English as a second language and naturalistic learning

Inhalt

1 Introduction

2 L1 Acquisition – An overview
2.1 Chomsky’s Universal Grammar
2.2 Parameters
2.3 Developmental sequences, stages and patterns
2.4 Motherese
2.5 Conclusion

3 UG in Second language acquisition
3.1 The acquisition of German negation in a naturalistic setting
3.2 The accessibility of UG in Second Language Acquisition
3.3 The effects of age on UG
3.4 Parameter- Resetting in L2 Acquisition
3.5 Conclusion

4 The Acquisition of the L2 Lexicon

5 L2 acquisition environments

6 Naturalistic learning vs. formal instruction
6.1 Implicit vs. explicit language knowledge
6.2 Monitoring and the vernacular / careful style
6.3 Transfer / Errors
6.4 Input, output and formulas
6.5 Communication in formal language learning contexts
6.6 Conclusion

7 Learner effects on L2 acquisition

8 Problems in L2 acquisition research

9 English as a Second Language in Germany
9.1 Teaching methods
9.2 Primary schools
9.3 Secondary schools
9.4 Conclusion

10 Final conclusion

1 Introduction

“How do we learn language ?” is both a very interesting and a very complex question. It has fascinated people for centuries to find an answer to that question, but so far only competing theories have come up. In 1965, Noam Chomsky invented the theory of a Universal Grammar underlying the structures of all languages and that babies were born with innate knowledge of this Universal Grammar. The idea has revolutionised language acquisition research. But only in the 1970s did researchers start to look at the acquisition of second languages as well as the first language acquisition. Since then, studies, theories and new research fields have sprung up at an amazing rate.

In this essay, I will try to give an account on what second language acquisition research has found out so far and in what way these findings could probably be applied to improve the way pupils learn English in Germany.

It will be necessary to differentiate between naturalistic and instructed language learning. Naturalistic language learning takes place if people are not explicitly taught the foreign language, but rather learn it by trying to communicate in it. Most studies relating to naturalistic language acquisition observe immigrants’ children who do not yet go to school. Instructed language learning takes place in environments in which the learner is explicitly taught grammar and vocabulary of the foreign language, as in a classroom.

To make second language acquisition in general understandable, chapter 2 will first take a look at the processes taking place in first language (L1) acquisition. Afterwards, it will be discussed whether the same processes take place in second language (L2) acquisition.

Everybody will agree that while L1 is a universal process (i.e. all children learn a language) and always endows children with full competence (i.e. they are all able to communicate fluently in that language without a foreign accent or any other inhibitions), L2 acquisition is more often than not a voluntary activity and usually L2 learners do not attain full competence. The reasons for this are to be found in the radically different language acquisition environments. An L2 may be learnt at school or ‘naturalistically’, i.e., without getting the language explained explicitly. While babies must acquire their language, pupils might not be especially motivated to learn a foreign language and might reject it.

These variables and many more have a great influence on the outcome of L2 acquisition, as chapters 6 and 7 will show. Chapter 6 will mainly deal with what formal instruction does to L2 acquisition, and chapter 7 illustrate the influences of the individual learner on it.

But before that, it will need to be established whether or not L1 and L2 acquisition have anything in common at all. Chapter 3 will consider that.

Facts put forward in this essay mostly remain theoretically hypothesised facts, supported by a huge body of evidence. It must nevertheless be mentioned that second language acquisition research experiences considerable problems in the collection and interpretation of data in its studies. Most of the data can be interpreted in more than one way, and thus every proposition proved by empirical study must still be taken with a grain of salt. I will briefly address research problems in chapter 8.

Chapter 9 will then consider in which way if any second language acquisition research can be practically applied for English teaching in Germany. I will take a short look at the current situation in this country and then propose a recommendation of how optimal second language teaching in Germany might work.

It will become apparent during the analysis that I have mostly refrained from explaining explicitly naturalistic second language acquisition. A sufficiently clear picture ought to be formed in juxtaposing the effects of formal instruction against it. An explicit analysis would not only exceed the scope of this essay, but it would also mean repeating the points made in the analysis of instructed language learning.

