Language production

1. Language Production vs Language Comprehension

Language Language Comprehension Production The two

systems make use of the same knowledge base,

Perception Planning including

Meaning Construction Structure knowledge about

Construction the language and

Utilization Execution the world.

Considerable

controversy has

arisen as to whether they share the same system of rules. Some people believe that grammatical rules are neutral between the generation and analysis of sentences. The same rules can be used both to assign syntactic structure in production, and to compute it in comprehension. Some others believe that comprehension and production have different lexicons, even grammatical rules. Still more people believe that although they share the same linguistic knowledge, it is unlikely that they share the same procedures of using that knowledge. In other words, the two systems use linguistic rules for different purposes, and hence require different kinds of processing.



The two systems are not just the reverse of each other. In language comprehension,

listeners try to construct meaning out of the syntactic structures, so as to understand the speaker's intentions. In language production, the speaker has some ideas and intentions and tries to make plans of expressing them. The two systems involve two different channels and biological organs which may have different neuropsyological basis. In AI, the process of transition from deep structures to surface structures is different from that from surface structures to deep structures. The former process involves in expressing one idea with one or more structures; whereas the latter involves in using one structure to express one or more ideas.

2. Methods of Studying Production

A. Sources

There are five types of speech that can be sources of studying production.

(1) Error-free, continuous speech, which is actually not natural language.

(2) Speech containing errors (including slips of the tongue and their corrections), which is normal in communication.

(3) Speech containing the discontinuities such as hesitations signals. (4) Speech during language acquisition period in young children. (5) Speech in “dissolution”, e.g. aphasia.

Language production has been studied in three main ways. The oldest and most popular method is to investigate cases in which it goes wrong, and to infer what happens in normal production. The second method is to carry out experiments with normal Ss, and the third is to write computer programs.

B Aphasia

In the mid-nineteenth century Broca and Wernicke established that injuries to certain areas of the brain cause gross speech deficits. In Broca’s aphasia, damage to part of the left frontal lobe leads to very hesitant and broken, though largely understandable, speech. In Wernicke’s aphasia, damage further back in the left hemisphere results in fluent, but meaningless speech. The Swiss neurologist, Lichtheim, constructed a model which has three processing systems, or “centres”: A, auditory word representations; M, motor-word representations, and B, a system where “concepts are elaborated.” These systems are connected by pathway, including auditory input to A-a, motor output from M-m.

Volitional or intelligent, speech involves centrifugal connection between B and M

Thus, broadly we find patients with lesions in Broca’s area (M) who have relatively good

comprehension but poor speech; patients with lesions in Wernicke’s area (A) show fluent speech but poor comprehension.

This kind of theory has been opposed on two grounds. First, apparently equivalent neural damage does not always lead to the same symptom picture. Secondly, the model does not explain a finer grained analysis of the syndromes.

C. Speech errors in normal speakers

The speech of aphasics is so different from that of normals, either in content or form, or both,

that it is difficult to draw conclusions about ordinary language production from it. An alternative approach is to investigate production errors in normal speakers and writers, The speech errors that psycholinguists study are the slips of the tongue:

D. Hesitations and pauses

A third case in which the language production system may not be working optimally is when speech contains hesitations or pauses. Although such disfluencies are not, strictly speaking, errors, they do indicate points at which processing load may be high. By determining when the production system becomes overloaded, the kinds of processing it performs can be inferred. There might be some connection between pauses and planning, but speakers need not pause every time they plan. Some pauses are primarily for breathing, and others are stylistic. They do not reflect difficulties in production.

E. Controlled production experiments

In these experiments Ss are given a stimulus -- perhaps a word or pair of words to

include in a sentence, or a picture to describe -- and they have to generate a word, a sentence or a longer stretch of speech. The rationale for this procedure is as follows. In language production a message is converted into words. Normally the message is selected by the speaker or writer. In controlled production experiments part of the message is provided. These studies address the question of how difficult it is to put certain ideas into words.

