Stephen van Vlack
Sookmyung Women`s University
Graduate School of TESOL
Introduction to Linguistics
Fall 2006
Week 13 - Answers for discussion
Radford et al (1999) pp. 378-405
1. What do children have to do to acquire the syntax of their L1?
According to standard Chomskian theory, and this basic belief has been around in many forms since the 1950s when Chomsky first developed the theory, all children need to do to acquire the syntax of their L-1 is to simply set parameters. The basic idea, and I think you are aware of this, is that UG is composed of two different parts: principles and parameters. Principals are the same for every language and children don't need to learn them or actually do anything with them. Chomsky claims the knowledge of principles is innate. They are just there simply present in the child's mind at birth. When a child begins to be exposed to language, so the theory goes, these principles automatically start to work. Parameters, on the other hand, are different. While they do exist in kind in all languages, not all languages have the same values for the same parameters. So, in order to acquire syntax, as Chomsky describes it, children simply need to decide how the parameters work in their particular language. Through a small amount of meaningful exposure parameters are set.
It is important to mention that syntax, in a very simple way, extends far beyond just UG. UG, logic would dictate, only accounts for a certain percentage of our overall grammatical system. Remember that parameters only regulate and control the use of functional categories. This means lexical categories, and everything that goes with them, are dealt with in the lexicon and more particularly in the lexical entries of lexical categories. The basic function of UG is used to, and a very general way, control the shape and configuration of languages in general. UG doesn't determine what goes into a clause it merely determines how the clause will be shaped. Thus, things like morphological marking are not determined exclusively by UG, but rather are seen as language specific types of operations which children would have to learn. Recently, Chomsky (2001) has claimed that even language specific operations, while not handled directly by UG are given a helping hand by UG, in that UG restricts possibilities thus making acquisition easier.
2. What does it mean, to set parameters?
According to Chomsky, setting parameters is easy. The basic idea is that parameters, like the distinctive features that we saw earlier in the theory of phonology, have one of two binary values. A particular parameter will have either a plus (+) value or a minus (-) value depending on the configuration and/or movement it allows. So, according to the theory there are not actually very many choices for a child to make. Based on this simple configuration it is therefore, easy for children to acquire parameter settings for their language with ease and immediacy. The idea is that since children have an innate knowledge of parameters they don't really need to figure out the parameters or know how the parameters work. All children need is a minimal amount of input in order to trigger the parameter setting. They simply need to recognize a particular structural element in context and this is said to happen very fast and with great ease. Once more, the claim goes, that once parameters are set children do not make any kind of errors in relation to parameters. They do not change them or experiment with them?
An important part of Chomsky`s claim about language is that native speakers all have the same competence and it is this competence which enables them to be infallible/error-free in their language use. If this is true then it should not be possible for people to change parameter settings, or indeed to ever use any other parameter settings than the ones they have. By logical necessity this claim must be extended to children, for if children are able to play with their parameter settings, or feel uncertain about them then why not adults and the idea of native speaker infallibility goes, therefore, up in smoke. Thus, many generativists, chief among them Andrew Radford, the chief author of the Blue Book, following Chomsky`s lead, have claimed that children`s UG is the same as the UG of adults. It has to be because UG is innate and therefore not subject to developmental change. Child language then varies form adult language due to other constraints on child thought and not linguistic differences. Having made this claim the authors of our textbook go to great lengths to try to defend the claim despite what seems to be a rather large amount of variant data coming from children. The basic idea is that if parameter settings happen immediately and if children have the same parameter settings as well as the same number of parameters as adults then we would not expect to see any variation in observed behavior relating to parameter settings.
