Dig a hole at the beach, then watch as surrounding sand pours into it, each grain trickling down and consolidating its new position.
In physics, this process is called "self-organised criticality"; each microscopic movement contributing to an ultimate pattern of adjustment and settling.
In one sense, this principle can be translated to the nearly 5000 aftershocks so far recorded in the wake of the Kaikoura Earthquake.
Each grain of sand represents a tremor, marking the spot where another part of earth's crust is accommodating itself to its new geological environment.
It's a process that can take a long time - in Kaikoura's case, scientists expect this seismic simmering will take many months at least.
But aftershock behaviour is more dynamic still.
While it might be possible to pinpoint the place in the hole where each grain of sand will slot into, it's beyond impossible to predict exactly where and when each aftershock - or the aftershock of that aftershock - will strike.
After all, each one can be viewed as just one more earthquake changing the way stress is distributed underground, and sometimes activating fresh spates of localised activity.
"So, if you have a magnitude 5.7 event in North Canterbury, then it can, on its own, kind of resuscitate activity there," Victoria University earthquake scientist Dr John Townend said.
"That's exactly what we saw in Canterbury, when everything kicked off with the 7.1 Darfield earthquake, then it went on for several months and died down, but then the sequence got another kick of energy from the Christchurch Earthquake."
Darfield's shaky legacy has continued with many more notable quakes - including the 5.7 Valentine's Day jolt that struck near New Brighton in February this year.
Aftershock activity also wasn't a case of only neighbouring faults shifting loads between each other.
A significant aftershock could change stresses on a fault far from the epicentre with the seismic waves it sent propagating through the earth.
This week, GNS scientists said they couldn't rule out the possibility that passing seismic waves from the 7.8 main shock at Culverden caused stress changes that triggered a new "slow-slip" silent quake off the North Island's east coast.
"So we have what are called static changes - what you see as the shape of the land around the earthquake site changes - and dynamic changes, those related to seismic waves."
The scale of all of these changes, however, could be related back to the scale of the initial earthquake.
It was also possible to say some places tended to generate more aftershocks than others.
This was especially true for the Marlborough Fault Zone, where seismic activity had been ever more heightened by the 6.5 quake that struck near Seddon on July 21, 2013, and the 6.6 quake that hit also hit near the town within a month later.
"There are a lot of quakes that have been happening in that area - and that probably relates to the structures of the faults that are there, and the prevailing activity.
"But I should emphasise that a huge area of the North and South Island has been affected by the big quake, so aftershock activity will take place anywhere within that wider area; it's just that some spots will be more productive than others."
At GNS Science's offices in Wellington, GeoNet officers have been mapping the Kaikoura Earthquake's aftershock sequence and feeding the data into statistical models that calculate forecasts.
GeoNet's quakesearch tool shows how the aftershocks have been clustered relatively tightly around those places in Marlborough where faults - the Kekerengu, the Hundalee, the Hope - violently ruptured.
The area nearest those faults had a probability of 80 per cent or more for damaging shaking in the next 30 days, while the probability of damaging shaking in the Wellington area was put at less than 10 per cent.
While this probability was considerably lower than in other regions, it was possible for shaking similar to what occurred last Monday to happen again in the capital.
The GeoNet probabilities also pointed to a virtual certainty of many more aftershocks large enough to be felt, although numbers were decreasing with each new forecast issued.
The latest gave a 99 per cent chance of an average 21 quakes between 5.0 and 5.9 this month and 54 over the next year.
Odds of up to 10 shakes between 6.0 and 6.9 striking over the next 12 months were just as high, but a plus-7.0 event was far less likely, with a probability of 18 per cent within a week and 38 per cent within a year.
By noon today, GeoNet had recorded a total 4879 aftershocks from the Kaikoura event.
But most of them had been small - 4828 earthquakes registered below 4.9 - and would have only been felt close to the epicentre.
As of Monday this week, there were also 47 aftershocks between magnitude 5.0 and 5.9 range, and three aftershocks in the magnitude 6.0 to 6.9 range.