3. An Outline of Language Production

A. The Generation of Messages

The construction of a preverbal message is a first step in the generation of speech. This step

is usually initiated by the conception of some communicative intention: the speaker wants to achieve some purpose by means of saying something, and he wants the addressee to recognize that intention from what is said. Given the communicative intention, the speaker will select information for expression that is expected to be instrumental in realizing the goal. But information can be instrumental only if it is relevant to the situation of interaction, and that situation is a continuously changing one. It will be suggested that the preparation of a message involves two steps:

(1) Macro-preparation, which consists in the elaborating of the communicative intention as a sequence of subgoals and the selection of information to be expressed in order to realize these communicative goals. This determines the content of the subsequent speech acts. (2) Micro-preparation, which is concerned with the further shaping of each speech act to bring it into the format required by a preverbal message.

In macro-preparation a speaker elaborates a communicative intention down to the level of the

content of individual speech acts (SA); in micro-preparation the content of each intended speech act is given informational perspective and is assigned all the features that are obligatory for a preverbal message (PM).

Both sides of the message-encoding process are heavily context-dependent. In order to elaborate his goals and to select effective information for expression, the speaker must take into account the precise discourse situation which is continuously changing. The processes of message encoding must refer to the records the speaker keeps of the ongoing discourse.

B. Bookkeeping

The defining characteristic of coherent discourse is that every new move of a speaker is in

some way related to whatever was said before. A co-operative speaker’s contributions are supposed to be relevant to the ongoing discourse. But this requires bookkeeping or storage on the part of the speaker. He will have to keep track of what was said and what was conveyed by himself and by the interlocutors. The sum total of the information about the discourse that is currently available or accessible to the speaker is called the speaker’s discourse record. A

discourse record is by nature a dynamic entity. It changes with each new contribution made to the conversation, whether by the speaker or by another participant. What is kept in the record can be classified in terms of:

(1) The type of discourse: informal everyday conversation, narrations, lectures, interviews, debates, etc.

(2) The topic of discourse: The discourse topic is what is being talked about, and thus mutually experienced, by the participants. A shift of topic can be initiated by explicit declaration (The next issue I want to address is ...), by suggesting a relation to the current topic (That reminds me of ...), a flashback (Whoops, I forgot to tell you that...), etc.

(3) The content of discourse. Interlocutors introduce and reintroduce referents (persons, things, events, etc.) and make predications about them. In doing so, they build “mental models” of these entities, their relations, and their properties. A discourse model is a speaker’s record of what he believes to be shared knowledge about the content of the discourse as it evolved. In the simplest case --- two-person interaction --- there are four knowledge structures involved in this part of a speaker’s bookkeeping.

a. The knowledge that the speaker believes he shares with the listener, independent of the present discourse interaction. Let us call this common ground. Xiao Li may believe that Lao Chen knows about Gaungzhou being the Southern Gate to China, and Ma lives there. b. The knowledge that the speaker believes to have successfully conveyed to the listeners during the discourse up to now. It is the shared knowledge arising from the speaker’s own contributions. Xiao Li talked about his experiences in Guangzhou, not about the Southern Gate. Rather, he discussed the food, especially eating of snakes, and other matters he believed Lao Chen didn’t know.

c. The knowledge the speaker believes the interlocutor to have intended to convey to him in the discourse as it developed. This shared knowledge emanating from the interlocutor is the interlocutor’s contribution. Xiao Li takes it that Lao Chen talked about GIFL (where their mutual friend Ma lives) and Dongfang Amusement Park ( a place he had never heard about). d. The knowledge which the speaker still intends to convey but which has not been up for expression yet. This is the information to be conveyed. Lao Chen wants to say more about the amusement park, and Xiao Li wants to relate what Lao Chen said about GIFL to his experiences in Guangzhou. The speaker’s discourse model can now be defined as “own contributions plus interlocutor’s contributions” -- i.e. the knowledge structure the speaker believes he has conveyed to the interlocutor plus the knowledge structure the speaker believes the interlocutor intended to convey in the discourse up to now.

(4) Focus The attention span of a speaker is fairly limited. He can work on only one or a few notions at a time in preparing his discourse. And the same holds for the listener who is interpreting the speaker’s utterances. The information to which the speaker is attending at a particular moment in time is called his focus. Similarly, there is a listener’s focus. If the speaker believes that their foci are shared, there is a unique part of the discourse model which speaker marks as “jointly attuned to.” In the conversation between Xiao Li and Lao Chen, this can be the food in Guangzhou, the amusement park, or any of the other discourse entities.

