Designing an Artificial Language: Lexical Semantics by Rick Morneau First Internet draft date: July 30, 1995 Current draft date: March 23, 1996 Copyright 1995, 1996, 1997 by Richard A. Morneau [Note: this monograph is still in the draft stage, and will be undergoing continuous revision. Ultimately, it will become the design document for an actual language that I plan to use as an interlingua for machine translation. A copy of the most recent version will always be available at ftp.eskimo.com in directory /u/r/ram/conlang as file LexicalSemantics. If you read and enjoy the monograph and have any helpful comments or suggestions, please feel free to contact me at ram@eskimo.com. If you do contact me, please quote sparingly from the monograph, since I have to pay long distance phone charges for my Internet connection.] 1.0 INTRODUCTION In the following sections, I would like to discuss word design in artificial languages (henceforth ALs). Specifically, how can a language designer apply semantic rigor to the design of words? One could spend a lot of time studying formal semantics (also called Montague, truth-conditional, or model-theoretic semantics), but I doubt if this will help much. Formal semantics as currently practiced is not only a pseudo- mathematical monstrosity, but it deals primarily with representation and analysis at the sentence level rather than at the word level. (One could also argue that this school of semantics doesn't deal with 'meaning' at ALL. But I didn't really say that. :-) Other semantic disciplines, such as the prototype theories of cognitive linguistics, can certainly be helpful, since they are inherently more suitable for dealing with the basic fuzziness of language. However, prototype theories do not provide us with the formal tools we need to design new words. And as for fuzzy logic itself, there are still too many wrinkles to be ironed out, especially as applied to natural language. A more practical solution, one that I feel is much better suited to designing new languages rather than analyzing old ones, is presented below. Basically, the solution I propose is to develop a simple but powerful derivational morphology that makes word design rigorous yet straightforward, while at the same time greatly reducing the number of basic morphemes (i.e. _primitives_) required by the language. Initially, I will not try to describe this method in abstract terms, since this discussion is intended for the non-linguist. Besides, I doubt that I would succeed. Instead, I will present the reader with many examples of various kinds of linguistic constructions, discuss the semantics of these constructions, introduce linguistic terminology where and as needed, and finally, try to derive some productive generalizations. I must warn you, though, that when all is said and done, this method is formal and rigorous. Those among you who are more interested in CREATIVE solutions to word design, or who wish to borrow words and meanings directly from existing natural languages, will not find this approach very useful or very interesting. 2.0 VERBS I'll start this exposition by looking first at verbs. Specifically, I will look at two of the most important criteria that go into defining a verb: its _valency_ (i.e. the number of basic arguments that it requires) and its _case requirements_ (i.e. the semantic roles played by the basic arguments). When combined, the valency and case requirements of a verb are usually referred to as the _argument structure_ of the verb. Before proceeding, though, let me give you a quick review of valency and case. Consider the following English sentence: The chimpanzee broke the window with a coconut. In this example, the verb "break" has a valency of two, since it requires two arguments: the subject "the chimpanzee" and the object "the window". The arguments are REQUIRED because, if either were missing, the resulting sentence would be ungrammatical (or, in the case of some verbs, would have a different meaning): *The chimpanzee broke. *Broke the window. [Please note that I am using the standard linguistic convention of indicating an unacceptable item by preceding it with an asterisk.] But the following is okay: The chimpanzee broke the window. For the verb "break", the case role of the subject is _agent_, and indicates the entity RESPONSIBLE for the event. The case role of the object is _patient_, and indicates the entity which EXPERIENCES the state or change of state described by the verb. In other words, the argument structure of the English verb "break" requires two arguments: the first argument (i.e. the subject) must be a semantic agent, and the second argument (i.e. the object) must be a semantic patient. Arguments required by a verb are called _core_ arguments. The phrase "with a coconut" is what is called an _oblique_ argument since it is not essential for the sentence to be grammatical. It simply provides additional peripheral information about what happened. In this sentence, it indicates the _instrument_ of the event. In other words, "a coconut" is the instrument used in carrying out the act indicated by the verb. If the sentence had been: The chimpanzee broke a thousand windows in Boston on Tuesday. "in Boston" would be a locative oblique argument, and "on Tuesday" would be a temporal oblique argument. [The case terminology that I am using here is fairly common, but not universal. Linguists who work with case grammar and thematic relations have yet to agree on the number and nature of case roles needed to adequately describe natural language. As it turns out, this lack of agreement is irrelevant to what we are trying to accomplish here. We will, in effect, create our own internally consistent, semantically precise, and easily expandable implementation of a case system.] In English, oblique arguments are usually _marked_ by preceding them with a preposition. Thus, the preposition is the marker which tells us the case role of whatever follows it. Agent and patient are almost always _unmarked_. The most common exception to this in English is in passive constructions, where the original subject is preceded by the preposition "by", as in "the window was broken BY the chimpanzee" or "the thieves were seen BY the children". Some verbs, such as English "put", have a third, required argument (i.e., it is part of the valency of the verb), which is marked by a preposition. For example: *He put the book. He put the book on the table. Here, the preposition "on" marks a _destination_ case role. Incidentally, natural languages often allow a speaker to omit a core argument if it is obvious from context. For example, a Japanese speaker often omits the agent of a verb as a sign of politeness. This usage, however, performs a discourse function - NOT a grammatical function - and the omitted argument is still assumed to be present. An additional case role that occurs within the valency of many verbs is what I will call _focus_. Linguists often call this case role _theme_, _object_, or _topic_, but there is no concensus, and their definitions often overlap other roles, especially patient. In all of the following examples, the direct object is the focus: The children saw the thief. The team needs a new coach. The woman remembered her father. The boys are playing baseball. The woman owns a beachhouse. The tarp covered the boxes. The fans enjoyed the game. The employees learned discretion. The man ignored his wife. The choir is singing a requiem. The boy loves his mother. The class is studying French. The fence surrounds three buildings. The old man told a story. Note that in each of the above sentences, the direct object provides a reference point or focus for the event, without causing or being changed by the event. It does this by pinpointing, narrowing down, or providing a reference for (i.e. 'focusing') the state or change of state indicated by the verb. Note that a focus does not play an active role in the event described by the verb, and is not obviously changed by the event. Thus, a focus can be best described as one of the following: 1. The entity on which the patient's attention or mental state is 'targeted' or 'focused'; e.g. to see, to play, to learn, to love, to tell, etc. 2. The referent of a relationship with the patient (i.e. the patient's state relative to the focus); e.g. to own, to surround, to include, to need, etc. 3. An elaboration of the event itself; e.g. to play, to sing, to tell, etc. Note that the concepts can overlap, as in "to need", "to avoid", "to know", and "to hate", since the object of such verbs can be considered the focus of a relationship OR of a mental state. In fact, without stretching the second definition too much, one could say that it applies to ALL focused events, even those involving perception or elaboration. For example, the sentence "John sees the forest" describes a relationship between "John" and "the forest", and the sentence "Louise sang a little ditty" describes a relationship between "Louise" and "a little ditty". Thus, we can say that the patient experiences a relationship whose referent is the focus. If the verb has an agent, then the agent is responsible for the relationship. The nature of the relationship is indicated by the meaning of the verb. It is important to keep in mind that the focus does not directly modify or interact with the patient. Perhaps the best and most useful generalization we can make is that the focus is the referent of a relationship with the patient, it is not affected by the event, and it is not responsible for the event. However, the precise meaning of the focus will ultimately depend on the meaning of the verb itself. Thus, it would appear that focus is not really a pure case role. Both agent and patient can be defined with semantic precision, while focus seems somewhat vague or even 'out-of-focus'. The reason for the vagueness is that it is possible to differentiate among the various senses of focus; e.g. the perceived entity ("to see"), the missing/lacking entity ("to need"), the locative reference point ("to surround"), an elaboration of the event itself ("to sing"), etc. But these senses never overlap for a particular verbal concept, and we would end up making distinctions that are never made in natural languages. Thus, focus IS a vague and general-purpose case role, but it is an essential one. In summary, the three major case roles that are capable of being included within the valency of a verb are: agent - the entity responsible for the event described by the verb patient - the entity which experiences the state or change of state described by the verb focus - the entity which acts as the referent of a relationship with the patient Thus, the agent is responsible for the event, the patient experiences the event, and the focus provides the referent for the state or change of state indicated by the event. [We will discuss the semantics of focus in more detail later on. First, though, we need to acquire a more substantial background in the semantics of verbs.] Note that an argument does not have to be a physical entity. It can also be an event. In the following examples, the direct object is the patient: We lengthened our trip. The police halted the procession. Bill chaired the seminar. Joe postponed the finance committee meeting. The army prevented the destruction of the village. The station repeated the broadcast. There are other case roles in addition to the ones I just mentioned, but they are all oblique (i.e., they are never required by a verb), or are only seen in a few verbs best categorized as oddballs ("to suit", "to elapse", "to postpone", etc.). I will discuss them as the need arises. For now, though, we have enough background to proceed with the discussion. In the following sections, I will discuss and classify a large and varied sample of English verbs, based on their semantics and their argument structures. While doing so, I will also introduce some of the terminology and the formal descriptive notation that I will be using throughout the remainder of this monograph. 2.1 VERB CLASSIFICATION - STATE VERBS Probably the largest group of verbs in English (or any language, for that matter) are called _state_ verbs, since they describe either an unchanging state of affairs or a change of state. Verbs which describe an unchanging or static situation are often called _stative_ verbs (do not confuse "stative" with "state"). Verbs which describe a changing or dynamic situation are often called either _process_ or _accomplishment_ verbs. Because linguists do not agree on the precise meanings of these terms, I will immediately abandon them and use the more generic expressions "static state verbs" and "dynamic state verbs". Let's start by looking at some static state verbs; i.e. verbs which describe a steady or ongoing state: The patients suffered. The boy sweated. The building shook. The baby slept. The fish stank. The stars twinkled. These verbs are all _intransitive_; i.e. they have a subject but no object. Also, each one describes the steady, ongoing state of the subject. Thus, the subject is the patient. From now on, I will refer to verbs of this type as "P-s", where "P" represents "patient" and "-s" indicates that the verb is a STATIC verb. Here are some more static state verbs with the form P-s: The plants were tall -> P-s verb = "to be tall" The door was closed -> P-s verb = "to be closed" The stew was salty -> P-s verb = "to be salty" The walls were blue -> P-s verb = "to be blue" The mouse was dead -> P-s verb = "to be dead" English speakers may be surprised to see adjectives and past participles being treated as descriptive verbs. However, words which describe steady states have just as much of a verbal nature as words which describe changes of state. The English verbs "to sleep", "to stink", "to twinkle", etc. illustrate this very well. In fact, many natural languages (e.g. Japanese, Korean, several Sino- Tibetan languages such as Mandarin Chinese, some Siouan languages, several Austronesian languages, and many native languages of Africa, Central America and South America) do not have true adjectives. Instead, these languages use words that are essentially intransitive verbs, and which can be inflected or otherwise used in the same way as any other intransitive verbs. Now, the above examples represent intransitive STATIC state verbs. Here are some examples of intransitive DYNAMIC state verbs: The window broke. The ice melted. The plants grew. The baby fell asleep. The mouse died. The stew cooled. The patient recuperated. The only difference between these and the previous examples is that the patient experiences a CHANGE of state rather than a steady state. Thus, these verbs are the dynamic counterparts of the intransitive static state verbs. From now on, I will refer to these verbs as "P-d", where "-d" indicates that the verb is a DYNAMIC verb. Next, let's look at some verbs which describe events in which the subject causes something to happen to the object. These verbs are all _transitive_; i.e. they have both a subject and an object. Here are a few examples: He cured the patient. He broke the window. He killed the mouse. He closed the door. He salted the stew. He captured the thief. In all of the above, the subject "He" is responsible for the event described by the verb. Also, in all cases, the event causes a CHANGE OF STATE to occur in the object. Thus, the subject is the agent and the object is the patient. In other words, these verbs are transitive dynamic state verbs. For verbs like these, I will use the notation "A/P-d", where "A" represents "agent", "P" represents "patient", a slash "/" separates subject from object, and "-d" indicates that the verb is a dynamic verb. Note that English, unlike almost all other languages, uses exactly the same word for SOME of its P-d and A/P-d verbs: P-d: The window broke. A/P-d: John broke the window. P-d: The patient healed. A/P-d: The doctor healed the patient. Note though, that this usage is highly idiosyncratic, and many words that you would expect to follow the pattern do not: A/P-d: The doctor cured the patient. P-d: *The patient cured. P-d: The patient recuperated. A/P-d: *The doctor recuperated the patient. A/P-d: The cat killed the mouse. P-d: *The mouse killed. P-d: The mouse died. A/P-d: *The cat died the mouse. In the design of your AL, you can, of course, mimic English. However, I do not recommend it, since it is extremely uncommon among the world's languages and it is almost certain to cause confusion among most non-English speakers. So far, we've seen P-s, P-d, and A/P-d verbs. Thus, an obvious question is: are there such things as A/P-s verbs? Yes. And as the designation implies, these verbs always indicate that the agent maintains the patient in some kind of steady state. Thus, all of these verbs imply that the agent somehow "controls" the patient. Here are some examples: He operated the lathe. He ruled the country. He conducted the orchestra. He chaired the symposium. He used the hammer. He managed the company. Note that, although these verbs may imply both an entry into and an exit from the event or situation, the major emphasis is on the process BETWEEN the endpoints. For these reasons, these verbs are static rather than dynamic. Now, for states that are normally rendered using adjectives, English uses the particle "keep" to distinguish between A/P-s and A/P-d verbs. Here are some examples: He kept the door open. A/P-s verb = "to keep open" He kept the girl alive. A/P-s verb = "to keep alive" He kept the thief captive. A/P-s verb = "to keep captive" He kept his mother happy. A/P-s verb = "to keep happy" All of the above are effectively A/P-s verbs. English simply uses the particle "keep" to achieve the desired effect. A good paraphrase of these 'verbs' is "agent causes patient to remain in a steady state". Next, let's look at some verbs that use the focus case role that we discussed earlier. Here are some examples: The student needs money. The boy misses his father. The company owns the yacht. The child has the coloring book. The report lacks a cover. The kids enjoy the game. The man loves his wife. The policeman sees the thief. The girls hear the music. In all of the above, the subject experiences a steady state relative to the object. Thus, the subject is a patient, the object is a focus, and the verb is a static state verb. For these verbs, I will use the notation "P/F-s", where "F" represents the focus. It is also possible to have verbs like these which also have an agent. Here are some examples: The boy wanted the money. The lady looked at the house. (Think of "to look at" as a single complex verb.) The boys obeyed the rules. The girls listened to the music. (Think of "to listen to" as a single complex verb.) The children followed their parents. The priest thought about his sins. (Think of "to think about" as a single complex verb.) In the above examples, the subject not only experiences the state indicated by the verb, but is also responsible for the state; i.e., the subject is also in control. Thus, the subject is both the agent AND the patient, and the object is the focus. I will refer to these verbs as AP/F-s. Incidentally, notice how some of the above complex verbs become simple verbs when they are