2 L1 Acquisition – An overview

L1 Acquisition – the study of first language acquisition – looks at how children learn to speak their mother tongue. This is an especially fascinating field of enquiry, as all children, regardless of where they live, their general cognitive capability, or their social circumstances, acquire their mother tongue completely. Moreover, L1 acquisition is not a voluntary undertaking, as the mother tongue must be learnt (cf. Pinker 1994; Felix 1982).

There are basically two major schools of thought in L1 acquisition, the empiristic and the nativistic. The empiristic school, whose main defendants were, for example, Skinner, Osgood, and Hull, proposes that language learning was one of many manifestations of a general learning ability or the adaptation of an organism’s behaviour. The nativistic position, which is often connected with names like Chomsky, McNeill, or Lenneberg, proposes a special, innate, language device in the biogenetic structure of a human being, which enables children to acquire their mother tongue in a unique way. Since it was formed, the nativistic position has massed a huge body of evidence to support it and has been accepted by more and more linguists. In this chapter, I will attempt to present a short overview of L1 acquisition, according to the nativistic position.

First language acquisition basically begins with the child’s birth. However, the baby will at first neither be able to speak nor to listen to language in a systematic way. It takes some training of the auditory system first to be able to distinguish the sounds of language. The same is true for the oral production system – babies are not able to produce distinct phonemes immediately after birth. They do have, on the other hand, the ability to produce any sound of any natural language in the world. It takes roughly six months for babies to turn from universal phoneticians to language-specified. (cf. Pinker 1994). Before starting to speak, babies usually go though a ‘silent period’ before a babble period. During this period they start making apparently uncontrolled noise, but in reality during this period they gain sufficient control over their voice-producing system that they are able to pronounce the sounds of their language correctly. An amazing fact about this period is that deaf children go through it in the same way that non-pathologic children do. Pinker (1994) even found out that deaf children may ‘babble’ with their hands in order to acquire a sign language. Moreover, as Felix (1982) states, deaf parents do not have an influence on the phonetic development of their child. Therefore, the ‘babble phase’ is independent of stimuli.

Only after the babble phase do researchers talk about L1 acquisition. Felix puts the time when L1 acquisition starts between the 18th and the 26th month of age. At that time, children start to produce words. At first, they start with grammarless one-word utterances only, but they may also produce language chunks like “What is that?” which they have heard frequently and the meaning of which is understood by them. According to Pinker (1994), the one-word-stage lasts from two months to one year. The words produced by children are usually words adhering to the here and now principle, that is, names for persons or objects the child can see and perhaps even touch. This has mainly to do with the cognitive state of development of the child; he cannot conceptualise abstract thoughts of objects yet.

The two-word-stage is the next big step for the child in L1 acquisition. During this period, children must start using grammar to make their utterances meaningful. Lexically speaking, children still mostly learn words and names for objects that are ‘here and now’, but the lexicon rapidly expands. Pinker (1994) found out that the words they learn are almost all word-for-word translatable into all languages. For him, the most important feature of the two-word-stage is, however, the grammaticality of the children’s utterances. The underlying word order agent → action → recipient → object → location was found to be true for more than 95% of all the children’s utterances. The children enunciate those words they find semantically most ‘unstable’, that is, the content of which they think cannot be guessed from context or extralingual devices such as mimes or gestures.

At the age of about three years, which is usually the time when the two-word-stage is over (Pinker 1994), children rapidly expand both their lexicon and their grammar. Lexically speaking, they are now able to acquire words that do not adhere to the here and now principle, as their state of their general cognitive abilities has developed further. At this point, lexical learning has reached its peak in learning rate and will continue at that rate until the children are about six years old. By that time, their lexicon will, on average, include 13,000 words (Pinker 1994: 150). That means that for more than three years children will have learnt one new word every waking hour.

