Question, Hypothesis, and Model

Builds testable questions, quantitative hypotheses, and models confronted with experimental evidence.

Conceptual context

Experimental physics does not begin with a formula but with a well-constructed question. Asking “what happens?” is insufficient unless one identifies what is changed, what is observed, and under which conditions. A useful question connects an $x$ to an $y$, proposes a direction of change, and allows measurements that could contradict the initial idea. A hypothesis turns that question into a concrete prediction; a model organizes the relation between quantities and separates essential features from secondary details. An experiment does not automatically “prove” the hypothesis. It compares $y_obs$ and $y_pred$, considers uncertainty, and decides whether discrepancies are compatible with the model or reveal a need for revision.

🟢 Essential level

A useful experimental question states which quantity is changed, which response is measured, and which conditions remain controlled.

The $x$ is the selected or modified condition, while the $y$ is the observed response. A hypothesis anticipates how that response should change, and a model makes the idea precise enough to generate predictions. If the data follow the expected trend, the hypothesis gains support; if they depart systematically, the model requires revision. The $m$ describes response sensitivity to changes in the condition, and the $b$ represents a baseline when the reference condition has physical meaning. Comparing $y_obs$ with $y_pred$ must include $u_y$, because a small difference may be ordinary while a repeated discrepancy may indicate a genuine failure of the proposed relation. A model that is too simple may fit a narrow range of data well and fail outside it, while an overly complex one can hide the true physical mechanism behind parameters with no clear meaning.

🔵 Formal level

Basic formulas

The simplest linear model relates the $y$ to the $x$ through a constant $m$ and an $b$.

Linear model for a quantitative hypothesis

The $m$ has units of $y$ per unit of $x$, whereas the $b$ has the same units as the $y$. For a selected $x$, the model calculates a $y_pred$. The experiment provides a $y_obs$, and their signed difference defines the $r$.

Residual between measurement and model

Real student questions

Is a hypothesis an opinion? No. It must produce observable consequences and allow possible data to contradict it.

Does a straight graph confirm the model? Not by itself. Uncertainties, domain, controls, and the $r$ pattern must also be checked.

Why does the $b$ matter? It may reveal a baseline response, an offset, or an unjustified extrapolation.

Does a large $z$ prove that the datum is wrong? No. The cause may be the datum, underestimated $u_y$, incorrect parameters, or an incomplete model.

May I change the model after seeing data? Yes, but the revised model should be tested with additional data or a validation method that separates fitting from prediction.

Final synthesis

An experimental question becomes testable when it identifies $x$, $y$, conditions, and domain. The model produces $y_pred$; the experiment provides $y_obs$ and $u_y$. Comparison through $r$ and $z$ does not deliver automatic truth, but reveals whether the hypothesis describes the data, where it works, and what should be investigated next.