Three pictures get confused constantly, so here they are stacked on one shared axis: the population you are sampling from, one sample you actually collected, and the sampling distribution — what the sample mean itself does when you repeat the whole study over and over. Draw samples and watch all three fill in.
Right-skewed populationn = 25 · press “Draw a sample” to begin
Pick a population to sample from
How big is each sample?
25
Population μ30.0
Population σ21.2
Std error σ ÷ √n4.24
Samples drawn0
What the collected means actually did
Every number on the top row is measured from the sample means you have piled up. Every number underneath it is what the theory predicted before a single sample was drawn.
Mean of the x̄'s—μ = 30.0
SD of the x̄'s—σ/√n = 4.24
Avg sample SD s—σ = 21.2
Skew of the x̄'s—γ₁/√n = 0.28
Three spreads, and only one of them shrinks. The population's spread is σ and it never moves — it is a fact about the population, not about your study. One sample's spread s hovers near σ no matter how big n gets; a larger sample estimates σ better, it does not make the data huddle together. The only thing that shrinks is the spread of x̄ across repeated samples, and it shrinks by √n. That is why a bigger sample buys you a more precise mean and not tidier data.
Read the formula carefully. Center and spread — mean of x̄ = μ, SD of x̄ = σ/√n — hold for any population shape as long as the observations are independent. The bell shape is a separate claim: it is exact only when the population itself is normal, and merely an approximation that improves with n otherwise. Also note what is missing from σ/√n: the size of the population. Precision is governed by how many you measured, not by what fraction of the group you reached — a sample of 1,000 is equally precise for a city or a country. (Sampling without replacement from a population only a few times larger than the sample does need a correction; that is the one exception.)