Graphs as pictures of relationships

In brief

A hiker's altitude over the course of a day, drawn as a line that climbs, flattens on a ridge, then drops steeply into a valley, tells the whole story of the hike without a single word of narration. A graph is not a decorative chart bolted onto data, it is a translation of a relationship between two quantities into the geometry of a plane, so that every visual feature, height, slope, curvature, peak, has a fixed meaning readable back into plain language. Once that translation rule is fixed, you can go in either direction: read a graph and recover the story, or hear a story and sketch its graph before doing any arithmetic.

The full treatment

First look: turning a story into marks on paper

Picture filling a bathtub. Water depth starts at zero, rises steadily while the tap runs, holds level while you sit in it, then falls quickly when you pull the plug. Draw time along a horizontal line and depth along a vertical one, and mark a dot for each moment: zero depth at time zero, rising dots as the tap fills the tub, a flat stretch of dots while you sit, then dots dropping fast toward zero. Connect the dots. The picture that results, rising, flat, falling, is not an illustration added after the fact, it is the bathtub story re-expressed as a shape. Anyone who has never seen your bathtub can read the depth story straight off the picture: where it rises, depth is increasing; where it is flat, depth is unchanged; where it falls fast, depth is draining quickly.

Building the idea: position as a coded fact

Two choices make this translation work, and both are conventions you must fix before a single dot means anything. First, pick which quantity goes on which axis: by convention the quantity you control or that changes independently, often time, goes on the horizontal axis, and the quantity that responds goes on the vertical axis. Second, pick a scale for each axis, how much distance on paper stands for one unit of the quantity. Once these two choices are fixed, a single point on the page is a coded fact: "when the horizontal quantity had this value, the vertical quantity had that value." A whole curve is a coded story, the complete record of how the response quantity behaved as the controlling quantity swept through its range.

This is why a graph without labeled axes and a stated scale is not really a graph, it is only a shape. The same rising-flat-falling silhouette could describe a bathtub filling or a population growing then leveling off. The picture only becomes a claim about a specific relationship once the axes are named and scaled.

The formal model: reading four features as statements

Once you accept that a graph is a plotted relationship between an input quantity (call it x) and an output quantity (call it y, with y depending on x), four visual features become directly readable statements:

Height at a point. The value of y when x equals some specific number is how high the curve sits above that x-value, answering "what was the output when the input was this."

Slope, or steepness. Between two nearby points, slope is the change in y divided by the change in x, how much the output moved for a given move in the input. A steep upward slope says the output is increasing quickly; a slope near zero says it is barely changing; a negative slope says it is decreasing. Slope is a ratio, not a raw height, which is why two graphs plotted at different scales can look equally steep while representing very different real rates.

Intercepts. Where the curve crosses the vertical axis (x equals zero) gives the output's value at the reference starting point. Where it crosses the horizontal axis (y equals zero) gives the input value at which the output vanishes or changes sign, for instance the exact time a tank runs dry.

Turning points. A peak marks a local maximum of the output; a valley marks a local minimum. At a turning point the graph is momentarily flat, meaning the output is neither increasing nor decreasing right at that instant, an idea made precise later once rate of change is formalized, but already visible here as "the top of the hill."

The mechanism in reverse: from story to sketch

The same reading rules run backward. Given a verbal story, "a car accelerates from rest, cruises at constant speed, then brakes hard to a stop," you can sketch a speed-versus-time graph without any numbers: start at height zero, rise for the acceleration phase, go flat for cruising, then drop steeply to zero for hard braking. The sketch need not be numerically exact to be correct; what makes it correct is that every qualitative feature, where it rises, where it is flat, where it drops, matches a feature of the story. This reverse move, story to shape, is exactly as legitimate as the forward move, shape to story, because both rest on the same fixed translation rule between position and fact.

Lineage

The idea of representing a relationship between quantities as a curve in a plane is usually dated to Rene Descartes' La Geometrie in 1637, which fused algebra with geometric position and made it possible to draw an equation. Pierre de Fermat developed closely related ideas around the same period. Before this formal fusion, graphical thinking already existed informally: the medieval scholar Nicole Oresme sketched qualitative diagrams of how a quantity like velocity varied over time, a proto-graph centuries before coordinate algebra existed. The two threads, Oresme's qualitative picture-of-change and Descartes' exact algebraic coordinate plane, are the ancestors of the modern graph.

The strongest case for it

Graphs earn their place because they compress an entire relationship, potentially infinitely many input-output pairs, into a single object the eye can scan at once. A table of a thousand rows hides trends that a single glance at the corresponding graph reveals instantly: where the relationship is roughly linear, where it curves, where it peaks. This is why graphs are the working tool of choice across every quantitative field, from tracking a fever's course to reading a stock's history to designing a bridge's load response. The translation is also two-way and lossless in the qualitative sense: any genuine feature of the underlying relationship, growth, decline, a threshold crossing, must show up as a corresponding visual feature, and any visual feature you can point to corresponds to a real fact about the relationship. That tight correspondence is what makes "reading" a graph a legitimate form of reasoning rather than a loose metaphor.

The strongest case against it

The correspondence between picture and fact is only as honest as the axes and scale that were chosen, and this is where graphs most often mislead. Truncating the vertical axis so it does not start at zero can make a small change look dramatic; using different scales when comparing two graphs side by side, a common trick in dishonest reporting, ruins the comparison entirely even though each individual graph is technically accurate. A second, subtler limitation is that a graph drawn from finitely many plotted points and connected by a smooth line assumes the relationship behaves smoothly between the points you actually measured; if the true relationship jumps or oscillates faster than your sample rate can catch, the graph will confidently show a false smooth story. A common misconception is treating steepness on the page as steepness in reality: redrawing the same relationship with a stretched horizontal axis changes how steep it looks without changing anything about the underlying relationship. Always ask what the axes and scale actually are before trusting what a shape seems to say.

Where it stands now

The practice of reading and constructing graphs is universal and stable; nothing about the correspondence between geometric features and quantitative facts is in dispute. What continues to develop is the craft of choosing honest scales and encodings, and the tools for producing graphs, from hand-drawn sketches to interactive digital plots that let a reader change the scale and see the same data from multiple views. The core skill this entry teaches, translating between a story and a shape, remains the same whether the plot is drawn with a stick in sand or generated by software.

Test yourself

A reservoir is fed by a river and drained by a spillway. Over one year, inflow is roughly constant except for a sharp spike during spring snowmelt, and outflow is held constant except for a short autumn shutdown for maintenance. Sketch a graph of reservoir water level versus time across the year, marking where the level rises, is flat, rises sharply, and where any peak occurs, justifying each feature from the story. Then do the reverse: given only a graph that starts at a mid-level, dips to a shallow minimum, rises steadily to a sharp maximum, then falls back to its start, invent a plausible real-world story that produces exactly that shape, and state which features of your story are genuinely forced by the graph and which are merely one possibility among several consistent with it.