C. Planning

Once ideas have been formulated, then the speaker will start to engage in two types of

activity -- planning and execution. Planning is made at three different levels, from discourse plans down to constituent plans.

(1) Discourse plans. There are essentially two types of discourse: dialogues (two or more people talking together) and monologues (one person talking alone).

Conversations:The planning of conversation is not so simple as most people believed. Each participant comes to a conversation with a particular goal in mind and is aware that the other participants have goals of their own. The problem they have to solve is how to coordinate their speech so that they can jointly reach their respective goals.

(a) Turn-takings. If the participants in a conversation are ever to achieve their goals jointly, they must agree implicitly on an orderly method for talking:

  Each participant should have a chance to talk.

  Only each person should talk at a time (so that he or she can be heard).      

The gaps between turns should be brief (for efficiency)

The order of speakers, and the amount they say, should not be fixed ahead of time. There must be techniques for deciding who should speak when.

The problem of coordinating talk is solved by the conventions captured in the following rules:

Rule 1: The next turn goes to the person addressed by the current speaker. Rule 2: The next turn goes to the person who speaks first.

Rule 3: The next turn goes to the current speaker, if he resumes before anyone else speaks.

These three rules are ordered. Rule 1 takes priority over Rules 2 and 3. If the current speaker A asks B a question, B is obliged to speak next, and C is not allowed to take the next turn merely by speaking first. And Rule 2 takes priority over Rule 3.

(b) Opening conversation. To start up a conversation, one person must get another’s attention and signal the desire for a conversation, and the other must show willingness to take part. For this purpose there is the summon-answer sequence. One person says “Hey, Bill,” and the other says “Yes?” In such summon-answer sequence, it is the summoner who is obliged to provide the first topic of conversation. Telephone conversations also begin with this sequence, but the telephone ring is the summon, and the “Hello” or “Extension 241” is the answer.

(c) Closing conversations

It is still more complicated to close a conversation. In a simple telephone conversation, the participants A and B have to figure out some way to coordinate their conversation so that it ends by mutual agreement. The process involves two steps: first, A and B agree to close; second, they actually close. It is the first step that poses the most difficult problem, and it is solved by use of yet another conversational device: the pre-closing statement and its response. When carried out properly, this statement initiates the closing section of the conversation.

A. O.K. (pre-closing statement) B. O.K. (response)

A. Bye (closing statement) B. Bye (response)

More often, even after A and B have agreed to end the conversation, they still have preparations to make, so there may be some more exchanges before they move to the closing statement.

A. Yeah. (pre-closing statement) B. Yeah. (response)

A. Alrighty. Well, I’ll give you a call before we decide to come down. O.K.? B. O.K. A. Alrighty. B. O.K.

A. We’ll see you then. B. O.K.

A. Bye bye. (closing statement) B. Bye (response)

Monologues:

In describing a landscape, novelists are faced with a series of problems:

(a) Level. At what level should they describe it? Should they merely say, “I saw a beautiful landscape,” or should they mention every last leaf and pebble?

(b) Content. Given the level, which parts should they include and which should they omit? (c) Order. Given the parts they have decided to include, what order should they put them in? Should they describe them from left to right, from nearest to farthest, or how?

(d) Relation. For the given level, content, and order, how should they relate the parts to each other? Is it enough for the parts to be listed as present, or should each be given a precise location with respect to the rest?

Linda and Labov asked about 100 New York City apartment dwellers the question:”Could you tell me the layout of your apartment? “ and tape-recorded the answers. The problem the respondents attempted to solve was this:”How can I describe every room in my apartment in such a way that the listener can imagine where everything is?” It was found that virtually everyone solved the problem in the same way. A typical description would be something like this:

You walked in the front door. There was a narrow hallway.

To the left, the first door you came to way a tiny room Then there was a kitchen, and then bathroom,

and then the main room was in the back, living room, I guess.

The tour solves the problems of content, order, and relations all at once. It proceeded according to a strict set of rules:

(a) The imaginary tour begins at the front door of the apartment.

(b) If the visitor comes to a one-room branch, he does not enter it.

(c) If he comes to a branch with rooms beyond the first room, he always enters.

(d) When he reaches the end of a branch, and there are other branches to be traversed, he does not turn around and go back; instead he is brought back instantaneously to the fork point where the other branches originate.