In addition to a rapidly expanding lexicon, children must also acquire the morphology, the syntax, and the pragmatics of their mother tongue. The first semantic rules children follow can be observed during the two-word-stage, but they start acquiring morphologic rules only after. As this is a fairly easy process to observe, many language acquisition studies have observed the changes in the children’s morphologic system. What the researchers found out is one of the main pillars of the nativistic position: children make errors that they cannot have heard from their parents. For example, they may use plural forms like ‘sheeps’ and ‘mouses’ which they will not have heard any native speaker say. In fact, those mistakes happen because as children acquire rules, such as the plural –s rule in English, they overgeneralise the use of the rule, applying it to any noun in their lexicon. The same phenomenon tends to be observable with any other rules the children acquire, and thus Ellis follows de Villiers and de Villiers in their assessment:

Thus, the acquisition of forms such as ‘went’ follows a U-shaped pattern of development, with children first using it correctly (for example, ‘went’) and then incorrectly (for example, ‘goed’) before they finally once again produce the correct form (‘went’). (de Villiers and de Villiers (1973) in Ellis (1994: 77))

Peltzer-Karpf and Zangl (1998) also illustrate this phenomenon, describing the different states of rule-acquisition:

Abbildung in dieser Leseprobe nicht enthalten

Table 2.1 The states of Rule-acquisition (Peltzer-Karpf and Zangl 1998: 7)

They state that during the Initial state and the Quasi stable state structures are used quite reliably, but that the utterances that do conform with the target-language norm are usually language chunks, i.e. unanalysed bits of language they may have heard in native speaker’s speech. The Turbulent state and the Intermittent state are mainly marked by many errors occurring in the children’s speech, which is caused by the instability of the learner’s rule system. In other words, children hypothesise rules in those states, trying to confirm them against the input they hear. If not quite random trial and error, the process is very similar to trial and error. I will later in this chapter explain the processes taking place in these states in more detail. The Partially ordered state, the Ordered state and the Coherent state, finally, are marked by a reduction in the learner’s errors, and a reduction in overgeneralisation. By then, the children will have found what is an acceptable rule system, and they gradually learn to use it and to apply it correctly. Simultaneously, they are able to identify exceptions to the rules (e.g. ‘went’) and to use exceptions correctly and coherently.

2.1 Chomsky’s Universal Grammar

This in itself does not automatically lead to a nativistic position, but in combination with the fact that learners do not typically receive negative evidence, i.e. correction, it makes a good argument for the nativistic position. If there were no special ‘language device’ to help children recognise that ‘goed’ or ‘sheeps’ are wrong, how could they so uniformly arrive at the target-language (TL) rules? To solve problems like, this, Chomsky invented the notion of a Universal Grammar in the 1960s. Universal Grammar (UG) is defined as a set of principles and parameters which enable human beings to acquire a language. If there existed a principle saying ‘There can only be one plural form of a morpheme’, the child would automatically by hearing ‘children’ know that ‘childs’ is wrong.

There is, of course, much more evidence pointing to the existence of a language device, and thus to UG. There is, for example, the poverty of stimulus argument. As Krosse (1992) points out, the linguistic competence of the L1 learner far exceeds the language he has been exposed to. This argument proposes UG, because it proves the creativity of child language despite the lack of any rule explanation. The process of rule-acquisition, therefore, must be subconscious and universal. Pinker (1994) also found out that rule-application is natural.^[1]

Another example of the principles proposed by UG researchers is the subset principle. According to Krosse (1992), children automatically assume the most restrictive grammar or rule to be the appropriate one if they are not exposed to positive evidence, i.e. input including a less restrictive grammar. In other words, the rules children create and test are rather like ‘To form a plural noun, add an –s at the end of the word’ than like ‘To form a plural, add an – s to the end of the word or do not add anything to it’. While the second variation of the rule would account for more words, and be more evident in the children’s input as they hear phrases like ‘two sheep’, they rather hypothesise the first rule to be the appropriate one. This principle is enormously important. If children thought the second variation of the above rule to be true, they could never deduce from input alone that ‘sheeps’ is wrong. A grammar including a rule like the second variation is called a superset grammar, always leaving the subset grammar a smaller part of the superset grammar.

