WHAT do by-elections tell you about the public mood? Probably not much. They are too small and too isolated to suggest defining moments in anything other than rare times of political transition. Yet there has been at least one by-election that did in fact change South African history.

On a warm February night in Potchefstroom in 1992, former president FW de Klerk's National Party lost the Potchefstroom by-election to Andries Treurnicht's right-wing Conservative Party.

De Klerk's party had won the seat decisively in the 1989 election. The 11% swing meant that in a future, whites-only national poll, Treurnicht could win. The result shredded De Klerk's claim that he could muster white support for a negotiated settlement with black leaders.

It was a crucial moment, because De Klerk could have responded in two ways: retreat and cower or launch a new offensive. He did the latter, demonstrating once again the lost art of assertive politics, by calling a referendum on the negotiations process, and winning decisively.

The Potchefstroom setback also had a sobering effect on his negotiating "interlocutors", as they were then known, who realised the key to all negotiations: to reach a conclusion, you have to finesse not only your own supporters but also your opponents' supporters.

Lots of water has flowed under the bridge since then, but it goes to show by-elections can tell a story that is significant but not necessarily decisive. So it seems odd that the media covered so few of the by- elections since this year's national election in any depth. The one that got the most coverage was September's municipal by-election in Phomolong, Tembisa, which the Congress of the People (COPE) won. COPE's Thando Mgaweni drew 50,6% of the vote against African National Congress (ANC) candidate Thomas Mabye's 45,4%.

The ANC has written this by-election off, by claiming that its candidate was a well-known drunk. This may or may not be, but if he was a well-known drunk why did the ANC put him forward as a candidate?

Was this really why he was defeated? It might have been one factor. Mail & Guardian writer Rapule Tabane, who visited the area, wrote it seemed voters had cited "problems with the candidate". The real problem was not that he drank, but that he didn't seem to care much about constituents' problems.

The interesting thing about this victory is that it runs counter to the popular notion that COPE is imploding under its chronic leadership problems. The point is that, like the history of Potchefstroom and its aftermath, confident and engaged politics can be very attractive to voters.

By no means does this single victory suggest a turning of the tide for COPE, which suffered crushing defeats in many other by-elections. But it does show that voters may be looking closer at the personal profiles of candidates and not simply voting like sheep for political parties as many claim.

Some other by-elections since the general election are equally intriguing. The Inkatha Freedom Party (IFP) lost three seats in KwaMbonambi and Newcastle to the ANC last month, and two in October, which suggests the election trend away from the IFP and towards the ANC is holding.

This raises the question of whether ethnicity is more of a factor than South Africans would care to acknowledge, given that the ANC is now Zulu led. Or is the IFP just crumbling?

On November 18, the Democratic Alliance (DA) won a tight race against an ANC candidate in Cape Town. The DA also won a seat from the ANC in Worcester in October. Does this mean that, like the trend towards the ANC in the national election in KwaZulu-Natal, the trend from the ANC towards the DA is also continuing in the Western Cape?

But the traffic has not been one way. For example, COPE also won a seat from the DA in Phillipstown, Northern Cape. Does this make many other DA seats potentially winnable by COPE?

By their nature, by-elections raise more questions than they answer. But the questions they raise can be crucial.

Cohen is a freelance writer.