Levelt asked Ss to describe spatial-grid-like networks, and he found that they followed the

next node to be describe one that has a direct connection to the current node. The following is a typical description:

Begin in the middle, a grey node. From there upward a red node. Then to the left, a pink node from the red. Then from pink again to the left a blue node. Then back to red. Then from red to the right a yellow node. And from yellow again to the right a green node.

the same pattern by adhering to the principle of connectivity: Whenever possible, choose as the

(2) Sentence Plans

In planning a sentence, a speaker has many options which fall into three categories. (a). Propositional content. What states or events does the speaker want to talk about?

(b) Illocutionary content. How does he want to deal with it? Does he want to make an

assertion, a request, a promise or what?

(c) Thematic structure. What does the speaker want as subject, what does he think the listener does and does not know, and what framework does he want to set his utterance in? Did Harry hit Bill? Was it Harry who hit Bill?

Propositional content: HIT(Harry, Bill)=S

Speech act: I request you to tell me whether S is true?

Thematic structure: Harry is the subject and theme; X hit Bill is given.

Harry is also the frame. The frame of a sentence is its first main phrase.

During the summer, Alison lives in Scotland.

The phrase during the summer is the frame, the setting within which one can understand the information that Alison lives in Scotland.

(3) Constituent plans.

Planning has so far been discussed as speakers decide on each sentence all at once -- as

if they select every last adjective, adverb, and article before they utter their first word. Common sense tells us that this cannot be right, and there is evidence to this effect. It seems equally outrageous to maintain the opposite extreme -- that speakers wait until they have said one word before selecting the next. The truth lies somewhere in between. Speakers plan more than one word at a time, but not the whole sentence at once. The obvious candidate for planning at this level is the constituent. In a noun phrase referring to a specific person, the choices of adjective and noun are interdependent. The same person could be referred to as an immense happy man or as a happy man, but not as an immense happy giant (thus making him too big). At the same time, these choices do not greatly affect word choices in the

preceding or following constituents -- at least not if the overall skeleton of the sentence has already planned. Drastic changes in the first noun phrase of The happy giant decided to buy himself a sandwich have little effect on word choices in the rest of the sentence.

The selection of specific words, therefore, might go like this. As part of the discourse, a speaker has built a skeleton plan for the whole sentence. He has chosen:

(a) to talk about some entity E92 hitting another entity E44,

(b) to place E92 as subject and frame and E92 hit X as given information, and

(c) to ask a yes/no question of this content.

The next step is to flesh out the skeleton with words. For example, he must decide whether to express E92 as Bob, Mr. Smith, or the sorry-looking man over there in the blue suit, all of which correctly describe E92. The idea here is that people generally plan the skeletal propositions before planning the propositions to be packed into each successive constituent.

D. Planning and Execution

It is one thing to plan what one is going to say and quite another another thing to execute that plan. As described in the previous section, speakers build up a general structure for the discourse, from a skeleton for the sentence to be uttered, and select words to fit this skeleton constituent by constituent. Their next task is to execute this plan, to get their articulatory organs -- larynx, mouth, and tongue -- to emit the sounds they intended. The fundamental problems are:

First, they have not always formulated their plans fully before they begin their execution. For this reason they often speak in fits and starts and make a variety of speech errors. Second, in final preparation for execution, they must built an “articulatory program,” a plan in working memory that tells the articulatory muscles what to do when. In forming this program they also make errors.

Thus, there are two fundamental issues to examine in the execution process: the alternation between planning and execution, and the formation of the articulatory program for pronunciation.

(1) The ideal delivery For

there to be a speech “error” there must a “correct” way of executing a

sentence, and this will be called the ideal delivery. When people know what they want to say and say it fluently, they are giving an ideal delivery. Actors says their lines, except when making deliberate errors, come close to ideal delivery, and so practised readers and orators. For theories of speech production the ideal delivery is of central importance. They all assume that people strive for the ideal delivery, and every deviation points to something that has gone wrong in planning or execution.