2.2 Parameters

I have now explained some UG principles, but UG also comprises a set of parameters. ‘Parameters may be defined as general organization constraints on possible structural variations across languages, their settings mediating between what is innately specified and what is environmentally determined.’ (Krosse 1992: 60) Thus, parameters refer to the individual language’s unique grammatical properties. A good example of a parameter is the so-called pro-drop parameter. ‘The pro-drop parameter, … , ‘determines’ whether the subject of a clause can be suppressed (Chomsky 1988:64). It has just two settings…’ (Ellis 1994: 431). In Spanish or Italian, the subject may be dropped, rendering grammatical sentences like Está el president des los Estados Unidos. In English or German, for example, a sentence like this, * Is the President of the United States, is ungrammatical. Thus, the pro-drop parameter must be set to the [+] value for Spanish or Italian children, while English or German children need to set it to [-].

There are, however, other parameters which have more than two values. A word-order-parameter, for example, would need to account for the values SVO, SOV or VSO. Values are ranked for their m arkedness. The values will range from the least marked to the most marked value. Markedness determines, for example, which parameter setting will be initially adopted by children until they switch to the appropriate value for their language. Krosse (1992) states that

The subset principle specifies that given a choice between two parameter values, where one value generates a language which is the subset of the other, a child will initially assume the value which generates the subset (i.e. smaller) language, for if a child were to pick the superset (i.e. broader) language, he would not be able to correct overgeneralisations, that is, he would not be able to ‘unlearn’ an overgeneral language generated by the marked value, given only positive data. For this reason, the child’s initial hypothesis, or unmarked case, is that the narrower language applies to any language being learnt; (Krosse 1992: 9)^[2]

In this case, unmarked values or structures are those which are at least near the UG core, which Ellis (1994) defines as including absolute universals and universal tendencies. Absolute universals are linguistic properties which can be found in all natural languages, universal tendencies are properties that can be found in most natural languages. (cf. Ellis 1994: 417) Absolute universals, therefore, represent the core and universal tendencies the periphery of UG. Ellis now proposes that the nearer a linguistic property, i.e. parameter value, is to the core, the more unmarked the structure is.

This outlook on parameters and markedness presents a new view on L1 grammar acquisition. For example, the rule-testing as described above is not random at all but rather a structured process. Felix (1982) states: ‘die Bandbreite der für das Kind a priori plausiblen Hypothesen scheint fest umrissen zu sein’ (1982: 66). According to Krosse’s (1992) statement that positive evidence is required in order to set a parameter to a more marked value, and the fact that the subset principle necessitates always the most unmarked setting to be the default setting, the process of hypothesis-testing appears to be as follows: The learners hypothesise the most unmarked setting of a parameter to underlie a rule subsumed under a more or less marked parameter setting. They then try out this specific rule and either accept or reject it. If they reject the rule subsumed under the most unmarked setting, they will create the rule according to second most unmarked setting, try this one out, and either accept or reject it. This process will repeat itself until the learners have found an appropriate rule. It can be a very fast process, though. If children have already recognised a certain structure in their input, they can reject some of the rules created according to more unmarked values on the basis of this at once. They merely need to test those hypotheses that would accept the structure they have heard in the input.

There is, however, an additional factor that must be calculated here. The mechanics I have described above can only account for those rules subsumed under settings of parameters concerning absolute universals. Rules like the plural –s in English, that is rules concerning language-specific features, will not be based on parameters the way core UG rules will be. However, as input hints strongly at how the rules must be formed, they might be hypothesised parameter-independent. Current language acquisition research cannot answer whether or not all rules must be based on parameters, i.e. the extent to which parameters can define a language. However, the mechanism of acquisition described above would account for rules deriving from, for example, the pro-drop parameter.^[3]

It remains to be investigated now whether or not the mechanics of rule hypothesising about language specific feature rules is in any way qualitatively different. Most importantly, there might be no parameter underlying a rule like plural-formation. If there were a parameter completely subsuming plural-formation, it would have to include all possible morphological affixes found in any natural language. This does not appear to be very plausible. Moreover, even if all affixes were parameter values of that specific parameter, it would be difficult to find out if and why the Italian plural suffix – i would be differently marked than the English – s plural marker. It is, therefore, probably safe to assume that surface-structures like plural formation might not be parameter-directed. However, the input children hear will be sufficient to hypothesise only a limited amount of rules they have to test. It does not appear plausible that English children should, for example, hypothesise ‘add an – i ’ to be an appropriate rule. Thus, the rule-finding process for surface-structure rules might be different from the deep-structure rule-finding process. The deep-structure rule-finding process, which is the more interesting one for language acquisition researchers, still appears to be systematically guided by innate principles.

