Sookmyung Women’s UniversityGraduate School of TESOL

Introduction to LinguisticsFall 2016

Week 4 - AnswerSounds and ExemplarsBybee (2010), Chapter 2Dirven & Verspoor (2004), Chapter 5

Bybee (2010), Chapter 2, Rich memory for language

1. What is the exemplar model and how does it contrast from traditional; generativist views.

Exemplar model is a way of explaining how language is stored in the brain. It basically claims that all instances of use are stored in the brain, as one would expect from a usage-based model of language. This is a very different model of language than we see in the generative theory, which claims that language is composed essentially of two elements; a lexicon and a computational element. In the generative model, the lexicon is the storehouse of idiosyncratic linguistic information. This is a view that has strongly permeated generative theory. Following the idea of Occam’s razor, the idea is that the theory should be ass simple as possible. This coupled with the idea, as mentioned in Bybee (2010), that limitations in memory are real led to the proposition that the lexicon includes only the bare minimum of information. For the generativists, there is no redundancy in the lexicon. As the earliest versions of generative linguistics saw the system as rule-based, anything that could be explained by rules, would not need to be stipulated in the lexicon. So, for example, the plural forms of nouns would not need to be listed along with the singular forms of nouns (with the exception of irregulars) because it was posited that the plural form was derived from the singular.

The exemplar model looks at things very differently. First of all, there is no lexicon because information related to language forms is scattered all over the brain. Again, As Hudson (2007) language on this view is nothing more and nothing less than a distributed network. So the network is language and it is not separate from other aspects of knowledge but integrated with them. In fact it is the integration of information that this model is supposed to work. In the exemplar model a huge amount of redundancy is built into the system. In fact the system is built out of redundancy. As we mentioned last week, in the usage-based models all instances of use are stored in the brain and each use contributes to the functioning of the network overall.

The distinction between the two models carries over the way categories are formed and used. In the generative model a traditional Aristotelian model of categorization is used. In this model categories are built out of features that allow for an idealized, abstract version to be created as the head of the category. These features are collected from our experiences with the world but those experiences are pruned away as they are not necessary, So, for example, the category for HEAD is made up of all the seemingly necessary features of HEAD, which may include size shape, location, function, hardness, etc. Based on

these features an idealized version of HEAD, which is not really existent in the world is created and is available for use in language. Remember, we think in categories. Categories in an exemplar model are composed not of features but of examples. The examples themselves embody these different features. Also, the examples themselves are linked to the context in which they were encountered. So, things that were experienced together are encoded together. This contextual information is not technically part of the network holding the form but is connected to it through the relevant exemplars.

There are at least a couple of reasons to prefer an exemplar model over the generative model of the restricted lexicon. The first of these comes from prototype theory (Rosch, 1973). Prototype theory basically states that for each category there is one example that is stronger and better than all other in the category. This is the prototype or best example of the category. Generative theory would have it that the prototype is the example that has the largest number of necessary features. So a prototypical bird is one with more features of Birdiness; e.g., feathers, a beak, two spindly legs, the ability to fly, wings, etc., But this does not seem to be the casein reality. A fairly bad example of a bird, an ostrich for example, can act as a prototype based on experience. Prototypes arise on the basis of and are simultaneously affected by frequency. An exemplar that is more frequently encountered will be the prototype regardless of the number of features that match it. We know that frequency plays an important role in how the brain functions (LeDoux, 2002). A system that ignores this is problematic. In the generative model there is no way that frequency information to be stored as categories are feature bundles. Another reason to favor an exemplar theory of categories is based on the way information in the category affects each other. In the generative model, simply put, it doesn’t. As Bybee (2010) mentions it is interesting to note that people don’t just have knowledge of what is in the category (features and members), but they also know how those things are there. This enables us to know intricate information about how elements in the category are interconnected; e.g. that small birds are the ones that tend to sing and big ones might not fly. We can see tendencies and of course make us of those tendencies because of this. One further and important advantage is that an exemplar model allows us to deal with gradience.