It is important to remember here to the authors of our textbook are trying to push a strong theoretical viewpoint and for that reason they are collecting a rather large amount of data which they hope will support what they have claimed. In reality the situation is more complex than they make it out to be. Simply put, there are many differing opinions on how children go about parameter setting. Some of the different viewpoints would be that children have limited access to parameters. The basic crux of this belief is that children basically start off with limited lexical categories and overtime, and we are assuming with more input, they are able to slowly acquire piece by piece functional categories and as a result the specific parameters that relate to these functional categories. This belief is held by many because it seems to explain in a very simple way some of the basic observations which that have come of late about variation in first language acquisition. Others would believe that children do have full access to all functional categories and therefore parameters but that parameters setting does not come instantly. These people believe that children go through a series of stages in the settings of parameters in which they waffle in their parameters settings. The idea is that it simply takes time for parameters to set. A third view is the Chomskian view that parameters are preset in the brain to the unmarked setting. So the basic idea is that parameters are set for all babies and all the same ways and this way would be the simplest or unmarked way. What the child that needs to do is observed how their language is actually used and if necessary make the appropriate parameters setting change. In this Chomskian view the role of input is extremely limited and that is an important point for us to consider. This is called the poverty of the stimulus argument.
Now, all as is very important for us as teachers of English in Korea because we would like to be able to assume that second language acquisition can, at least theoretically, occur in the same way as first language acquisition. If we are to follow the view of ease and immediacy in parameters setting which the authors of our textbook give us then it is hard to find parallels in second language acquisition because we know that learners of the second language especially those who have reached a certain age do not acquire parameters settings or much of anything else with ease and immediacy. We need to take a critical view of what the authors in this book are claiming and is possible for us to do this because we all have special insight into this process as language learners ourselves in addition to being language teachers. We have access to information that people like Radford may not have.
3. How does the parameter of finiteness explain differences in child and adult grammars?
The idea of finiteness combined with the idea of the Optimal Infinitive Stage can be used to predict differences between adult and child language. The bottom line is that children seem to be prone to create non-finite clauses where adults would use finite ones. That at least is the claim that the authors of our textbook make. It is not necessarily important that you believe this claim. As I mentioned above a very large amount of counter evidence and a tremendous amount of disagreement on this area still exists. In fact there is too much dissension to faithfully believe in any one theory or quick fix idea. There is no definitive theory in relation to first language acquisition and what is interesting is that the more information which comes in overtime the more clouded the issue becomes. What is important to get from this argument is to see how linguists form theories and argue for those theories because this is what you need to do as well as graduate students.
The authors of our blue book have made a very strong and very simple claim which is that children set parameters very early with ease immediacy and no waffling back and forth. This means that once they set parameters the parameters setting are cemented so they are never make mistakes in creating utterances which reflect different, or incorrect parameter settings for that particular language. Once they have made this claim and they have to try to defend it and we can see that it's not necessarily that easy to do. What they wind up with in the end is the claim of finiteness being different for children and adults. The bottom line is the children seem to be nonfinite while adults can to be finite. Unfortunately, however, there is no claim about why this might occur. Also, the claim is circular because they are interpreting the child utterances of being non-finite from an adult point of view. One thing researchers of child language have learned is that trying to interpret the intent of a child`s linguistic utterance from an adults point of view can be highly problematic. It may very well be the case that the authors of the blue book are trying to impose an interpretation on the utterances which is simply not valid. It is circular because this interpretation is coming form their adult system which is the point they are trying to prove in the first place.
Again, the authors are moving from certain theoretical point of view and they're trying to prove a set of beliefs so we always need to be careful and critical when we read these kind of things. One of their main goals is to try to prove that there is a very strong link between child language and adult language and of course why there should be at the nature of that link is highly debatable. We're not going to find any particular answers in discussing this because they're simply too many possibilities right now but I want you just to be critical and develop critical thought.
4. Can this difference in finiteness be related to seconds language problems?
Maybe in order to determine this you would need to run through particular data from second language learners and see if there are any kind parallels between second language learners and the children. I personally have not done this in any systematic way so I can't actually say anything definitively but on simple observation there do indeed seem to be some parallels. Simply noticing parallels is not enough though because there might be a tremendous number of different reasons why second language learners behave in certain ways. Thus the big question is why do second language learners often leave out TENSE and AGREEMENT features on verbs and other marking features on nouns. It all revolves around inflections. It could be a linguistic problem (they are not aware of what they need to do) or it could simply be a processing problem (they are too busy trying to get the right words to be able to deal with little form issues like that). We could always argue that avoidance is a simple strategy which second-language learners apply and this might very easily account for the higher amount of non-finiteness in the speech of second-language learners. Through extension, the same argument could be applied to first language learners. They too may be using strategies. In any case there is no simple solution and many factors might be involved. My personal feeling is that the reason for the prevalence of non-finiteness in first language and second language acquisition processes is lexical. I won`t explain that here as it would take a very long time to do so, but just remember there are always other explanations. They lexical explanation has been largely ignored because it requires a view of the lexicon which generativists do not presently allow.
