Often, not having an opinion on a candidate is an indicator that the person responding to the poll hasn't heard of or isn't familiar with the politician. In other words, it serves as something of a proxy for name recognition.
Several of the Republican candidates who aren't highlighted with circles in that second graphic have raised the point that, this far out, polling often is solely a function of name recognition. A consultant to Bobby Jindal made the case in the Wall Street Journal on Wednesday; Sen. Lindsey Graham has called it the "Brad Pitt" debate, since Pitt would likely be in the top ten in polling. (Maybe not among Republicans.)
That Monmouth poll, one of the few to spend the time asking for favorability assessments of all of the candidates, lets us compare how familiar people are with candidates to how those candidates do in polling. The July poll was conducted just as Donald Trump was surging to the front of the pack; he was second. But if we plot percentage of support against percentage unfamiliar with the candidate, the correlation is clear.
In fact, it's stronger than the correlation that existed in June -- in part thanks to Trump having increased his levels of support. (If you care, the July r-squared is 0.55; Junes was 0.26.)
If only Graham had the foresight to call it the "Donald Trump" debate. Would have been much more accurate.
That's one polling firm, mind you. But even a rougher comparison holds up. This is favorability among voters in Colorado, Iowa and Virginia, as surveyed by Quinnipiac, versus the Real Clear Politics polling average. Same deal.
It's important to note that this is by no means exclusively a function of the current Republican field. Early polling is often linked to name recognition. But for candidates like Graham and Jindal -- statewide elected officials -- to have to watch Trump in the first debate from their campaign offices must be particularly frustrating.