In ideal delivery, most types of clauses are executed in a single fluent speech train under one smooth intonation contour. On the other hand, the “grammatical junctures” between these clauses may contain momentary pauses. Language has been designed this way so that speakers may breathe without interrupting fluent speech. In the ideal delivery they can breathe at junctures, but not within clauses. So the ideal delivery has the following characteristics. The

execution of each clause between junctures takes a fixed amount of time; any pauses that appear within the clause are obligatory and vary little from one execution to the next. But the junctures themselves may vary considerably in length depending on whether or not the speaker takes the opportunity to breathe or to stop momentarily.

(2) Common types of speech errors

Although most are self-explanatory, repeats, unretraced false starts, and retraced false starts are more complicated. Repeats are repetitions of one or more words in a row, whereas false starts are corrections of a word. Retraced false starts also include the repetition of one or more words before the corrected word, whereas unretraced false starts do not. Name Silent pause Filled pause Repeat False start (retraced) False start (unretraced) Correction Interjection Stutter Slips of the tongue

Examples Turn on the // heater switch. Turn on, uh, the heater switch Turn on the heater/the heater switch. Turn on the stove/the heater switch. Turn on the stove/heater switch Turn on the stove switch -- I mean the heater switch. Turn on, oh, the heater switch. Turn on the h-h-h-heater switch Turn on the sweeter hitch. (3) Hesitation points

There are three major points at which speakers are liable to stop for planning.

(a) Grammatical junctures. This is the logical place to stop to plan the selection and first constituent of the upcoming sentence. Pauses at these junctures tend to be long and frequent. (b) Other constituent boundaries. Within sentences these boundaries are the appropriate place to plan details of the next major constituent -- precisely what noun phrase, prepositional phrase, or adverbial phrase is to fit next into the sentence skeleton.

(c) Before the first content word within a constituent. This is a point after speakers have committed themselves to the syntactic form of the constituent being executed, but before they have planned the precise words to fill it out. This stopping place gives speakers time to plan the very next major constituent. It is typically marked by a silent pause (the /ｄirty cups) or by a repeat of the beginning of the constituent (the/the dirty cups).

(4) Sources of planning difficulty

(a) Cognitive difficulty When people were asked to produce, as quickly as possible, a sentence on a topic like “car,” “joy,” “kaleidoscope,” or ”dominance,” it took them longer to produce the first word of the sentence for an abstract than for a concrete topic. In another experiment, Ss were shown some New Yorker cartoons and asked to talk about them. One group was asked to describe each cartoon, while another was asked to explain why each one was funny. There were more hesitations scattered through the explanations than through the descriptions, presumably because it was harder to come up with explanations and the right words to express them. All these results show that planning could become difficult for cognitive

reasons.

(b) Psychological factors. A second source of speech errors is situational anxiety. When people talk about topics they are anxious about, they tend to produce more silent pauses and certain other speech errors. Why? One possibility is that anxiety disrupts the planning and execution processes generally. Speakers become tense, and their planning and execution become less efficient.

(c) Social factors. Under the press of a conversation, speakers must make clear when they still have something to say and when they are finished. If they hesitate too long at a point, someone else may take over the conversation. Earlier it was noted how this might push speakers into starting the first word of the next constituent before having the constituent all planned out. It might also push them to use more filled pauses -- uh, for example -- to fill spaces where other speakers might possibly take over. Indeed, speakers do tend to use more uhs in dialogues than monologues

E. The articulatory program

Speech consists of a sequence of articulatory gestures, a coordinated succession of muscular contractions in and around the mouth. Speech execution like this requires a plan -- a plan to direct the order and timing of these articulatory gestures, a plan to command what muscles to move when.

(1) Slips of the tongue

In the articulatory program there are several lower, finer levels of planning and execution that deal with the formation of words, syllables, and sounds. What is the articulatory program? How is it formed? How is it executed? The answers to these questions can be found in part in an extraordinary source of evidence: slips of the tongue.

The First Law of Slips of the Tongue: Each speech error results in a sequence of sounds that is permitted in the language being spoken. Thus a speech error in English should never result in a sound such as /btir/, since English words cannot begin with /b/ followed by /t/. The Second Law of Slips of the Tongue: For segmental errors, the beginning of a syllable can exchange only with the beginning of another syllable, and similarly for middles and ends of syllables.

The Third Law of Slips of the Tongue: Segmental errors tend to occur within major syntactic constituents, and word errors across syntactic boundaries.

