Other languages, such as Turkish and many Native American languages, are called ‘agglutinative’. This means that each part of each word often has a single meaning, unlike Romance languages, where each part of the word can have several meanings, as in the ‘o’ of falo. A single word in Turkish can be long and have many parts, but each part has but a single meaning:
Çekoslovakyalılaştıramadıklarımızdanmışsınızcasına ‘as if you were one of those whom we could not make resemble the Czechoslovakian people’.
Some inflectional languages even have special kinds of morphemes called circumfixes. In German the past tense of the verb spielen ‘to play’ is gespielt, where ge-and -t jointly express the past tense, circumscribing the verb they affect.
Other languages use pitch to add meaning to words. This produces what are called simulfixes. In Pirahã the nearly identical words ?áagá ‘permanent quality’ vs ?aagá ‘temporary quality’ are distinguished only by the high tone on the first vowel á. So I could say, Ti báaʔáí ?áagá ‘I am nice (always)’ or, Ti báaʔáí ?aagá ‘I am nice (at present)’.
Or one could express some meaning on the consonants and another part of the meaning of the word on the vowels, in which case we have a non-concatenative system. Arabic languages are of this type. But English also has examples. So foot is singular, but feet is plural, where we have kept the consonants f and t the same but changed the vowels.
It is unlikely that any language has a ‘pure’ system of word structure, using only one of these strategies just exemplified. Languages tend to mix different approaches to building words. The mixing is often caused by historical accidents, remnants of earlier stages of the language’s development or contact with other languages. But what this brief summary of word formation shows is that if we look at all the morphological systems of the world the basic principles are easy. This is summarised in Figure 30.#
Figure 30: Types of language by word type
These are the choices that every culture has to make. It could choose multiples of these, but simplicity (for memory’s sake) will favour less rather than more mixing of the systems. From the beginnings of the invention of symbols, at least 1.9 million years ago by Homo erectus, there has been sufficient time to discover this small range of possibilities and, via the grammar, meaning, pitches and gestures of language, to build morphological systems from them.**
Arguably the greatest contribution Noam Chomsky has made to the understanding of human languages is his classification of different kinds of grammars, based on their computational and mathematical properties.2 This classification has become known as the ‘Chomsky hierarchy of grammars’, though it was heavily influenced by the work of Emile Post and Marcel Schützenberger.
Yet while Chomsky’s work is insightful and has been used for decades by computer scientists, psychologists and linguists, it denies that language is a system for communication. Therefore, in spite of its influence, it is ignored here in order to discuss a less complicated but arguably more effective way of looking at grammar’s place in the evolution of language as a communicative tool. The claim is that, contrary to some theories, there are various kinds of grammars available to the world’s languages and cultures (linear grammars, hierarchical grammars and recursive hierarchical grammars). These systems are available to all languages and they are the only systems for organising grammar. There are only three organisational templates for human syntax. In principle, this is not excessively difficult.
A linear grammar would be an arrangement of words left to right in a culturally specified order. In other words, linear grammars are not merely the stringing of words without thought. A language might stipulate that the basic order of its words is subject noun + predicate verb + direct object noun, producing sentences like ‘Johnsubject noun hitpredicate verb Billdirect object noun.’ Or, if we look at the Amazonian language Hixkaryana, the order would be object noun + predicate verb + subject noun. This would produce ‘Billobject noun hitpredicate verb Johnsubject noun,’ and this sentence would be translated into English from Hixkaryana, in spite of its order of words, as ‘John hit Bill.’ These word orders, like all constituents in human languages, are not mysterious grammatical processes dissociated from communication. On the contrary, the data suggest that every bit of grammar evolves to aid short-term memory and the understanding of utterances. Throughout all languages heretofore studied, grammatical strategies are used to keep track of which words are more closely related.
One common strategy of linking related words is to place words closer to words whose meaning they most affect. Another (usually in conjunction with the first) is to place words in phrases hierarchically, as in one possible grammatical structure for the phrase John’s very big book shown in Figure 31.
In this phrase there are chunks or constituents within chunks. The constituent ‘very big’ is a chunk of the larger phrase ‘very big book’.
One way to keep track of which words are most closely related is to mark them with case or agreement. Greek and Latin allow words that are related to be separated by other words in a sentence, so long as they are all marked with the same case (genitive, ablative, accusative, nominative and so on).
