"Automatic thoughts" are a key concept in this approach; rising like bubbles from our underlying beliefs, and interfering adversely with our emotional lives. Note the proposed order of events: I have an automatic thought and then I feel a certain way. This is not uncontroversial; surely it is sometimes the case that we think something because of how we feel. However, I'm not concerned with that argument here. Classic CBT seeks to intervene at the level of thoughts (although more recent "3rd wave" approaches get interested in affect too). Whether that process has a downstream effect on feelings (as per this model), or whether it manages to work backwards (a modified thought changing the mood from which it was derived), if it helps it helps. There is a more interesting problem for CBT, not necessarily a barrier to its use, but a conceptual tangle which is hard to resolve: what is a thought?
We all have thoughts, and all of us will have had them automatically "popping into our heads". Often they are verbal (these linguistically articulated inner experiences are what CBT trades in), but often they are not, and here lies a problem that reveals the messiness of reality when compared to the simplicity of a treatment manual. We all know a propositional statement when we see one, but when it comes to defining a thought things get very tricky indeed.
CBT instructional texts often rely on the neatly articulated automatic thoughts of straightforward seeming cases: "I'm no good"; "I will never get a job" and so on. These can form hypotheses which we test (and inevitably fail to confirm) in the process of "collaborative empiricism". However, in reality, when we ask ourselves (or the people we are working with) the central question--"what just went through your mind?"-- we need to be prepared for the answer "nothing in particular", or more commonly to hear that, whatever it was can't really be articulated.
What then are we asking people to do? Thoughts don't straightforwardly exist. You can't see them, and often when you try to write them down they are entirely elusive. What do we mean when we say "thought"? It looks like there is a philosophical tangle at the heart of CBT that isn't being addressed*. This conceptual difficulty is not (as far as I can see) a deep practical flaw in CBT, though it is a superficially unnerving question in the context of a task that takes linguistic content so seriously. Thinking about its implications carefully might have positive ramifications for practice.
To help think this one out, here is a list (not comprehensive and not in any particular order) of things that can reasonably be said to exist (stolen entirely from Paul Meehl's ontology, about 70 minutes and 15 seconds into video #4 here):
Minds are abstract entities, but to the extent that we have them, they must arise out of things in the world like this. Which of them corresponds to a thought? A "thought" doesn't seem as though it can be mapped onto a specific substance or structure. Equally it doesn't seem quite plausible that a thought is the result of a specific brain-state either; thoughts seem dynamic in a way that states are not. It might then be reasonable to say that any given thought could be re-described as something like an event (meanwhile, another of CBT's concepts, "attitudes", sound like they could be re-described as dispositions).
The Numbskulls: Thoughts as Instructions
We have the experience of a verbal, or pictorial (or whatever) "thought", and what is going on at that moment is an event in our mind. So when we ask a person "what went through your mind just then", perhaps we are asking them a version of "how would you describe your subjective experience of the event which just took place?". This helps us around the potential objection that there was no obvious automatic-linguistic-proposition ("I will fail the exam") to report.
My re-description is a bit over-the top, and won't usually be necessary in practice (despite the conceptual complexities, most of us have a working definition of "thought"), but I wonder if it helps us to get around the apparently perplexing challenge raised by the question guiding this post. The failure to locate neatly verbalised "thoughts" could be unsettling for someone new to CBT, therapist or client, raising the spectre that the process won't "work" because it's not proceeding as it does in the book. Having a more flexible sense of what is meant opens up the space for productive questions within the therapy about just what both parties are trying to get at.
-Richard Gipps has posted a fascinating and beautifully written whole book chapter offering a philosophical critique of CBT on his Clinical Philosophy blog.
*I have heard it suggested that CBT is more generally silent on these broad conceptual questions, and that it only appropriated the moniker "cognitive" as a way of trading on the enormous success of the "cognitive science" framework that came to dominate psychology after the late 1950s.