2.3 Developmental sequences, stages and patterns

The grammatical development of L1 learners does not proceed at a steady rate or at a steady pace. Felix claims that naturalistic language acquisition underlies systematic developmental sequences at different stages. Peltzer-Karpf and Zangl (1998) identify three phases in a sequence: (1) memorized, unanalysed chunks, (2) incomplete analysis, and (3) complete analysis. These phases involve five different mechanisms: (1) figure-background separation, (2) segmenting the input into groups, (3) extraction of marked features, (4) discovery of rules and categories, and (5) organization of function-dependent hierarchies. Two main processes are used in the mechanisms: (1) top-down (i.e. the splitting up of memorized chains (longer chunks)), and (2) bottom-up (i.e. successive construction of single elements to lengthier combinations). The three phases Peltzer-Karpf and Zangl propose can be subdivided further into the seven states in table 2.1. The five mechanisms all have to work hand-in-hand to allow the learner to arrive at a working rule system, as they all do. The children have to recognise what constitutes a phrase with semantic content. They must then seek out the grammatical features that express the special semantic property. Moreover, they have to compare many of those features to arrive at a hypothesis they can test. The fifth mechanism Peltzer-Karpf and Zangl mention determines in which order the features discovered are applied (for example, the plural – s is applied before adding a genitive case marker – s).

The sequences are called systematic because they only allow a certain order of appearance for the structures. In trying to get to a target-language structure c, the sequence a → b → c must have been run through. There may be children who can almost ‘jump’ from a to c, but the sequence itself cannot be altered.

The different stages, however, allow for some flexibility. The term stage incorporates two dimensions: a temporal and a structural dimension. This means that for some time during the acquisition process only certain structures can be acquired. Thus, during a stage X, in which structures 1, 2 and 3 are acquired, structure 4 cannot be acquired if 4 is to be acquired in stage Y. The order of acquisition of 1, 2 and 3 may vary, however. After 1, 2 and 3 have been acquired, the learner will be ready to proceed to stage Y. Proceeding from one stage to another can more adequately be described as a ‘jump’ than a smooth transition. Still, children do not proceed from one stage to the next as soon as the acquisition process in the first stage is accomplished. Marchman and Bates (1994) proposed in their critical mass hypothesis that stage-transition could only take place if a certain amount of input pointing to the features to be acquired in the next stage had been collected. This would account for some ‘stops’ in the acquisition process and further underline the importance of rich input.

In conclusion, the process of first language acquisition can be split up into discrete stages, during which the learner acquires certain structures going through systematic developmental sequences. The order of acquisition may vary within a stage, but never among different stages. After having acquired all structures of one stage, the learner rapidly proceeds to the next stage. The process of language acquisition proceeding along discrete stages alone makes language acquisition a unique learning process unlike any other learning processes which are marked by a continuous growth of knowledge and ability. As a result of the theory of developmental stages, developmental patterns can be found in the natural acquisition of a language. Ellis (1994) gives a good example of the developmental pattern based on information provided in Johnston and Pienemann (1986):

Abbildung in dieser Leseprobe nicht enthalten

Table 2.2: A developmental pattern (Ellis 1994: 105).^[4]

