Making a conjecture starts with observing a pattern, then stating a precise claim you believe is true but haven’t yet proven. A conjecture is essentially an educated guess built on incomplete information. It sits between a random hunch and a proven fact: you have evidence suggesting it’s true, but no airtight proof. The process of forming one is surprisingly systematic, and it’s a skill you can practice whether you’re in a middle school math class or exploring open problems in number theory.
What a Conjecture Actually Is
A conjecture is an opinion or claim based on incomplete information. It differs from a hypothesis in one important way: a hypothesis is a conjecture that can be tested through experiment or observation. In mathematics, the line between the two blurs because “testing” often means searching for counterexamples or attempting a proof rather than running a lab experiment. But the core idea holds. You notice something that seems to be true, you state it clearly, and then you (or someone else) try to prove or disprove it.
Some conjectures survive centuries of scrutiny without anyone proving them. The Collatz conjecture, for instance, claims that if you take any positive integer, apply a simple rule (halve it if it’s even, triple it and add one if it’s odd), you’ll always eventually reach 1. It has been verified by computer for every number up to 10 raised to the 21st power, a number with 22 digits. Nobody has found a counterexample, and nobody has proven it true. That’s the nature of an open conjecture: strong evidence, no certainty.
Step 1: Gather Data and Look for Patterns
Every conjecture begins with raw material. Cognitive science research on how people solve inductive reasoning problems identifies three core activities: data gathering, pattern finding, and hypothesis generation. You can’t skip straight to a claim. You need examples first.
Start by generating specific cases. If you’re exploring a number theory question, compute results for small values. If you’re investigating a geometric relationship, draw several versions and measure. If you’re working with a function, build a table of inputs and outputs. The goal is to accumulate enough concrete instances that patterns start to emerge on their own.
For example, suppose you’re adding consecutive odd numbers starting from 1. You notice: 1 = 1, 1+3 = 4, 1+3+5 = 9, 1+3+5+7 = 16. Those results (1, 4, 9, 16) are perfect squares. You now have a pattern worth investigating.
Step 2: Pursue the Pattern Systematically
Spotting a pattern isn’t enough. Researchers describe a strategy called “pursuit,” where you take a detected pattern and chase it through several deliberate steps. First, you decide the pattern is worth following. Then you investigate it as its own subproblem, gathering more data about the pattern itself. Next, you try to express the pattern in terms of your input variable. Finally, you build that expression into a full, testable statement.
This is where the real work happens, and it’s rarely a single flash of insight. The conjecture gets constructed piece by piece, like assembling a puzzle. Each smaller relationship you uncover becomes a building block. In the odd-numbers example, you might notice that the sum of the first n odd numbers always equals n squared. That relationship between n (your input) and the sum (your output) is the algebraic piece you’ve been looking for. You assemble it into a clear statement: “The sum of the first n odd numbers equals n².”
Step 3: State It Precisely
A conjecture needs to be specific enough that someone could prove it wrong. Vague claims like “prime numbers have interesting spacing” aren’t conjectures. A good conjecture names exactly what it applies to, what it claims, and under what conditions.
The best conjectures share several qualities. They should be nontrivial, meaning not too easy to prove. If you can verify your claim in a few lines of algebra, it’s a theorem, not a conjecture worth announcing. They should have substantial evidence behind them, not just two or three examples. And ideally, they should be concise. As mathematicians sometimes put it, a good conjecture should fit on a T-shirt.
The most compelling conjectures also carry a quality of surprise. They connect ideas from domains that don’t obviously belong together, or they make claims that seem almost too clean to be true. Mathematician John Conway once said the best conjectures should be “outrageous,” with unforeseen range and consequences that feel a little fantastic.
In practice, write your conjecture using clear language with explicit conditions. “For all positive integers n, the sum of the first n odd numbers equals n²” is well-formed. It tells the reader exactly what’s being claimed and exactly where it applies.
Step 4: Test It With Counterexamples
Once you have a statement, your job shifts from pattern-hunter to critic. Try to break your own conjecture. Search for counterexamples, the single case that would prove the claim false. Mathematicians do this constantly during the conjecturing process, even though it’s rarely visible in the polished final product.
Start with edge cases. If your conjecture involves positive integers, test it at 1, at 0, at very large numbers. If it involves geometric shapes, try degenerate cases like a triangle with a zero-length side. The philosopher Imre Lakatos argued that mathematics doesn’t grow through a steady accumulation of proven facts. It grows through “the incessant improvement of guesses by speculation and criticism, by the logic of proofs and refutations.” Counterexamples are the engine of that improvement.
When you do find a counterexample, don’t throw out everything. Analyze why it fails. Often, the counterexample reveals that your conjecture is almost right but needs a tighter condition. Maybe it holds for all odd numbers but not even ones. Maybe it works for primes greater than 3 but not for 2. Research with students learning to conjecture shows that confronting a counterexample drives people to refine both their claim and their understanding. They identify the part of their reasoning that still works and modify the conjecture to account for the failure. This refinement loop, conjecture then counter-example then revised conjecture, is how most good conjectures actually take shape.
Step 5: Strengthen the Evidence
If your conjecture survives your best attempts to disprove it, build more evidence. Test additional cases, especially cases far from the ones you originally used to spot the pattern. Look for structural reasons why the conjecture might be true, even if you can’t construct a full proof. Can you verify it in a related but simpler setting? Does it follow logically from results that are already proven?
Computational tools have become essential at this stage. The Collatz conjecture’s credibility rests partly on the fact that computers have checked it across an astronomically large range of numbers without finding a failure. For your own conjectures, writing a simple program to test thousands or millions of cases can either build your confidence or uncover a counterexample you’d never find by hand. AI tools and large language models are now being used to generate, test, and iteratively refine conjectures, automating parts of the cycle that once required pure human intuition.
A Practical Template
Putting it all together, here’s the process in its simplest form:
- Explore: Generate many specific examples of the situation you’re investigating.
- Detect: Look for a recurring pattern or relationship in your examples.
- Pursue: Investigate the pattern as its own subproblem. Try to express it algebraically or in precise language.
- State: Write a clear, specific, falsifiable claim. Include the exact conditions under which you believe it holds.
- Attack: Search aggressively for counterexamples, focusing on edge cases and unusual inputs.
- Refine: If a counterexample appears, adjust your conjecture to account for it and restart the testing cycle.
- Accumulate: Once the conjecture survives repeated testing, gather as much supporting evidence as possible before sharing it.
The cycle between pattern finding and hypothesis generation isn’t a straight line. You’ll loop back and forth, refining your data, spotting subtler patterns, and reshaping your claim multiple times. That’s normal. The hypothesis is not discovered all at once. It’s built step by step, from smaller algebraic or logical pieces that you assemble into a coherent whole. The messiness of the process is exactly what makes the final, clean conjecture feel so satisfying.