de-focused: The lady is looking. The lady is looking at the house. The girls are listening. The girls are listening to the music. The priest is thinking. The priest is thinking about his sins. Thus, the unfocused verbs would be described as AP-s. Later, we'll have much more to say about the difference in semantics between focused and unfocused verbs. It is also possible for AP/F verbs to indicate a change of state. Here are some examples: Louise befriended her classmate. Mike joined the party. John memorized the poem. John entered the room. Mary divorced the jerk two years ago. The man disowned his oldest son. Bill left the building. These verbs describe a situation in which the agent causes HIMSELF to undergo a change of state relative to the focus. Thus, they are all AP/F-d. Since all of this may be confusing, let me paraphrase the relationships in a way that illustrates the states and how they are focused: P/F-s: John saw the mouse. = John experienced a visually perceptive state focused on the mouse. AP/F-s: John looked at the mouse. = John maintained himself in a visually perceptive state focused on the mouse. P/F-d: John noticed the mouse. = John entered a visually perceptive state focused on the mouse. AP/F-d: John glanced at the mouse. = John caused himself to enter a visually perceptive state focused on the mouse. P/F-s: The platoon heard the music. = The platoon experienced an aurally perceptive state focused on the music. AP/F-s: The platoon listened to the music. = The platoon maintained itself in an aurally perceptive state focused on the music. P/F-d: John remembered the party. = John entered a state of remembrance focused on the party. AP/F-d: The platoon surrounded the village. = The platoon caused itself to be in a state of 'around' focused on the village. P/F-s: He loved his father. = He experienced a state of loving focused on his father. P/F-d: She learned discretion. = She entered a state of knowledge focused on discretion. Overall then, verbs in this group can be generalized as follows: P/F-s: Static, subject = patient only (to hear, to love) X experienced a steady state focused on Y AP/F-s: Static, subject = agent & patient (to look at, to listen to) X maintained himself in a steady state focused on Y P/F-d: Dynamic, subject = patient only (to remember, to learn) X underwent a change of state focused on Y AP/F-d: Dynamic, subject = agent & patient (to glance at, to surround) X caused himself to undergo a change of state focused on Y Note that in all of the above paraphrases, the words "focused on" could be replaced by the words "relative to", emphasizing that the focus is the referent of a relationship with the patient. Now, some verbs involve the exchange of one item for another, usually between two people. Here are some examples: John swapped an apple for an orange with Bill. John sold Bill a book for $10. Bill bought a book from John for $10. John loaned Bill his tiller for $10. Bill rented a tiller from John for $10. In each case, two transfers of possession take place. John loses possession of one item while gaining possession of another, and the reverse change of possession occurs for Bill. Thus, we have, in effect, two patients and two foci, where the foci are the items being exchanged. We can also regard these verbs as composites; i.e. useful abbreviated versions of two distinct verbs, as in "John gave me his apple and I gave him my orange". Since both patients are equally responsible for the exchange, each one functions as both agent and patient. However, the subject in the above exchanges plays a more important or 'primary' role as agent than the other patient, and the first item plays a more important or 'primary' role as focus. Thus, for example, in the case of "sell", the seller is the primary agent- patient, while the buyer is the 'secondary' agent-patient. The object sold is the primary focus, and the amount paid is the 'secondary' focus. [This is not the only possible analysis, but I feel that it is the most practical. It also eliminates the need for any special treatment of exchange verbs that do not need a secondary focus, such as "to lend/borrow".] Finally, there are some cases where the subject is the ONLY agent-patient, as in "John swapped his brown tie for a blue one". Here, John causes himself to undergo a change of relationship with two different items, without the involvement of anyone else. In this example, "a blue one" is the secondary focus. There are also state verbs which are used to describe the weather and other environmental phenomena. Here are some examples: It's raining. It stinks in here. It's windy. It's cold outside. It's snowing. It's scary in there. It's humid today. It's dark in there. It's getting hot = it's heating up. It's getting cloudy = it's clouding up. It's quiet when the kids are at school. In this group of verbs, the subject is the null place holder "it". English verbs always require a subject in the indicative, but this is not true of most languages. Whether or not you require it in your AL is up to you, but it is not semantically meaningful, and I recommend against it. Note that verbs in this class can be either static or dynamic. Also note that, since these verbs describe states or changes of state, they have an IMPLIED patient which is obvious from the context (i.e. the local environment or current situation). In effect, English uses the pronoun "it" to represent the implied patient. I will not describe the argument structure of these verbs right now, because we do not yet have a sufficient background to treat them properly. Instead, I will postpone their discussion until after we discuss grammatical voice changes. 2.2 VERB CLASSIFICATION - DYNAMISM AND TELICITY So far, all of the verbs we have discussed are state verbs. That is, the basic concept represented by such a verb is some kind of state, and that this state applies only to the patient. The states can be focused or unfocused, and they can be brought about or maintained with or without an agent. Also, the states themselves can be categorized by their dynamism; i.e. a state can be "energetic" (e.g. 'alive', 'twinkling', 'sleeping', 'smelly', etc.) or "non-energetic" (e.g. 'dead', 'green', 'tall', etc.). In general, an energetic state can be described using an English present participle, and a non-energetic state can be described using an English adjective or past participle, but there are many exceptions. Verbs may also be categorized according to their _telicity_. Dynamic verbs that have a built-in endpoint are called _telic_, as in "The violinist played a dirge". Dynamic verbs that do NOT have a built-in endpoint are called _atelic_, as in "The violinist played with the local orchestra". Unfortunately, distinctions in dynamism and telicity are not very useful, and I know of no natural languages that mark these distinctions. Whether a concept is energetic or not is a basic part of the nature of the concept and has nothing to do with how the concept is applied. In other words, it is an inherent part of meaning of the verb root, and there is no need to mark it or express it externally. Also, the telicity of a verb often depends on the meaning of its arguments rather than on the form of the verb. Thus, in a derivational system such as I am proposing here, telic distinctions are useless. [Incidentally, this entire section is 'for your information only'. I felt that it was important to mention dynamism and telicity only because linguists attribute so much importance to these concepts in their theoretical discussions about verbs. In the process, they often end up weaving complex webs of rationalization that end up leading nowhere. I got caught in this web in my earlier studies in lexical semantics, but managed to extricate myself after a considerable waste of time. In my opinion, distinctions in dynamism and telicity are interesting but useless. And, as I will illustrate below, there is a much more important and useful distinction: the distinction between agent- oriented concepts and patient-oriented concepts.] 2.3 VERB CLASSIFICATION - ACTION VERBS State verbs are not the only kind of verbs that languages employ. There is one other class of verbs, which I will refer to as _action_ verbs, which differ significantly from state verbs. Let's look at a few examples and then see if we can deduce some useful generalizations: Louise told Bill a joke. Louise kicked Bill. Louise teased Bill. In each of the above examples, the subject "Louise" is clearly the agent. Also, in the first example, the second object is clearly the focus. But what is the object "Bill"? In each case, Louise is trying to have some kind of effect on Bill, but the final result is not clear. For example, when Louise kicks Bill, we know that something happens to Bill, but the actual outcome depends on many things that are left unstated, such as how hard she kicked, what kind of shoes she was wearing, where she kicked Bill, and so on. This is quite different from state verbs, where the final state is always clearly indicated by the meaning of the verb. For example, the sentence "He broke the window" makes it very clear what the final state of the window is; i.e. 'broken'. It doesn't tell us anything about the act itself or how it was accomplished. Now, we could say that Bill's final state is 'kicked', but this does not tell us about his condition - it simply tells us how it was accomplished. The reason why the final outcome of the above examples is not clear is because these verbs tell us about the act itself rather than the outcome of the act. In other words, these verbs emphasize what the agent is doing rather than emphasizing what is happening to the patient. Thus, state verbs are _patient-oriented_, since they highlight what the patient experiences. Action verbs are _agent-oriented_, since they emphasize what the agent is doing. If a root concept is patient-oriented, then the verb will indicate what the patient experiences. Patient-oriented verbs may or may not have agents. If the root concept is agent-oriented, then the verb will indicate what the agent is doing. An agent-oriented verb MUST have an agent. All patient-oriented verbs are state verbs. All agent-oriented verbs are action verbs. The most common action verbs are _speech acts_. Here are some examples: He advised his clients. He blessed the crowd. He told me a joke. He mocked them. He answered the teacher. He called me an idiot. He blamed John for the accident. He dared me to try it. In all of the above the first object is the patient, since it is the entity which the agent is trying to affect. For the verbs which have two objects, the second object is the focus. Thus, in the sentence "He told me a joke", "He" is the agent, "me" is the patient, and "a joke" is the focus. [Incidentally, verbs which have two objects are called _ditransitive_.] Now, a verb like "to mock" is different from a verb like "to kick" because "mock" is a speech act while "kick" is a physical act. However, this difference is not important for our purposes. A much more important difference is that "kick" implies success, while "mock" does not imply success. In other words, if you are kicked, then you experience a change of state, even though the verb itself does not tell us what the final state is. If you are mocked, however, the verb does not indicate if the mocking was successful. In fact, most speech acts do not even guarantee that the patient hears the agent. Thus, the speech acts involve an agent ATTEMPTING to affect, influence, or transfer information to a patient using speech. In effect, the patient can POTENTIALLY undergo a change of state, but the outcome is not certain. I will use the designator "-p" ("p" for "potential") for verbs which indicate that the patient may POTENTIALLY experience a change-of-state in the event described by the verb. Thus, the complete designation of the above speech acts are as follows: A/P-p: He advised his clients. A/P-p: He blessed the crowd. A/P/F-p: He told me a joke. A/P-p: He mocked them. A/P-p: He answered the teacher. A/P/F-p: He called me an idiot. Note that I'm using TWO slashes to separate the arguments of ditransitive verbs. Now, the astute reader will be asking why I chose to define "-p" verbs as I did above; i.e. as verbs which indicate that the patient is POTENTIALLY affected. After all, I spent a lot of time defining the difference between agent-oriented and patient-oriented concepts, but I never mentioned this difference when I defined the new "-p" class of verbs. The reason is that the orientation of a verb is an inherent part of the verb's root concept, and has nothing to do with how the root is used. For example, a verb root representing the agent-oriented concept 'kick' could be used to create the following verbs: A/P-d: to kick A/P-p: to kick at Here, the "-d" verb indicates that the patient experienced a change of state, even though the actual final state can only be guessed at. The "-p" verb, however, simply indicates that the agent TRIED to affect the patient, and says nothing about whether the agent was successful. Note though, that both the "-d" and "-p" versions are action verbs. The root concept is simply being applied differently. Similarly, a verb root representing the patient-oriented concept 'scratched' could be used to create the following verbs: P-s: to have a scratch or scratches P-d: to get/become scratched A/P-d: to scratch A/P-p: to scratch at In fact, some English verbs can be converted from a "-d" state verb to a "-p" verb using the preposition "at" and occasionally "on". Here are some examples: He cut the rope. -> dynamic state verb He cut at the rope. -> potential state verb He grabbed the rope. -> dynamic state verb He grabbed at the rope. -> potential state verb He shot the deer. -> dynamic state verb He shot at the deer. -> potential state verb He pulled the rope. -> dynamic state verb He pulled on the rope. -> potential state verb He tugged the rope. -> potential state verb Note that the dynamic state verbs imply that a final state is achieved as well as what the final state is. The potential counterparts indicate only that an attempt is made with no guarantee of the final outcome. In English, the above "at" construction is not always productive. However, the generic verbs "to have at", "to try", or "to work at" can be used if the combination of dynamic state verb plus "at" or "on" is either unacceptable or has an inappropriate meaning, as in the following examples: He opened the door. -> dynamic state verb He broke the door. -> dynamic state verb He painted the door. -> dynamic state verb He tore down the door. -> dynamic state verb He had at the door. -> potential state verb He tried the door. -> potential state verb He worked at the door. -> potential state verb In summary, it's important to be able to distinguish between "-s", "-d", and "-p" verbs, whether they represent patient-oriented or agent-oriented root concepts. In effect, a particular verb (either a simple verb, such as "kick", or a complex one, such as "kick at") can be categorized as one of six possible types: Actions verbs: -s: "to dictate" -d: "to kick" -p: "to tell" State verbs: -s: "to stink" -d: "to break" -p: "to tug" Finally, we mentioned earlier that the focus of a verb can be one of the following: 1. The entity on which the patient's attention or mental state is 'targeted' or 'focused'; e.g. to see, to play, to learn, to love, to tell, etc. 2. The referent of a relationship with the patient (i.e. the patient's state relative to the focus); e.g. to own, to surround, to include, to need, etc. 3. An elaboration of the event itself; e.g. to play, to sing, to tell, etc. We can now state a very important observation regarding the focus of action verbs: The focus of ALL action verbs MUST be (3) above. However, there can still be an overlap. Thus, although an action verb MUST be in category (3), it can also be in another category. For example, because "sing" is an action concept, the focus must elaborate the event, as in "John sang a little ditty". However, it can also fall into category (1), since the object "a little ditty" can be considered the focus of the mental state of the patient. There is another group of action verbs that are typically referred to as _activities_. Here are some examples: The children played (hide and seek). The athletes ran (the marathon). The guests danced (the polka). The old hag smoked (a pipe). The boy read (a good book). These verbs describe situations in which the agent maintains itself in an ongoing, energetic state. As a result, these verbs are all static AP/F-s verbs, and can be paraphrased as "Agent does something to maintain itself in a steady, active state". In effect, since the agent and the patient are the same, and since an action verb tells us what the agent is doing, it also tells us the state of the patient. In other words, the action and the state are essentially the same. [Incidentally, some readers might argue that the objects of the verbs "smoke" and "read" in the above examples do not elaborate the event, as is required of all action verbs. However, I disagree. Just because the objects "a pipe" and "a good book" are noun phrases representing physical objects does not mean that they cannot represent events. In fact, these simple noun phrases actually evoke complete events because of their inherent natures. A book does not serve its main purpose if it is not read and a pipe does not serve its main purpose if it is not smoked.] Now, many activity verbs CAN take an explicit patient that is not also the agent. Here are some examples: John played Bill three games of chess. The athletes ran their sneakers threadbare. His wife danced him into a stupor. She smoked us out of the house (i.e., her smoking caused us to leave). The boy read his sister a story. In these examples, we are still saying what the agent is doing while placing more emphasis on what is being done to someone/something else. Thus, these verbs are the A/P versions of the basic activities. [Incidentally, the word "threadbare" in the "run" example, and the expressions "into a stupor" in the "dance" example and "out of the house" in the "smoke" example are called _resultatives_, since they indicate the final or 'result' state of the patient. We'll have more to say about resultatives later.] It's important to emphasize that, when dealing with action concepts, we cannot treat AP derivations as we did with state verbs. In an AP state derivation, the agent is causing itself to experience the state that normally applies only to the patient. In an AP action derivation, the agent is causing the patient to perform the action that is normally performed only by the agent. In other words, in an AP state derivation, the agent EXPERIENCES the same thing (i.e. state) as the patient. In an AP action derivation, the patient DOES the same thing (i.e. action) as the agent. Thus, an AP-s version of a verb such as "to kick" does NOT mean that the agent kicks himself. Instead, it means that the agent is simply "kicking"; i.e., he is involved in the activity of "kicking" with no specified or discernable target. This is a subtle distinction, but it is an extremely important one. [Incidentally, this distinction could also be handled by designating the above verb as simply A-s rather than AP-s. However, I have chosen to keep the AP notation because of the inherent symmetry of the distinction, and because it emphasizes that the agent is causing itself to experience what is essentially an energetic "state".] 