5. How do people go about processing sentences?
This is somewhat unclear from the description in the book but sentence processing seems to make at least some use of the principles that we studied in this class. Again, it is important to note that there are many different ways of looking at sentence processing. Obviously certain structural elements are possible in sentence processing but the burning question is are they a necessary and are they a part of the general way of sentence processing. Unfortunately we can't really answer this question because sentence processing is generally done in a decontextualized laboratory like situation and we would expect people's processing to actually be different in such situations. In the laboratory, where sentences are presented not only decontextualized but also in forms not normally present in normal speech situations, people need to process in the way the researchers would like them to; linguistically. This is what the researchers want but it also means there are some serious limitations in the way they conduct their experiments which may just have them arrive at the conclusions they expect due to design.
Through this type research what people are finding out is that in relation to language processing goes there doesn't seem to be just one system. There is the regular system linguistic system which people seem to use in regular basic conversational contexts when there is no difficulty and then there seems to be a fallback system which people seem to use when people need to process things differently, for example, if they run into any problems. What the authors of the book are describing here in sentence processing is what I would believe to be the fallback mechanism. It would seem that in usual situations in which context is probably the most important aspect (only would argue that the context is situation is actually more important than the particular language which is used and many conversational contexts) people would probably not have to fallback on structures and knowledge of structures at least not completely when they're processing language. Another interesting difference could be between the medium in which language is coming in. If somebody is listening we might expect them to process differently than if they are reading because there is a major difference in both time constraints and the type of language being used. Written texts tend to be much more linguistically complex which at the same time substantially less contextualized than spoken texts. Unfortunately none of this is mentioned in the blue book, but again this is just an introduction so we can not expect the authors to bring people deep or at least deeper into the topic area.
What we get in this section then is a view of language processing based almost entirely on linguistic principles. The authors have set out to prove and indeed do show to some extent that language processing reveals the importance of linguistic structures similar to those we have been doing in our tree diagrams. In this view structural dependency seems to be key to language processing. We do seem to process in units not unlike the phrases and clauses identified in grammar. Such phrases and clauses are treated as groups. Once more, units that are structurally dependent on each other need to be close to each other. Thus, adjuncts could be claimed to be processed than complements in their relation to their heads. The dependencies between certain grammatical objects seem to be real as shown by the authors. Also, the authors claim and unsuccessfully I believe, that empty categories also have some psychological reality through processing studies, but this is because the sentence types used vary. So their experimental deign was flawed.
All in all, there do seem to be strong linguistic effects in sentence processing along the lines of the generative system. The findings reported here would work, however, for virtually all grammatical models proposed, even for cognitive ones as no linguists question the validity of phrases or units in language, in fact a lexical model is built on it.
6. Is all processing based on the linguistic principles we have studied in this class? Explain!
In addition to the linguistic principles discussed above, there seem to be extra linguistic principles that apply quite normally to sentence processing. Things such as recency effects as well as local attachment (and dare we even say lexical effects in the form of collocation in a grand sense) all seem to play an important role in sentence processing. Again, falling back on what was mentioned in the previous question, it is important to realize the people probably process differently depending on the situation. We know from research on listening that people do not listen the same way all the time. There is much more focused careful listening and much more general listening and the average listener seems to move back and forth between careful listening processes and more general or less careful listening. The same holds true for reading. Thus, we would assume that there would be processing parallels to this distinction in types of listening in all types of processing. Also talking about the different mediums that are speech or written text you definitely expect people to process them differently not just because of the different actual forms that the language is coming in but for the most part because of time constraints. We can therefore argue that the extra linguistic elements to language processing, most notably memory limitations imposed by time constraints, probably outweigh the linguistic elements in most situations simply because in most situations we don't need to break everything down completely. We need to process very fast and very efficiently and certainly extra-linguistic principles like short-term memory are the key to that. Simply looking at collocations we can see how collocations would allow a native speaker who has a tremendous amount of knowledge of collocations via the mental lexicon would be able to use this to process language and much faster and much more efficiently. It's important to take these things in the consideration and form your own thoughts on the matter based on your own tremendous wealth of experience.