2. How does the exemplar model work and what affects it?The exemplar model is deceptively simple. It works by storing and connecting information based on its use in the real world. So, the experience we have with language is preserved in memory. This pertains to all levels from individual sounds to constructions. Thus, in an utterance such as, the bad boy had beer for breakfast, the entire construction is preserved together based on its similarity to other HAVE constructions. It should be noted that this might also fall into a subset of HAVE FOR constructions. So the entire construction is mapped to the situation in which it was used. Each of the words are also stored in a network of the same word. So, the word breakfast is simultaneously linked to all other instances of breakfast the speaker has heard and stored. And it moves down here to morphemes and then to the individual sounds. The individual sounds are stored together sequentially to form words but each is also linked to a network that holds all the variants of the sound. That sound, the /b/ in /bæd/, is linked to all b’s but is also linked to the environmental conditions (word initially following a vowel) as well as the social context of its use. Looking at my own language use, I have several principled ways that I pronounce the

word bad. There is the New York [bœd], standard [bæd] and Irish [bæ:d]. These differences are important to me and are recorded as different exemplars of the word but each of those is also linked to the setting in which it is encountered. So we can see that the context in which a form at any level is encountered affects how it is stored. Frequency also plays an important role. Exemplars that are not frequently encountered will be weakened or lost. It is important to bear in mind that the connections between elements vary in strength and frequency is one of the things that affects this strongly.

Dirven & Verspoor (2004), Chapter 5, the Sounds of Language

3. How are speech sounds made?All sounds have a physical representation. Like any kind of movement they require a specific combination of movements in a particular sequence. In some ways, then, individual sounds can be characterized by those sets of movements and those movements can be seen as the features of a given sound. Sounds require air and for the sounds we are talking about here that air needs to come from the lungs. Pressure is applied to the lungs and this forces them to expel air. Air is pumped out of the lungs and enters your trachea. From there it moves up into the larynx, a short pipe in your throat for air (the esophagus, running parallel, handles the food). As the air moves up the larynx, it moves through your voice box. Your voice box is that lump which sticks out

on your neck. It is larger for men than for women. Children have small voice boxes, which is why they generally have such high pitch voices. It is in the voice box that air is first changed into sound. Air enters the voice box and moves across the vocal cords. Your vocal cords are stringy muscles that are stretched across the inside of your voice box. As the air moves across the vocal cords, friction is created which creates resonance. It is this resonance that creates the potential for a specific sound. Airflow out of the voice box is controlled by the vocal folds, muscles which cover the voice box and are able to open or close. Think of the vocal folds as two doors which, when open, allow air to pass and when closed stop the flow of air.

Once the air gets out of the voice box it winds up in one of two places; the mouth or the nasal cavity. Often, it goes to both. Where the air goes is

determined by the velum/uvula. When the uvula is lowered, air passes into the nasal cavity and out the nose. When the uvula is raised, airflow to the nose is blocked and it must move through the mouth alone.

Inside the mouth, there are several parts that determine exactly how a particular sound is made. These are called articulators. The most important articulator is the tongue. What part of the mouth the tongue touches or occupies determines which sound will be made. The lips are also important articulators as is the glottis. It is important to remember that the articulators in our mouths are able to move. Teeth are not, therefore, a good articulator. A good definition for articulators would be the parts of the mouth that move to occupy a certain position and/or touch other parts in order to create specific sounds. Strangely this is also true for sounds made with airflow in the nasal cavity (nasal sounds). Nasals are distinguished by articulators in the mouth even though air is flowing out of the nose.

4. How are vowels and consonants different?Returning to the basic question, by looking at a set of articulatory properties

we can distinguish one linguistic sound from another. These articulatory properties are divided into two major concerns: Manner of Articulation (MoA) and Place of Articulation (PoA). MoA centers on the flow of air as determined by the configuration(s) of the vocal folds. PoA is all about the position of the articulators (tongue, lips, glottis) and how they are used to distinguish sounds. Let’s look at each of these and the way they affect/create variant sounds.

The two most basic MoAs in languages, are consonants and vowels. It makes sense to introduce these separately. Consonant Sounds - In these sounds the vocal folds are not completely open

and the flow of air is somehow restricted. There are several different types of consonants based on MoA.

Stop - In these sounds the vocal folds are entirely closed before the sound is made. Air pushes against the closed vocal folds until the pressure becomes too great and forces them open. The vocal folds burst open suddenly releasing a large amount of air and then close again.