2.4 GENERALIZATIONS ABOUT VERBS Now, let's look at some of the distinctions that exist among these categories, and see if we can make some generalizations about verbs. In looking over the above groupings, we can draw the following conclusions: 1. All verb concepts are either: a. Patient-oriented -> the root describes the ongoing or final state of the patient. b. Agent-oriented -> the root describes what the agent is doing. 2. All verbs are either: a. Static verbs -> these indicate that the patient experiences a steady state. b. Dynamic verbs -> these indicate that the patient experiences a change of state. c. Potential verbs -> these indicate that the patient MAY experience a change of state. 3. The subject of a verb can be any of the following: a. Agent b. Patient c. Both agent and patient d. Nothing 4. The object of a verb can be any of the following: a. Patient b. Focus c. Nothing 5. Some English verbs take three arguments. In these cases, the subject is the agent, the first object is the patient, and the second object is the focus. 6. All verbs have a patient, whether stated or implied. As mentioned earlier, there are a few oddballs which have unusual argument structures, but these are rare and tend to be irregular or idiosyncratic. For the time being, we will limit our discussion to the larger, more regular categories. [Actually, as we will see throughout this monograph, the so-called 'oddballs' can ALWAYS be derived from more regular verbs via some form of grammatical voice change or re-classification.] From the above list, we might be tempted to create a matrix of 2x3x4x3x2 = 144 elements. However, most combinations never appear. Note, for example, that the orientation of the verb is an inherent part of the meaning of the root, and we will never find two verbs that differ ONLY in this characteristic. Also, a patient can be the subject OR the object - not both - which, of course, makes sense. And if the first argument is both agent and patient, then the second argument cannot be a patient. Also, it serves no useful purpose to have a verb with an object but with no subject. And so on. With all of the above in mind, we can construct a chart of the possible forms that verbs can take: ARGUMENTS STATIC DYNAMIC POTENTIAL -------------------------------------------------------------------- A/P/F to conduct to teach to tell A/P to manage to cure to shoot at AP/F to ignore to memorize to study AP to behave to escape P/F to see to recall P to stink to recuperate none to be cloudy to cloud up Note that I left most of the "potential" category empty. We WILL be creating many of these verbs, but to list them now would only cause confusion. Note also that I have excluded verbs that take instrumental subjects (e.g. "The hammer broke the window"). English is one of the very few languages that allows constructions like this. And those few that do allow this generally mark the verb to indicate that the subject is instrumental (e.g. Malagasy, many Bantu languages, many Philippine languages, etc.). I'll have more to say about this later. I have also excluded verbs that can take a _beneficiary_ as an indirect object, as in "He baked the kids a cake". A beneficiary is someone who may only be indirectly affected by the event, unlike the patient which is directly affected. I'll discuss the beneficiary case role later. 2.5 VERB DESIGN ALGORITHM So, how do we apply these generalizations to the practical problem of verb design? Answer: we do it by CLASSIFYING and MARKING our verbs (in some way or other) to indicate their valency, case requirements, and whether or not they reflect a steady state, change of state, or potential change of state. The easiest way to do this, in my opinion, is to design the morphology of the language to reflect these differences. For example, the following English verbs will all be derived from the same root but will have different markers to indicate their different argument structures: AP-d to escape = Agent causes self to become free AP/F-d to escape from = Agent causes self to become free relative to focus A/P-d to release, to free, to liberate = Agent causes patient to become free A/P/F-d to release from, to free from = Agent causes patient to become free relative to focus P-d to get loose, to become free = Patient becomes free P/F-d to get loose from, to become free of = Patient becomes free relative to focus AP-s to stay free, to remain free = Agent keeps self free P-s to be free = Patient is free And so on. For all of the above, we can use a state root with the meaning 'free' or 'unrestrained', and can apply a different marker to indicate whether the result is AP-s, A/P-d, etc. Now, let's illustrate this approach in greater detail by precisely defining the morphology that we will use, and providing lots of examples of how to use it. Here is a formal description of the sample morphology that we will use: Word ::= Root + ( Classifier ) + Part-of-speech Root ::= Morpheme Classifier ::= Morpheme Morpheme ::= C D | C V { X } Part-of-speech ::= Terminator Terminator ::= da | no | pe | si C = any consonant (p, b, t, d, k, g, c, j, l, m, n, f, v, s, z, x) D = any diphthong (ai, au, eu, ia, ie, io, iu, oi, ua, ue, ui, uo) V = any vowel (a, e, i, o, u) S = any semi-vowel (w, y) X = extension = S V | C C V | = logical 'or' () = enclosed item is optional {} = enclosed item may appear zero or more times Lower case letters represent themselves If you have difficulty understanding the above formal description, I suggest that you read my separate essay entitled "Morphology". It provides a brief and simple tutorial on how to describe the shapes of words and morphemes. The above is a portion of the complete morphological system that I will be using throughout this monograph. Additional features will be introduced as needed. Appendix A contains a complete description, and Appendix B contains a list of all of the morphemes that will be created and used in the monograph. Pronounce vowels as in Italian, Swahili, or Japanese (i.e. the five cardinal vowels /a/, /e/, /i/, /o/, and /u/). Pronounce consonants as in English, except for the following: "c" is like "ch" in "church", and "x" is like "sh" in "she". In two-vowel clusters, you may optionally pronounce "i" like the semi- vowel "y", and "u" like the semi-vowel "w". The diphthong "ui" should be pronounced /wi/ and "iu" should be pronounced /yu/. Main stress should be applied to the next-to-last syllable. Secondary stress should be applied to the second syllable if there is at least one unstressed syllable between it and the syllable that receives the main stress. [Incidentally, a diphthong is simply a VSV in which the semi-vowel has been deleted; the semi-vowel 'w' is deleted if either V is 'u', and the semi-vowel 'y' is deleted if either V is 'i'. Thus, for example, "-io-" is actually "-iyo-", "-eu-" is actually "-ewu-", "-ui-" is actually "-uwi-", "-iu-" is actually "-iyu-", and so on.] For verbs, I will use the terminator "-si" to indicate the part-of-speech. Thus, examples of well-formed verbs are: baposi xendisi moyakuesi taskesi guyondetayusi ba-po-si xendi-si moya-kue-si taske-si guyonde-tayu-si where hyphens show the boundaries between roots, optional classifiers, and terminators. For example, the word "baposi" consists of the root "ba", the classifier "po", and the terminator "si". The word "xendisi" consists of only the root "xendi" and the terminator "si" - it does not have a classifier. Incidentally, note that a word with the above morphology can always be parsed unambiguously into its component morphemes, and that a stream of words can always be divided unambiguously into individual words even if there are no spaces between words. Thus, the boundaries between morphemes and words is never in doubt. This feature of word morphology is usually referred to as either _self-segregation_ or _auto-isolation_. Because of the self-segregation rules, a morpheme cannot have the form of a terminator. For example, a root with the form "-da-" is illegal because "da" is a terminator. However, "-daye-", "-daspo-", and "-sunda-" ARE legal because the extensions "-ye-", "-spo-", and "-nda-" create unique morphemes. For verb classifiers, let's use the following morphemes: A/P/F-s: -tue- A/P/F-d: -ko- A/P/F-p: -nio- A/P-s: -zoya- A/P-d: -pu- A/P-p: -ce- AP/F-s: -fi- AP/F-d: -sua- AP/F-p: -mi- AP-s: -panji- AP-d: -za- AP-p: -diu- P/F-s: -ma- P/F-d: -do- P/F-p: -gui- P-s: -se- P-d: -pia- P-p: -moncu- Now, before proceeding, let's briefly review the semantics behind the notation we are using. All verbs have a patient, whether stated or implied. If a verb has an agent, then the agent is responsible for the event described by the verb. If a verb has a focus, then the focus is the referent of a relationship (or potential relationship) with the patient. This referent can be either another entity, as in "John needs a pencil", or it can elaborate the event itself, as in "John told a joke". A verb is either an agent-oriented action verb or a patient-oriented state verb. An action verb emphasizes what the agent is doing rather than what the patient is experiencing. A state verb emphasizes the ongoing or final state of the patient rather than how it came about or how the agent, if any, brought it about. An action verb MUST have an agent. A state verb may or may not have an agent. If an action verb has a focus, then the focus MUST elaborate the event. Finally, it is extremely important to note that many states are so strongly scalar that it makes no sense to think of them in absolute terms. For example, if I say that something is 'long', I mean that it is long relative to some referent understood by both the speaker and the listener. When deriving dynamic verbs, we will always use state concepts such as these in a relative sense, rather than in an absolute sense. Thus, an A/P-d verb formed from the state concept 'long' will mean 'to lengthen' - it will NOT mean 'to make long'. This convention will be followed rigorously in the derivation of ALL state verbs that indicate a change in an inherently scalar state. 2.5.1 VERB DESIGN EXAMPLES For these examples, I'm going to start with an English verb, analyze it to determine its argument structure, and create a word for it in our sample language. I will then try to create as many other verbs as possible from the SAME root by using different classifiers. Let's start with the verb "to know", in the sense of 'having knowledge'. Typical sentences using this verb could be: He knows the answer. or He knows calculus. Here, the subject is the patient and the object is the focus. The subject experiences a steady state of 'knowledgeable' focused on the object. Thus, this verb is a patient-oriented state verb and is classified as P/F-s. Now, in our sample language, I will assign it the root morpheme "-teyo-". Thus, the root "-teyo-" will represent the state concept that means 'knowing' or 'knowledgeable'. Next, if we add the classifier "-ma-" to indicate that the argument structure is P/F-s, and add the terminator "-si" to indicate that the word is a verb, then the resulting word "teyomasi" is the verb meaning 'to know'. We will also adopt the convention that each root will have a default class if none is explicitly provided, and that the default class will depend on the meaning of the root. Since 'knowing' is inherently relational, the default class for "-teyo-" will be P/F-s. Thus, the words "teyosi" and "teyomasi" are synonymous. [Later, we will discuss how to shorten many words even further when we make the distinction between 'high' and 'low' versions of the language. The high language will be used in formal occasions and when talking to computers (because computers require a self-segregating morphology). The low language can be used when talking informally to people. In the low language, the word "teyomasi" can be further shortened to simply "teyo".] Next, let's take the same root and see what happens when we apply different classifiers to it. Note that the English glosses are approximations - I'll have more to say about this later. We will deal first with focused verbs, since the concept of 'knowing' is inherently focused: A/P/F-s: "teyotuesi" = 'to keep (someone else) current in (something)' Agent maintains patient's knowledge. e.g. He keeps them up-to-date in company procedures. A/P/F-d: "teyokosi" = 'to teach (someone) (something)' Agent causes patient to gain knowledge. e.g. He taught them French. A/P/F-p: "teyoniosi" = 'to instruct (someone) in (something)' Agent attempts to cause the patient to gain knowledge. e.g. He instructed them in table manners. AP/F-s: "teyofisi" = 'to review', 'to keep oneself current in (something)' Patient maintains his knowledge. e.g. He reviewed the day's lessons every evening. [Note that this verb implies that the patient is successful in maintaining his knowledge. Thus, the English word "review" is not a perfect match.] AP/F-d: "teyosuasi" = 'to self-teach (something)', 'to determine/learn' Patient causes himself to gain knowledge. e.g. He taught himself French OR He learned French on his own. He determined the meaning with the help of a dictionary. AP/F-p: "teyomisi" = 'to study', 'to attempt to teach oneself (something)', 'to try to determine/learn' Agent attempts to cause himself to gain knowledge. e.g. He studied the new lesson. P/F-s: "teyosi = teyomasi" = 'to know (something)', 'to understand (something)' Patient is knowledgable. e.g. He knows the rules of the game. P/F-d: "teyodosi" = 'to learn (something)', 'to come to know' Patient gains knowledge. e.g. He learned the rules by watching the others. P/F-p: "teyoguisi" = 'to be exposed to knowledge of (something)' Patient has an experience that may potentially increase his knowledge. e.g. He was exposed to French for three years, but never learned a word of it. Keep in mind that the above English glosses are approximations, and that the real meaning should be determined from the root plus its argument structure. For example, "teach" implies that the teaching was successful because a dynamic state verb indicates that a change-of-state did, in fact, occur, while "instruct" does not imply success, but only that there was potential for success. With the precisely defined semantics used above, there is no doubt. Also, keep in mind that the paraphrases cannot capture the IMMEDIACY of the involvement of the participants. This immediacy can only be represented by the single word - NOT by the paraphrase. For example, a paraphrase of the verb "to kill" is 'to cause to die', even though the two are not synonymous. The paraphrase is simply the closest we can get to the true meaning using multiple words. Please keep this in mind, since we will be using paraphrases throughout this monograph. Note that all of the above derivations are focused. Focused derivations are the most useful simply because the concept 'knowing' is most often applied this way. But the unfocused derivations are also very useful, as we'll see later, when we discuss the difference between focused and unfocused concepts. Before we can discuss these differences, though, we need to acquire a little more background in the semantics of verbal concepts. Now, let's wrap up this section. In the above examples, we managed to derive the following useful, non-synonymic English verb lexemes from a single root morpheme: teach, study, review, learn, determine, know, and instruct We also created the basis for deriving the following non-verb lexemes: knowledge, education, pedagogy, erudition, determination, and several others that can be further derived, such as knowledgeable, pedagogical, heuristic, etc. We'll see how to do these later. In all, though, we will be able to derive a large number of non-synonymic English lexemes from a single root morpheme. Not bad, huh? Also, we were able to deal with concepts such as 'to keep/stay up-to-date', 'to self-teach', etc. for which English must use periphrasis, metaphors, or even idiom. 2.5.2 MORE ON DESIGN PHILOSOPHY Well, I hope you're still with me. And if you think I've gone off the deep end, then I suggest that you stop reading now, because I plan to carry this basic approach even farther. In self-defense, however, I might remind you that ALL of the constructions above have functional counterparts in natural languages. For example, Japanese, Swahili, Hungarian, Korean, Turkish, and many, many other languages use an inflectional morpheme to indicate causation ('kill' vs. 'die'). Romance languages use different auxiliaries and/or reflexive pronouns to indicate transitivity. Some Semitic languages (e.g. Arabic), quechuan languages (e.g. quechua), Malayo-Polynesian languages (e.g. Indonesian), and Bantu languages (e.g. Swahili) employ a rich set of inflectional or derivational morphemes to indicate transitivity, reflexivity, passivity, causation, reciprocity, etc. Iroquoian languages such as Cherokee and Mohawk explicitly mark their verbs to indicate transitivity, as well as whether the arguments of a verb are agent, patient, or both. [In fact, there are many languages from completely unrelated language families, such as Lakhota (Siouan), Acehnese (Sumatran), and Batsbi (Caucasian) that mark their verbs to indicate which argument of the verb is the semantic agent and which is the semantic patient in ways that are almost identical to what we are doing here.] Other languages, such as Swahili, Mokilese, Ainu, Fijian, Tagalog, and Somali, mark their verbs inflectionally and/or derivationally to change their argument structure (i.e. to add or change a thematic role required by the verb). Quite a large number of languages (e.g. Caucasian languages of Central Asia, Salishan languages of North America, and others) go so far as to mark a verb when it does not take an argument that it normally takes (e.g. "He is smoking" vs. "He is smoking a cigar"). Russian has two derived forms for most of its verbs which, when applied to many speech acts, captures the distinction we are making here between A/P/x-d and A/P/x-p verbs. Mayan and related languages mark their verbs for agency, transitivity, and stativity in a manner that is remarkably similar to what we are doing here. The difference, of course, is that no single natural language uses such a precise, comprehensive and consistent derivational morphology for ALL of its verbs. Keep in mind, though, that we are talking about the design of an ARTIFICIAL language, and the more regularity we build into it, the easier it will be to learn. Another thing to keep in mind is that this approach is derivational, NOT inflectional, and only the word designer has to spend the time needed to master the system. Thus, the language learner does not have to memorize inflections or derivations. Also, even though some of the above constructions may appear useless (a debatable judgement, in my opinion), we still managed to create several useful words from a single root morpheme. And, as we will see below, the actual number of useful creations will be MUCH higher. Finally, this rigorous approach to verb design has some interesting consequences that may not be immediately obvious. In using this kind of approach, you will find that many of the words you create have close (but not quite exact) counterparts in your native language. For example, the English verb "study" is the agentive/potential counterpart of the verb "learn", yet the following two sentences are pragmatically identical: 1. I went to MIT to learn math. 2. I went to MIT to study math. By "pragmatic", I mean that a listener hearing one of the above sentences would not come away with a different understanding if he had heard the other. However, there are cases where the two words cannot be interchanged, as in the following examples: 3. As I got older, I gradually learned discretion. 