Consider the Greek sentences below, transcribed in English orthography, all of which mean ‘Antigone adores Surrealism’:
(a)
latrevi
ton iperealismo
i Antigoni
Verb
Object
Subject
adoreverb
the surrealismaccusative
the Antigoninominative
(b)
latrevi i Antigoninominative
ton iperealismoaccusative
Verb
Subject
Object
adoreverb
the Antigoninominative
the surrealismaccusative
(c)
i Antigoni
latrevi
ton iperealismo
Subject
Verb
Object
the Antigoninominative
adoreverb
the surrealismaccusative
(d)
ton iperrealismo
latrevi
i Antigoni
Object
Verb
Subject
the surrealismaccusative
adoreverb
the Antigoninominative
(e)
i Antigoni
ton iperealismo
latrevi
Subject
Object
Verb
the Antigoninominative
the surrealismaccusative
adoreverb
(f)
ton iperealismo
i Antigoni
latrevi
Object
Subject
Verb
the surrealismaccusative
the Antigoninominative
adoreverb
Figure 31: Grammatical structure
All of these sentences are grammatical in Greek. All are regularly found. The Greek language, like Latin and many other languages, allows freer word order than, say, English, because the cases, such as ‘nominative’ and ‘accusative’ distinguish the object from the subject. In English, since we rarely use case, word order, not case, has been adopted by our culture as the way to keep straight who did what to whom.
But even English has case to a small degree: Inominative saw himobjective, Henominative saw meobjective, but not *Me saw he or *Him saw I. English pronouns are marked for case (‘I’ and ‘he’ are nominative; ‘me’ and ‘him’ are objective).
We see agreement in all English sentences, as in ‘He likes John.’ Here, ‘he’ is third person singular and so the verb carries the third person singular ‘s’ ending on ‘likes’ to agree with the subject, he. But if we say instead, ‘I like John,’ the verb is not ‘likes’ but ‘like’ because the pronoun ‘I’ is not third person. Agreement is simply another way to keep
track of word relationships in a sentence.
Placing words together in linear order without additional structure is an option used by some languages to organise their grammars, avoiding tree structures and recursion. Even primates such as Koko the gorilla have been able to master such word order-based structures. It certainly would have been within the capabilities of Homo erectus to master such a language and would have been the first strategy to be used (and is still used in a few modern languages, such as perhaps Pirahã and the Indonesian language Riau).
One thing that all grammars must have is a way to assemble the meanings of the parts of an utterance to form the meaning of the whole utterance. How do the three words, ‘he’, ‘John’ and ‘see’, come to mean ‘He sees John’ as a sentence? The meaning of each of the words is put together in phrases and sentences. This is called compositionality, without which there can be no language. This property can simply be built off the phrases in a tree structure directly, which is the case for most languages, such as English, Spanish, Turkish and thousands of others. But it can also be built in a linear grammar. In fact, even English shows that compositionality doesn’t need complex grammar. The utterance ‘Eat. Drink. Man. Woman.’ is perfectly intelligible. But as the meaning is put together, it relies heavily on our cultural knowledge. The absence of complex (or any) syntax doesn’t scuttle the meaning.
Many languages, if not all, have examples of meaningful sentences with little or no syntax. (Another example from English would be ‘You drink. You drive. You go to jail.’ These three sentences have the same interpretation as the grammatically distinct and more complex construction, ‘If you drink and drive, then you will go to jail.’) Note that the interpretation of separate sentences as a single sentence does not require prior knowledge of other syntactic possibilities, since we are able to use this kind of cultural understanding to interpret entire stories in multiple ways. One might object that sentences without sufficient syntax are ambiguous – they have multiple meanings and confuse the hearer. But, as noted in chapter 4, even structured sentences like ‘Flying planes can be dangerous’ are ambiguous, in spite of English’s elaborate syntax. One could eliminate or at least reduce ambiguity in a given language, but the means for doing so always include making the grammar more difficult or the list of words longer. And these strategies only make the language more complicated than it needs to be, by and large. So this is why, in all languages, ambiguous sentences are interpreted through native speaker knowledge of the context, the speaker and their culture.
The more closely the syntax matches the meaning, usually the easier the interpretation is. But much of grammar is a cultural choice. The form of a grammar is not genetically predestined. Linear grammars are not only the probable initial stage of syntax, but also still viable grammars for several modern languages, as we have seen. A linear grammar with symbols, intonation and gestures is all a G1 language – the simplest full human language – requires.
The next type of languages are G2 languages. These are languages that have hierarchical structures (like the ‘tree diagrams’ of sentences and phrases), but lack recursion. Some have claimed that Pirahã and Riau are examples. Yet we need not focus exclusively on isolated or rarely spoken languages such as these. One researcher, Fred Karlsson, claims that most European languages have hierarchy but not recursion.
Karlsson bases his claim on the observation: ‘No genuine triple initial embeddings nor any quadruple centre-embeddings are on record (“genuine” here means sentences produced in natural non-linguistic contexts, not sentences produced by professional linguists in the course of their theoretical argumentation).’3 ‘Embedding’ refers to placing one element inside another. Thus one can take the phrase ‘John’s father’ and place this inside the larger phrase ‘John’s father’s uncle’. Both of these are embedding. The difference between embedding (which is common in languages that, according to researchers like Karlsson, have hierarchical structures without recursion) and recursion is simply that there is no bound on recursion, it can keep on going. So what Karlsson is saying is that in Standard Average European (SAE) languages he only found sentences like:
John said that Bill said that Bob is nice, or even
John said that Bill said that Mary said that Bob is nice, but never sentences like
John said that Bill said that Mary said that Irving said that Bob is nice.
