Learning to Communicate (3)

How Does It Happen?

A few days after birth, infants show signs of being able to discriminate between their native language and a foreign language and even between some pairs of foreign languages. Tests such as low-pass filtering and syllable scrambling have shown that utterance-level prosodic information (such as changing pitch or volume) allows infants to make such distinctions (Guasti 2004). Apparently, the earliest linguistic information that children acquire pertains to whole constructions, generally at the utterance level. No information about constituent words or phonemes is present at this point, only a very basic representation of constructions as tone contours. This makes sense as such constructions vary a great deal in their words and phonemes but stay relatively consistent in their tone contours. As weeks go by, infants seem to lose their ability to distinguish foreign language pairs (Guasti 2004). Early on, the infants probably lacked a sufficient amount of linguistic input to identify their native language with much certainty. If given a choice between two streams of audio, an infant might choose the one that sounds like it could be the target language.

Babbling is an important first step in language production. It gives infants a chance to practice forming the phonemes, syllables, and tone contours of the language they hear around them. Many adult (or semi-adult) words can be produced as a simple by-product of babbling ("mama", "dada", "baba", etc.). According to O'Grady (2005), "[. . .] 'mama'-like sounds have been detected in children's vocalizations starting from as early as two weeks of age up to around five months, usually in a 'wanting' context (wanting to be picked up, wanting food, and so on)." It seems that as young as two weeks old, many children are able to understand adult speech for its most basic intentional meaning: "Pay attention to me!" The simple act of parroting back known syllables with the goal of gaining someone's attention represents a very basic form of semantic knowledge.

In order for children to gain a better idea of what exactly is being said, certain strategies, or "operating principles", have been suggested (Aitchison 1983). Among them are instructions to associate only one form with each unit of meaning, pay attentions to the order of words, and avoid interruptions. Another obvious cue is prosodic stress; the most salient words in a sentence are often the most vital to its meaning. Young children seem especially good at understanding words referring to objects in their current attentional frame, which account for many of their first semantic associations (Tomasello 2003). There seems to be an inborn bias urging children to look for objects, as described by Elizabeth Spelke (Bloom 2002). Such objects possess cohesion, continuity, solidity, and (for inanimate objects) contact-driven motion. This object bias even seems to exist in such verb-heavy languages as Japanese and Korean. Tomasello feels that the object bias actually comes from the child's theory of mind, and Quine goes so far as to suggest that we get our very idea of objects from language (Bloom 2002). This seems a bit strong, but language certainly seems to influence our perception of reality. Bloom gives the example of one third of a line being called a "zoop" and the rest of it a "moop". If you were then shown three equally spaced dots under the line and told to put them into two groups, you would most likely group the two "moop" dots together, though no perceptual rationale exists for this decision.

Based on children's natural speech, there seem to be two basic styles of word learning: analytic and gestalt (O'Grady 2005). Children using the analytic style look for salient words in the utterances they hear and often assign them more meaning than they would possess in normal adult speech. For example, the word "up" could be used to mean, "Pick me up!" Those of the gestalt style attempt to mimic entire adult utterances. This results in unanalyzed chunks, which may sound fairly adult-like, as in /gimidat/ for "Give me that!" The difference is that such an utterance consists of a single morpheme in the child's lexicon. It may be a long time before the child figures out that /gimidat/ can be broken down as "Gimme that!" or even "Give me that!"

According to Tomasello (2003), children start by learning a few performatives (such as "hello", "please", and "no"), add a large number of nouns, and finally move on to verbs and modifiers. It is actually not true "nouns" and "verbs" that the child is learning but merely words referring to objects and words referring to actions. Adult-like categories emerge only later after a sufficient number of constructions are acquired. And even in adult language, many words, such as "kiss", actually belong to two or more of these categories at once. Children between the ages of 12 and 18 months are generally capable of uttering only one word or holophrase at a time (Tomasello 2003). In choosing which word to produce they follow the "Informativeness Principle" and choose the word that would most clearly convey their intended message (O'Grady 2005). Because subjects often consist of information understood by context, they are generally omitted.

Early constructions are item-based, the earliest containing only one variable element. This might be called the "two-word stage", but in the idiosyncratic language of young children, "words" are especially difficult to identify. The length of a child's utterance is traditionally measured by counting its morphemes (as defined by the adult language), which simply doesn't work for unanalyzed chunks and the gestalt style of learning. A better indication of the number of morphemes present in a given utterance comes from careful analysis of the child's naive grammar. Full utterances clearly possess meaning, so anything uttered in isolation is sure to comprise at least one morpheme. Furthermore, any element of a construction that the child replaces or omits with the apparent intention of affecting the meaning of the utterance as a whole can be said to possess meaning and comprise at least one morpheme. It is not a given that when a string of syllables meets the criteria for morpheme-hood in one construction, it will in other constructions.

Item-based constructions (usually centered around verbs) linger on in child speech long after the two-word stage. Children under three generally have trouble with tasks requiring the use of generalized verb patterns, but at around the age of four, children come to realize that verbs in English tend to follow the general pattern of "subject + verb + direct object" (O'Grady 2005). Many words such as "because", "tomorrow", "morning", and "of" are used only in frozen or at least formulaic phrases long before their meaning is extracted through generalization across such phrases (Tomasello 2003).

Once the basic constructions of a language have been learned and major patterns have been discovered through analysis of these constructions, all that remains is to increase vocabulary size and correct the many minor imperfections undoubtedly present in the child's speech. One method of correction that parents seem to find intuitively appealing involves repeating what the child has just said but adjusting it to be grammatically acceptable. However, such recasts have been shown to be effective only when the child is already using the desired form at least half the time (O'Grady 2005). A type of error correction that might be able to get children to this halfway mark is "indirect negative evidence" (Pinker 1996). The conspicuous absence of a particular form may be enough to make children avoid using it. To some degree, this process of error correction continues into adulthood.

Another way a person's language might be adjusted is through an imbalance in linguistic trust. During a lecture to a full classroom, one of my professors recently produced the sentence "They developed that own code of behavior themselves." I found this to be an interesting construction and decided to write it down. As I did so, a student in the class responded to what the professor had just said with "According to that own philosophy . . ." followed by a jumble of words, which I can't remember. This situation illustrates how even an adult speaker's grammar can be subconsciously affected by others. The professor most likely wanted to convey the idea that a certain group of people developed their own code of behavior while stressing that it was the same code of behavior just mentioned. In general, the phrase "that own" is simply not used in this way. "Own" must be preceded by a possessive such as "their", "his", or "our" to make sense to the average listener. The addition of the word "themselves" to the end of the sentence served to give "own" meaning but felt clumsy and unplanned. However, because the professor was in a position of intellectual authority compared to the student, the "that own" construction was not viewed as a mistake by the responding student and was in fact (if only for a short while) incorporated into the student's own grammar.

If language acquisition is described as the process of learning the words and constructions of a given language, then a language is never fully acquired. Not only are there simply too many words in a given natural language for any one person to know them all, but new words, and occasionally new syntactic constructions, are added all the time. And because of constant shifts in accepted usage, even the formal speech of well-educated adults may be challenged as ungrammatical. For a person to speak any language well, ey must never stop learning to communicate.


Labels: , , , , , ,


Post a Comment