4. *As I got older, I gradually studied discretion. In other words, even though the meanings of the two words can overlap, their prototypical interpretations are distinct. With the rigorous approach to word design discussed here, these distinctions will be enforced to the degree that the first two sentences above (1 and 2) would have distinctly different meanings - number 1 would imply passivity since the subject is NOT an agent, while number 2 would imply industry since the subject IS an agent. This approach has interesting and highly desirable consequences. If you do NOT use a rigorous approach to word design, and attempt to create a vocabulary that duplicates every nuance of every natural language, your vocabulary will be essentially infinite in size (also, the task will be quite impossible). If you create your vocabulary based only on what you know of English or some other natural language, your vocabulary will simply be the original in disguise, with all of its idiosyncrasies. If, however, you design your vocabulary using the semantic techniques discussed here (or something equally rigorous), you will not only maximize its inherent neutrality, but you will also make it much easier to learn. The penalty, of course, is that the words you create may not precisely overlap, in meaning and usage, their closest equivalents in your native vocabulary. But this lack of precise overlap is exactly what you ALWAYS experience whenever you study a different language. So, do you want to create a clone of an existing vocabulary with all of its idiosyncracies? Or do you want to maximize the neutrality and ease-of-learning of the vocabulary of your AL? You can't have it both ways. In fact, it is this aspect of vocabulary design that seems to frustrate so many AL designers, who feel that they must capture all of the subtleties of their native language. In doing so, they merely end up creating a clone of the vocabulary of their natural language. The result is inherently biased, semantically imprecise, and difficult to learn for speakers of other natural languages. It is extremely important to keep in mind that words from different languages that are essentially equivalent in meaning RARELY overlap completely. Fortunately, all of this does NOT mean that your AL will lack subtlety. In fact, with a powerful and predictable derivational morphology, your AL can capture a great deal of subtlety, and can go considerably beyond any natural language. The only difference is that, unlike a natural language, the subtleties will be predictable rather than idiosyncratic. Another major advantage of a system like this is that it will make the vocabulary building process faster and easier. In effect, we will be able to design our vocabulary by using a 'back door' approach; i.e., we will start with a powerful derivational system (of which we've only seen a tiny part so far), and iteratively decompose words from a natural language and apply all possible derivations to the resulting root morphemes. In doing so, MANY additional useful words will be automatically created, making it unnecessary to decompose a large fraction of the remaining natural language vocabulary. In the above example, we started with the verb "to know" and derived several useful words from its basic root morpheme. This is only the tip of the iceberg. The rest of this monograph will illustrate how to derive many more additional words from the same root morpheme. 2.5.3 FROM BASIC VERB TO NOUN The sample morphology I am using here requires that all verbs end with "-si". Let's extend the morphology so that all nouns end with "-da", and so that any verb can be converted to a noun by simply changing the final "-si" to "-da". The question then becomes: how do we interpret the result? Here is the rule that I feel is most productive: When converting a basic verb to a noun, the noun will represent a PROTOTYPICAL GENERIC SUBJECT of an event indicated by the verb. Here's a short list of the most useful derivations (I leave the detailed analyses as an exercise for the reader): A/P/F-d: "teyokoda" = 'teacher' A/P/F-p: "teyonioda" = 'instructor' AP/F-p: "teyomida" = 'student' P/F-s: "teyoda" = 'knower', 'the one in the know' P/F-d: "teyododa" = 'learner' It would also be useful to have passive-like forms such as 'generic-someone- who-is-taught' = "pupil", 'generic-something-which-is-learned (without agency)' = "experience", 'generic-something-which-is-taught' = "subject/course/ curriculum", 'generic-something-which-is-known' = "knowledge", etc. To obtain these, we can apply a class-changing morpheme to the verb to make it passive, and then change the final "si" to "da". I'll have more to say about how to do this later. An instrumental derivation could be used to create a 'generic-instrument- used-to-teach' = "teaching materials", 'generic-instrument-used-to-maintain- knowledge' = "review materials", 'generic-instrument-used-to-teach-oneself' = "self-study materials", etc. Later, I'll have much more to say about these additional derivations. 2.5.4 FROM BASIC VERB TO ADJECTIVE Again, let's extend the morphology so that all adjectives end with "-no", and so that any verb can be converted to an adjective by simply changing the final "-si" to "-no". The question again becomes: how do we interpret the result? Here is the rule that I feel is most productive: When converting a basic verb to an adjective, the adjective will represent the prototypical QUALITIES of a generic subject, expressed attributively. This meaning can be best paraphrased as "having the attributes of one who VERBs or of something which VERBs". Here's a short list of the most useful derivations: A/P/F-d: "teyokono" = 'having the attributes of one who teaches' = 'teaching' (literally: 'having the attributes of one who causes others to increase in knowledge') (example: "He was a teaching locksmith") AP/F-p: "teyomino" = 'having the attributes of one who studies' = 'student/studying' (literally: 'having the attributes of one who attempts to cause self to increase in knowledge') (example: "He was a student geologist") AP/F-d: "teyosuano" = 'self-taught' P/F-s: "teyono" = 'knowing', 'in the know' It is important to note that the use of present participles (e.g. "teaching") and past participles (e.g. "self-taught") to represent the actual meanings is misleading, because English participles have strong implications of tense and aspect. For non-participial renderings, this is not a problem as in "a student geologist". Whereas "a studying geologist" carries an implication of present tense and imperfective aspect that should not actually exist in the adjective. Also, for similar reasons, do not confuse adjectives with relative clauses. For example, a "student geologist" is not quite the same as a "geologist who studies/studied/etc" since the relative clause DEFINITELY specifies tense and aspect. As with nouns, it will be useful to perform class derivations to create even more words. For example, an instrumental derivation could be used to create 'quality-of-a-generic-instrument-used-to-teach' = "pedagogical", 'quality- of-a-generic-instrument-used-to-study' = "heuristic", etc. We'll discuss these additional derivations later. 2.5.5 FROM BASIC VERB TO ADVERB To continue along the same lines as above, we will adopt the rule that all adverbs end with "-pe". However, before we can put this to use, we must first digress for a while and discuss the semantics of case tags and adverbs. 2.5.5.1 FROM BASIC VERB TO VERBAL MODIFIERS In this section, I would like to discuss the semantics of adverbs (especially those that correspond to English adverbs that end in "-ly") and most case tags (such as English prepositions, Japanese postpositions, Hungarian case inflections, etc.), and I will try to show how verbs can be converted to adverbs and case tags. The final result will be a system that can replace many complex, idiosyncratic and periphrastic constructions of natural languages with constructions that are syntactically simple and semantically transparent. 2.5.5.2 SEMANTICS OF CASE TAGS AND RELATED ADVERBS First, let me illustrate how verbs can, in fact, represent the semantics of English prepositions, adverbs, and particles by giving examples from other languages. In these languages, some verbs are actually used in the same way as English prepositions, adverbs, and particles. Consider the following from Vietnamese: (1) Toi di lai nha bang. I go to bank I'm going to the bank. (3) Nha bang o Hanoi... bank in Hanoi The bank in Hanoi... In the first example, the word "lai" is actually the verb 'to come'. When used transitively, it takes a destination as a direct object (like the English verb 'to enter'). In the second example, the word "o" is actualy the verb 'to be located at' and takes a location as a direct object. (Thus, the second example could also stand alone as a complete sentence meaning 'The bank IS IN Hanoi'.) Many other languages, such as Igbo, Ewe, Twi, and Yoruba (Niger-Congo languages of west Africa), Indonesian, Chinese, Cambodian, and many pidgins and creoles have similar constructions. Also, these constructions are not limited to locatives. In Chinese, for example, the word "yung" is the verb meaning 'to use'. It is also the preposition meaning instrumental 'with', as in the sentence "He broke the window WITH a hammer". It's also possible to create adverbs, particles, and completely new verbs in this manner. In Hindi, for example, "to run go" means 'to run away', and "to cook take" means 'to cook for oneself'. In Yoruba, "to carry come" means 'to bring', and "to carry go" means 'to take away'. Linguists have a name for this type of construction, in which two or more verbs are linked without the use of coordinating conjunctions or subordinators. They are called _serial verbs_. There are two major types of serial verb constructions: the events indicated by the verbs are either simultaneous or consecutive. In this discussion, we are only interested in the first category, where the two verbs represent events that occur simultaneously. Other useful serial verb constructions are those in which two or more verbs are linked, all taking the same subject and object. In these cases, the lack of a conjunction or subordinator often implies a certain 'immediacy'; i.e., that the event is a single entity, rather than a combination of unrelated or sequential events. Some languages, such as Chinese and Yoruba, allow any combinations that make semantic sense, and even allow noun phrases to split the verbs, creating an effect similar to relative clauses, but where the events indicated by the verbs are often much more tightly linked. Note that these types of constructions are not idiomatic - they are actually quite productive and their meanings are predictable from syntax and context. What most serial verb constructions have in common is that they are taken by speakers as representing parts of the same event. English has a few verbs that can be used in this way, such as "to go look", "to come see", "to let go", "to stir-fry", and "to test-fly" but note that the first two represent consecutive events, which is not what we are interested in here. Most of the time, English uses participles to achieve a simultaneous effect. Here are some examples, where the first sentence of each triplet indicates simultaneity: The child ran screaming to his mother. vs. The child who ran to his mother was screaming. vs. The child who was screaming ran to his mother. The man woke up shivering. vs. The man who woke up was shivering. vs. The man who was shivering woke up. The boy stumbled, knocking over several chairs. vs. The boy who stumbled knocked over several chairs. vs. The boy who knocked over several chairs stumbled. The girl slept, dreaming of unicorns. vs. The girl who slept dreamt of unicorns. vs. The girl who dreamt of unicorns slept. What seems to be happening here is that the participial phrase is more closely linked to the verb rather than to the noun it ostensibly modifies. As a result, we can create what are essentially compound verbs without subjects, and the results make perfectly good sense: to run screaming to wake shivering to stumble knocking over several chairs to sleep dreaming of unicorns In effect, the words "screaming" and "shivering" behave exactly like adverbs, and the words "knocking over" and "dreaming of" behave exactly like case tags (i.e. English prepositions) that introduce phrases that modify the verb. Thus, we should be able to create adverbs and case tags from verbs by applying the same semantic logic. Here's are some examples: I broke the window using a hammer. I broke the window with a hammer. to use: A/P-s The kids ran, crossing the road. The kids ran across the road. to cross: AP/F-d They came, tagging along (i.e. accompanying an unspecified focus). They came along. to tag along: AP-s The army positioned itself, surrounding/encircling the town. The army positioned itself around the town. to surround: AP/F-d or P/F-s or AP/F-s The car moved slowly, backing up. The car moved slowly backwards. to back up: P-d He visited his parents, staying three days. He visited his parents for three days. to stay: P-s or AP-s Additionally, if English had a verb like Vietnamese "o", Chinese "zai", Cambodian "niw", or Hausa "yana" (all of which mean 'to be located at or in'), we could create the locative senses of the prepositions "in" and "at" from it. For example, if the English word "bain" meant 'to be located in/at', we would have: The children were playing, baining the backyard. The children were playing in the backyard. to bain: P/F-s This really should not be all that strange to English speakers, since this is essentially how several English prepositions came about; i.e. a combination of the verb 'to be' plus a locative morpheme. Some examples are "before", "behind", "between", "beneath", etc. Some adverbs were also formed in this way, such as "below" and "beyond". In summary, speakers of languages with serial verb constructions effectively make up new 'prepositions' as they are needed. If a preposition with a desired literal meaning is not available, English speakers will either use existing prepositions metaphorically, or will use participial constructions as illustrated above. In this monograph, we will implement a system that has the flexibility of the serial verb constructions (but which is semantically and morphologically precise), and thus avoid the need for potentially untranslatable metaphor. 2.5.5.3 DESIGNING CASE TAGS AND RELATED ADVERBS As an example of the adverb/case tag creation process, let's continue where we left off when we started this digression, and create a set of adverbs and case tags from the state concept of 'knowledgeable'. As mentioned earlier, we will do this by changing the final "-si" of the verb to "-pe". Those whose verb forms do NOT take direct objects will become adverbs, and those which DO take direct objects will become case tags (i.e. English prepositions) adding a new oblique argument to the main verb. Thus, in effect, the case tag will LINK its argument to the verb. In the following examples, I will use English for all words except the new case tag/adverb. I will also use English word order. Here are the results: A/P/F-s: "teyotuepe" = 'keeping (someone else) current in' e.g. The company spends a lot of money teyotuepe its employees the latest technology. [Note that "teyotuepe" has two objects. Thus, there is no need for the preposition "in".] A/P/F-d: "teyokope" = 'teaching (someone) (something)' e.g. The man stood in front of the class teyokope the boys geometry. A/P/F-p: "teyoniope" = 'instructing (someone) (in/on something)' e.g. He spoke softly teyoniope the audience the new company policies. AP/F-s: "teyofipe" = 'reviewing', 'keeping oneself current in' e.g. They spent the night at John's house teyofipe the lessons for the next day's exam. AP/F-d: "teyosuape" = 'teaching oneself (something)' e.g. He stayed up late every night teyosuape French. AP/F-p: "teyomipe" = 'studying (something)' e.g. He stayed at that university for three years teyomipe physics. P/F-s: "teyope = teyomape" = 'knowing (something)' e.g. Joe quietly left the room teyope he would be called on next. P/F-d: "teyodope" = 'learning (something)', 'coming to know' e.g. He watched their activity for three hours teyodope valuable information. P/F-p: "teyoguipe" = 'being exposed to knowledge of (something)' e.g. He lived in a jungle teyoguipe how to survive. In all cases, note how the derived case tag or adverb modifies the whole sentence, just as if it were an oblique argument of the main verb. Note also that the case tag or adverb is usually tightly bound to one of the core arguments of the main verb, and that this core argument is predictable from the argument structure of the case tag. For example, in the sentence: Joe quietly left the room teyope (= 'knowing') he would be called on next. the subject of the case tag "teyope" is P and links to the patient of the main verb "to leave" which itself is AP/F-d. Thus, the effective subject of the case tag "teyope" is "Joe". And in the sentence: The man stood in front of the class teyokope (= 'teaching') the boys geometry. the subject of the case tag "teyokope" is A and links to the agent of the main verb "to stand" which is AP-s. Thus, the effective subject of the case tag "teyokope" is "the man". If the subject of a case tag is AP, it will link to the agent of the main verb; i.e., the effective subject of an AP case tag or adverb will always be the agent of the main verb. This is simply because agents are inherently more salient than patients or foci. If the main verb does not have an agent, then it will link to the patient. Finally, it is important to note that linkage to an argument of the main verb is common but not universal. Since it is possible for the argument of a verb to be an event (as opposed to an entity), it is also possible that a case tag can provide a linkage between its argument and the 'event' represented by the main verb. In other words, the case tag could link to the main verb PLUS its core arguments rather than to just one of its arguments. We'll have more to say about this later. In this section, we discussed how to convert existing verbs into adverbs and case tags. Later, we will discuss how to systematically create the many case tags required by a language, such as those needed to represent English prepositions. 