The first sentence of this triplet has one level of embedding, the second has two and the third has three. But the claim is that standard European languages never allow more than two levels of embedding. Therefore, they have hierarchy (one element inside another), but they do not have recursion (one element inside another inside another … ad infinitum).
According to Karlsson’s research it does not appear that any SAE language is recursive. He acknowledges that one could argue that they are recursive in an abstract sense, or that they are generated by a recursive process. But such an approach doesn’t seem to match the facts in his analysis. In other words, Karlsson claims, to use my terms, that all of these languages are G2 languages. Karlsson’s work is therefore interesting support from modern European languages that G2 languages exist, as per the semiotic progression. Again, these are languages that are hierarchical but not recursive. To show beyond doubt that a language is recursive we need to show ‘centre-embedding’. Other forms are subject to alternative analyses:
(a)Centre-embedding: ‘A man that a woman that a child that a bird that I heard saw knows loves sugar.’
(b)Non-centre-embedding: ‘John said that the woman loves sugar.’
What distinguishes (a) is that the subordinate clauses, ‘that a woman that a child that a bird I heard saw knows’ are surrounded by constituents of the main clause, ‘A man loves sugar’. And ‘that a child’ is surrounded by bits of the clause it is embedded within ‘a woman knows’, and ‘that a bird’ is surrounded by parts of its ‘matrix’ clause ‘a child saw’ and so on. These kinds of clauses are rare, though, because they are so hard to understand. In fact, some claim that they exist only in the mind of the linguist, though I think that is too strong.
In (b), however, we have one sentence following another. ‘John said’ is followed by ‘that the woman loves sugar’. Another possible analysis is that ‘John said that. The woman loves sugar.’ This analysis (proposed by philosopher Donald Davidson) makes English look more like Pirahã.4
The final type of language I am proposing in line with Peirce’s ideas, G3, must have both hierarchy and recursion. English is often claimed to be exactly this type of language, as previous examples show. As we saw earlier some linguists, such as Noam Chomsky, claim that all human languages are G3 languages, in other words, that all languages have both hierarchy and recursion. He even claims that there could not be a human language without recursion. The idea is that recursion is what separates in his mind the communication systems of early humans and other animals from the language of Homo sapiens. An earlier human language without recursion would be a subhuman ‘protolanguage’ according to Chomsky.††
However, the fact remains that no language has been documented in which any sentence is endless. There may be theoretical reasons for claiming that recursion underwrites all modern human languages, but it simply does not match the facts of either modern or prehistoric languages or our understanding of the evolution of languages.
There are many examples in all languages that show non-direct connection between syntax and semantics. In discourses and conversations, moreover, the meanings are composed by speakers from disconnected sentences, partial sentences and so on. This supports the idea that the ability to compose the larger meanings of conversations from their parts (or ‘constituents’) is mediated by culture. As with G1 languages and people suffering from aphasia, among many other cases, meaning is imputed to different utterances by loosely applying cultural and individual knowledge. Culture is always present in interpreting sentences, though, regardless of the type of grammar, G1, G2, or G3. We know not exa
ctly which of these three grammars, or all three, were spoken by Homo erectus communities, only that the simplest one, G1, would have been and still is adequate as a language, without qualification.
It should ultimately be no surprise that Homo erectus was capable of language, nor that a G1 language would have been capable of taking them across the open sea or leading them around the world. We are not the only animals that think, after all. And the more we can understand and appreciate what non-human animals are mentally capable of, the more we can respect our own Homo erectus ancestors. One example of how animals think is Carl Safina’s book Beyond Words: What Animals Think and Feel. Safina makes a convincing case that animal communication far exceeds what researchers have commonly noticed in the past. And other researchers have shown that animals have emotions very similar to human emotions. And emotions are crucial in interpreting others and wanting to communicate with them, wanting to form a community. Still, although animals make use of indexes regularly and perhaps some make sense of icons (such as a dog barking at a television screen when other dogs are ‘on’ the screen), there is no evidence of animals using symbols in the wild.‡‡
Among humans, however, as we have seen, there is evidence that both Homo erectus and Homo neanderthalensis used symbols. And with symbols plus linear order we have language. Adding duality of patterning to this mix, an easy couple of baby steps get us to ever more efficient languages. Thus possession of symbols, especially in the presence of evidence that culture existed, evidence strong for both erectus and neanderthalensis, indicates that it is highly likely that language was in use in their communities.
How Language Began Page 25
