A naive observer might say this is a group of people strung out one after another, but someone more familiar with lines will know there is some kind of more complex social hierarchy here. Getting to handle such things was part of our history..
The current issue of the online journal Biolinguistics is devoted to the subject of recursion, a procedure that enables the creation of sentences of indeterminate length. Last week's post discussed an article that doubts "full-blown recursion" is part of language. Another important paper in the issue is The Neural Basis of Recursion and Complex Syntactic Hierarchy by Angela D. Friederici, Jörg Bahlmann, Roland Friedrich & Michiru Makuuchi. Based on a series of experiments, it concludes that hierarchical thinking is common to many tasks besides syntax. Most striking is the evidence that mathematical thinking uses different parts of the brain from syntactical thinking and therefore (says I) evidence of unbounded recursion in mathematical reasoning cannot be used to support unbounded recursion in "the language faculty."
- Hello Mister Bolles
- Howdy Pard
- Good afternoon Miss Louisa
- Glad to see you
and so on. There are a limited number of sentences this rule can produce (number of greetings in dictionary times number of titles times number of names). There is no hierarchy here; words come one category after another. To understand the output a listener can take words one at a time.
Hierarchies use phrases that can be embedded in other phrases. Have as a rule S = subject [+S…] + verb. That can generate:
- The cat died.
- The cat the dog chased died.
- The cat the dog the neighbor owned chased died.
- The cat the dog the neighbor my wife kissed owned chased died.
and so on. Logically this nesting process could go on forever although there is some dispute over whether these multiple embeddings really produce English sentences. Anyway, these kinds of sentences are hierarchical. A listener has to reunite broken phrases in order to understand the sentence.
Does it take a different brain system to follow a hierarchical utterance than a non-hierarchical one? In other words, did we have to evolve a new system to produce and understand hierarchical sentences?
Experiments used an artificial language. A rule generates strings of letters (not words), e.g., a stringlike mvwwamx. They learn some strings that are correctly generated and then have to determine whether or not other strings are correct or not. Believe it or not, people can do this at better than chance levels even without knowing the generative rules.
We seem to be able to spot the patterns without being able to explain them. Once participants in experiments gain some practice they are given a test. to determine if a string is correct or not? Brain scans determine which area in the brain is used to solve the problem. This approach gets rid of any complications arising from meaning.
All the studies looked at prefrontal cortex (PFC) regions. The paper asserts as a general principle that there is a gradient running from back to front of the PFC over which hierarchical thinking grows more intricate. I find that news interesting in itself, and more interesting because, the oldest known language area of the brain, Broca's area, lies toward the middle of the PFC. Thus, there are likely to be more complex hierarchical thoughts than those used to support ordinary language.
This news also suggests that as the PFC grew during our transition from Homo habilis to erectus and on to sapiens the steady growth in size of the PFC carried with it an increasing ability to think hierarchically. Recursive syntax therefore might well be a side effect of those interesting forces that for two and a half million years kept our brains, especially our PFC, growing.
I'm running ahead of the authors here, in my usual way, feeling inspired to mention possibilities that undermine old assumptions. My point is not to toss out those old assumptions but to remind us all that they stand on uncertain ground. Recursive expression may not have reflected any great leap forward, but part of a steady change over millions of years.
Back to the paper—I'll summarize what the reported experiments found:
- Finite-state grammar: presentation of correct strings first activated an older region (the frontal operculum) while presentation of incorrect strings first activated both the frontal opeculum and the younger Broca's area.
- Recursive strings: presentation activates Broca's area "suggestin that Broca's area is involved in the processing of complex hierarchically structured sequences" [p. 89]
- Visual-Spatial sequences: this test uses strings of abstract elements that were categorized by shapes and textures. These comparisons involved Broca's area and other regions too, indicating that Broca's area is not limited to the linguistic domain. "We may conclude that Broca's area receives its domain-specificity as a part of a particular neural network which differs from domain to domain." 
- Mathematical Formulae: A group of math and physics students who were deeply familiar with the use of equations and mathematical reasoning were shown a series of algebraic formulae without any numbers and asked to judge their syntactical correctness. This work most strongly activated a region a bit forward from Broca's area known as Broadman's Area [BA] 47. The authors report "that even for mathematicians, the processing of mathematical formula could be less automatic,, requiring more cognitive control than the processing of language hierarchies" .
The full data is a bit more ambiguous than the highlights I've specified here, but the basic picture of a steady march to the front as the hierarchies grow more demanding seems correct. The last detail is especially interesting because we can be reasonably confident that our ancestors of 200 thousand years ago were not solving algebraic formulas. So what was the pressure to make us masters of complex hierarchy?
I don't know but I would be interested in learning what kind of cognitive demands it takes to understand the ins and outs of a hunter-gatherer band.