2.5.6 DESIGNING ACTION VERBS All of the previous examples involved patient-oriented root concepts; i.e. the derivations were applied only to STATE verbs. In doing so, we always described the semantics of the derivations in terms of the ongoing or final state of the patient. The semantics of ACTION verbs, however, are quite different, since agent-oriented root concepts place emphasis on what the agent is doing rather than on what the patient is experiencing. Thus, a paraphrase of the semantics of a state verb will be different from a paraphrase of the semantics of an action verb. The following chart lists and clarifies these differences: State verbs: -s -> The patient experiences a steady state -d -> The patient experiences a change of state -p -> The patient potentially experiences a change of state If a state verb has an agent, then the agent is directly responsible for the event. Action verbs: -s -> The agent does something that maintains the patient in an unspecified steady state. (The agent is in control of the patient.) -d -> The agent does something that changes the state of the patient, but the final state is not explicitly indicated. -p -> The agent does something that potentially changes the state of the patient. We've already spent a lot of time on the analysis of state verbs. Let's briefly look at a few action verbs: A/P-d: "to kick" Agent 'kicks', having an unspecified effect on the patient. A/P-p: "to kick at" Agent 'kicks', attempting to affect the patient. A/P/F-p: "to tell" Agent 'speaks' something (= the focus), potentially having an unspecified effect on the patient. A/P/F-p: "to sing" Agent 'sings' something (= the focus), attempting to affect the patient. A/P/F-s: "to sing" Agent 'sings' something (= the focus), and DEFINITELY controls or maintains the patient in an unspecified steady state. A/P/F-d: "to sing" Agent 'sings' something (= the focus), and DEFINITELY caused the patient to experience an unspecified change of state. Of the last two examples, the "-s" form would be used to emphasize the control the agent had over the audience while singing, as if the audience were somehow mesmerized. The "-d" form would emphasize the change that the singer causes the listener to experience. For example, if the "-d" form were used in the sentence "She sang the child a lullaby", it would imply that the child did, in fact, fall asleep. In other words, the "-s" form implies successful control over the patient, while the "-d" form implies successful change of the patient. Note that these not-so-subtle distinctions can be made easily and with perfect regularity using the semantic system I am proposing here. However, to make these distinctions in a natural language like English requires either explicit elaboration or idiosyncratic periphrasis. At first glance, it would seem that non-agentive versions of action verbs would be useless. However, if we treat a P-x or P/F-x action derivation as the non-agentive equivalent of the AP derivation (as opposed to the non-agentive version of the A/P or A/P/F derivation), then it will have the added meaning that the activity was done by the patient, but was done accidentally, without actually trying, or without intent. In other words, the patient performed the activity but was not fully responsible or in control. In general, action concepts are not as productive as state concepts, because only a small number of argument structures can be usefully applied to an action concept. For example, an action verb must always have an agent. Thus, all of the agentless argument structures, such as P-s or P/F-d, are not going to be very useful (except as discussed above). Also, most of the physical actions, such as 'kick', indicate a very rapid change of state. Thus, the "-s" forms will not be very useful. [Actually, even 'kick' can have a useful "-s" derivation, as in "He kicked the ball all the way home". Here, "all the way" does not need to be explicitly stated because the "-s" form of the verb implies that he controlled the ball over a period of time by kicking it.] Fortunately, this will not have a serious effect on the overall productivity of the semantic system being presented here, because action verbs are quite rare compared to state verbs. Also, we will find that many verbs that appear to be actions at first glance, can actually be derived as "-p" state verbs. The net result is that we will be able to derive most verbs of a language from a surprisingly small number of roots. In spite of the above, I have no intention of abandoning action verbs, and, throughout the remainder of this monograph, I will be deriving many action verbs and doing further derivations on these verbs. (In fact, we will derive some very useful action verbs in the very next section.) Finally, there will be times when the orientation of a root concept is not clear. When this happens, do a complete derivation assuming BOTH orientations, and use the one that is most productive. For example, is 'work' an action or a state verb? I first assumed that it was inherently agent-oriented. However, very few useful derivations resulted, which I thought was strange for such a universal concept. However, by treating it as the patient-oriented state 'working/having a job/employed', I got several useful results: A/P-s: to employ The company employs 22 engineers. A/P-d: to hire The company hired 22 engineers. A/P-p: to recruit for We're recruiting for 22 engineers. P-s: to work/have a job He has a job at the butcher shop. P-d: to get a job He got a job last Friday. P/F-s: to work as He works here as a foreman. P/F-d: to get a job as He got a job as a foreman. AP-s: to be self-employed He's self-employed. AP/F-s: to be self-employed as He's self-employed as a plumber. [The English preposition "for", as in "he works FOR the company", would be the beneficiary case tag, which we will discuss later.] Finally, it's important to keep in mind that an AP version of an action verb is an _activity_ that emphasizes what the agent is doing without specifying an external patient. Thus, an AP/F-s version of the verb "to sing" would be used in a context such as "What's she doing? She's singing a little ditty". 2.6 GENERIC VERBS All of the basic verb classifiers represent concepts that are useful in their own right; i.e., they can be used to create useful verbs WITHOUT root morphemes to indicate states or actions. For this purpose, we can attach the terminator directly to the classifier. In effect, the classifier becomes a generic state root with the same class as the classifier. Here are examples of the most useful derivations: A/P-s: zoyasi - 'to keep/maintain' e.g. The warlord KEPT the village in a state of terror. [Note that the compound preposition "in a state of" marks what is called an oblique 'state' case role. This role is the static counterpart of the dynamic resultative case role that I mentioned earlier. I'll have much more to say about state case roles later.] A/P-d: pusi - 'to change' (transitive), 'to have an effect on', 'to affect' e.g. The divorce HAD AN EFFECT ON him. The wizard CHANGED the prince into a frog. [Here, the preposition 'into' marks the oblique dynamic 'state' case role; i.e. the _resultative_ case role.] A/P-p: cesi - 'to try to affect/change' e.g. He TRIED TO CHANGE her attitude, but she remained stubborn. AP/F-s: fisi - 'to keep oneself in an unspecified steady state with respect to', 'to stick with', e.g. He STUCK WITH the project. AP-s: panjisi - 'to persevere', 'to remain steadfast' e.g. He REMAINED STEADFAST/PERSEVERED until the end. AP-d: zasi - 'to change one's ways', 'to transform oneself', 'to cause oneself to change' e.g. He CHANGED HIS WAYS after the trial. AP-p: diusi - 'to try to change (oneself)' e.g. He TRIED TO CHANGE but remained a nerd. P/F-s: masi - 'to be in a relationship with', 'to be involved with', 'to have something to do with', 'to be in an unspecified steady state with respect to' e.g. Bill HAS SOMETHING TO DO WITH the new project. P/F-d: dosi - 'to enter a relationship with', 'to become involved with', 'to become associated with', 'to come to have something to do with' e.g. He BECAME INVOLVED WITH the new project. P-s: sesi - 'to be in an unspecified steady state', 'to experience something', 'something is happening/going on with P' e.g. SOMETHING'S HAPPENING/GOING ON WITH Bill. P-d: piasi - 'to change' (intransitive), 'to undergo a change' e.g. The office CHANGED since I was last here. Now, we can also derive generic ACTION verbs using a generic agent-oriented root morpheme. Our sample language will use the morpheme "-ze-" for this purpose; i.e., to represent a generic action. Keep in mind that action verbs emphasize what the agent is doing rather than what the patient is experiencing. Also, the focus of action verbs ALWAYS elaborates the event itself. Here are some of the more useful derivations of generic action verbs (in the sample language, "-ze-" will be A/P-s by default): A/P-s: zesi = zezoyasi - 'to run/operate/use/control' e.g. He CONTROLLED the company. John USED the hammer to break the window. [Note that the English word "use" often has a sense of 'abuse', especially if the object is sentient. The generic "zesi" does not have this sense, unless the context makes it inevitable.] A/P-d: zepusi - 'to do something to' e.g. Billy DID SOMETHING TO the cat. A/P-p: zecesi - 'to try (something)', 'to have a go at', 'to have at' e.g. He TRIED the stuck door. AP/F-s: zefisi - 'to do (something)', 'to keep oneself busy/occupied with (something)' e.g. He IS DOING his homework. AP-s: zepanjisi - 'to be doing something', 'to be busy/occupied' e.g. He IS BUSY right now. As we will see later, many of the above generic verbs can undergo additional derivation to produce some very useful words. 2.7 GRAMMATICAL VOICE So far, we've only talked about verbs in the active voice; i.e., where all of the arguments of a verb are present and appear in the proper order. For example, the A/P-d verb "to break" has an agent subject and a patient direct object. However, natural languages have many ways of changing the relative positions of the arguments of a verb in order to change their relative importance or topicality. Languages can also remove arguments from the argument structure, while implying that they still exist, and make the missing arguments either obliquely expressable or not expressable at all. Finally, languages can also incorporate normally oblique arguments, making them part of the argument structure of the verb. For example, consider the following: John broke the window. = active voice The window was broken. = passive voice, implied agent The window was broken by John. = passive voice, oblique agent The window was broken with a hammer. = passive voice, oblique instrument, implied agent A hammer broke the window. = incorporated instrument, agent cannot be expressed at all (*by John), new structure is something like I/P-d, where I = instrument. Those windows broke easily. = middle voice, implied agent, agent cannot be expressed at all (*by John). The windows broke. = P-d verb. This is sometimes confused with middle voice. In the system described in this monograph, this verb is a basic verb and the example is in the active voice. No agent is expressed or implied. John is the breaker (of the window) or John did the breaking (of the windows). = anti-passive (this is an approximation - English does not have a true morphological anti-passive construction). The agent alone is prominent. The patient loses its prominence but may be expressed obliquely. However, even when not expressed obliquely, a patient is always implied. The window broke John. (poetic license needed here) or The window, John broke it. = inverse voice (again, these are approximations - English does not have a regular inverse construction). Patient becomes subject, agent becomes object and MUST appear. Different languages handle these distinctions in different ways. As you can see from the above examples, English uses combinations of syntax, morphology, periphrasis, and even poetic license. Other languages are more regular, some using inflections for some voices, while others may use derivations or a combination of both. In addition, some languages allow the incorporation of other case roles into the argument structure of a verb. In fact, the number of possible voice variations among the world's languages is quite large. Since grammatical voice has different meanings to different people (with middle voice being the most confused/confusing), let me precisely define the meaning that I am using here. Specifically, A grammatical voice change starts with a basic verb and increases the topicality of one core argument relative to another. In the process, an existing argument may be deleted. A deleted argument may be expressed obliquely (e.g. passive) or may not be expressable at all (e.g. middle). However, the role of the deleted argument is ALWAYS implied. Thus, even though the original subject may not be expressed in a middle voice construction, it is still implied. For example, in "Mice kill easily", someone or something is responsible for the killing even though it cannot be expressed. In "Mice die easily", no agent is expressed or implied. Thus, the former is an example of a grammatical voice change, while the latter is not. An argument that increases in relative topicality is said to be _promoted_, and an argument that decreases in relative topicality is said to be _demoted_. Demoted arguments continue to play their original semantic roles, but are somehow less important or less involved. The following examples illustrate this effect: Active: The enemy bombed the city. Passive: The city was bombed. <- no agent or The city was bombed by the enemy. <- oblique agent Active: She sewed the dress. Anti-passive: She did the sewing. <- no patient or She did the sewing on the dress. <- oblique patient Although the number of possible voice combinations is large, there are a few that crop up often among the world's languages. Here are the most common ones: Active - transitive: The subject is slightly more important or topical than the object. Both must be expressed. This is by far the most common form used in almost all languages. [The only exceptions I know of are Fijian and the Salish languages of northwestern North America. In these languages, all transitive verbs are derived by addition of an affix to the intransitive form. Also, in Fijian, the most commonly used verb form is active INTRANSITIVE. (Shades of Sapir-Whorf!)] Passive: The original object becomes the subject and becomes considerably more topical than the original subject. The original subject is no longer part of the verb's argument structure, and does not have to be expressed. However, it is always implied and may be expressed obliquely (in English, typically using the preposition "by"). Middle: The original object is made more topical and becomes the subject. The original subject is deleted from the verb's argument structure. The original subject is implied, but is so unimportant that it can NOT be expressed obliquely. Only the original object and the event indicated by the verb remain important. Anti-passive: The subject is made considerably more salient than the object. The original object is no longer part of the verb's argument structure, and does not have to be expressed. However, it is always implied and may be expressed obliquely. Inverse: The arguments of the active verb are simply reversed. The original object becomes the subject, gaining in importance; and the original subject becomes the object, losing importance. Unlike passive, the original subject is not oblique and MUST appear. Keep in mind that the above are generalizations. Individual languages vary both in the ways that the various voices are implemented as well as in their semantics. Also, keep in mind that the list contains just the most common voice systems. Many other combinations are possible, especially those involving normally oblique case roles. As we saw above, a language like English, which does not have this ability, must resort to complex and idiosyncratic constructions to achieve the same effect. Always keep in mind, though, that a voice change simply re-arranges the topicality of some of the participants in a sentence. Our goal should be to achieve the same results in a consistent and easy-to-understand manner. Also, English rarely uses the same strategies to handle these needs. For example, an effect similar to (but not exactly the same as) that of passives and anti-passives can be achieved by using impersonal constructions: "Johnson punched someone" (anti-passive) or "They don't make good cars anymore" (passive). An effect similar to (but not exactly the same as) the inverse can often be accomplished by fronting or left dislocation, as in "(As for) the car, John wrecked it". However, true inverse effects can sometimes be obtained by periphrasis, as in: Active: A 10-year old can read the book. Inverse: The book is readable by a 10-year old. Active: The cup is full of water. Inverse: Water fills the cup. Finally, inverse and middle effects are sometimes achieved in English by using completely different root morphemes, as in "I enjoyed the show" vs. "The show pleased me" (inverse), or by use of metaphor or idiom, as in "He remembered the answer" vs. "The answer came to mind" (middle). [Incidentally, the inverse voice comes in two varieties. In the first, which is sometimes called a _semantic inverse_, an inverse operation may be required in order to properly assign case roles to the arguments of a verb. Semantic inverse constructions are especially common in the native languages of North America. For example, in Plains Cree (Algonquian), a more animate argument is inherently more topical than a less animate argument, and neither word order nor case marking of nouns can change the interpretation. Thus, if "man" and "dog" appear as the main arguments of the verb "bite", then it will always be interpreted as "man bites dog", regardless of word order. An inverse marking on the verb simply reverses the relative topicality, making "dog" more topical than "man", and is REQUIRED to obtain the sense "dog bites man". I do not consider this usage a true voice alteration. It is simply an uncommon way of marking semantic case roles in a sentence. Similarly, some Sino-Tibetan languages have an inverse voice based on the relative topicality of 1st, 2nd, and 3rd person, rather than animacy. Note though, that although an inverse operation may at times be required, it can also be used when it is not required in order to achieve the changes in topicality that we are describing here. In these cases, such an operation is called a _pragmatic inverse_. True pragmatic inverses can be found in languages such as Maasai (Nilo- Saharan), Sahaptian languages (Penutian, western North America; e.g. Nez Perce), Caucasian languages (e.g. Georgian), and Chamorro (Austronesian, Guam). (In fact, Maasai and Sahaptian languages have both semantic AND pragmatic inverses.) Finally, a combination of word order changes and direct case marking of nouns can sometimes be used to achieve an inverse effect (e.g. Korean). However, other languages which have this ability (e.g. Russian) frequently use it for quite different purposes. As for true inverse systems, recent research indicates that such systems are actually much more common among the world's languages than had been previously supposed.] 