2.4 Motherese

I have so far neglected to discuss one major variable in first language acquisition which is nevertheless enormously important, the input. Input can be any language the learner gets to hear (in second language acquisition, to read as well). Basically, any contact with the target language may be called input. There are, however, many drawbacks on so wide a definition of input. First of all, evidence from Long (1983), Larsen-Freeman and Long (1991) and Krashen (1985; 1989) ‘supports the commonsense assumption that learners need to understand input in order to learn from it.’ (Ellis 1994: 278). Moreover, a child cannot retain and use all language he is confronted with for language acquisition. Therefore, Krashen (1985) has started to distinguish between input and intake. Intake is that part of the input the child might use for language acquisition. Especially in L1 acquisition, this is usually a special kind of language. ‘The register that results has been referred to variously as ‘baby-talk’, ‘motherese’, ‘caretaker talk’ … and ‘child-directed language’.’ (Ellis 1994: 247) Adults talking to babies and young children normally adapt their language to use simpler grammar and words that are thought to be easy to understand. Caretaker talk has long been thought of as poor language, but Ellis found ‘Three main features of caretaker talk: (1) It is more grammatical than speech addressed to adults, (2) It is simpler and (3) It is more redundant.’ (Ellis 1994: 251) This means that motherese or caretaker talk actually enhances possible intake for the child. Pinker (1994) underlines the necessity of motherese and goes on saying that TV, for example, cannot supplant human ‘input sources’ as has been studied in families with deaf parents.

Earlier in this chapter I also stated that children obtain little negative evidence. It is true that children are hardly ever corrected for grammatical errors, but ‘Brown and Hanlon (1970) have shown that mothers do correct children to ensure that what they say is true. … Mothers also pay attention to their children’s pronunciation of words and draw their attention to politeness, formulae (for example, when to say ‘thank you’).’ (Ellis 1994: 250). Therefore, children need input that fulfils the three features of caretaker talk. Higher grammaticality means that motherese is a reliable source of input for the children. Its being more simple means that it is richer in simple grammatical structures and features and thus will provide the child with more grammatical structures he is ready to acquire or should start to acquire soon, as seen in the developmental stages plan. Motherese being more redundant is an important factor as well. As I have stated above, a child needs to understand input to learn from it. Motherese, therefore, is exactly the kind of input that helps the child acquire the mother tongue.

2.5 Conclusion

First Language Acquisition is a unique process during which every child (in non-pathologic cases) learns his mother tongue and, in that tongue, achieves full competence. The result is a language that is at least as productive, communicative and rich in grammar as the parents’ language.^[5]

In the process of L1 acquisition the learner proceeds along developmental stages. In one of these stages, only certain structures are acquired and only when all of them have been acquired will the child proceed to the next developmental stage. In order to acquire the different structures in a stage, the child runs through developmental sequences. The sequences are marked by their systemacity and rigidity. They do not vary in their order.

During one of those sequences, a child starts at the Initial state, and then goes on to a turbulent phase. This phase is marked by high error rates in the child’s speech. In that phase the child creates rules and tests those rules against the intake he gets and the language he already knows. The rules the learner hypothesises to be true are not randomly selected rules, but they are different parameter value settings for deep-structure rules. Initially starting with the most unmarked setting, the child will then try out whether the setting is appropriate in the language it is about to acquire. If this is not the case, then the child will try out the second most unmarked setting and try it out. This process goes on until the correct setting has been found.

As the most unmarked setting of a parameter is the setting which will create a subset grammar of the grammars created by the other settings, the child has no need of negative evidence. Positive evidence alone, that is listening to grammatically correct language, is sufficient for the child to acquire the L1. And the way adults usually talk to children, in a language variety that is often called motherese, provides the child with exactly the input they require.

The lexicon, as it grows in a linear function takes up a special position in L1 acquisition, can be described as a ‘pace-setter’ (Peltzer-Karpf and Zangl, 1992: 107). Language production, which plays as important a role as input, would not be possible without the continuously extending lexicon. This, however, depends both on cognitive development and encyclopaedic knowledge.

