a failed limit.
A useful sanity check: the size of the jump equals the difference between the one-sided limits. A jump of 3 units means the two sides disagree by exactly 3, which is why no single L can satisfy both sides. Discrete jumps like this appear in real modeling — tax brackets, shipping tiers, and postage rates all jump at thresholds — which is why piecewise graphs are common in applied problems.
Holes, jumps, and asymptotes: a visual comparison
The three discontinuity types each produce a distinct limit outcome, and a side-by-side view makes the distinctions stick. The table below assumes a is the x-value under investigation.
| Graph feature at x = a | Left-hand limit | Right-hand limit | Two-sided limit | Function value f(a) |
|---|---|---|---|---|
| Hole (removable) | L | L | L (exists) | Undefined |
| Hole plus displaced dot | L | L | L (exists) | Some other value |
| Jump | L₁ | L₂, with L₁ ≠ L₂ | DNE | Defined at one side's value |
| Vertical asymptote | ±∞ | ±∞ or ∓∞ | DNE | Undefined |
| Smooth curve, no break | L | L | L (exists) | Equals L |
Read the table down the "Two-sided limit" column and the pattern is stark: only the hole rows and the smooth row produce a limit. The jump and asymptote rows fail, and they fail for different reasons — disagreement versus unboundedness.
The hole case deserves emphasis because it feels paradoxical. A hole means the function has no value there, yet the limit exists. That's not a contradiction; it's the definition working as designed. The curve approaches L from both sides whether or not anyone bothered to fill in the point.
An asymptote behaves oppositely in spirit: the curve doesn't approach any finite height, so no L exists. If the left side dives to negative infinity and the right climbs to positive infinity, the disagreement is total. Write DNE for the limit and describe the asymptote separately.
What DNE looks like on paper and why it isn't a failure
"DNE" reads like a cop-out to students who expect every problem to have a number for an answer. It isn't. Nonexistence is a substantive mathematical result, and graphs make it visible in at least four ways: a jump, a vertical asymptote, wild oscillation, and a mismatch between the one-sided limits at an endpoint.
Oscillation is the strangest case. The classic example is f(x) = sin(1/x) near x = 0 — the curve wiggles faster and faster between −1 and 1 as x shrinks, never settling on a height. Plot it on a graphing calculator and the ink fills a vertical band near the origin. No single L works, so the limit is DNE. Most textbook graph problems avoid this case because it's hard to draw, but recognizing the pattern matters.
At the endpoints of a domain, only one side exists, so the two-sided limit technically fails while the one-sided limit is perfectly well defined. A graph of a function defined on a closed interval [0, 5], for instance, has a right-hand limit at x = 0 but no left-hand limit, because there's nothing to the left. Instructors often accept the one-sided limit as the answer there — check your course's convention.
The habit worth building: when you conclude DNE, say why in one clause. "DNE — jump at x = 3" or "DNE — vertical asymptote at x = −1" earns full credit on most rubrics, while a bare DNE can lose a point for insufficient justification.
Limits at infinity and end behavior on the same graph
Horizontal behavior uses the same left-right logic, just aimed at the far edges of the grid. The limit as x approaches infinity asks what height the curve settles toward as you trace rightward indefinitely. The limit as x approaches negative infinity asks the same going leftward.
A rational function like f(x) = (3x² + 1)/(x² − 4) flattens toward y = 3 on both far edges — the ratio of leading coefficients, 3/1. The graph shows a horizontal asymptote at y = 3, and both end-behavior limits equal 3. That's a case where limits at infinity exist and are easy to read: find the flat height the curve hugs at the left and right margins.
Polynomials behave differently. A cubic like f(x) = x³ climbs without bound to the right and plunges without bound to the left. Both end-behavior limits are infinite — positive on the right, negative on the left — and the graph has no horizontal asymptote. Reading this from a graph means watching whether the curve exits the top or bottom of the viewing window; a curve that vanishes off the top is heading to positive infinity within the window shown.
The distinction that matters for the table above: a horizontal asymptote is a limit that exists, while a vertical asymptote is a limit that fails. Same word, opposite verdicts. Confusing the two is a common error, and the axis of the asymptote — horizontal versus vertical — is the tell.
Worked example: a graph with every feature at once
Put the pieces together on a single imagined graph, the kind you'd see on a midterm. Suppose the curve has a hole at x = 1 with the approach height y = 2, a jump at x = 3 where the left side lands at 1 and the right at 4, a vertical asymptote at x = 5, and a horizontal asymptote at y = 0 as x runs to both infinities.
