下载文档原格式
Chapter one1.What is language?Language is a system of arbitrary vocal symbols used for human communication. System means it is rule-governed; arbitrary means there is no logical relationship between language elements and their meaning; vocal means speech is primary; symbols related to arbitrariness, it means language elements are only the symbols for the meaning they express. Human, it means language is uniquely human; communication is the primary function of language.2.What are the design features of language, give their definitions and examples. Arbitrary: arbitrary is the core feature of language, which means that there is no logical relation between meanings and sounds. Arbitrary is a matter of degree, language is not entirely arbitrary, first, the onomatopoeia are words that sound like the sound they describe, to some extent, they have natural basis. Second, some compounds are not entirely arbitrary either, “snow” and“storm” are arbitrary words, but the compound word “snowstorm”is less so. Thirdly, some surnames, such as Longfellow, Johnson.Examples: a rose by other name would smell as sweetDuality: is meant the property of having two levels of structures, such that units of the primary level are composed of elements of the secondary level and each of the two levels has its own principles of organization .we call sounds secondary units as opposed to primary units as words, since the secondary are meaningless and the primary unit have distinct and identifiable meaning.Creativity: language users can understand and produce new sentences to express new meanings. By creativity, we mean language is resourceful because of its duality and recursiveness. By duality the speaker is able to combine the basic linguistic units to form an infinite set of sentences, most of which are never produced or heard before. Language is creativity in another sense, that is, itspotential to create endless sentences. The recursive nature of language provides a theoretical basis of this possibility.Example: He bought a book which was written by a teacher who taught in a school which was known for its graduates who……Displacement: means that human languages enable their users to symbolize objects, events and concepts, which are not present at the moment of communication.Examples : we can refer to Confucius and Bush at the moment of communication , but the former lived thousands of years ago and the latter live far away from us .3.Functions of the language and their examples⑴informative function: when language is used to express human experience and knowledge about the world. It is predominantly the major role of language.⑵interpersonal function is the most important sociological use of language by which people establish and maintain their status in a society. Language marks our identity, physically in terms of age, sex, and voiceprints; psychologically in terms of language, personality, and intelligence; geographically in terms of accents and dialects; ethnically and socially in terms of social stratification, class, status, role, solidarity, and distance.(3) Performative function: is primarily to change the social status of persons, as in marriage ceremonies, the sentencing of criminals, the blessing of children, the naming of a ship at a launching ceremony and the cursing of enemies. The kind of language employed in performative verbal acts is usually quite formal and even ritualized.(4) Emotive function: changes the emotional status of an audience for or against someone or something.(5) Phatic communion: occurs when language is used for pure interpersonal purposes, e.g. greetings, farewell, etc.(6) Recreational function: when language is used for the pure joy of using it, such as baby’s babbling and chanter’s chanting.(7) Metalingual function: when language is used to discuss itself. For example, to be honest, to make a long story short, come to think of it, on second thought.4. What is linguistics?Linguistics is the scientific study of language.4.Main branches of the linguistics1)Phonetics: studies speech sounds, it is the description, classification, andtranscription of speech sounds. It includes articulatory phonetics, acoustic phonetics and auditory phonetics.2)Phonology: the study of speech sounds as a system: the relations betweenspeech sounds, the way in which speech sounds are related to meaning, the rules governing the structure, distribution and sequencing of speech sounds. 3)Morphology: the internal structures of words. It studies the minimal units ofmeaning-morphemes and word-formation processes.4)Syntax: studies the internal structure of sentences5)Semantics: the study of meaning as encoded in language6)Pragmatics: the study of language use, meaning in context5. Macrolinguistics1) Psycholinguistics investigates the interrelation of language and mind, in processing and producing utterances and in language acquisition. The psycholinguistics constraints on the form of grammar are studied. It also studies language development in the child, biological foundations of languages, and the relationship between language and cognition.2) Social linguistics: is an umbrella term, which covers a variety of different interest in language and society, including the social functions of language and the social characteristics of its users.3) Anthropological linguistics: anthropology and linguistics combined: the relationship between language