Garnham and others suggest a seven-way classification: anticipations, perseverations,

omissions, additions, exchanges, substitutions and blends. They even provide a conservative estimate of the frequency of speech errors -- about one for every 1800 words.

Name Attested Examples Anticipations (提前) take my bike>bake my bike Perseverations (延续) practical classes >practical krasses Reversals (倒置) Katz and Fodor>fats and kodor omission(省略) British>Britsch addition(增加) better off than>better off-wise than Blends (混合) grizzly+ghastly>grastly Haplologies (合一) never lets>nets Misderivations (误导) an intervening node>an intervenient node Word substitutions (代替) before the place open>before the place closes

(2) Units in the articulatory program a. Segments and features

with this ring I thee wed > with this wing I thee red left hemisphere > heft lemisphere a phonological rule > a phonological fool b. Syllables

harp-si-chord > carp-si-hord

a-ni-mal > a-mi-nal c. Words

shout+yell > sh/ell

hilarity OR hysterics > hilarics d. Larger units

a tank of gas > a gas of tank

wine is being served at dinner > dinner is being served at wine (3) Formation of the articulatory program

From clues like these, Fromkin has pieced together a picture of how the speaker forms the articulatory program in memory before executing it. The process has roughly five steps: a. Meaning selection. The first step is to decide on the meaning of the present constituent is to have.

b. Selection of a syntactic outline. The next step is to build a syntactic outline of the constituent. It specifies a succession of word slots and indicates which slots are to get primary, secondary, and zero stress.

c. Content word selection. The third step is to select nouns, verbs, adjectives, and adverbs to fit into the appropriate slots.

d. Affix and function word formation. With the content word decided on, the next step is to spell out the phonological shape of the function words (like articles, conjunctions, and prepositions), prefixes, and suffixes.

e. Specification of phonetic segments. The final step is to build up fully specified phonetic segments syllable by syllable.

Evidence for the model

a weekend for MANIACS > a maniac for WEEKENDS

Step 1: The speaker decides to build a constituent referring to a time period, Time41. Step 2: The speaker builds a syntactic outline for this constituent:

indefinite-article+noun+[2 STRESS]+preposition +noun+[PLURAL]+[1 STRESS]

Step 3: The speaker, using his semantic plan, selects weekend and maniac to fit into two noun slots. But he makes a mistake and insert them into the wrong slots.

indefinite-article+maniac+[2 STRESS] + preposition+weekend+[PLURAL]+ [1 STRESS]

Step 4: The speaker spells out the phonological shape of the function words, suffixes, and stresses. The results like this:

a + maniac + for + WEEKEND+[z]

Step 5: The speaker specifies the individual phonetic segments in terms of their distinctive features.

The basic idea is that maniac and weekend have been interchanged while everything else -- articles, plurality, and stress -- have stayed out.

(4) Word formation and word representation

There is a fundamental question here that is still to be answered. How are words represented in permanent memory? In building up an articulatory program, speakers obviously draw on their phonological knowledge of words in memory. Consider the word act. At some level in memory it is surely represented as a single unit, with a coherent meaning, as a string of phonological segments ready to be inserted into the articulatory program. But what about acted, the past tense of act? There are at least two ways it might be represented. It might be a single unit, acted, to which is attached the semantic information that it is the past tense of act. Or it might consist of two units, act+[PAST], which some later process turns into the articulatory program. These two views might be called the concrete and abstract theories of word formation.

The above-mentioned example suggests that word representation is fairly abstract. Fromkin has argued that words like imprecise, disregard, unclear, and nothing may be represented with the negative prefix [NEG], as evidence she cites the following tongue-slips:

I regard this as imprecise > I disregard this as precise

if there was anything that was unclear > if there was nothing that was clear.

Another interesting phenomenon is that not all words are equally accessible for insertion into the articulatory program, so people often the “tip of the tongue” (TOT) experience. Brown and NcNeill successfully recorded TOT by asking Ss to give the corresponding word of a

concept that has very low frequency by listening its definition, e.g.

A flat-bottom boat used along the coasts and rivers of China.

The target word is sampan, but not all Ss could remember it, the word seem to be at the tip of the tongue. And they were asked four questions: (i) How many syllables does it have? (ii) What letter does it start with? (iii) What words does it sound like? (iv) What words are similar in meaning?