2.7.1 IMPLEMENTATION OF A GRAMMATICAL VOICE SYSTEM Most European languages (including English) use cumbersome rules involving auxiliaries, participles, reflexives, context, word-order, and even complete lexical changes to indicate voice. More heavily inflected languages (Arabic, Latin, Japanese, Ainu, etc.) use the very simple expedient of inflecting the verb for most indications of voice. Many South American lowland languages and some isolating (i.e. uninflected) languages such as Chinese and Vietnamese do not have a formal morphology or syntax to cover voice, although they can achieve similar effects via explicit topicalization and/or periphrasis. Finally, other languages such as the Bantu languages of Africa (e.g. Swahili) and Austronesian languages (e.g. Indonesian) use derivational morphemes (which is essentially what we are doing here) to achieve most voice effects. In other words, they create a completely different verb from the same root as the active verb, but the new verb has a different topicalization and argument structure. So, how should an artificial language implement grammatical voice? Ideally, we would like to create a system that can handle any voicing needs, while being both simple and consistent. I will say no more about syntactic approaches, since this monograph is strictly about morphology and semantics. (Besides, I do not feel that grammatical voice change should be implemented in syntax - syntax is not nearly as flexible as morphology.) Instead, I recommend that all grammatical voice changes be implemented using derivational morphology. One possible way to accomplish this would be to allocate a single _CLASS-CHANGING MORPHEME_ for each voice, which would be used in addition to the active classifier. The resulting verbs will, of course, have a different argument structure. In the sample language, here is the new morphology that I will use: Word ::= Root + ( Classifier ) + ( Class-Changing Morpheme ) + Part-of-speech Root ::= Morpheme Classifier ::= Morpheme Class-Changing ::= Morpheme Morpheme Morpheme ::= C D | C V { X } Part-of-speech ::= Terminator Terminator ::= da | no | pe | si C = any consonant (p, b, t, d, k, g, c, j, l, m, n, f, v, s, z, x) D = any diphthong (ai, au, eu, ia, ie, io, iu, oi, ua, ue, ui, uo) V = any vowel (a, e, i, o, u) S = any semi-vowel (w, y) X = extension = S V | C C V | = logical 'or' () = enclosed item is optional {} = enclosed item may appear zero or more times Lower case letters represent themselves For example, if the state root meaning 'closed/shut/unopened' is "-benzo-" (default class = A/P-d), then the word for the A/P-d verb 'to close/shut' is "benzopusi = benzosi". We can implement the other voices as follows: middle: -de- -> benzodesi e.g. Windows benzodesi easily = Windows close easily. passive: -nu- -> benzonusi e.g. The window benzonusi (by the thief) = The window was closed (by the thief). anti-passive: -ga- -> benzogasi e.g. The thief benzogasi (of the window) = The thief did the closing (of the window) or = The thief was the closer (of the window). inverse: -vi- -> benzovisi e.g. The window benzovisi the thief = The window - the thief closed it. where optional oblique arguments are shown in parentheses. In the above examples, the inverse paraphrase is only approximate, and actually increases the topicality of the fronted item more than it should. A better example of a true inverse effect in English would be: Active: John owns the book. Inverse: The book belongs to John. where "to belong to" should be thought of as a single complex verb (rather than as a verb plus the case marker "to"). Note that the second sentence is a true inverse of the first, and is only roughly approximated by the paraphrase "The book - John owns it". A useful notational scheme would be to put an implied case role in square brackets, with a plus "+" or minus "-" sign to indicate whether it can be expressed obliquely. Thus, middle: -de- changes A/P-x to P-x [-A] AP/F-x to F-x [-AP] P/F-x to F-x [-P] passive: -nu- changes A/P-x to P-x [+A] AP/F-x to F-x [+AP] P/F-x to F-x [+P] anti-passive: -ga- changes A/P-x to A-x [+P] AP/F-x to AP-x [+F] P/F-x to P-x [+F] inverse: -vi- changes A/P-x to P/A-x AP/F-x to F/AP-x P/F-x to F/P-x where "-x" represents either "-s", "-d", or "-p". For verbs that take three arguments, I feel it is most useful to do the following: middle: -de- changes A/P/F-x to P/F-x [-A] e.g. *The students taught French easily. [This is ungrammatical in English with the intended meaning, but grammatical in the sample language.] passive: -nu- changes A/P/F-x to P/F-x [+A] e.g. The students were taught French (by Mr. Johnson). anti-passive: -ga- changes A/P/F-x to A/F-x [+P] e.g. He shouted obscenities (at the crowd). [Note that the English verb "to shout" is inherently anti-passive. Thus, in the system proposed here, we would start by creating an A/P/F-p version of this verb, and then perform an anti-passive operation to derive an exact equivalent of the English verb "to shout".] inverse: -vi- changes A/P/F-x to P/A/F-x e.g. The student - John taught him geometry. In addition, some languages, such as Latin, Shona (Bantu), Turkish, Classical Greek, and German allow _impersonal_ passives, in which an intransitive verb is passivized becoming a zero-argument verb. For example, the AP-s verb "to run" could be passivized to 0-s [+AP] or 0-s [-AP], depending on the language, where "0" is used to indicate that the verb has no arguments. It would be interpreted as something like 'running took place' or 'there was running'. A verb like P-d "to grow" could become 0-d [+/-P], and would mean something like 'growing took place' or 'there was growth'. You could allow your language to use the same voice changing morphemes to achieve similar results. Don't let your native language be a straightjacket! Try to accomplish as much as possible with as small a vocabulary as possible, and with as simple a syntax as possible. Another useful derivation would be to take an A/P/F verb and reduce the topicality of the THIRD argument. (Remember, the anti-passive discussed above reduces the topicality of the SECOND argument.) In the sample language, this derivation will be as follows: anti-anti-passive: -miu- changes A/P/F-x to A/P-x [+F] We'll see examples of how to use this later. Grammatical voice alterations are also useful for creating speech act verbs which never take a patient as a direct object, such as the A/F-s [+P] verb "to dictate", as in "He dictated the letter (to his aide)". For verbs like these, we can create a verb that DOES allow a direct object patient, and promote a focus to direct object by means of the anti-passive alteration. We can also derive exact equivalents of intransitive English verbs such as "to complain" (A-p [+P] [+F]) via a DOUBLE anti-passive voice alteration of the basic A/P/F-p form. For example, using the basic form, a sentence like "He complained me the new equipment" would be grammatical. A SINGLE anti-passive alteration would allow a sentence like "He complained the new equipment TO me", and a DOUBLE anti-passive alteration would create an exact equivalent of the English verb, as in "He complained TO me ABOUT the new equipment", where "TO" is an oblique patient case tag and "ABOUT" is an oblique focus case tag. In fact, this double anti-passive operation is so useful that I have created a single voice-changing morpheme to implement it in the sample language, rather than using the anti-passive morpheme twice: double anti-passive: -kua- changes A/P/F-x to A [+P] [+F] At the same time, we can introduce a morpheme to perform a double PASSIVE: double passive: -jau- changes A/P/F-x to F [+A] [+P] For example, the single passive would be used to create the equivalent of the English word "bespoken", while the double passive would be used for the word "spoken". The double passive will be especially useful with speech act verbs. Finally, I know of a few languages such as Finnish, Ika (Lower Niger), Breton, and Alabama-Koasati (Muskogean) which implement impersonal constructions in morphology. In an impersonal construction, an argument is REDUCED in topicality while the topicality of the other argument is left unchanged. Most languages, however, implement impersonal constructions using impersonal pronouns, as in the following English examples: THEY don't make cars like THEY used to. ONE should never assume anything. SOMEONE broke the window. THEY gave him SOMETHING for the pain. YOU/ONE should always drive carefully. If you decide to implement impersonal constructions in morphology, you will need class-changing morphemes to perform the impersonal SUBJECT alteration: X/Y-x -> 0/Y-x [-X] and the impersonal OBJECT alteration: X/Y-x -> X/0-x [-Y] Note the notational distinction between the impersonal subject alteration result "0/Y-x [-X]" and the middle voice alteration result "Y-x [-X]". Later, I will discuss how to implement impersonal constructions that use impersonal pronouns. 2.7.2 MORE ON MIDDLE VOICE The passive, anti-passive, and inverse voices are easy to understand, and I'll say no more about them. Middle voice, however, is so frequently confused with basic intransitivity that I'd like to say a little more about it. English does not have a formal morphosyntax for middle constructions, unlike many other languages (Persian, Swahili, Basque, Somali, Hausa, Turkish, and many, many others - middle forms in these languages often go by other names, such as statives or agentless passives, but they often function semantically as middles). English does not even have a reflexive clitic construction, as do several other European languages, which often performs additional duty for middle voice. This is unfortunate, since, as we will see, it can be extremely useful and productive. English SOMETIMES allows an active verb to be used without modification in a middle construction, as long as the context forbids an active interpretation. Thus, we can say "The joke did not translate well" or "The plane landed ten minutes ago". But even when the meaning is clear, English can be quite idiosyncratic as in "*The mountains see in the distance" or "*The boxes are covering in the storeroom". And in cases where context and semantics do not make it clear, English is forced to use periphrastic constructions, completely different words, metaphor, and even idiom. Consider the following examples: ACTIVE MIDDLE I see the mountains. *The mountains see. The mountains are in view. Thus, from the verb "to see", P/F-s, we can derive: "to be in view", F-s [-P] The gang terrorized the *The neighborhood terrorized for three neighborhood for three years. years. The neighborhood lived in a state of of terror for three years. Thus, from the verb "to terrorize", A/P-s, we can derive: "to live in a state of terror", P-s [-A] That woman buys caviar only *Caviar buys only when it's on sale. when it's on sale. Caviar sells only when it's on sale. Thus, from the verb "to buy", AP/F-d, we can derive: "to sell (intransitive sense only)", F-d [-AP] He threw the rock at the window. *The rock threw at the window. The rock went flying at the window. Thus, from the verb "to throw", A/P-d, we can derive: "to go flying (metaphorically)", P-d [-A] I remembered her face. *Her face remembered. Her face came to mind. Thus, from the verb "to remember", P/F-d, we can derive: "to come to mind", F-d [-P] He swallowed the pills *The pills swallowed with difficulty. with difficulty. The pills went down with difficulty. Thus, from the verb "to swallow", A/P-d, we can derive: "to go down", P-d [-A] And so forth. The number of possible examples is almost unlimited. Thus, English CAN deal with middle concepts, although the forms are usually highly irregular, unpredictable, periphrastic, and often either metaphoric or idiomatic. Finally, middle verbs are often confused with basic P-x state verbs (English- speaking linguists often make this mistake). The reason is that the patient is the subject of an intransitive verb, and it is often uncertain whether or not the transitive subject is implied. In languages which have a formal middle voice, however, there is never any doubt. Unfortunately, speakers of languages like English will have to be a little more careful. When in doubt, the basic P-x form should always be used instead of the middle form unless a transitive subject is clearly implied. Middle verbs are also often confused with reciprocals and reflexives because some languages (especially European languages) use the same forms for more than one voice. In the semantic system being proposed here, middles, reflexives, and reciprocals are completely different. [We will discuss reflexives and reciprocals in great detail later, in the section on class-changing morphemes.] 2.7.3 INCORPORATING OBLIQUE CASE ROLES Some natural languages can make almost any case role a subject or object of the verb (e.g. Malagasy, some Mayan languages, and most Philippine languages). In fact, among the Philippine languages, verbs almost always have an explicit morpheme that indicates the case role of the subject, and almost any case role can be promoted to subject. Many Bantu languages of Africa (e.g. Swahili) and some Australian languages (e.g. Dyirbal) allow an instrumental case role to be promoted to object. Many Bantu languages also allow a locative case role to be promoted to subject. Indonesian allows a beneficiary case role to be promoted to object. And so on. Obviously, the above system could be easily extended to add normally oblique case roles to the argument structure of a verb. However, I will NOT be doing this in the sample language for the following reasons: 1. It is extremely rare among natural langauges. 2. The number of possible combinations of argument position and case role is very large, and would require a large number of class-changing morphemes that would rarely be used. 3. Most languages that allow promotion of normally oblique case roles have special reasons for doing so. For example, many languages allow relativization of only certain core arguments, and thus a voice change is REQUIRED before other arguments can be relativized. For all of the above reasons, I feel that it is not necessary to implement grammatical voice changes that would promote normally oblique case roles to core positions. In fact, as we'll see later, ALL oblique case markers will be derived from verbs, and these verbs can be used directly instead of case tags (i.e. prepositions) when we need to increase the topicality of the associated role. Thus, while there must be a way to modify the relative topicalities of core arguments, there is simply no need to create special morphemes to promote normally oblique arguments. 2.7.4 NOMENCLATURE OF GRAMMATICAL VOICE There are two voice changing operations that demote an argument: passive and middle. A passive voice change demotes an argument but allows it to be expressed obliquely. A middle voice change demotes an argument but does NOT allow it to be expressed obliquely. If the prefix "anti-" is NOT used, the first argument (i.e., the subject) is demoted. If a single "anti-" prefix is used, the second argument (i.e., the first object) is demoted. If two "anti-" prefixes are used, the third argument (i.e., the second object) is demoted. The modifier "double" indicates that the same voice operation is performed twice. For example, an anti-middle demotes the second argument, and suppresses its salience so much that it cannot be expressed obliquely. An anti-anti-passive demotes the third argument and allows it to be expressed obliquely. And so on. 2.7.5 ENVIRONMENTAL VERBS One of the major uses of a grammatical voice change is to take an existing argument and make it implied but inexpressible. Since environmental verbs often have an implied but non-statable patient, they are ideal candidates for voice derivation from a more basic form. Consider the following (assume that the root morpheme for 'hot' is "-xau-", and that the default class is P-s): They heated the room. transitive = A/P-d = xaupusi 'to heat', 'to raise the temperature of' The room heated up. intransitive = P-d = xaupiasi 'to heat up', 'to rise in temperature' It heated up (outside), It got hotter out, It warmed up. 'it heated up', middle voice = 0-d [-P] = xaupiadesi The room is hot. 'to be hot', intransitive = P-s = xausi = xausesi It's hot (out). 'it is hot (in the implied environment)', middle voice = 0-s [-P] = xaudesi = xausedesi Note that it is also possible to create passive forms, such as: The room was heated up (by the janitor) => xaupunusi = P-d [+A] It got hotter (in the room) => xaupianusi = 0-d [+P] And so forth. Other environmental verbs can be formed in the same way, including verbs that explicity indicate the presence of some kind of physical entity, such as "to rain" or "to snow". [We'll have more to say about how to derive such verbs later.] 2.7.6 DISJUNCTS When using verbs, we must be careful not to confuse case roles. It is sometimes easy to mistake an event for a patient. Consider the following example: It's sad that John died. It is tempting to treat the embedded sentence "John died" as if it were the patient in a P-s state verb formed from the root meaning 'sad'. However, an event cannot be "sad" in the sense that it can experience sadness. What we are really describing are the feelings of the speaker (and perhaps others) towards the situation. Thus, when we say "it's sad that ...", we are really describing our feelings or beliefs about the situation. In effect, the speaker and those he may be speaking to are the real, implied patients. Thus, in a sentence like the above, the true patient is implied, and the mental state of the patient is 'focused' on the event indicated by the embedded sentence. Thus, the embedded sentence is the FOCUS of the main state verb. We can easily create a basic P/F-s verb meaning 'to be sad about', as in the sentence "Bill is sad about his parents' divorce". Using this basic verb, we can perform a middle voice alteration to create the F-s [-P] form meaning 'it is sad that'. It is also possible for an event to be the agent or cause of the sadness. For this, we would need an A/P-s verb, since the event itself causes the patient to be sad. Thus, we really have several possible forms, as illustrated below: A/P-s John's death makes (i.e. keeps) me sad. A/P-d John's death saddened me. P/F-s I am sad that John died. F-s [-P] Sadly, John died. F-s [+P] It's sad for everyone that John died. A similar analysis can be done using the state concept 'hoping': P/F-s I hope that I'll win. F-s [-P] Hopefully I'll win. where "hopefully" is actually a verb that takes a complete embedded sentence as an argument - it is NOT an adverb as in English. Words and expressions like these are called _disjuncts_, and many other examples can be derived in the same way: "to presume" -> "presumably", "to be interesting" -> "interestingly", "to be possible" -> "possibly", "to be incidental" -> "incidentally, by the way", "to be necessary" -> "necessarily", "to be fortunate" -> "fortunately", etc. Also, many of these concepts reflect the ATTITUDE of the patient about an event. Thus, they are essentially _modal_ in nature. I'll have a lot more to say about _modality_ later. 