Bother to read through the article, and you find the following admission:
At this juncture it might be wise to enter a major caveat. The above figures were culled from the regional breakdown of YouGov's UK-wide survey of the opinions of 1,877 voters, of which 167 were Scottish voters. As such it cannot be expected to as accurately reflect voter intentions as could a bigger sample.Now, lets just break this down a bit. 4% of 167 is about 7 people. 16% would be 27 people (the percentage we had at the last election). A difference of 20 people. In national polls where the sample is a good 1000+, people usually take a margin of error of a couple of percent as read, which means that the numbers for a given party might be off by up to, well, yes, 20 people.
Now, I'm sure there is an SNP honeymoon, building on their squeeze of the "tired of Labour" vote, and that is going to make life uncomfortable for us. But I don't think this poll should make us panic. Apart from anything else, if this were a continuing trend and not a statistical outlier, wouldn't we have heard something about this before?
Sadly, I cannot find any data on YouGov's website, suggesting that they would not themselves want to put their professional reputation behind any interpretation of region-specific numbers. They leave that to Telegraph opinion pieces during silly season.