It is impossible to put a number on this because it depends on specifics of the device that are hard to know. How often does the device sample the GPS position? What is the noise on those positions? Does the device implement a minimum distance to new position? Does the minimum distance change over time or with accumulated error?
For example, lets say a device records the position every second, and every reading has up to ±5 meters of random noise on it. If you draw a point on a map for all the readings in one hour, then in theory you get a random distribution of 3600 dots within a 10 m diameter circle. The distance you are asking about is the length of a "connect the dots" line between the points, in the order the points were received.
I'm not going to do the math to find what the average distance is between any two consecutive dots, but it's probably around a couple of meters or so. Let's say for sake of example that this average distance is 2 m. In this example the stationary tracking distance would be 7.2 km/hour!
Real GPS units don't do that for several reasons:
- For hiking, points are usually sampled much slower than once every second. I use Back Country Navigator on my Android tablet for GPS tracking, and the fastest setting is 5 seconds if I remember right.
- There is usually a minimum distance required to declare a new point. Sometimes this is fixed, sometimes adjustable. You probably won't find a hiking GPS that declares a new point without it being at least 2 meters from the last point.
In addition to the above, the overall position error per reading obviously makes a large difference too. This is very hard to predict since you said nothing about how good a view of the sky the unit has.
Just being "outdoors" isn't good enough to guarantee low error. I do a lot of trail mapping. One way I get a rough handle on error confidence is to walk each trail segment at least twice, once in each direction. I can see significant differences in discrepancies between multiple tracks on open ground, under trees, and whether the trees have leaves on them. This is to the point that I just don't do mapping when the trees are leafed out. Trail mapping season for me here in north-central Massachusetts starts in late October and ends by May.
In addition to variables like canopy cover and geography, the GPS satellites also come and go into view. Sometimes the configuration visible to your GPS unit is more optimum than at other times.
Under reasonable conditions with a good GPS, you can usually say the error won't exceed 10 m, and is usually less than 5 m. But there is no guarantees, and local conditions matter.