and culture.4) Computational linguistics: the use of computers to process or produce human language: machine-translation, information retrieval, expert systems.6. Important distinction in linguistics1) Descriptive VS PrescriptiveThe distinction lies in prescribing how things ought to be and describing how things are. Describing language as it is used by its native speakers is descriptive; trying to lay down language rules for correct uses of language is prescriptive. To say linguistics is a descriptive science is to say that linguist tries to discover and record the rules to which the members of a language community actually conform and does not seek to impose upon them rules, or norms, of correctness.2) Synchronic VS diachronicStudying language as it is used at a particular point in time is synchronic study; studying language as it changes over time is a diachronic study.3) Langue Vs paroleSaussure distinguished the linguistic competence of the speaker and the actual phenomena or data of linguistics as langue and parole. Langue is the language system, which is social, essential, and stable while parole is the actual use of the language system, which is individual, accidental, and unstable.4) Competence VS performanceA language user’s underlying knowledge about the system of rules is called his linguistic competence, and performance refers to the actual use of language in concrete situation. (Chomsky)5) Syntagmatic VS paradigmatic relationsSyntagmatic relations are relations between units present in the same sequence or construction, syntagmatically related elements form structures.Paradigmatic relations are relations between a unit and other units that can replace it in a given sequence. Paradigmatically related units form systems.how he can get to his destination. Then you tell him all that you know about the course, including which bus or subway he should take. This is the very process in which you use your experience and knowledge to help him. If you have no idea about his destination, you would say “Sorry, I can’t help you”. At this moment, he may buy a map of the city. This is the reason why map came into being. Map provides the detailed information including not only the complicated and crossing roads, but also the course of bus and subway. Map is an instrument by that peopleuse language to show their experience and knowledge. By analogy, there are also many other instruments for people to express content, such as books, papers, magazines and dictionaries and so on. These forms are utilized by human to record the facts of social development; conversely, the recorded facts is a prerequisite of social development. Having talked so much, I believe there are two important points to be captured. 一是，人这个主体通过使用语言将内容传递出去；另一个是语言通过报纸、杂志、书、字典等的外在形式把知识和经验传递出去。
我对al的理解英语作文The Essence and Significance of AL in the Modern Era.In the ever-evolving landscape of technology and education, the acronym "AL" often surfaces as a pivotal concept, carrying with it a depth of meaning thattranscends its mere letters. While the interpretation of AL can vary depending on the context, this essay aims to explore the multifaceted understanding of AL, particularlyin the realms of artificial intelligence, adaptive learning, and academic literacy, and its implications for our society.Artificial Intelligence (AI) and AL.In the realm of artificial intelligence, AL is often associated with the concept of "Artificial Life" or "Artificial Learning". Artificial Life explores the simulation of natural life processes using computational techniques, aiming to understand the fundamental principles of life itself. On the other hand, Artificial Learning, ormachine learning, refers to the ability of computer systems to learn from data, identify patterns, and make predictions or decisions without being explicitly programmed.The intersection of AI and AL is particularly fascinating, as it heralds a new era of intelligent machines capable of.。
Anticipatory coarticulation:If the sound becomes more like the following sound, as in lamb, it is known as anticipatory coarticulation.Affix: An affix is the collective term for the type of morpheme that can be used only when added to another morpheme. Antonymy: Anthnymy is the name for oppositeness relation.Acronym: Acronym is made up from the first letters of the name of an organization, which has a heavily modified headword. Broad transcription: When we use a simple set of symbols in our transcription, is is called a broad transcription. Categorization Categorization is the process of classifying our experiences into different categories based on commonalities and differences.Closed syllable: A syllable with coda.Classification of Words: 1. Variable and invariable words 2. Grammatical words and lexical words 3. Closed-class words and open-class words 4. Word classConsonants: Consonants are produced by a closure in the vocal tracy, or by a narrowing which is so marked that air cannot escape without producing audible friction.Coarticulation: When such simultaneous or overlapping articulations are involved, we call the process coarticulation. Conjoining: conjoining refers to the process where one clause is coordinated or conjoined with another.Cognitive linguistics: Cognitive linguistics is a newly established approach to the study of language as a reaction against the dominant generative paradigm which pursues an autonomous view of language.Cohesion: Cohesion is a concept to do with discourse or text rather than with syntax, it refers to relations of meaning that exist within the text, and defines it as a text.Design Features: 1.Arbitrariness: the forms of linguistic signs bear no natural relationship to their meaning. 