Hopefully, the first three parts of this thread on metalinguistic tasks has shown that reliance on data from such tasks can lead one astray from an accurate understanding of speech processing in the context of more ecologically valid situations. There seems to be a prominent historical precedent for language research being misled by metalinguistic tasks. In particular, it seems to be Lichtheim's use of such a task that contributed to the downfall of the classical model of aphasia...
In Wernicke’s original model, volitional speech production consisted of two parallel pathways. A direct pathway from conceptual representations to motor word memories, and an indirect pathway from conceptual representations to auditory word memories to motor word memories. The indirect pathway, was thought to exert a “corrective” influence on the selection of motor word memories. As such this pathway explained the occurrence of selection errors (paraphasias) in the speech production of sensory (Wernicke’s) and conduction aphasics.
Lichtheim, in his 1885 development of Wernicke's model, concurred with his predecessor that auditory word representations were indeed activated during speech production, and that this activation helped constrain motor word selection. He even devised a (metalinguistic) task to assess the ability to activate these auditory representations in patients who could not speak.
"…this is the method I use: I ask the patient to press my hand as often as there are syllables in the word to which an object corresponds. Those who have not lost the auditory representations can do this, even if their intelligence be limited, as I have been able to satisfy myself even under the least favourable circumstances. For instance, a patient who, besides a focal lesion of the right hemisphere, had had a haemorrhage in the left half of the pons, and suffered, among other pseudo-bulbar symptoms, from complete speechlessness, preserved the faculty of fulfilling the test to the very last." p. 441.
He applied this test to Broca’s aphasics (although he admits they were not pure forms), and found that they could not perform the task. From this he concluded that these patients “lost the innervation of the auditory word-representations” (p. 441), and therefore that in Broca’s aphasia “the path from concept- to sound-centre must be interrupted” (p. 441). This conclusion forced Lichtheim to reject Wernicke’s position that conceptual representations can directly activate auditory word representations. Instead, Lichtheim proposed that the auditory activation in volitional speech production must pass through Broca’s region; i.e., concept → motor → auditory. An alternative interpretation, which Lichtheim did not consider, is that it is Broca's region that somehow supports the ability to perform his syllables task. On this interpretation, he would not have had to reject Wernicke's original position that conceptual systems can activate both motor and sensory representations of speech directly, and in parallel.
Lichtheim's conclusions from his metalinguistic task led him to several awkward claims and logical contradictions, for example, how to explain paraphasias in transcortical sensory aphasia. These problems were targeted by subsequent authors (e.g., Freud) as serious problems for the general connectionist (i.e., classical) approach. If Lichtheim had stuck with Wernicke’s original claim, these problems would not have arisen, and perhaps the classical models would not have fallen out of favor.