Normal probability plot

Each observation is plotted against the z-score (standardized score) a normal model would expect for a value at that rank, so data that really is normal produces points hugging a straight line. Systematic curvature is the signal to read: a bend in one consistent direction points to skew, while an S-shape points to tails heavier or lighter than normal. Which way it bends depends on which axis holds the data, so fix the setup first: with the data on the horizontal axis and the expected z-scores on the vertical, a right-skewed set of reaction times curves away below the line at the high end, because its largest values sit farther out than a normal model predicts. Treat the plot as a diagnostic rather than a proof, since small samples wobble off the line by chance and no plot can confirm normality.