This language acquisition process is a unique way of learning. It is different from any other way of cognitive learning. It is aided by the humans’ special, innate language device or UG, a set of principles and parameters that enable children to learn to speak the way they do. Ellis (1985) sums up the conclusions regarding L1 acquisition:

1. Language is a human-specific faculty.
2. Language exists as an independent faculty in the human mind (i.e. it is sep­arate from the general cognitive mechanisms responsible for intellectual development).
3. The primary determinant of acquisition is the child's 'language acquisition device', which is genetically endowed and provides the child with a general set of principles about language which can be used to discover the rules of a particular TL.
4. Input data are required to trigger the process by which the 'language ac­quisition device' discovers the rules of the TL.” (Ellis 1985, cited in Ellis 1994: 81)

3 UG in Naturalistic Second Language Acquisition

The study of second language acquisition is a much wider field of study than the study of first language acquisition. While first language acquisition happens regardless of motivation, L2 acquisition is a voluntary undertaking in many cases. Moreover, L1 acquisition happens during early childhood (with very few exceptions), while L2 acquisition occurs across the whole age spectrum. Furthermore, second language acquisition can take place in a naturalistic way or in an instructed way, or a mixture of both. All of these differences make the study of second language acquisition a much broader field than the study of first language acquisition.

The most striking difference to be found between L1 acquisition and L2 acquisition is the outcome: while all L1 learners achieve full competence, most L2 learners will not reach high levels of competence. It will be the central part of this chapter to find out at least some reasons for this fact. It would far exceed the scope of this paper to try and examine all of the factors playing a role in second language acquisition, and thus I will confine the investigation to the following points:

- The role of age in second language acquisition,
- The effects of instruction on second language acquisition,
- The learner-internal effects on second language acquisition, and
- The effects of input on second language acquisition.

It will be necessary, though, to try to find out how L2 acquisition works first in order to examine the effects of the different factors on it. To accomplish this, I will start by briefly describing the acquisition of L2 German negation in a naturalistic setting. I will then give an account of the discussion of UG availability in second language acquisition.

3.1 The acquisition of German negation

Felix (1982) describes the progress of David, a five-year-old American boy living in Germany and acquiring German in a naturalistic setting. David had already lived in Germany for about one month before the study commenced. The first examples given were obtained between the sixth and the ninth day of the investigation:

F: wollen wir ein Spiel machen?

D: nein.

F: warum nicht?

D: was?

F: warum kein Spiel.

D: nein.

F. clears up David’s toys.

D: nein! (protest)

The negation David uses here is holophrastic, that is, they consist of a negator only. Wode found out that in holophrastic negation you will always only find the negator which is used as independent form in the target language: nein, no, non, etc. (Wode 1976b, 1977b). Following the holophrastic negation is the sentence-external negation. David used this form of negation one or two weeks after the holophrastic negation.

F: ist das deins?

D: nein, Barbara.

F: ist das dein Ranzen?

D: nein, das ist Tasche.

As you can see here, David merely added more language to the holophrastic negation. This kind of negation is very much closer to the target language (TL), though. The anaphoric use of the negator can be found in TL as well. The consecutive step, sentence-internal negation, is the next big step towards TL norms. However, the negator in these sentences is still the negator found in holophrastic negation:

D. rhythmically bangs his head against the wall.

F: lass das doch, du tust dir ja weh.

D: das nein aua.

F: was machst du denn da? Isst du Schokolade?

D: ich nein essen.

In these cases, the German negator nicht would be appropriate, but that will be the next stage of acquisition. At that time, there is mainly a lexical transition from nein to nicht, but the pre-VP negator word order still prevails:

D: du nicht spielen Keller.

D: ich nicht essen mehr.

At the same time, though, in copula and auxiliary sentences, the negator already appears in a postverbal place, in agreement with the TL norm:

J: ich kann nicht.

J: das ist nicht so. (Julie is another 5-year-old child Felix observed in his study)

However, post-verbal placement is not yet differentiated any further, independent of a possible object-NP. Therefore, errors like

Du kannst nicht das.

Du musst nicht das.

still occur. Wode (1978b), Ravem (1969) and Kruse (1976) found out that in English L2 acquisition another phenomenon occurred at this stage: The undifferentiated use of the sentence-internal negators no, not, and don’t. The children he observed used these morphemes regardless of their function, but rather randomly:

You not shut up.

I don’t talking to you.

I no like that.