2.7.7 GENERIC VOICE DERIVATIONS Voice derivations can also be performed on generic verbs. Again, we will use the generic action root "-ze-" for generic action verbs, and will assume a generic state root when a root is not present. The results are very useful: Anti-passive A-d [+P] "pugasi" - 'to work/cause a change' e.g. The new job CAUSED A CHANGE (in his behavior). Middle P-s [-A] "zedesi = zezoyadesi" - 'to be under control' e.g. The reactor IS now UNDER CONTROL. Inverse P/A-s "zevisi = zezoyavisi" - 'to be under the control of' e.g. The project IS now UNDER THE CONTROL OF the engineering department. Anti-passive A-s [+P] "zegasi = zezoyagasi" - 'to be in control/charge' e.g. John IS IN CONTROL (of the project). Middle P/F-s [-A] "zetuedesi" - 'to experience' e.g. John EXPERIENCED a new kind of freedom. Inverse F/P-d "zedovisi" - 'to befall' e.g. Tragedy BEFELL the entire crew. Anti-middle A/F-p [-P] "zenioxisi" - 'to try/attempt' e.g. John TRIED to open the door. In the last example, the patient cannot be obliquely expressed because it appears within the focus; i.e. because it is a direct patient of the embedded sentence that elaborates the 'trying' event. [Remember, the focus of an action verb must always elaborate the event indicated by the verb.] Note that we also introduced the class-changing morpheme "-xi-" to perform the anti-middle voice operation. As we discussed earlier, the prefix "anti-" is used when the second argument of the verb is demoted. 2.8 INDIRECT CAUSATION AND VERBAL COMPLEMENTS In many of our verb derivations, we used the word "cause" in our paraphrases of the semantics of verbs which have an agent in their argument structure. Unfortunately, these paraphrases are approximate and often imply some distance between the agent and the event. However, I must emphasize that the agent argument of a verb is the entity that is DIRECTLY responsible for the event indicated by the verb. Thus, there is a definite semantic difference between 'kill' and 'cause to die', even though our paraphrases may imply otherwise. If we wish to intentionally put distance between an agent and an event, we must design words that are equivalent to English "cause", "make", etc. Consider the following sentences: He MADE his son wash the dishes. I HAD Bill deliver the package. He CAUSED his wife to have a miscarriage. In the above examples, the patient (if that is what it really is) cannot be expressed directly: *He made his son. *I had Bill. *He caused his wife. However, the English verb "to cause" can be used without this quasi patient: John caused the accident. Thus, these verbs indicate that an INDIRECT agent is responsible for an event which itself may have a DIRECT agent - the quasi patient is not at all a true patient of the verb "cause/make/have" (although it may be the true patient of the embedded sentence). Also, the English distinction between "cause", "have", and "make" is somewhat idiosyncratic. Semantically, there is no significant difference between them. [Actually, "to have" is a more polite version of "to make", but this distinction is not important to us here. We will discuss how to derive more polite forms of words in the section on register variations. More importantly, though, we will derive a verb with this sense of 'to have' in the section on _modality_.] The most neutral paraphrase of indirect causation is simply 'to cause to exist' or 'to cause to be real'. In our sample language, I will use the state root "-veya-" to represent the concept 'real/actual/existent' (default class = A/P-d). Here are some useful derivations: A/P-d: "veyasi=veyapusi" - 'to cause/make/create', 'to cause to come into existence', 'to cause to become real/actual' e.g. John CAUSED the accident. John MADE Billy wash the dishes. John MADE some apple cider. A/P-s: "veyazoyasi" - 'to insure/guarantee', 'to keep/maintain a reality' e.g. The contract GUARANTEED their compliance. P-s: "veyasesi" - 'to be real/actual', 'to exist', 'there be' e.g. The particles EXIST for only ten nanoseconds. THERE ARE ten people at the party. THE SITUATION IS SUCH THAT only ten people came. P-d: "veyapiasi" - 'to come into existence', 'there came to be', 'it came to be that' e.g. The new policy CAME INTO BEING after he resigned. THERE CAME TO BE fewer people willing to help. IT CAME TO BE THAT fewer people were willing to help. Note that English uses different forms depending on whether the focus is a noun phrase (i.e. an entity) or an embedded sentence (i.e. an event). Another type of indirect causation can be indicated by using a generic verb. As we mentioned earlier, the focus of an action verb MUST elaborate the event described by the verb. In effect, the agent is responsible for causing or bringing about the event. For example, in a sentence like "John told a joke", the word "joke" elaborates what John did; i.e. an elaboration of a 'telling' event describes the specific method or means used in the 'telling'. Thus, a GENERIC A/P/F-d action verb would indicate that the agent successfully affected the patient BY MEANS OF the focus. Now, if we used the unmodified A/P/F-d generic action form "zekosi", we could create a sentence like this: A P F John zekosi (the project) (he resigned). = John did something to the project by resigning. = John affected the project by resigning. This sentence sounds odd in English, but is perfectly acceptable in our sample language. To implement it differently, we could simply used a nominalized verb which indicates the process of "resigning": A P F John zekosi (the project) (his resignation). = John did something to the project with his resignation. = John affected the project with his resignation. We'll discuss how to create process nouns later. The anti-middle A/F-d [-P] form is also useful: A/F-d [-P] "zekoxisi" - 'to affect someone/something unspecified by means of' e.g. John CAUSED SOMETHING TO HAPPEN BY resigning. If we now perform an additional normal middle voice operation, we get: F-d [-A] [-P] "zekoxidesi" - 'something was affected by doing X' e.g. SOMETHING HAPPENED WHEN John resigned. This last example could have also been accomplished by means of a double middle construction; i.e. "zekodedesi". In fact, this double construction will be useful enough to allocate a single CCM: double middle -voi- changes A/P/F-x to F-x [-A] [-P] Thus, the new word is "zekovoisi". Note how we neatly capture some important functions which are often highly idiosyncratic in natural languages. Finally, basic verbs which already have a focus as a direct object can also take an embedded sentence as an argument, as the following examples illustrate: John wanted the book vs. John wanted Bill to leave I saw the mountains vs. I saw them working They liked her vs. They liked her portrayal of Juliet We know the answer vs. We know he wants her to buy the car Note that the English embedded sentences are idiosyncratic in that they require either infinitives, participles, nominalizations, or complete finite sentences, depending on the particular verb. By using an embedded sentence with the same form as a normal sentence (i.e. a complete finite sentence), you can achieve the same effect with a simpler morphology and syntax. Here is how the above examples would look (the complete embedded sentence is in parentheses): John wanted (Bill leave). I saw (they were working). They liked (she portrayed Juliet). We know (he wants (she buy the car)). They seem awkward in English, but they're linguistically sound, syntactically simpler, and totally lacking in idiosyncracy. Also, this approach is used in MANY natural languages. [Incidentally, an embedded sentence which appears as the argument of a verb (subject or object) is called a _complement_.] 2.9 IS IT REALLY AN ACTION VERB? Some verbs that appear to be actions, especially speech acts, imply an attempt to put the patient in a particular, known state. For example, the verb "to apologize" indicates that the agent is attempting to cause the patient to be 'forgiving'. Similarly, the verb "to lie" indicates that the agent is attempting to cause the patient to be 'deceived'. Since we know the intended final state, we might be tempted to use it in converting an action concept to what is effectively a state verb. For example, we could perform the following derivations: A/P-p: to scold => A/P-d: to shame, to cause remorse P-s: to be ashamed, remorseful A-p [+P]: to lie => A/P-d: to deceive A-p [+P]: to apologize => A/P-d: to achieve forgiveness from And so forth. In other words, the "-p" verbs imply a POTENTIAL change of state while the "-d" verb indicates an ACTUAL or SUCCESSFUL change of state. However, even though the above derivations imply it, the root meanings of the "-p" and "-d" verbs are not really the same. For example, an effective scolding is not the only thing that can cause shame or remorse; an effective lie is not the only thing that can cause one to be deceived; and an effective apology is not the only way to achieve forgiveness. In other words, the action and the state are not truly reversible. One problem here is that the English state verbs do not imply the use of speech, whereas the speech act verbs do. This suggests that a better approach would be to derive what appears to be a speech act from a STATE concept and add a morpheme to indicate that the event involves speech. Here is a sample derivation using the concept of 'forgiving': [Note that the basic state here is 'forgiving' - NOT 'forgiven'. A 'forgiving' state is a true mental state. However, the state of being 'forgiven' is not a true state since it merely indicates the attitude or feeling of someone ELSE towards the forgiven person. Note also that the concept of 'forgiving' implies a relationship with the person forgiven. Thus, the forgiven person is the focus, NOT the patient.] P-s: to be a forgiving person, to be forgiving/magnanimous AP/F-d: to forgive (i.e. to cause oneself to enter a forgiving state focused on someone else) A/P/F-p: to attempt to cause the patient to be forgiving towards the focus. AF/P-p: to attempt to achieve forgiveness for oneself from someone. I.e., the agent attempts to cause the patient to be forgiving to the agent. AF-p [+P]: to attempt to achieve forgiveness - 'forgiver' can be expressed obliquely Now, if we add a morpheme to the AF-p [+P] derivation which indicates that the verb implies the use of speech, we get: speech morpheme + AF-p [+P]: to apologize Note that we had to introduce the new and unusual case role "AF". This can be implemented using a class-changing morpheme similar to what we did earlier when we discussed grammatical voice operations. In this case, the new class- changing morpheme would simply perform a _reflexive_ function, indicating that the agent and focus are the same entity. Finally, on first glance, it appears that the verb "to apologize" has a focus, as in "I apologized for my rude behavior". However, the preposition "for" does not introduce a true focus - it actually marks the REASON case role. [We'll have more to say about reflexives and the reason case role later. Also, the actual implementation of the speech morpheme mentioned here will be discussed later in the section on modality.] The above works well with the concept 'forgiving', and should also work well for other verb pairs such as 'lie/deceive'. In other words, many words which apear to be actions (especially speech acts) can be derived from state concept root morphemes, and are not true actions. However, this will not work in all cases. For example, if we tried to derive the verb "to scold" from the state concept of 'ashamed/remorseful', we end up with: A/P-d: to shame/embarrass A/P-p: to attempt to embarrass speech morpheme + A/P-p: to scold Unfortunately, the English word "scold" implies much more than simply an attempt to cause shame using speech. It also implies that the patient is either young or immature, and that the speech act is somewhat sharp and loud. This does not mean, however, that the above derivation is useless. In fact, it is a very good derivation of the English word "reprove". This still leaves us with the problem of words like "scold". Since the concept behind 'scold' is complex, and since the word is a very useful one, we would probably want to create a unique action root for it. And, as with other action derivations, the "-s" and "-d" forms will indicate successful control or change of the patient: A/P-p: to scold A/P-d: to effectively and successfully scold A/P-s: to keep under control by scolding ("to henpeck"???) In other words, when an action root is used with a "-s" or "-d" classifier, the resulting state can be defined as whatever complex and/or vague state that a patient experiences when the action is effective or successful. In many cases, we'll find that such derivations can be quite useful, often replacing periphrastic, metaphoric, or even idiomatic expressions in English. Here are some examples: A/P-p: to speak to A/P-d: to have a good/productive talk with A/P-p: to hit A/P-d: to give something a good whack A/P-p: to scold A/P-d: to set someone straight, to give someone a good dressing-down A/P-p: to kick A/P-d: to give something a good, swift kick AP/F-s: to play (a game) AP/F-d: to win A/P/F-p: to tell someone something A/P/F-d: to inform someone of something A/P/F-p: to call someone something A/P/F-d: to name/christen someone something [Actually, "to call/name/christen" can also be implemented as a state verb rather than as an action verb. At this point in time, I can't make up my mind which approach is better.] And so forth. In summary, we should always try to derive potential actions from state concepts, if possible. If not, then we should create basic action roots and perform additional derivations as described above. Now, it is extremely important to keep in mind that a word with subtle shades of meaning like the English word "scold" is not likely to have exact counterparts in other natural languages, and deriving a close (but not quite exact) counterpart using a state root rather than a specific action root will, in the long run, be much more productive. Thus, in spite of all the above, my personal preference is to derive an equivalent of the English word "scold" using a state root meaning 'ashamed' or 'embarrassed', the A/P-p classifier, and the special speech morpheme. This will not capture the precise meaning of the English word, but it will be very close and just as effective and useful. Also, if we insist on capturing the subtleties of every English word, then in fairness, we must do the same for all other natural languages, which is quite impossible. Finally, the astute reader may have realized from the start that verbs like "to lie" and "to apologize" cannot possibly be action verbs for two reasons: First, since the final, desired state of the patient is precisely known, the concept is inherently stative in nature. Second, a concept like 'forgiving' implies a relationship with the 'forgiven' entity. Thus, since the focus of an action MUST elaborate the event itself, the verb cannot possibly be an action. However, some verbs, especially speech acts, can be deceptive, and I felt that it was important to spend a little time clarifying their semantics. 2.10 FOCUSED VERSUS UNFOCUSED At first glance, it would seem that the only difference between most focused and unfocused verbs is simply syntactic, and that there is little or no difference in the actual MEANINGS. As it turns out, though, there is a significant semantic distinction between focused and unfocused verbs, as I hope to illustrate now. For starters, consider the following: AP/F-s: I think about Joan a lot. AP-s: I think therefore I am. The focused verb refers to specific situations, while unfocused verbs seem to indicate both an ability as well as an actuality. Thus, the unfocused verb can be paraphrased as "I not only have the ability to 'think', but I also put this ability to use". Another way of looking at the distinction is that unfocused verbs do not make any reference at all to a specific focus (either implied or stated). Thus, the AP-s verb "to think" could also be paraphrased as 'to be a thinker'. Note that this is very similar to the anti-passive. The difference, though, is that the anti-passive simply reduces the topicality of the object while implying that it is still taking part in the event. A completely unfocused verb eliminates the object completely. Thus, the anti-passive version of the above example would mean something like "I am THE thinker (in the current context)", while the completely unfocused version means "I am A thinker" (regardless of the current context). Also, do not confuse an unfocused verb for a focused one whose focus is strongly implied by the context but not mentioned. For example: Boss: "I want to make sure that no one screws up." Worker: "I understand." The verb "understand" must be either P/F-s with an omitted object or the anti- passive P-s [+F] because the focus is still strongly implied, even though it is not specified. Here are some other examples. In each case, try to extract the sense of the unfocused verb with paraphrases that use the focused verb. These paraphrases should sound something like "I not only have the ability to xxx, but I also put this ability to use", "I am an xxx-er", or "I am an xxx-ing kind of person": P/F-s: to understand P-s: to be discerning or astute (Note how this is similar to the distinction between "to know" and "to be intelligent".) AP/F-s: to imitate AP-s: to be a copycat/imitator P/F-s: to see P-s: to have sight/vision, to be a sighted person AP/F-d: to befriend AP-s: to be amiable/friendly P/F-s: to enjoy P-s: to enjoy oneself, to have a good time P/F-s: to need P-s: to have needs A/P/F-p: to order/command A/P-p: to give orders to A/P/F-d: to order (successfully) A/P-s: to be in charge of (someone) And so on. Note that the English versions of the unfocused verbs are almost always idiosyncratic or periphrastic. In summary, there is a definite semantic difference between focused and unfocused verbs. Also, by making this distinction, we achieve some highly useful results in a manner that is simple, consistent, and semantically precise. 