Primitive stages in language development
In an earlier chapter have discussed language development with respect to age and onset. Now we must search for regularities within stages of development as well as regularities in the strategy of language acquisition regardless of the age at which this happens.
Our language is such an intimate aspect of our thoughts and behavior that it is difficult for us to obtain an objective appreciation of its mechanisms. The common description of its structure is a woeful oversimplification.
“Language consists of elemental units, that is, speech-sounds ; these are joined together into morphemes or words ; and words are put together into sentences. Speech-sounds have no meaning, but words do. They acquire meanings by an associative process in which the visual image of an object is linked to the sound of a word,”
The attractive simplicity of statements such as these has prevented many students of human behavior from recognizing the deep problems posed by the true structure of language and by the infant’s amazing ability to acquire these skills within two years time. A few examples may help to show some of the hidden problems.
Let us look first at phonological aspects. Suppose we build a typewriter that could print out (for instance, in international phonetic alphabet) anything that is spoken into it, hopefully in any language. What are the tasks of this machine ? In order for it to be successful it would have to be capable of a highly peculiar form of pattern recognition. For instance, it is not enough that it can recognize isolated examples of English speech sounds. The acoustic peculiarities of a given phoneme are never twice the same, and there are instances in which two acoustically very different sounds represent the same phoneme. Interpretation of a sound as one or another phoneme often depends on the acoustic context. Furthermore , phonemes are not simply linked as in a chain during production, but they affect each other , and the influence may work both ways : a later phoneme may be influenced by an earlier one or, because of anticipation, an earlier phoneme may be influenced by a later by later one. Acoustic analysis also shoes that speech sounds, instead of neatly following one another, may actually overlap either partly or wholly; by giving a vowel /o/ an/R/ coloring, or nasalizing it, or palatalizing it, an R-, η-, or y-quality is produced together with the vowel, but it may be interpreted, under certain conditions, as if it were following it. In addition, every speaker influences his utterances by idiosyncrasies due to the peculiar shape of his vocal tract and the peculiarities of his own motor-skills; this means that every utterance he produces undergoes acoustic transformations.
A machine that prints a graphic symbol for every acoustic feature heard in the sounds of any language would produce very unreadable records. Speech-recognition of a natural language involves focusing attention on specific types of phonetic relations or contrasts, and on ignoring many other acoustic phenomena as irrelevant for transmission of information in a particular language. Every natural language selects its own set of contrasts, and recognition of the sound patterns of one language is governed by its own peculiar set of rules. (Although all such sets of rules have common formal properties; see Chapter 9 and Chomsky’s Appendix A.) In order to be useful, the typewriter would have to be able to select the right set of rules for every input, in other words, it would have to be able to recognize one of several natural languages and then behave in accordance with this selection. Clearly, pattern-recognition of an acoustically operated typewriter is dependent upon rules which must take into consideration many factor and which must operate on much longer sound configurations than single or short speech sound segments. The relational patterns, even on the level of speech sounds, are very intricate. The reason why the many attempts to build such a typewriter have not been successful(except for very limited input) is probably not entirely due to technical limitations; it may be because of our ignorance of the proper rules that regulate speech perception.
Suppose we had solved all of these problems and had succeeded in constructing a typewriter that at least transcribes English. The device could not be considered a model of an infant who is developing speech. The problems to be solved by the infant whose task it is to “crack” the society’s communication code are much more complicated; he is biologically constituted to learn to recognize any type of natural language, and the acoustic input is usually not limited to one narrow topic.
When the intricacies of patterns and relations in the structure of language are discussed abstractly, it is often interjected that the child’s task might be simplified because he does not learn an abstract or formal system but is learning meanings at the same time, and this might facilitate the situation. Let us look into this more closely.
First, it is necessary to make a distinction between reference and meaning. Reference deals with the relationship between an individual word and some aspect or object of the physical environment. The problem of reference need not be considered here as they constituted the subject matter of Chapter 8. Meaning deals with the semantic interpretation of utterances (or written material), and these come invariably in the form of sentences in various states of completion or degrees of imperfection; this includes the special case of one-word sentences which we shall discuss presently. Because discourse is not encountered in anything but the essential form of sentences, and sentences can be interpreted only through grammatical analysis, meaning cannot be divorced form grammatical structure (Chomsky, 1957; Chomsky and Miller, 1963; Miller and Chomsky. 1963 )
Grammatical structure an not be understood as a phenomenon of transitional probability of specific lexical items or parts of speech. Something more intricate must be involved as may be seen from the following examples, suggested mostly by Chomsky. Consider the following strings of words:
The sentence is incomplete. How can you stop here without citing the examples?
語言由最基本單位構成,說話聲音和字詞跟音素的結合,然後字詞在結合成句子。雖然聲音是沒意義的,但是字有。藉由這兩著連結,使我們可以由聲音與字表達我們的意思。這讓我們好奇,人的最開始-嬰兒是如何獲的這些語言能力的?! 首先注視著音系學方面。特定音素的音響特色在不同次時是從未相同的，並且有二音響上非常不同的聲音代表同一音素的事例。 聲音的解釋或其他音素經常取決於上下文,他們之間彼此影響。且在字詞中,後一個音素容易受前面音素的影響̀,並非是完整的,原本的聲音的。語音聽學也顯示,說話的聲音並非完整的出現,會有部分獲完整的覆蓋。其中像是R-, η-, or y-quality就是利用特別的發音嘴型跟聲帶。語言識別自然語言介入的注意在於語音聯繫或對比的具體類型和在忽略許多其他音響現象如毫不相關為信息傳輸在一種特殊語言。任一個語言都有自己獨特的規則,然而我們想要用一個機器去,去辨認不同的語言,但是事實上,那是一件很困難,十分不容易的一件事,因為現在語言有太多種了,而每個語言的特色眾多,無法一一分別。這就是到現在許多嘗試修造這樣機器不是成功的(除了非常有限的輸入)。