Reading each feature in order: at x = 1, both sides approach 2, so the limit is 2, and the function value there is undefined (a bare hole). At x = 3, the one-sided limits are 1 and 4, so the two-sided limit is DNE — a jump of 3 units. At x = 5, the curve diverges, so the limit is DNE with a vertical asymptote. At both infinities, the limit is 0.
Now compute something the graph lets you do that a formula sometimes hides. The jump at x = 3 spans 3 units of height, from 1 to 4. If the vertical axis is scaled at 2 units per grid square, that jump covers one and a half squares — a quick visual check that your reading is calibrated to the grid. Misjudging the scale is the single most common estimation error on graph problems, per the guidance in standard precalculus lab manuals.
For a second computation, suppose the vertical asymptote at x = 5 shows the curve reaching y = 100 at x = 4.9 and y = −100 at x = 5.1. The function has crossed from large positive to large negative with no intermediate value near zero, which is exactly what "no finite limit" means — the curve has no height to settle on.
Three mistakes that sink graph-reading scores
The first mistake is answering with the function value instead of the limit. Cover the dot; trust the curve. Systematic practice — twenty graphs with deliberately displaced points — trains the eye within an afternoon.
The second is forgetting the right-hand check. Students trace from the left, see a clean approach, and stop. The right side is where jumps hide. Always run both traces, in that fixed order, before writing anything.
The third is writing DNE without justification. On the AP Calculus exam, free-response rubrics award points for stating the reason — a jump, an asymptote, a mismatch. "The limit does not exist because the left-hand limit is 1 and the right-hand limit is 4" earns the point; a bare DNE may not.
A fourth, subtler error involves reading the graph off its labels rather than its shape. A curve drawn to pass exactly through a grid intersection invites you to assume the limit is that integer, when the true approach height might be 2.4 and the drawing slightly off. Use the labeled points and the axis scale, and when the graph offers no labels, state your estimate with the scale you used.
FAQ
Does a limit exist at a hole in the graph?
Yes — a hole does not prevent the limit from existing, because the limit only concerns nearby values. If both sides of the curve approach the same height L at the hole, the limit is L, and only the function value f(a) is undefined. This is the most common misconception-correction on the topic. If the two sides approach different heights, the limit fails regardless of the hole.
What if the left and right sides disagree by just a tiny amount?
Even a 0.1-unit disagreement means the limit does not exist. The definition demands one single L that both sides converge on, and proximity between two distinct values doesn't create that. On a hand-drawn graph, a small visible gap signals a jump you should report as DNE. Precision of the drawing doesn't loosen the rule.
Can I estimate a limit to a decimal from a graph without gridlines?
Only if the graph supplies labeled points or a clear scale. Without either, any decimal you write is a guess dressed as a measurement. State the estimate as approximate and name the reference — "about 2, judging by the labeled point (1, 2)" — rather than inventing precision. Confirm against the function's formula when one is given.
Why does my calculator graph show a vertical line at a hole?
Many graphing calculators connect plotted points with line segments, and a hole can render as a solid vertical streak or a spurious line. The Texas Instruments TI-84, for instance, draws by sampling and connecting. Use the table feature or zoom in; the hole reappears as a gap at sufficient magnification, and the limit is unaffected by the rendering artifact.
Is the limit at an endpoint different from the interior limit?
At an endpoint, only a one-sided limit exists, so the strict two-sided limit is DNE. Most courses accept the one-sided value as the answer at endpoints, and you should check your instructor's convention. For interior points, both sides exist and must agree. This distinction shows up most often on domain-restricted modeling problems.
How many points do I lose for confusing a limit with a function value?
On typical AP-style rubrics, a limit-versus-value confusion costs the answer point but not the reasoning points if your work shows correct tracing. The fix is procedural: always write the limit and the function value as two separate lines. Naming both explicitly prevents the substitution error from surviving to the final answer.
Does a horizontal asymptote count as a limit that exists?
Yes — the end-behavior limit toward a horizontal asymptote exists and equals the asymptote's height. Keep the vocabulary straight: horizontal asymptote means the limit exists at infinity; vertical asymptote means the limit at that x-value fails. Same word, opposite limit verdicts, and the axis orientation is the tell.
What should I write when the graph runs off the top of the grid?
Trace both sides first and describe what each does. If both sides climb without bound, the limit is infinite rather than a finite number, and many courses still write DNE for the two-sided limit while noting the behavior. If the sides diverge in opposite directions, DNE is unambiguous. Name the vertical line where the divergence occurs so the reader knows where you looked.