3.0 NOUNS By now, it should be obvious that word design can be extremely productive in a language possessing a rich classificational morphology. This kind of morphology allows the AL designer to create a large vocabulary with semantic precision, while minimizing the number of root morphemes needed. However, so far we've only used this approach to design basic VERBS. We now need to see if a similar approach can be used to design basic NOUNS. I began my discussion of verbs by providing a large number of examples that I placed into groups based on their argument structures. I felt that this was necessary because my approach to classifying verbs is unusual (and probably unique). For nouns, though, I don't think that large numbers of examples will be needed, simply because the classes and their semantics are fairly obvious. [Incidentally, I am not aware of any other work that classifies verbs as I have done here. Initially, I was tempted to adopt the more widely accepted Vendlerian analysis which classifies all verbs into the four major categories: _state_ (e.g. "to know", "to love"), _activity_ (e.g. "to run", "to sing"), _accomplishment_ (e.g. "to sing a song", "to write a book") and _achievement_ (e.g. "to die", "to find"). However, although I experimented with these four categories, I was very unhappy with the results. The standard categories seemed too vague, and I often had difficulty deciding which category a verb belonged to. In any case, I felt that I needed a more productive system, and eventually ended up with the approach that I am presenting here.] 3.1 BASIC NOUN CLASSES Before starting, let's precisely define what we mean by the expression "basic noun". Here is the definition that I will use: A basic noun will represent an entity that has an actual physical existence (including entities from fantasy, mythology, etc.). Thus, such an entity must be composed of matter, energy, a combination of both, or time. Furthermore, characteristics which distinguish it from other entities must be verifiably physical (as opposed to functional, social, cognitive, etc). Note that my definition is purely semantic and has nothing to do with how a word is actually used in a sentence. Thus, for example, we will derive the word for "window" as a basic noun, while the word for "teacher" will be derived from a basic verb (as we illustrated earlier), even though both are used as nouns in a sentence. The word "window" is a basic noun because it can be uniquely described using only its physical properties. The word "teacher", however, must be derived from a verb, even though it represents a physical entity, because it does not differ from other related entities (such as "student" or "learner") in a verifiably physical way. In other words, we cannot distinguish a "teacher" from a "student" or determine their respective natures by examining only their physical traits. Their differences lie in what they DO, not in what they ARE. At any rate, if the above definition causes problems, we can use the following simpler (and perhaps more practical) alternative: a basic noun represents an entity that cannot be easily or logically derived from a basic verb. As it turns out, the derivational system that I am proposing here is completely reversible, making the actual definition irrelevent. In effect, a "basic verb" is any word whose class is indicated by a verb classifier, and a "basic noun" is any word whose class is indicated by a noun classifier. I will classify most basic nouns as follows: 1. An entity represented by a basic noun must consist of matter, energy, a combination of both, or time. 2. An entity of matter and/or energy represented by a basic noun must be either living or non-living. 3. A non-living entity represented by a basic noun must be either natural or artificial. Note that this is not the only way that one can classify nouns. For example, some classificational schemes make distinctions based on composition, shape, consistency, size, etc. Other schemes, especially those that have evolved naturally in some human languages, make distinctions based on concepts such as animacy and function, which often require reference to culture, religion, history, etc. However, I feel that the system proposed here is not only easier to work with and inherently neutral, but is also much more productive in its ability to generate as many words as possible from as few roots as possible. So, using this approach, we can create the following basic noun classes: matter & energy: living -> man, lizard, demon, grass, tree, bacteria non-living, natural -> storm, tornado, geyser, earthquake non-living, artificial -> computer, airplane, oven, fountain matter: living -> hand, leaf, branch, liver, acorn, ear non-living, natural -> water, salt, rock, cliff, river, island non-living, artificial -> keyboard, statue, ax, book, wharf, kitchen energy: living -> lifeforce, ghost, god, energy creature non-living -> heat, thunder, light, noise, energy, photon time: -> winter, sunset, equinox, morning, midnight I am not making a distinction between natural and artificial, non-living energy because we would be forced to make useless distinctions. For example, "light" from the sun would require a different classifier than "light" from a light- bulb. Note an important relationship that can be seen from the above list: items of pure matter or pure energy can be COMPONENTS of items that consist of both matter and energy. For example: matter & energy matter energy --------------- ------ ------ man hand lifeforce steam water heat computer wire electricity In other words, complete energized or living entities that can function independently are treated as both matter and energy, while their components are treated as just matter or just energy. I believe that the above classes are fundamental, and that any useful system should contain at least these eight classes. However, I also feel that we should provide additional sub-classes for classes that are likely to have a large number of members. For example, in the 'matter & energy, living' class, it will be useful to distinguish between plants and animals. In fact, I feel that we should create even finer distinctions, such as between 'mammal', 'bird', 'fish', 'insect', etc. In the 'matter, non-living, artificial' class, it will be useful to distinguish between substances (e.g. "plastic"), locatives (e.g. "wharf") and others (e.g. "hammer"). Finally, the substance/locative/ other distinction should also be applied to the 'matter, non-living, natural' class to allow us to distinguish between words such as "water" (substance), "cliff" (locative), and "boulder" (other). If we make these additions, our chart will look like this: matter & energy: living, vertebrates: mammals -> man, tiger, mouse, elf, unicorn birds -> hawk, ostrich, canary, penguin, duck reptiles -> lizard, frog, turtle, snake, newt fish -> trout, halibut, perch, lamprey, herring insects -> ant, crab, mosquito, grasshopper, fly other animals -> jellyfish, octopus, worm, clam, starfish plants: trees & shrubs -> tree, shrub, oak, blueberry, apple, pine other sperma- tophytes -> grass, flower, carrot, seaweed, lily other plants -> algae, bacteria, moss, fern non-living, natural -> tornado, geyser, storm, earthquake non-living, artificial -> airplane, computer, oven, robot matter: living -> hand, leaf, branch, liver, acorn, ear non-living, natural, substance -> water, sand, salt, ivory, urine, air locative -> cliff, river, island, mountain, bay, sky other -> boulder, fang, stalagmite, shell, hair non-living, artificial, substance -> plastic, benzene, whisky, sauce, glue locative -> wharf, building, kitchen, city, road other -> window, statue, ax, book, nail, button energy: living -> ghost, spirit, god, energy creature non-living -> heat, thunder, photon, noise, light time: -> winter, sunset, equinox, morning, midnight [Technically, the "insect" group includes the entire arthropoda phylum, the "reptile" group includes both the amphibia and reptilia classes, and the "fish" group includes not only pisces but all other vertebrate classes such as selachii (sharks), marsipobranchii (lampreys), etc.] Note that I use the word "locative" in the following sense: a locative noun represents an entity which typically is built in place or evolves naturally in a single location, which is extremely difficult (if not impossible) to move to a different location, which is relatively permanent, and which is typically considered a place where one can go to, remain at, or depart from. By the way, I think that most readers will agree that the words for 'living', 'non-living', 'natural', and 'artificial' should be derived from basic verbs. I also recommend that the words for 'matter' and 'energy' be derived from basic verbs since these do not describe specific entities or substances, but instead indicate particular states of actual items and substances. Thus, for example, the word for 'matter' would be the noun version of the verb meaning 'to be material' (i.e. 'to consist of matter'). We can do the same for words that are too general for the existing classifiers, such as 'plant' and 'animal'. 3.2 NOUN DESIGN ALGORITHM & EXAMPLES In order to illustrate the noun derivation system, I will use the following classifiers: matter & energy: living, animals -nembi- vertebrates, mammals -mo- birds -su- reptiles -pusta- fish -sai- insects (all arthropods) -zio- living, plants -kaya- trees & shrubs -po- other spermatophytes -tonze- non-living, natural -ji- non-living, artificial -fiu- matter: living -vau- non-living, natural, substance -fa- locative -nai- other -le- non-living, artificial, substance -niu- locative -te- other -ki- energy: living -dengi- non-living -pai- time: -be- The classifiers "-nembi-" and "-kaya-" will be used for animals and plants, respectively, that have not been assigned more specific classifiers. In the derivations that follow, we will refer to these categories as "other animals" and "other plants". Thus, the more specific categories are actually subsets of the super-categories "-nembi-" and "-kaya-", but the super-categories will be used only if a more specific category is not available. Now, let's design some words. We'll start with a word for 'water' and use the root morpheme "-gua-". Since 'water' is a natural, non-living substance, the classifier is "-fa-". Thus the word for water is "guafada". Now, what happens when we apply OTHER noun classifiers to the root "-gua-"? For example, what is the meaning of the 'matter & energy, living, mammal' form "guamoda"? Is there such a thing as a creature that is the mammalian equivalent of water? No - at least not if we wish to ensure semantic precision in our derivations. Furthermore, with a little experimenting, you'll quickly discover that it is essentially impossible to cross noun class boundaries with any degree of semantic precision - something that was very easy to do with verb classes. The reason for this is simple: nouns represent truly unique entities, and entities in different classes are seldom related in precisely definable ways. Thus, it is not practical to do with nouns what we did earlier with verbs. Instead, we have two choices: 1. Do not use classifiers with nouns at all. In effect, each noun root will belong to one and only one class, and will not be able to change class. 2. Compromise. Use roots for their MNEMONIC value, rather than for their literal meaning. In effect, the combination of classifier plus root will become a new, unique, de-facto root. Using the compromise approach, root morphemes can be combined with unrelated noun classifiers in a way that is semantically imprecise, INTENTIONALLY, but which is mnemonically useful. However, the classifier ITSELF will be semantically precise, and the combination of classifier PLUS root will also be semantically precise. For example, the root morpheme "-gua-" with the basic sense of 'water' could be used to create the following nouns: matter & energy: living, mammals -> guamoda - cetacean birds -> guasuda - duck fish -> guasaida - puffer/swellfish/blowfish reptiles -> guapustada - water snake insects -> guazioda - crab other animals -> guanembida - jellyfish trees & shrubs -> guapoda - water willow other sperma- tophytes -> guatonzeda - kelp/seaweed other plants -> guakayada - algae non-living, natural -> guajida - hot spring non-living, artificial -> guafiuda - jacuzzi (i.e. hot tub) matter: living -> guavauda - bladder (i.e. urinary) non-living, natural, substance -> guafada - water locative -> guanaida - lake other -> gualeda - puddle non-living, artificial, substance -> guaniuda - broth/soup locative -> guateda - reservoir other -> guakida - aquarium energy: living -> guadengida - water spirit, undine non-living -> guapaida - hydropower, water power time: -> guabeda - monsoon, wet/rainy season Note that the above are just suggestions. With a little effort, it may be possible to come up with even better ones. It should even be possible to develop rough rules or guidelines that would make the results of the derivational process more predictable (even though the process can never be TOTALLY predictable). For example, one very useful technique is to always state a concept using an English compound or descriptive phrase that is appropriate to each class. Here are some examples: water plant -> seaweed water mammal -> cetacean water bird -> duck water creature -> jellyfish natural energetic water -> hot spring artificial energetic water -> hot tub natural water location -> lake natural water thing -> puddle artificial water location -> reservoir artificial water thing -> aquarium living water energy -> undine non-living water energy -> hydropower water time -> wet season And so on. Another possible rule would be to always use the same root for a plant and its fruit or seed. The plant would take a plant classifier, and the fruit would take the 'matter, living' classifier. For example, if the root for 'banana' is "-gelba-", then "gelbapoda" would mean 'banana tree' and "gelbavauda" would mean 'banana (the fruit)'. The approach I'm suggesting here is appealing because it's not possible to use noun roots with unrelated classifiers in a way that is semantically precise. Instead, we must either create an extremely large set of root morphemes for nouns, or re-use the roots for their mnemonic value. I feel that the latter choice is much more preferable. In effect, the combination of classifier-plus-root becomes a de facto NEW root, even though it has the morphology of classifier-plus-root. And when looked at in this way, there is really nothing imprecise about this approach. We only need to keep in mind that the root morpheme is just a mnemonic aid. To me, it seems like a great way to re-use roots that would otherwise be underutilized. It will be especially useful when you first come across a new word, because the classifier is SEMANTICALLY PRECISE and provides a lot of information about the word. The mnemonic value of the root provides even more information, making it easier to guess the meaning of the word, and to remember its meaning in the future. Also, there is nothing typologically unnatural about this scheme. In fact, this approach is somewhat akin to English compounds such as "whitefish", "highland", "seahorse", etc. However, the noun classifiers that we are using here are sometimes vaguer and more generic than English headwords such as "land" and "horse". Thus, what I am proposing is actually much closer to what is done in some of the Bantu languages of Africa or some of the aboriginal languages of Australia and New Guinea. The main difference, though, is that the classifiers in these languages are even VAGUER than the classifiers we are using here. Thus, this approach fits in quite snugly between the opposite poles of classificational possibility. [The AL designer also has the option of creating classifiers that are even more specific than those above. For example, the 'mammal' class could be further sub-divided into 'primates', 'carnivores', 'marsupials', 'rodents', etc.] In summary, I am suggesting that we use semantic precision only when it is practical. We should re-use root morphemes as mnemonic aids when semantic precision is not practical. The alternative is to create many thousands of additional root morphemes which will have to be learned by the student and which will have little or no usefulness in the creation of additional words. 3.3 FROM BASIC NOUN TO OTHER PARTS OF SPEECH The simplest kind of derivation is to change the part-of-speech. For these, I suggest that the verb form have the meaning 'to be X', and that the other forms be interpreted in the usual way. Thus, for example, the word "guasusi" would be a P-s verb meaning 'to be duck'. The adjective form, "guasuno", could be used in expressions such as "duck chick", "duck egg", "duck colony", or "duck wing". The adverb form "guasupe" would have the meanings 'being duck', 'since it is (a) duck', 'since they are duck', etc. Note that this approach is perfectly consistent with the rules we adopted for basic verbs. Although the above derivation is useful, it will be much more productive to add verb classifiers to basic nouns. Before doing this, though, we need to define the semantics of the noun-to-verb conversion. To this end, I suggest that we first derive an illustrative 'prototype' verb form. The semantics of this prototype, which I will describe below, is intended to maximize the number of useful words that we can derive from each basic noun with semantic precision. Next, we can extract the basic state concept from the prototype and use it to derive all possible verb forms. Once these other verbs have been derived, we can then derive adjectives, adverbs, and case tags by applying exactly the same rules that we applied earlier to basic verbs. After a lot of experimentation, I decided that the best prototype form would be an A/P-d verb with the following semantics: The agent causes the patient and the entity indicated by the basic noun to physically COME TOGETHER in a manner that is characteristic of both entities. Here are some English examples: Prototype Noun A/P-d verb Semantics ---- ---------- --------- salt to salt to cause salt to come together with P e.g. He salted the stew. seed to seed to cause seed to come together with P e.g. He seeded the garden. bed to 'bed' to cause P to come together with bed(s) e.g. She 'bedded' the children. = She put the children to bed. tree to 'tree' to cause tree(s) to come together with P e.g. They 'treed' the pasture. = They planted trees in the pasture. OR They brought trees to the pasture. storm to 'storm' to cause storm(s) to come together with P e.g. Mother Nature 'stormed' us. = Mother Nature hit us with a storm. brain to 'brain' to cause brain(s) to come together with P e.g. The scientist 'brained' the android. = The scientist gave the android a brain. oven to 'oven' to cause P to come together with oven(s) e.g. He 'ovened' a cake. = He baked a cake. land to land to cause P to come together with land e.g. The pilot landed the airplane. glue to glue to cause glue to come together with P e.g. He glued the envelope. boat to 'boat' to cause P to come together with boat(s) e.g. The crew 'boated' the cargo. = The crew loaded the cargo onto (the) boat(s). pencil to pencil to cause pencil(s) to come together with P e.g. He penciled the sign with graffiti. wharf to 'wharf' to cause P to come together with wharf(s) e.g. He 'wharfed' the rowboat. = He moored the rowboat (to a wharf). ghost to 'ghost' to cause ghost(s) to come together with P e.g. The sorcerer 'ghosted' the house. = The sorcerer caused the house to become haunted. Note that, if a noun N is physically larger or inherently less movable than the patient P, it makes more sense to state the semantics as 'to cause P to come together with N'. However, if the patient is physically larger or inherently less movable than the noun, it makes more sense to reverse the order; i.e. 'to cause N to come together with P'. As you can see from the above derivations, the paraphrase 'to come together' is quite vague. In most cases, the semantics can be more precisely stated as one or more of the following: A causes P to undergo a change of state by USING N on P.