GeoNet spokesperson Sara McBride explained that, when a large quake like the 7.8 Kaikoura event occurred, aftershocks happened "thick and fast".
"When that happens, our seismic detection network can get overloaded, so not all aftershocks are detected," McBride said.
"Our seismic network is very sensitive and typically picks up even the smallest of shakes."
But now, due to the big earthquakes coming through, it was more difficult to detect all of the quakes.
"Imagine a clear lake. Most of the time, when you skip small pebbles on the clear lake, you can spot the associated ripples easily.
"But then a giant rock is thrown into the clear lake.
"The splash and ripples it creates can cause the smaller pebbles to go unnoticed, but pebbles are still being skipped on the lake.
"Once the giant rock's waves subside, the smaller pebbles and their ripples become noticeable again. That is what is similar to what is happening here."
At this stage, GeoNet haven't been able to detect all of the smaller aftershocks in amongst the waves of the larger earthquakes.
Therefore, the total number of aftershocks in the earthquake catalogue below magnitude 5 is currently lower than what has actually happened.
"The total of 4879 aftershocks is the bare minimum of what we've detected," McBride said.
"When we have more time for data processing we will likely find further small aftershocks in the seismic waves of the main shock - our earthquake catalogue will be changed to reflect this in future."
More small aftershocks would continue to occur than big ones.
"As a rule of thumb, there is a tenfold increase in the number of earthquakes for every one-magnitude decrease," McBride said.
"For example, for one M6.0 earthquake, we expect around 10 earthquakes of M5.0 to 5.9 and around 100 earthquakes of 4.0 to 4.9 on average."
This applied to all seismicity experienced, as well as that occurring as part of the aftershock sequence.
"In summary, the aftershocks are at the lower end of the forecasted range," McBride said.
"It is a bit puzzling and we are scratching our heads at this one.
"What we can say is that just because we are in lower end of the forecast, it doesn't mean that this will stay that way."
Scientists could really only be sure of one thing: aftershock sequences generally fade away over time - something already starting to be seen in this event - with spikes of activity and occasional larger earthquakes as the earth corrects itself.
"With probabilities, there are two faces to it: one is making observations about what aftershocks are happening and estimating the statistical properties of how many we can expect, and the other is the shaking that's likely to be produced by a particular aftershock," Townend said.
"So, if you take that whole sequence of possible aftershocks, and, for each one, try to calculate the shaking for different parts of the country, that is what these maps can show us.
"This is why it's still important to study aftershocks; not just to work out what sort of things we can expect, but to make probabilistic forecasts of shaking intensities, because that's what buildings respond to and that's what people respond to."
GNS Science seismologist Dr Anna Kaiser said aftershocks were also useful at offering clues to how truly seismically active a region was.
"Before an earthquake, you sometimes can't identify fault structures because the seismicity is so sparse, and you need a long period of time to be able to see a structure by associating aftershocks with it," she said.
Aftershocks could also give away previously secret ways in which faults behaved.
"They won't tell you everything - you really need to rely on the other data to do that in detail - but it can give you some ideas."
Was it possible Kaikoura's aftershock sequence would move elsewhere?
In the Canterbury quakes, scientists had watched the pattern shift eastward.
There were also historical cases of quakes "travelling" along a fault zone.
One of the most famous was a westward moving pattern along Turkey's North Anatolian Fault, which culminated in the 7.6 Izmit quake in 1999, which killed around 17,000 people, and the 7.2 Duzce quake several months later, that left at least 845 people dead.
"Basically what can happen in these cases is one big quake happens and reduces the stress acting on that fault, but it loads the stresses acting on the adjacent fault," Townend said.
"So once you get a big quake sequence moving, it's plausible that it would keep moving in the same direction.
"At the moment, however, we still need to understand some of the basics of it, and, really, it's not easy to make useful statements with any particular quake about which way the aftershocks are going to move."
As the Christchurch Earthquake highlighted - a tragically-located aftershock powerful enough to 185 people - the take-home message remained people should be prepared for anything, at anytime, anywhere.