Section 24 - Children`s sentences
The basic idea here is that children somehow acquire the structural rules of their language. How they do this is a continuing debate. In the past, researchers have believed that children move through a series of stages in which their structural system becomes increasingly more similar to the structural system of adults. This means that children`s structure is actually different than adults. It has been argued in the past that the systems were different. Recent research by generativists, however, has indicated that this might not be true.
The authors of our text believe that much of the structural properties of language are universal. This means that as children hear their L1 in meaningful context they are able to figure out the structure of the language based on the principles of UG. All they have to do after this is figure out the specific factors of their language. The basic idea is that by the time children get to the two word stage, this is the point when structure first becomes relevant, they already have all the basic components of structure. That is, they have the universal components. Their structure is based on the same principles as adult structure via universal constraints and information. An example of this is setting parameters
The principles of Universal Grammar allows the children access to the parameters that govern all structural systems in all languages. Examples of these would be the head parameter, INFL parameter (AGRP parameter), wh-parameter, and null subject parameter. These parameters are universal, but the values for the parameters are language specific and must be acquired. The argument is that they are acquired basically by the time the children have entered the two word stage. In fact, we might even be able to turn the argument on its head and state that maybe children cannot enter the two word stage unless they have gotten the values for these parameters. All they have to do now to completely acquire the language is to figure out the lower order, language specific elements like inflections if the target language has them.
To help you better understand this it could help to liken syntax to phonology. Of course, it would seem logical to assume that the processes are quite similar. In phonology, we have the distinctive features (as well as some basic phonological operations, such as assimilation, blending, etc.) as the universal parameters. Distinctive features are part of UG and need not be learned for they are there in the innate UG system, but the values for the features themselves in relation to the specific sounds of the language must be acquired. Syntax is the same except that acquisition of lexemes and structure is more strongly linked.
The authors of the text clearly show that children in the age group of 1:6 - 2:0 have access to the different parameters. One of the two main views are that children vacillate between the values of the parameters. For example, for the Wh-parameter, we see that children produce both types of sentences; ones with wh-fronting and ones without. The fact that they sometimes front the Wh-sequences proves that they have access to the rule (presumably through UG), but they are not yet quite sure about the values of the wh-parameter for their language in all instances. This is what takes time to learn. The other view is that the system of children is entirely the same as adults but they show different linguistic behavior due to extra-linguistic variables predominantly processing limitations.
The bottom line, then, is that children have adult-like systems of structure in their heads from the very first, but they need time to figure out the values of the parameters for their language in addition to dealing to language specific phenomenon, such as verb and noun classes.
Section 25- Sentence processing
Sentence processing is quite a bit more complex and we know much less about it than lexical processing. One of the main things we know is that a certain amount of sentence processing does, indeed, follow closely to the type of syntax we have learned in this course. One example of this is parsing. Native speakers of a language are able to identify phrasal and, especially, clausal boundaries. This means that they have access to, and use in processing, knowledge of the surface level structure of utterances. Empty categories and traces are also shown to be involved in processing but the results shown here are to be questioned due to problems in the experimental design. Reaction times in sentences where chunks have been fronted and a trace has been left behind are the same as in sentences where no fronting has occurred. This means that the value we put on traces as carriers of important information might be true. So, it seems obvious that our syntactic descriptions and the way they are formulated are correct or at least we are on the right track.