They provide a number of words like Saipan, Siam, Cheyenne (indigenous people of the Plains in US) , sarong (phonetically similar), barge, housboat (semantically similar). Phonetically similar words were greater in number, showing the Ss had a sound lexicon. Of the 224 words similar in sounds, 48% had similar syllables as the target word. Only 20% of the words similar in meaning had similar syllables as the target word.

(5) Rhythm and timing

Since speech occurs in real time, the articulatory program has to specify not only the phonetic segments and their order, but also their timing and rhythm. In English, a “stress timed” language, the major carriers of rhythm are the stressed syllable, which plays an important part in two aspects of the articulatory program: its regularity and its speed.

As anyone who has stumbled through tough tongue twisters can attest, regular articulatory patterns are easier to pronounce than irregular ones. What tongue twisters do is to sabotage the articulatory program by mixing regular with irregular patterns in a way that confuses.

she sells sea shells (a regular pattern ABAB : ee, ells, ee, ells for the vowels, but the initial consonants follow another pattern ABBA: sh, s, s, sh)

The next aspect of the articulatory program in which rhythm is important is speed. When people try to talk fast, they have to cut corners, and the way they cut corners involves rhythm. In formal situation, or talking to foreigners, people tend to speak slowly and articulate every phonetic segment as it is “supposed” to be pronounced. They would say:

I want to get you on his catamaran.

But in casual everyday speech people speed up and have to take shortcuts. The first shortcut is to reduce the vowels in unstressed syllables to schwa, the neutral uh.

I wanta get you on his catamaran. I wanna gecha oniz catamaran.

F. Articulation

The final step in executing speech is the actual articulation of the phonetic segments. Fluent

articulation is probably man’s most complex motor skill. It involves the coordinated use of approximately 100 muscles, such that speech sounds are produced at a rate of about 15 per second. These muscles are distributed over three automatically distinct structures: the respiratory, the laryngeal, and the supralaryngeal. Almost all vocal organs, from the lungs to the lips, subserve other functions than speech alone. The respiratory system’s main function is breathing, the larynx protects the respiratory system from intrusions of food. The supralaryngeal structures are used in the mastication and swallowing of food. Though largely the same musculature is involved in the production of speech, the pattern of coordination is totally different. Theories of speech articulation have to account for this “speech mode” of coordination.

(1)The motor command hypothesis The phonemic input to the motor system consists of sequences of subphonemic units corresponding to distinctive features.Thus, each phoneme in the input contains a number of lower-level units.. These, in turn, correspond to invariant motor commands, neurological signals to particular articulators to perform specific movement. According to this hypothesis, then, there is a one-to-one correspondence between the set of subphonemic units making up a phoneme and the corresponding set of neurological commands to the articulatory muscles involved in producing that phoneme. Thus, the invariance and discreteness of the input phonemes is preserved until very near the end of the production process.

The motor command hypothesis has stimulated a number of investigations of the muscle activity occurring during speech. Many of these studies have involved implanting electrodes in various speech muscles. The electrodes are used to record the patterns of muscle contractions accompanying a phoneme when it is in different contexts.

(2) The vocal tract target hypothesis According to the hypothesis, the phonemic input to the speech production is translated into a motor representation of the vocal tract configuration that will produce that phoneme (this is the vocal tract target). The movements are “calculated” by the production system so that they will reach the specified targets from wherever the articulators are at the moment the movements are initiated.

The vocal tract targets are, therefore, represented in the brain of the speaker. In order for the targets to be represented, the speaker must have an internal representation of the vocal tract itself.

MacNeilage makes an analogy to the positioning of a door. If one wants to have a door open at an angle of forty-five degrees, one moves it one way if the door is closed or if it is open less than forty-five degrees, but another way if it is open more than that amount. And one moves it different distances depending on the discrepancy between the starting position and the target position. The same goes for articulation. The articulatory program specifies the target position of the articulatory organs, and the articulatory muscles compute the discrepancy between their present position and the target position and initiate a ballistic movement to eliminate that discrepancy. When speech is speeded up, it becomes more and more difficult to attain each of the successive targets, and so it becomes sensible to omit certain “distant” targets (like the [t] in wanta) and to coalesce certain adjacent target (like [t] and [y] in get ya) into a single movement.