So, I'd say there are three modes of thinking I can identify:

• Normal thinking, what I'm doing the vast majority of the time. I'm thinking by manipulating concepts, which are just, well, things.
• Introspective thinking, where I'm doing the first kind of thinking, and thinking about it. Because the map can't be the territory, when I'm thinking about thinking the concepts I'm thinking about are represented by something simpler than themselves - if you're thinking about thinking about sheep then the sheep you're thinking about thinking about can't be as complex as the sheep you're thinking about. In fact they're represented either by words, or by something isomorphic to words - labels for concepts. So when I'm thinking about thinking, the thinking-about-thinking is verbal - but the thinking isn't (although there's a light-in-the-fridge effect that might make one think it was).
• Auditory thinking, where I'm thinking in words in my head, planning a speech (or more likely a piece of writing - and most of the time I never actually write or say it). This is the only kind of thinking I'm conscious of doing that really feels verbal, but it feels sensory rather than thinking in words; I'm hearing a voice in my cartesian theater.