The reason for this persistent gap is not difficult to divine: Sanders is viewed much more favorably than Clinton in the broader electorate. Pollster.com calculates his current favorability rating at +9 (49 percent favorable to 40 percent unfavorable); Clinton is currently at –16 (40 percent favorable to 56 percent unfavorable) and Trump at –21 (37 percent favorable to 58 percent unfavorable).
Purist insurgencies like the Sanders campaign usually lack direct evidence that they might be stronger general-election nominees than their primary opponents, which causes them to fall back on "hidden vote" theories claiming that they would inspire large numbers of previously inert citizens to flock to the polls to their behalf—thus compensating for their comparatively weaker standing among habitual voters. (This was Barry Goldwater's professed path to victory in 1964, an election that he lost by 23 points.) But Sanders and his supporters can currently point to national surveys suggesting that he would be a more electable candidate for the Democratic Party, which have become incorporated into his current pitch for why superdelegates should reverse the results of the pledged delegate count in order to award him the nomination.
Of course, popularity in June does not guarantee popularity in November. The flaw in Sanders's argument is that he has yet to demonstrate that he can survive a sustained blast of negative attacks. His advocacy of higher taxes on the middle class, support for government-run universal health care, past dalliances into leftist foreign policy, and self-identification as a "socialist" all represent potentially serious political vulnerabilities for Sanders, who represents a small, left-leaning state where he hasn't faced a serious electoral challenge in over 20 years and who therefore has little experience in defending his record against well-financed opposition in an unfriendly political constituency.
The fact that Sanders is currently polling well in a hypothetical matchup against Trump therefore don't necessarily predict how a general election between the two would turn out. Sanders would begin the race unburdened by the negative popular evaluations that now weigh down the Clinton campaign; on the other hand, November is a long way off, and there would be plenty of time for him to lose his current relative appeal—while Clinton, as more of a known quantity, may be less vulnerable to further Republican attacks and may in fact make a modest recovery in the polls once the Democratic nomination contest has concluded. Most Democratic leaders view Clinton, for all her flaws, as still representing a safer choice for the party's nomination (putting aside the question of whether it would be appropriate for the superdelegates to overturn the results of the pledged delegate count).
But these polls do tell us something useful about the election all the same. Some observers who view Trump as a potentially formidable general election candidate have suggested that he could receive a significant crossover vote from Democrats. "Because Trump isn’t a doctrinaire conservative—because he appeals on emotion and not policy—the theory is that he can win white working-class Democrats and other disaffected voters in the Democratic coalition," wrote Jamelle Bouie of Slate in a March critical appraisal.
If Trump had a unique appeal to non-Republicans, however, we'd expect to see evidence of it in trial heat polls against Sanders, who may be viewed as more or less a generic Democratic candidate by a general electorate who is not yet very familiar with him. Sanders's fairly wide lead at this stage indicates that there is no substantial population of Democrats who are sufficiently attracted to the Trump candidacy that they would support him over either candidate of their own party.
Perhaps things will change, and Trump will succeed in crafting a message that appeals to Democrats and Democratic-leaning independents—or perhaps the Democratic nominee will alienate a sufficiently large sector of the electorate that such an achievement won't be necessary. But while the polls can't tell us for sure whether Clinton or Sanders would be a stronger general-election candidate, they do continue to demonstrate that the Trump fan club has very few non-Republican members. Regardless of the final outcome in 2016, it is therefore unlikely that the Trump candidacy will fundamentally redefine the two parties' mass bases of support.