Mendeleev ordered the elements by mass but let chemical behaviour fix the columns

1. Dmitri Mendeleev arranged the elements known in the 1860s using their properties and the properties of their compounds.
2. He worked mostly in order of increasing relative atomic mass, starting a new row whenever similar chemical behaviour appeared again.
3. Elements that reacted in similar ways and formed compounds of similar formulae were placed in the same column.
4. The columns showed that chemical properties repeat at regular intervals, which is what makes the arrangement periodic.
5. Where an element did not fit, he trusted the chemical evidence rather than force it into the wrong column.
6. He left a gap instead of filling a position with an element whose chemistry did not match its neighbours.

Gaps in the table became testable predictions

1. Properties change in a regular way across a row and then repeat when the next row begins.
2. A gap therefore sits between elements whose properties bracket the missing one.
3. Mendeleev estimated a missing element's relative atomic mass from the elements on either side of the gap.
4. He predicted physical properties such as density and melting point from the same pattern.
5. He predicted the formulae of the compounds that the missing element would form.
6. He named two predicted elements eka-aluminium and eka-silicon, after the elements directly above them.
7. Gallium and germanium were discovered later and fell into exactly those two positions.

Isotope abundance explains the pairs that sit out of mass order

1. A sample of an element contains its isotopes in fixed proportions, and those proportions set the value of $A_r$.
2. A common isotope pulls the average towards its own mass far more strongly than a rare one does.
3. For a few neighbouring pairs the averages come out in the opposite order to the one their chemistry demands.
4. In today's table argon has $A_r\approx 39.9$ and potassium $A_r\approx 39.1$, yet argon is an unreactive gas and potassium a reactive metal.
5. Argon was not isolated until 1894, so that pair belongs to the modern table rather than to Mendeleev's.
6. Mendeleev himself reversed two such pairs, putting cobalt ($58.9$) before nickel ($58.7$) and tellurium ($127.6$) before iodine ($126.9$).
7. Ordering by atomic number removes every one of these exceptions, but that idea arrived after Mendeleev's work.

Why the later discoveries counted as evidence

1. The predictions were published before the elements were found, so they could not have been fitted to the results afterwards.
2. Gallium and germanium matched the predicted masses and densities closely.
3. A table that predicts unknown facts carries more weight than one that only organises known facts.
4. The gaps closed one by one as more elements were isolated.
5. The pairs still out of mass order were explained once atomic number replaced relative atomic mass as the ordering rule.