At the same time, however, there are things that have been observed in sentence processing which are not accountable in our current system of syntactic description. These three areas of processing difficulty are ambiguity, center-embedding, and garden-path sentences. The idea in ambiguous sentences is that there is no syntactic way of determining which meaning is intended. The listener, therefore, has to fall back on semantic or contextual information to be able to syntactically process the utterance.
There also seem to be constraints on processing as imposed by the short term memory. As we all know, the short term memory is not very efficient. We can only remember things as long as we thing about them or generally about 4 or 5 seconds. Since we use our short term memory for processing it is important that utterances are so constructed that they are easily retained by the short term memory. This means that there are such rules as local attachment and center embedding which must be obeyed for effective processing. Linear order with enough function words to clearly point to the syntactic relationships is very important. Related words, such as verbs and their proposition in two part (phrasal) verbs are also more easily processed if they are in close proximity to one another.
The final thing is that certain utterances fool us as we go to process them due to their structure. Such utterances are called garden path sentences.
(1) The armadillo eating ice cream tasted delicious.
Most native speakers looking at this sentence would think it unacceptable in its present form. They would probably want to change it to the sentence in (2) below even though it is fine in its present form.
(2) The armadillo was eating ice cream which tasted delicious.
This happens because speakers seem to have a strong preference for closing clauses as quickly as possible. The structure of the first utterance contains an embedded clause. It is the armadillo, not the ice cream, which is delicious. In the second, more easily processed utterance the second clause is a subordinate clause and it is the ice cream, not the armadillo, which is delicious.
Again, we could assume that utterance (1) is straining constraints on the short term memory, but it is shorter than utterance (2). Hmmm. Maybe simple easily directed sentences are the best. Simple clauses and not much distance between the AGRP/TP and the VP might make things easier to process.
Lee (2001) Chapter 12, Constructivist Processes in Discourse
7.What is constructivism and where does it come from?
The idea of constructivism basically claims that in the process of creating discourse what people are really doing is creating a specific version of reality. In this process they are not only using concepts but also constructing concepts at the same time. Thus, in the discourse process people use and create propositions and it is these propositions which are used to frame discourse. These prepositions come from what they know and how they are thinking about or using what they know. Thus, a piece of discourse is just a long line of propositions created and thrown back and forth from side to side. For the listener who first receives the proposition needs to somehow respond to it. According to discourse analysis a conversation is often revolved around one train of thought at a time. This train of thought is a proposition. So a conversation revolves at any one time stretch around a single proposition which is put forth and then altered and thrown back at the original speaker who will alter their own proposition again in lieu of what the other person said. In this way the two speakers are working together to construct new and possibly complementary propositions. That is really want conversation is in the constructivist point of view. For a comprehensive view of how this works in generating new ideas see (Mercer, 2000)
8. How does constructivism relate to certain aspects of the cognitive model?
There are basically two levels upon which the cognitive model of language relates to discourse processes as explained by constuctivism. In the first of these constructivism relates to the cognitive model in that it is prototypes of radial categories which are used to organize these constructions. In a general sense we speak using words, the meanings of which are arranged in radial categories. At the center of these radial categories are prototypes and it is the prototype which first pops in the people's heads when they go to say something or when they hear a certain word being spoken. At the same time while people start with prototypes a large part of the actual cognitive process of language production is extending from these prototypes and using different meanings which may be related to the prototype but are actually quite different. While you're speaking then even though we are using a meaning which is not prototypical for a particular word the prototype will certainly color the meaning for the person who is listening to us. In this way we can create distinct new language while still making it comprehensible to the other person. People can therefore be using the same words but actually coming up with different meanings. This allows new meanings to be constructed in people's brains which is the basic crux of constructivism. Thus, it is the flexible condition of radial categories (the basic idea that categories are not composed of a single set of features and that the presence of these features is necessary for category membership) along with the idea of prototype theory which allows new ideas to be built along with new linguistic forms and meanings.