Fricative - In these sounds the vocal folds are open, but only slightly. Airflow is impeded but never completely stopped. The small opening in the vocal folds allows only a small amount of air to exit the voice box. Air pressure in the voice box is high and this causes a large amount of friction on the sides of the vocal folds as the air passes by and that is why these sounds are called fricatives.

Nasal - These are sounds in which the uvula is lowered causing the air to move through the nasal cavity instead of the mouth. Both fricative consonants and vowels can be nasal, but there are no distinctive nasal vowels in Standard American English.

Affricate - These sounds are a mix of two sounds; one stop and one fricative, and share elements of both. Affricates begin with the vocal folds closed. As the air begins to put pressure on them, they open up slightly as in a fricative and the air is allowed to slowly escape. There is a much less violent explosion of air than in stop sounds.

Glide - The flow of air is almost completely free. For this reason these sounds are sometimes called `Semi-vowels`. They differentiate themselves from other sounds by having some sort of movement going on inside the mouth.

Lateral - These sounds also fall under the Semi-vowel classification but are different because the air moves along the sides of the tongue.

Vowel Sounds - These sounds have no obstruction of the airflow. This means that the vocal folds are completely open as the sounds are made. For this reason, they are louder than the consonants. Vowels themselves are an MoA in their own right. As a special MoA, they also have different manner features separate from that of consonants, which create different vowels sounds.Roundness – This feature has to do with whether the lips are rounded or not during the articulation of the vowel sound. Back vowels in English tend to be rounded while front vowels generally are not. Vowels are designated as being either round (+round) or unround (–round).Tension – This feature has to do with the muscles in the throat during articulation. If the muscles are tightened then the vowel is called tense. If the muscles are loose then the vowel is called lax.

Consonants and vowels have different PoAs, or really the PoAs are explained and defined differently. This is generally done so because with consonants the tongue usually touches a part of the mouth, but with vowels tongue position is important, yet it does not touch anything. For this reason the PoA of vowels sounds is often harder to pin down.Consonants – In these sounds, one or more of the articulators is touching some other part of the mouth as the sounds is made. This is the PoA.Bi-labial - These sounds are made using both lips. The lips are either closed

completely and open up or are rounded, like you are kissing something a little bit far away from you.

Labio-dental - In these sounds the upper teeth are touching the lower lip.Inter-dental - The tip of the tongue sticks out from between the two front teeth. Dental - The tip, blade, or front of the tongue touches the back of the teeth as

the sound is made.Alveolar - The blade or front of the tongue touches the alveolar ridge as the

sound is made. Palatal - In these sounds, the middle of the tongue is pushed up again the

palate. Velar - The back of the tongue pushes against the velum in these sounds.Glottal - These sounds are articulated in the glottis, the space between the two

vocal folds. A very small opening in the glottis allows for a small puff of air to escape. As the air escapes it creates friction that makes the sound.

Vowels – in these sounds the articulators do not touch any other part of the mouth as the sound is being made. Thus, their POA is not always easy to feel or discern without instruments. POA in vowels is measured across two planes: a horizontal (relative height) and a vertical (relative frontness) plane. High – In these sounds the tongue is standing up like a snake ready to attack.Mid - The body of the tongue is situated in the central portion of the mouth. Low – Here the tongue is held low in the moth, near the floor of the mouth.Front – The tongue is located in the front part of the mouth near the teeth.Mid – The tongue is located in the central portion of the mouth.Back – The tongue is located in the back of the mouth near the velum.

These vowel sounds are generally placed in a grid that represents the mouth as viewed from the side and shows their PoA. Each of these sounds also has an accompanying symbol in the IPA (International Phonetic Alphabet). The beauty of the IPA is that it provides us with a uniform way of identifying the sounds used in language and in a systematic way based on PoA and MoA.

5. How do phonemes show variation?There are different types of variation in relation to sounds. These variations are quite dramatic from slight alteration of the form; e.g., nasalization of vowels, to a complete deletion. It is also important to note that there is a lexical feature to this in that these variations do not occur at the individual sound level but are situated in words and constructions. So, really it does not make any sense to talk about changes at the individual sound level despite the fact that sounds do change. Conditioned/environmental variationThese are sound changes that occur because of where the sound is located in a word or construction in addition in relation to the sounds that surround it. So, for example the /t/ in stop is different than the /t/ in top, with is also different than the /t/ on spat. There is a wide range of these kinds of variations across all languages. Socio-linguistic variationThese are sound changes based on usage in different social groups or under different social conditions. This could involve shifting sounds based on dialectical differences. It also involves speed and carefulness in the production of sounds (in words). These changes are typically never below the word level. This means that individual sounds are affected only in so far as being parts of words and constructions.