2.Duality: the property of having two levels of structures, such that units of the primary level are composed of elements of the secondary level and each of the two levels has its own principles of organization. 3.Creativity: language is resourceful because of its duality and its recursiveness. 4.Displacement: human languages enable their users to symbolize objects, events and concepts which are not present at the moment of communication.Endocentric:endocentric construction is one whose distribution is functionally equivalent to that of one or more of its constituents, i.e.Functions of Language: 1. Informative 2. Interpersonal Function 3. Performative 4. Emotive Function 5. Phatic Communion 6. Recreational Function 7. Metalingual FunctionHeuristics: 1. What isn’t said, isn’t. 2. What is simply described is stereotypically exemplified. 3. What’s said in an abnormal way, isn’t normal; or Marked message indicates marked situation.Language: 1.human speech 2.the ability to communicate by this means 3.a system of vocal sounds and combinations of such sounds to which meaning is attributed, used for the expression or communication of thoughts and feelings 4.the written representation of such a systemLinguistics: Linguistics is usually defined as the science of language or, alternatively, as the scientific study of language. Main Branches of Linguistics: 1. Phonetics: phonetics studies speech sounds, including the production of speech, that is, how speech sounds are actually made, transmitted and received, the description and classification of speech sounds, words and connected speech, etc 2. Phonology: Phonology studies the rules governing the structure, distribution, and sequencing of speech sounds and the shape of syllables. 3. Morphology: Morphology is concerned with the internal organization of words. It studies the minimal units of meaning-morphemes and word-formation processes. 4. Syntax: Syntax is about principles of forming and understanding correct English sentences. 5. Semantics: Semantics examines how meaning is encoded in a language. 6.Pragmatics: Pragmatics is the study of meaning in context.Macrolinguistics: 1. Psycholinguistics: Psycholinguistics investigates the interrelation of language and mind, in processing and producing utterances and language acquisition for example. 2. Sociolinguistics: Sociolinguistics is an umbrella term which covers a variety of different interests in language and society, including the social functions of language and the social characteristics of its users. 3. Anthropological Linguistics: As a science, the study of language is somewhat older than anthropology. 4. Computational Linguistics: Computational linguistics is an interdisciplinary field which centers around the use of computers to process or produce human language.Morphology:The systematic study of morpheme is a branch of linguistics called morphology.Morpheme: Morpheme is the smallest unit of language in regard to the relationship between sounding and meaning, a unit that cannot be dicided into further smaller units without destroying or drastically altering the meaning.Metaphor: 1. Ontological metaphors 2. Structural metaphors 3. Orientional metaphorsNarrow transcription:The use of more specific symbols to show more phonetic detail is referred to as a narrow transcription.Open syllable: A syllable that has no coda.Origin of Language: 1. The “bow wow” theory: In primitive times people imitated the sounds of the animal calls in the wild environment they lived and speech developed from that. 2. The “pooh-pooh” theory: In the hard life of our primitive ancestors they utter instinctive sounds of pain, anger and joy. 3. The “yo-he-ho”theory: As primitive people worked together, they produced some rhythmic grunts which gradually developed into chants and then into language. Phonetics: Phonetics studies how speech sounds are produced, transmitted, and perceived.Psycholinguistics:Psycholinguistics is the study of psychological aspects of language; it usually studies the psychological states and mental activity associated withe the use of language.Phonology: Phonology is the study of the sound patterns and sound systems of languages.Phonetic similarity:means that the allophones of a phoneme must bear some phonetic resemblance.Phrase:phrase is a single element of structure containing more than one word, and lacking the subject-predicate structure typical of clauses.Recursiveness: Thought it mainly means that a phrasal constituent can be embedded within another constituent having the same category.Referential theory: The theory of meaning which relates the meaning of a word to the thing it refers, or stands for, is known as the referential theory.Stem: A stem is any morpheme or combination of morphemes to which an inflectional affix can be added.Synonymy: synonymy is the technical name for the sameness relation.Sentence: sentence is the minimum part of language that expresses a complete thought.Subordination: subordination refers to the process or result of linking linguistic units so that they have different syntactic status, one being