Have linguists been hunting for a pot of gold at the end of the rainbow?
The argument in favor of language beginning as personal thought is now dead and should be buried. What? You have always assumed language began as a tool for telling things to one another? You must be new to this blog. At the start of this month I posted a report on a paper by Bolhuis et al. that rehashes the argument that language began as a new and improved way of thinking. Speech and signing came only later when internal thinking was "externalized." Technically speaking, there was a mutation that improved the way an ancestor thought, making things like planning a whole lot easier. This mutation proved adaptive and was passed on to the descendants. Finally, one or more mutations led to the creation of an interface connecting thinking to motor actions so that the thoughts could be expressed publicly. It is tempting to admire the boldness of asserting such hooey, but in the end it is still hooey. Now someone has taken the trouble of refuting the thought-first hypothesis and the refutation is a doozy. I expect this and tomorrow's follow up to be my last posts discussing the thought-first hypothesis.
Dr. Maggie Tallerman has published a paper in Language Sciences titled, "No syntax saltation in language evolution" (abstract here). The title is a bit of misdirection. It means that syntax did not evolve in a single mutation (a great leap forward, or saltation), but the paper really argues that the evidence is overwhelmingly in favor of externalization first. Tallerman boldly declares, "I challenge both the idea that there was a recent syntax saltation, and the idea that the language faculty originally arose to support thought." [page 208]
The obvious question that the thought-first hypothesis raises is: how do you think linguistically without words? It is clear that animals can do some thinking without benefit of language, but their thoughts are perceptual. They recognize things discovered by the senses, probably through a process of association. Anybody who has ever walked their dog knows that the dog is busy nosing its way through the world. But while they recognize scents, they presumably don't have names for what they recognize. A hawk swooping down toward a wren can maneuver beautifully, but it does not think, "I can get it when it moves toward that branch." When I try to imagine such a life, sports come to mind. I never was much of an athlete, but I can recall swinging at a tempting pitch entirely on the basis of what I saw. As Yogi Berra probably never said, "You can't think and hit at the same time." Driving a car is also pretty much perceptually-based.
So I've been dubious, but Tallerman goes to the trouble of showing that you cannot use a word until it has been externalized.
The thought-firsters are perfectly aware of the problem. Their solution is to assume the presence of concepts and say that linguistic thinking began when a mutation allowed two concepts to be joined together. Can concepts be linked like words so that the concept of redness and the concept of elephants can be linked to form the new concept of a red elephant? Thought-firsters say yes, and Tallerman quotes Chomsky's statement that "we have no reason to believe that there's any difference between lexical items [words] and concepts."  Tallerman then goes on to demonstrate how words and concepts do differ.
- Words are learned: The concepts the thought-firsters depend on appear to be innate. They have not been externalized or passed on through some communal process. Perhaps some clever fellow develops a new concept through insight, but that concept will not be shared. Words, however, are all learned. That means that words must be passed on externally. If in my cleverness I coin a new word, I can use it and perhaps it will win a foothold.
- Words combine according to rules: The technical term for these rules is and "edge feature." You cannot combine just any two words. Running stalactite does not make much sense. Words acquire their rules via public usage. In the sentence I ran from the falling stalactite, the word order (falling stalactite, not stalactite falling) is determined by convention and varies from language to language. In English we hope for the best because hope cannot take a direct object, so you cannot combine hope + the best. In Welsh, however, hope the best is the correct combination. After running through a few more examples, Tallerman concludes that the rules of combination "are a property of externalized language, and … they become established and undergo changes only in the usage of lexical items. Thus, concepts are not equivalent to (and cannot simply turn into) lexical items." 
- Coining new words requires a knowledge of existing words: I'm not going to dwell on this point because I don't believe it. Oh, I'm sure it is mostly true, but I have seen babies come up with words that usually fade away. (Sometimes, however, the word survives as the child's nickname. Say "boog" often enough and folks may start calling you Boog.) The best known examples of spontaneous coinages are the homesigns that deaf children invent and teach their families. These examples, however, also show the limits of coining. Homesigns do not keep growing to become a full language unless the user becomes part of a deaf community where a real process of communal externalization can take place. Tallerman proposes that language growth requires a "speech community … [where] entries [in the lexicon] only subsequently [gain] lexical properties (including edge features) through usage." 
Tallerman insists: "Externalization is critical in the gradual growth of the ability to learn, store and retrieve items from the lexicon…. The very fact of engaging with other speakers starts to develop complex human concepts." [210-211]
Mild-mannered as this remark seems, it strikes a dagger at the focus of the linguistic efforts of the past 60 years. The original ambition was to come up with a set of rules that could allow a machine to generate "all and only" the sentences of any particular language. Tallerman is showing that no such rules are possible. I can understand sentences I cannot generate and can alter my speech in response. That process is what allows me to acquire language.
Thought-firsters pay so little attention to external processes that they insist the process of language acquisition can effectively be considered an instantaneous event. After all the process is virtually the same in every child and leads every speaker to the same result. Even if we accept the process argument, Tallerman has just blown the result argument out of the water. There does not come a time when a person stops acquiring the language and just gets on with using it. Need proof? A week ago I could not have written this blog piece, and not just because of getting a bit more data to pass on. It's the implications of the data that have spun my head around. It is not just that it is hard to come up with the complete set of the internal rules for generating all and only the sentences of a language. Turns out there is no such rule book. The thought-firsters have been chasing after a leprechaun's pot of gold.
Tallerman also discusses the importance of externalization and syntax, but I will save considering that for tomorrow's post.