1. The straightforward "thought preceeds emotion" model is a common misunderstanding or oversimplification of CT's view. What Beck really says is that the brain is constantly and automatically appraising perceived stimuli. These appraisals (aka "automatic thoughts") elicit emotional and behavioral responses. However, the chain of events doesn't stop at that point. The individual's mood biases cognition in mood-congruent ways. This tends to result in additional mood-congruent thoughts which elicit additional emotional and behavioral responses, ... ad infinitum.
   
   A good example is Beck's discussion of "the downward spiral of depression." Negative automatic thoughts tend to elicit a depressed mood, a depressed mood tends to bias perception and recall in a way that results in vigilance for negative experiences, enhanced recall of prior negative experiences, and a more pessimistic outlook. This tends to result in additional negative thoughts, which tend to elicit more of a depressed mood, ...
   
   
2. Beck has endorsed a more complex view than "thoughts cause feelings" at least since the early 1970's if not longer. However,some authors have misunderstood his view and some have presented the model more simplistically when writing for non-professional audiences.
   
   Debate over whether cognition always preceeds emotion heated up when Zajonc published a major article in the early 1980's that seemed to provide empirical evidence that emotional responses occurred before cognitive responses did and some interpreted this as contradicting CT. Actually, Zajonc's research didn't show that emotional responses preceed cognitive responses, it showed that simple appraisals such as good/bad or threat/non-threat can occur before a stimulus is fully perceived and can elicit emotional responses. This is an interesting finding but is doesn't contradict CT. I'm not aware of any solid evidence that emotion precedes appraisal in clinically-relevant situations.
   
   
3. If emotion really did preceed thought, this would present a variety of conceptual problems such as "If emotion preceeds recognition and appraisal of the stimulus, then what determines the emotional response?" However, even then I'd argue that we have good evidence that cognitive interventions are useful with a variety of serious problems (especially when used in the context of a good therapeutic relationship and accompanied by experiential and behavioral interventions).

...

One of the shortcomings of current CT accounts of the relationship between cognition and emotion is that we tend to talk as though automatic thoughts happen in isolation, one at a time, followed by emotional responses, one at a time. In reality, a constant stream of automatic thoughts is accompanied by a constant stream of emotions, each influencing the other. We tend to talk as though it is simply a matter of thoughts causing feelings because this simplified view is easier to put into words and works well enough to lead to clinically useful interventions, not because it is a comprehensive model of the relationship between thoughts and feelings.