To solve a system of inequalities graphically, plot each inequality's boundary line on the same coordinate plane, then shade the half-plane that satisfies it; the solution is the region where all shaded areas overlap. Dashed boundaries exclude points on the line, solid boundaries include them, and any point inside the final overlap satisfies every inequality at once.
This method works because each inequality divides the plane into two infinite half-planes, and a system's solution is their intersection. The approach is fast for two variables, transparent for checking work, and immune to the algebra errors that plague substitution on nonlinear systems. What follows covers the mechanism, its quantified limits, the edge cases that break naive graphs, common misconceptions, a practical procedure, and where the method stops being useful.
How boundary lines and half-planes encode inequality solutions
Every two-variable inequality can be rearranged into a form where one side is a boundary curve. For a linear inequality like 3x + 2y < 12, the boundary is the line 3x + 2y = 12. The inequality symbol tells you which side of that line contains the solutions: the strict symbol (<) excludes the boundary, while ≤ includes it. Why the split works is a monotonicity argument. Moving perpendicular to the line changes the value of 3x + 2y in one direction only, so once you find one side where the expression stays below 12, every point there qualifies.
The test-point shortcut exploits this. Pick any point not on the boundary, usually (0, 0), and substitute it. For 3x + 2y < 12, the origin gives 0 < 12, which is true, so you shade the side containing the origin. If the test fails, shade the opposite side. This is valid because the boundary line is the only place where the expression equals the threshold; on either side the sign is constant.
For the system, you repeat the process for each inequality on the same axes. The solution set is the intersection of all half-planes, which is always a convex region for linear systems. That convexity is the geometric reason a linear program's optimum sits at a corner point. One plain-language version: each inequality votes for a side of its line, and only points that win every vote survive.
Why the overlap is often a polygon and how to quantify its extent
Two linear inequalities typically produce an unbounded wedge. Three or more, if their boundary lines form a closed arrangement, produce a bounded polygon whose vertices are the pairwise intersections of boundary lines. The vertex count generally tracks the number of constraints: a three-inequality system in standard position can yield a triangle, while four constraints arranged around a feasible region can yield a quadrilateral. The vertices matter because they are testable candidates for optimization.
Sizing the region is straightforward once vertices are known. Take the system x ≥ 0, y ≥ 0, x + y ≤ 4, 2x + y ≤ 6. Its feasible region is a quadrilateral with corners at (0, 0), (0, 4), (2, 2), and (3, 0). The vertices come from intersecting boundary pairs: x + y = 4 and 2x + y = 6 meet at (2, 2). The exact overlap area is 7 square units, computed by splitting the quadrilateral or applying the shoelace formula to the ordered vertices.
That area figure is also a useful self-check. If your shading claims an area larger than the polygon your vertices define, you shaded a side incorrectly. A ten-second verification: substitute each vertex into every inequality and confirm all hold. If one vertex fails, an adjacent boundary is misclassified.
Edge cases that break the standard procedure
Four situations reliably trip people up, and each has a clean fix.
Dashed versus solid is a real distinction, not decoration. A point on a dashed boundary fails a strict inequality even though it satisfies the equation. In a system that mixes < and ≤, the shared boundary is solid only where every inequality allows equality. Parallel boundaries can shrink the solution to nothing. The system y > x + 1 and y < x − 2 has no solution because the half-planes point away from each other. Checking whether boundary lines have identical slopes tells you this before you shade anything. Isolating y flips the inequality when you divide or multiply by a negative number. Dividing −2y < 6 by −2 gives y > −3, not y < −3. This single sign error is responsible for a large share of wrong shaded regions. The origin lies on a boundary. If your test point (0, 0) makes the boundary expression zero, the test is inconclusive; choose (1, 1) or (0, 1) instead and retest.
A counterintuitive detail worth internalizing: shading every half-plane is not the goal. The mathematical object is only the overlap. Ink spent outside the intersection is wasted and, in exams, actively penalized.
Misconceptions worth correcting before practice
The most common false belief is that graphing is a lesser method, suitable only for drawing pictures. In fact, the graphical solution is exact. Every vertex coordinate is a precise algebraic intersection, and the region is an exact set, not an approximation.
A second misconception: the solution must be a single connected region. For linear systems it always is, because half-planes are convex and intersections of convex sets are convex. For nonlinear inequalities, such as y > x² and y < 4, the overlap can be a connected lens shape, but systems like y > x² and y > (x − 5)² plus a height cap can produce disconnected components. Do not assume connectivity when the boundaries curve.
A third error is treating the dashed line as automatically excluded everywhere. If two inequalities share the same boundary line but one is strict and one is inclusive, equality points remain excluded from the strict one regardless of the other. The rule is the intersection of the inclusion conditions, applied pointwise.
A practical procedure with thresholds and checks
Follow these steps in order. They convert a system into a verified shaded region in about two minutes for linear cases.
Rewrite every inequality in slope-intercept or standard form so its boundary is explicit. Record the strict or inclusive symbol next to each. Compute each boundary line's intercepts by setting x = 0 then y = 0. Two points per line are enough to draw it accurately. Draw each boundary: solid for ≤ or ≥, dashed for < or >. Test one point per inequality, never reusing the origin if it sits on a boundary. Shade only the qualifying side, lightly for each line. Darken only the intersection of all shaded regions. Mark each vertex where boundary lines cross. Verify each vertex against every inequality before trusting the region or an optimization result.
A useful accuracy threshold: if your shaded region has more than six vertices for a system of three or fewer inequalities, you have almost certainly misclassified a side. Recheck the sign flips and the dashed-solid assignments, since those two errors account for most bad regions.
Where graphical solutions stop being reliable
The method degrades when the number of variables exceeds two, because the feasible region becomes a polyhedron in higher-dimensional space that cannot be drawn precisely on paper. For three variables you can sketch a plane arrangement, but overlap ambiguity rises quickly. For four or more variables, use the simplex method or a numerical solver instead; the graphical insight carries over but the drawing does not.
Nonlinear systems complicate matters because boundary intersections can outnumber the visible vertices. Two circles cross at most twice, but intersecting a parabola with a line can create two vertices while the feasible region remains unbounded in a direction your sketch compresses. Resolution matters as well: without careful plotting, a small feasible triangle two units wide on a scale from −10 to 10 is easy to miss entirely.
A professional caveat worth stating plainly: the graphical solution is exact as a set, but the drawing is only as precise as your construction. Trust algebra to confirm vertices and use the graph to build intuition and catch sign errors. The two methods verify each other, and in practice you want both when stakes are more than casual.
There is no high-stakes professional domain implied by this topic beyond classroom mathematics and routine modeling, so no clinical, legal, or financial disclaimer applies. The method is stable and its limits are the ones stated: two variables, exact boundary classification, and verification at every vertex.