Archaeological evidence suggests that spoken or sign language is a relatively recent phenomenon that emerged right after our divergence with Neanderthals. Gathered data show that the life of early Homo sapiens was replete with external symbolic representations. However, such findings provide little to no insight as to when the relevant semantic and syntactic features evolved, as well as what pressures were responsible for their emergence.
Biological evolution of language is peculiar due to the fact that humans are the only species capable of such a developed form of communication. Most species are capable of communicating in one way or another. Such communication can be compared to that of other related species and in this way explained in evolutionary terms.
An example of successful evolutionary analysis of intra-species communication is the case of túngara frogs. Most male frogs whine during the period of their sexual advertising, however túngara frogs are peculiar in that their males add chucks to their whining. How did they come to do this?
In order to answer this question, one must compare sister species of the túngara frog and determine, which differences could have caused chucks in their communication. It turned out that females of the sister species do not have sufficient auditory mechanisms to recognize the chucks by male túngara. This, in turn, aids in explaining, why túngara males have evolved the larynx allowing them to produce chucks.
But evolutionary analysis of such kind is mostly irrelevant, when investigating the origins of language development. Homo sapiens have no sister species that would be capable of anything even remotely similar to human language. This may sound like a bold or even arrogant statement, however the empirical data suggests that no other species including nonhuman primates is able to perform high level computations required for language processing.
A language consists of abstract information units that are organized and combined following some specific computational procedures. Such computation allows for distinguishing between many ambiguities that are prevalent in any natural language. As an example, the word “unlockable” is ambiguous because it can be understood both as something which is not possible to lock ([un-[lock-able]]), as well as something that is fairly easy to unlock ([[un-lock]-able]). The order in which the word is processed, which is in turn grammatically dependent, lets us distinguish between phonetically identical words according to the context.
Another example of high-level computational capacity, needed for natural language processing, comes from research on language acquisition. It has been observed that a child first makes use of incorrect but biologically possible grammars, later narrowing them down to the target grammar of their mother tongue. For example, many kids omit the subject and say “tickles me” instead of “he tickles me”.
The former is ungrammatical for English, though fully consistent with grammars such as Mandarin. As the child acquires more grammatical rules, forms like that are gradually eliminated in favor of grammatical forms in English. This suggests that children are endowed with a capacity to acquire a wide range of possible grammars, which are then selected by the linguistic data in the specific environment.
In comparative studies, humans are often compared to songbirds or nonhuman primates. Songbird communication is a highly specialized and intricate system. However, it is also finite and linked to a single (acoustic) sensory channel. Human language, on the other hand, has recursively definable syntax with an infinite number of possible combinations, as well as gesticular, symbolic or other forms of visual representation. What is more, when song syllables comprise longer structures, new combinations have little or no impact on the meaning of the song, whereas human language semantics is highly dependent on the way sentences are formed.
Comparison of human and nonhuman primate communication does not provide much insight neither. Even though our vocal apparatus is nearly identical to that of a chimpanzee, there are no parallels between human language and chimpanzee communication. For example, primate children do not acquire their communicational capabilities in the same way we do – they do not babble, they do not learn vocally and they do not induce new syntactic structures from the ones they already know.
Given the data above, researchers conclude that there is no empirical basis to suppose that a nonhuman animal form of communication served as a precursor to the modern human form. The evidence from comparative animal behavior provides little insight into how our language phenotype evolved.
Another method prevalent in evolutionary analysis is modelling. In language evolution research it is common to model a population of individuals communicating by means of their particular languages. Some kind of fitness measure is introduced which differentially affects the following generations of the modelled individuals. However, the vast majority of modelling efforts assumes an already existing language phenotype, thus shedding no light on how it emerged in the first place.
Since our pre-linguistic past is not available for observation, the only reliable method is to observe the current changes in languages and test the assumptions according to what is called the uniformitarian principle of historical science: if something has played a role in the evolution, the force of this phenomenon should be observable in currently observable processes as well.
Among the assumptions to be tested by means of evolutionary modelling is the fitness metric – what makes one form of language better than another in terms of survival of a species. A leading proposal is to identify language fitness with communicative success: if individuals can communicate better and learn languages more efficiently, then they have a greater reproductory success. However, even in this case evidence is perplexing. For example, the final consonant in the word “walked” is usually omitted, making it difficult to discern it from the present tense “walk”.
However, it is easier to recognize “walked” as a past participle because of its syntactical properties (i.e. it is usually accompanied with the words “have” or “by”), whereas simple past tense “walked” is phonetically indistinguishable from “walk” when used as a past tense verb. If the communication hypothesis were true, language evolution should favor easily discernible forms to those that are ambiguous, but the deletion rates for the past participle and for past tense do not differ in relevant contexts.
Therefore, modelling, just like comparative analysis, shows how limited our methods for language evolution analysis are, and how poor is our knowledge of our most distinguishing feature.
Original research article: Hauser MD, Yang C, Berwick RC, Tattersall I, Ryan MJ, Watumull J, Chomsky N and Lewontin RC (2014) The mystery of language evolution. Front. Psychol. 5:401. doi: 10.3389/fpsyg.2014.00401