The same thing to be said to be occurring at a higher level and this would be the level of frames in the cognitive model. The frames are used to organize higher bits of information. While words expressed single ideas we can say that frames with expressed complicating concepts in the cognitive model. These concepts consist of several different ideas organized in sequence. Like the radial categories, frames are also highly flexible in nature. This means that not only do different speakers have different frames of the same concepts but the actual frames at a speaker has will change, and possibly quite drastically, overtime. It can, therefore, be said that frames as well are also subject to prototype effects. For each frame there is a prototypical sequence or scenario which immediately springs to the mind of speaker, but in the process of conversation is normal for people to alter or change these scenarios, thus causing new information to enter into a specific frame. This idea of frames in the cognitive model also fits extremely well into the constructivist view of conversation or discourse because it allows people to construct new and different views and particularly based on views which come from others.
From the above discussion we can see that it is the flexibility of the cognitive model which gives it its power. The real beauty is not just the flexibility but also the fact that the cognitive model is also constrained due to the importance of prototypes in managing how people both create and interpret linguistic discourse. This cognitive model falls in nicely with other models of discourse in that it has become clear that in a discourse situation ideas are exchanged accepted, altered, repackaged and reused. All this is made clearly possible by the mechanisms claimed to be part of a cognitive model. Any model less flexible would certainly not be able to adequately describe the data which people get on the construction of discourse.
9. How does constructivism work in creating discourse?
Constructivism seems to be connected with the cogitative idea of frames. Much of our discourse comes about as a result of frames which help us to organize things in the brain but are also extremely useful in the productive process of language. The most important aspect of this and particularly in relation to the cognitive model of language is that these frames are alterable. In fact they are constantly being changed. In fact this is what discourse is really about. Think about it. Why do we talk to other people? Well, one of the main reasons is to get their views on a certain issue, but really people tend to give us their views when we asked for them are not and it is this putting forth of views or propositions which makes discourse what it is. What is amazing bell is that as people put forth new ideas these ideas interact with the ideas we are to have in our head thus creating new slightly altered versions of reality in our head every time we talk. This is what discourse does for us cognitively and this was already discussed to a certain extent in question 7. What we really want to discuss here is how this actually works. It was mentioned in the answer to the previous question some of the basic mechanisms of how this might work in relation to a cognitive model, but I think we know little bit further than that actually discussing repercussions which Lee (2001) should have mentioned but fails to do so.
In order for discourse to take place we need to open up certain mental spaces. This is pretty clear from our previous discussion and mental spaces which took place quite some time ago. Each piece of discourse is going to be composed of several different mental spaces which need to be linked together in order to create the intended meaning. In this way to theory of mental spaces is somewhat akin to dependency grammar but occurs as a result of a semantically driven and not grammatically driven system. Thus, things are connected across mental spaces for meaning purposes and not for purposes of syntax or grammar. In this view the grammar is incidental. Having gone over this quick review we can now see how mental spaces might need to relate to discourse. We open up mental spaces which contain, basically propositions. The person we're talking to need to somehow also recognize these propositions and to the way that these propositions are linked together. This of course is helped along or to a certain extent regulated by the idea of frames. Frames helpless figure out how things might be related across different mental spaces, for indeed there are mental space connections which are prototypical and these would be embedded within a specific frame. Now in conversation not only are people throwing out different frames for a listener to break apart and understand with a listener is throwing is frames back at them. In the cognitive point of view what we basically have is a very long link or chain of different frames which are connected over long spaces.
Discourse is not a neat chain of events. People throw out propositions retract them change them and in the process new propositions are created and new ideas thrown out into the discussion. Discourse is very often a cluttered mess yet people are able to continue on the process sometimes for hours and hours and hours. It can also be claimed that their discourses which continue over an entire lifetime. Every time we talk to a person we constantly bring up old pieces of discourse or really just frames thereof. What all this means is that we somehow have to have a memory of these discourses. All these past discourses that we've had most somehow be packaged somewhere in the brain and a ready for future use. For indeed we do use these discourses in the future. It is in this way that we can see the huge connection between discourse and thought/cognition. We need certain mechanisms in the brain much allowed this to actually occur in and in the view of Lee (2001) it is this cognitive model and that certainly doesn't make sense to us.