Based on these brief descriptions it seems that an exemplar-based explanation makes more sense in explaining how speakers deal with this kind of sound variation. Here is some further explanation of sounds changes.Alteration - place, manner, voicing (quality)In this type of change some part of a sound, or class of sounds, for we have seen that often a whole class of sounds will change in the same way, changes in a specific linguistic environment. This does not result in any change to the phonemic inventory of a language, and often results in increased allophonic variation. An example would be the aspiration of /t/in syllable initial positions (t---->th/s_).    MergeThis is a type of change where a sound or group of sounds join with another preexisting group of sounds to become one sound or one group of sounds. An example would be the inter-dental fricatives. In some dialects of English they are changing. For example in Irish English they are merged with dental-alveolar stops, but in other dialects they are merged with labio-dental fricatives. In merger, no new sounds are created to the inventory, but a sound might be dropped. Another example is the /hw/, which is spoken in only some dialects. In the distant past /hw/ and /w/ were distinct sounds. But now the former is an allophone of the latter in most dialects and in many North American dialects it has disappeared entirely. SplitThis is when one sound or group of sounds split into one or more sounds or groups of sounds. This has happened in English in the past in respect to the sound /f/. At one time there was only one sound, but pressure from French borrowings and the overriding system created the sound /v/ first as an allophone and then as distinctive sound in its own right. This can be compared

to Dutch (one of the closest relatives to English) where the relationship between /f/ and /v/ is not as neat or clean and change is still going on in relation to these two sounds. This type of change is often seen in vowels as well. At some point in time we could posit that the sounds /i/ and /I/ were one sound (/i/), but that the sound split into the two sounds we now have in English in the Germanic languages. Most other Indo-European languages (eg. Spanish) do not have both these sounds.      ReplacementThis a change by which one sound or group of sounds is replaced by another sound. The distinction here with merge is that the sound which replaces another can result in the creation of a totally new sound. DroppingThis is a change where a sound or group of sounds will be dropped in a certain linguistic environment.

InsertionThis is a change in which a certain sound will be inserted into a certain linguistic environment.

Here are some specific types of consonant change.flappingglottalizationyod-droppinginsertion

Here are some specific types of vowel changesdiphthongizationmonophthongizationThe overall systematic nature of sound change tells us that there must be some integration with the mental lexicon. The clearest examples of this are the push and pull chains that we see in vowel change. Unless the overall system of vowels were not encoded in the mental lexicon we would not expect to see such types of changes. Sound changes, as we see from looking especially at diachronic sound change, often occur residually to preserve the shape of a system.      Also, the location of certain sounds with in a word bear importance in the mental lexicon. There are many ways in which words are arranged in the mental lexicon, but one of the major ones is based on the beginning sequences of a word. When listening, or when going to produce an utterance, a person will access the necessary word by moving down a list. This list is based, in part, on the first sounds in the word. Research has shown that words which begin with the same sound sequence are located together in the lexicon. Likewise, the final sequences of words are also of some importance. We have seen that phonological rules make reference to word boundaries, both initial and final. This means that such boundaries have psychological reality, not only in phonology, but in the mental lexicon as well. From a more practical point of view, we can say that people need to pay closer attention to word boundaries. What is in the middle of a word is often of less importance. (Not surprisingly, it

is vowel sounds that are usually centrally located in words. We have seen that vowels show more variation than consonants and are subject to change at a more rapid pace as well. This only lends more credence to what had been posited above. 6. What are some of the phonological properties above the phoneme level?Until now we have looked at how segments (individual sounds) are created and distinguished, but there are also several things that operate on units larger than that of the segment. These are called suprasegmentals. Before looking at the two major types of suprasegmentals, we need to first look at the different size units upon which they operate. The smaller of these is the syllable. A syllable is a unit of sounds clustered around a vowel. It may have a consonant (at) or a consonant cluster (als) following the vowel and it must have a consonant (?a) or a consonant cluster (spla) preceding the vowel. The vowel is the heart of the syllable and as a result is referred to as the nucleus. Different languages vary quite a bit in the types of syllables they allow. This is called phonotactics and English has quite limited phonotactic constraints meaning there a wide range of different syllable types while languages like Japanese and Hawaiian have strong phonotactic constraints meaning they have only a very limited set of syllable types. Korean is somewhere in the middle which means that Korean often have difficulty with some of the larger syllable types. The second unit important for suprasegmentals is thought groups. Basically, thought groups are units of a word or words that contain some meaning. They vary based on the intent of the speaker and the only constraints that apply are syntactic. From the syntactic perspective as we shall see, thought groups are made up of constituents. They cannot violate constituency.There are two main suprasegmental features of English. They are stress and intonation. There are two main types of stress in English.

