Verified by 
"How can you not know yourself?" Perry said.
Mr. Berg threw up his hands in mock despair. "Too much evidence. Too much information. Most of it contradictory. Evidence here suggests one thing; other information suggests the opposite. No clear picture emerges. Won't hold steady."
-From Bridge of Sighs by Richard Russo, p. 378
I suspect most of us feel that our own behavior is incredibly variable. Sometimes you act like an extreme extravert, sometimes you're the paragon of introversion, right? All of us feel that way to some extent, like there is no clear pattern. Won't hold steady.
Some research supports this intuition. Will Fleeson's work on within-person variability in personality and behavior suggests that each person's personality fluctuates quite dramatically over time, even within the span of just a few weeks or even days. So all of us are right, in a way, we do not have just one personality.
However, there is also evidence that we might be exaggerating our own flexibility. Research on actor-observer asymmetries (that is, the difference between observing yourself and observing someone else) suggests that we are much more sensitive to fluctuations in our own behavior than in others', but we're better at detecting general patterns in others than in ourselves. That is, we have a hard time seeing the forest for the trees when it comes to our own behavior.
This research is usually taken to show that we overestimate the consistency in other people's behavior (e.g., Schwarz and Wellens, 1997, make the point that polling data based on observer reports (e.g., spouses' ratings of each other) probably overestimate the consistency in people's behavior). This is a very good point - if you want to know how someone's behavior fluctuates from one situation to another, it's probably best to ask them, not their friends or family. We are much more attuned to the situational influences on our own behavior than others are. For example, I can tell you that the reason I was grumpy last night is because the Celtics lost, not necessarily because I'm always a grumpy person.
However, there is also a flip side to this story. If others are worse at seeing the fluctuations in our behavior, they are also better at seeing the generalities. I may notice that I'm more pleasant when the Celtics win and grumpier when they lose, but those around me have a better sense than I do about whether I'm generally a pleasant or grumpy person. They're not plagued by knowledge of my thoughts, feelings, intentions, moods, etc., and can just look at my behavior and see a trend.
Why is it so hard for us to calculate an average based on all the specific information we have? I suspect, as Russo's character does, that it's because we are overwhelmed by the numerous examples and counterexamples. Ask me if I'm a grumpy person, and I can think of ten examples supporting either answer. But I don't think it's just that. As I implied above, I also think that our awareness of our internal states (moods, intentions, thoughts, etc.) can lead us astray - we forget how we actually behaved because we are so wrapped in how we felt, how we meant to behave, etc.
In addition to these informational burdens, we are also less motivated to do the arithmetic for a variety of reasons. For the person living in the body of the grumpy person, it's more important to know what will make us more or less grumpy, so we can choose the right situations, than to know whether we're overall more grumpy than the average person. And of course there's the potential cost of finding out something negative about ourselves - maybe not knowing is safer than doing the calculation and coming up short. Finally, there's the danger that even if we try to count all the trees, we'll get it wrong, and then we'll end up thinking that we have self-knowledge when we don't. In a world with so much contradictory evidence, sometimes it seems wisest to do as Socrates did and just admit that we know very little.