On the whole, it is amazing that the children’s utterances show regularities that they cannot have heard in TL speech. It is the children’s own language acquisition performance. Another interesting point Felix makes is that German and English acquisition orders are remarkably similar to the point of sentence-internal negation, when English L2 learners start varying the negative morphemes. He goes on to point out that this is not only true for L2 acquisition, but also for L1 acquisition. Considering the analysis of developmental stages and sequences in L1 acquisition, it becomes apparent that the same phenomenon seems to take place in L2 acquisition. This led to the L1 = L2 hypothesis (Ellis 1994). Researchers tried to find out whether the same language device which nativists claim to be responsible for L1 acquisition is available for L2 learners. There are, however, too many differences between L1 and L2 acquisition to let studies as the acquisition of the negation draw any conclusions at once. (cf. Ellis 1994; table 3.10: 107). Bley-Vroman et al. therefore believe:

These general characteristics of foreign language learning tend to the conclusions that the domain-specific language acquisition of children ceases to operate in adults, and in addition, that foreign language acquisition resembles general adult learning in fields for which no domain-specific learning system is believed to exist. (1988: 25, cited in Ellis 1994: 108)

Ervin-Tripp (1974), cited in Hatch (1978: 205), still concludes that

We found that the functions of early sentences, and their form, their semantic redundancy, their reliance on ease of short-term memory, their overgeneralisation of lexical forms, their use of simple order strategies were similar to processes we have seen in first language acquisition. In broad outlines, then, the conclusion is tenable that first and second language learning is similar in natural situations.

The interesting distinction which Bley-Vroman in particular makes is the distinction between adults and children. While children in a naturalistic setting appear to acquire an L2 in very much the same way as an L1, adults appear to learn a foreign language in a very different manner. One possible explanation for this phenomenon is much-discussed between L2 acquisition researchers: Does UG atrophy with age? But before turning to this question it must first be established whether UG is accessible in L2 acquisition at all.

3.2 The accessibility of UG in Second Language Acquisition

In her doctoral thesis, Monique Krosse (1992) compared four different hypotheses regarding the access to UG in second language acquisition.

The Direct UG Accessibility Hypothesis (DAH) claims that UG is accessible in full to the language learner as if L1 had never been acquired. However, there are many problems with this hypothesis. If UG was accessible in full, the acquisition processes of L1 and L2 would have to be exactly alike, and L1 would play no role in L2 acquisition. Therefore, transfer of L1 structures to L2 would not be observable. I will discuss transfer in chapter 6.3 in detail, but there is general agreement among L2 researchers that transfer does take place in L2 acquisition. Therefore, the Direct UG Accessibility Hypothesis does not seem to be plausible.

[...]

---

^[1] He tested children’s ability to recognise compounds as regular or irregular. He proposed that while all children would accept ‘mice-eater’ as the correct form, none accepted ‘rats-eater’, despite the fact that none of them may ever have been told that ‘rats-eater’ was wrong or never have heard ‘rat-eater’ before. Therefore, the input the children have heard would not be sufficient to judge ‘rats-eater’ ungrammatical, and there is no other explanation than UG for the test results.

^[2] Markedness can also be defined in a different way: ‘In Chomskyan linguistics, unmarked structures are those that are governed by UG and which, therefore, require only minimal evidence for acquisition.’ (Ellis 1998: 70)

^[3] Current language acquisition research does not in any way describe any relationship between parameter setting and the rule-finding process, but taking into consideration Ellis’ (1994), Felix’ (1982) or Krosse’s (1992) description of the processes involved it does not appear to be a rash conclusion.

^[4] This table in Ellis’ work is a table describing a ‘Generalized pattern of acquisition for L2 English (based on information provided in Johnston and Pienemann (1986)’. Ellis himself describes it, however, as ‘ … one of the most successful attempts to identify a general pattern of development (1994: 105). It may, therefore, also be used as a description of the English L1 acquisition developmental pattern. The pattern offered in Felix (1982) distinguishes between syntactic and morphological acquisition in a very strict way, so that I thought Ellis’ account of a general pattern to be more appropriate here.

^[5] An expansion in communicativeness, creativeness and productiveness of the language is found, for example, as children of Pidgin-speaking parents extend their parents’ Pidgin to a Creole. (cf. Pinker 1994)