Word-level StressThis is the stress that falls on one syllable in each word. Each word has one main stress and stress is marked on the vowel in the stressed syllable. Stressed syllables are produced, higher, longer and louder than other syllables. For one syllable words, the stress will be easy to find. In multi-syllable words stress can be difficult.If you have been studying English for a while, you will begin to develop feelings for what sounds good and what may be not so good. Stress is a great place to try to use your intuition. Say the word to yourself using different stress patterns. Your good intuition should tell you which one sounds better and is, therefore, acceptable. Sentence-level StressJust like some syllables in words are stressed so are some words within a sentence. Stressed words with a sentence are also produced higher and longer than the other words. The basic patterns of which words within a sentence receive stress are easy. Basically, content words are stressed, while function words are not. Content words are words which carry some sort of semantic meaning, like nouns, verbs, and modifiers. Function words are words which have grammatical or functional significance, like prepositions, articles, and pronouns.

Content Words Function WordsMain Verbs go, talk, writing Pronouns I, you, he ,theyNouns student, desk Prepositions on, under, withAdjectives big, clever Articles the, a, someAdverbs quickly, loudly Conjunctions but, and, soNegative Aux. Verbs

can’t, don’t, aren’t

Auxiliary Verbs can, should, must

Demonstratives this, that, those Verb “to be” is, was, amQuestion Words who, which,

whereSource: http://www.englishraven.com/files/TeachingSentenceStress_1_.doc

Sentence stress is not that difficult. You just have to practice in order to get used to following the simple patterns. One of the best ways to practice pronunciation is to read aloud. Take any material and read it out loud as fast or as slowly as you want. Try to play a little with your voice as you read. Alter your stress patterns as you read and see which sounds better. Have fun with it.

IntonationEnglish has three different tones. These tones are like musical notes and are differentiated by pitch. We, thus, have high, middle, and low tones in English. Intonation is the patterns of use these tones enjoy in language.

ExampleHigh _______________ areMid __________ HowLow ____________________ you?

In English there are two main intonation patterns: rising and falling. Rising is when the tones rise at the end of a sentence.

High ___________________ fat? Mid _________ Is Low _____________John

Falling is when the tones fall at the end of the sentence.

High ___________ areMid _______ WeLow _________________thin.

In English, rising intonation is used in yes/no questions. Yes/no questions are questions to which you must answer either yes or no. Falling intonation is used in all other types of sentences, from direct statements to wh-questions, and finally commands.

7. How would the exemplar model work with sounds in context?Sounds are represented as exemplars stored together with a range of contextual information derived from actual language use. These instances of use are stored simultaneously in all the different levels from visual and

contextual information up to the entire construction, followed by each of the individual words in the construction, then up the specific sounds from each word. So that exemplar of a sound is put in a network of similar sounds, while simultaneously being linked to a network for the word in which the sound was embedded, which in turn is linked to a network for the entire construction, and finally is linked to networks for the contextual aspects, such as location, co-interlocutor, topic, the speaker’s internal state (cognitive, emotional, physical). It is this rich representation of language in this kind of massive, redundant distributed network that makes language comprehension and use possible. Before one even hears what someone else is saying or will say, perceived aspects of the contextual situation have already led to the activation of certain form networks based on past experience. This makes language processing much more efficient.

References

Lakoff, G. (1986). Women, fire and dangerous things: What categories reveal about the mind. Chicago: University of Chicago Press.

LeDoux, J. (2002). Synaptic self: How are brains become who we are. New York: Viking.

Rosch, E. H. (1973). Natural categories. Cognitive Psychology 4, 328-350.