dependent upon the other, and usually a constituent of the other.Syntax: sentences are constructed according to the grammar of arrangement.Syntactic function: The syntactic function shows the relationship between a linguistic form and other parts of the linguistic pattern in which it is used.Type of Morphemes:1. Free morpheme and bound morpheme 2. Root, affix and stemVoiceless: When the vocal folds are apart, the air can pass through easily and the sound produced is said to be voiceless. Voiced: When they are close together, the airstream causes them to vibrate against each other and the resultant sound is said to be voiced.Perseverative coarticulation: If the sound shows the influence of the preceding sound, it is perseverative coarticulation. Word: Word is a typical grammatical unit between morpheme and word group.Relation of substitutability: the relation of substitutability refers to classes or sets of words substitutable for each other grammatically in sentences with the same structure.。
语言学学名词解释1)Linguistics is generally defined as the scientific or systematic study of (human) language.a.The word language preceded by zero article in English implies that linguistics studies not any particular language, e.g. English , Chinese , French and Japanese, but languages in general.b.The word study does not mean “learn” but “investigate”.c.The word scientific refers to the way in whichlanguage is studied.It is a science in the sense that it scientifically studies the rules,systems and principles of human languages. It deals with a wide range of linguistic phenomena,analyzes them,and makes general statements about them.2)Linguistics is always guided by the 3 canons of science:(e c e)exhaustiveness: it strives for thorough-goingness in the examination of relevant materials;consistency: there should be no contradiction between different parts of the total statementeconomy: other things being equal, a shorter statement or analysis is to be preferred to one that is longer or more complex. (e c e)3) The subject matter of linguistics is all natural language, living or dead.4) Linguistics has 2 main purposes:it studies the nature of language and tries to establish a theory of language, and describes languages in the light of the theory established.It examines all the forms of language in general and seeks a scientific understanding of the ways in which it is organized to fulfill the needs it serves and the functions it performs in human lifelinguistics differs from traditional grammar at least in 3 basic ways: Linguistics describes languages and does not lay down rules of correctness. Linguists are interested in what is said. So they are often said to be descriptive.Linguists regard the spoken language as primary. It is believed that speech came into being first for any human language and the writing system came along much later. Traditional grammar is based on Latin and it tries to impose the Latin categories and structures on other languages, while linguistics describes each language on its own merits.For a student of language, some knowledge of linguistics is of both interest and importance.For a teacher of foreign languages, he will definitely a great deal from the knowledge of linguistics.For a researcher, there is even more scopefor displaying his abilities.Why study linguistics ?1. Linguistics takes an analyticalapproach to the study of language, and focus on developing skills in data analysis, problem solving, and logical thinking that can be applied to many fields.2.It is an interdisciplinary subject.3.Linguistics is a science that is stillin its infancy but undergoing rapid development, and it is “a pilot science”.What and how linguists study language?1. nature of language (focus on languageitself)2. nature of acquisition (focus onlearners)3. nature of teaching (focus on teachers)The process of linguistic study can be summarized as follows:between language and society: how social factors influence the structure and use of language. Another name for sociolinguistics is the sociology of language. Psycholinguistics is the study of language and mind: the mental structures and processes which are involved in the acquisition, comprehension and production of language.Neurolinguistics or neurological linguistics is the study of language processing and language representation in the brain.Stylistics is the study of how literary effects can be related to linguistic features. It usually refers to the study of written language, including literary texts, but it also investigates spoken language sometimes.Discourse analysis, or text linguistics is the study of the relationship between language and the contexts in which language is used.Computational linguistics is an approach to linguistics which employs mathematical techniques, often with the help of a computer.Cognitive linguistics is an approach to the analysis of natural language that focuses on language as an instrument for organizing, processing, and conveying information. Apart from the different branches of linguistics, there are some distinctions of linguistics, such as: functional linguistics vs formal linguistics; theoretical linguistics vs applied linguistics.Applied linguistics is primarily concerned with the application of linguistic theories, methods and findings to the elucidation of language problems which have arisen in other areas of experience.Phonetics（语音学）:Phonetics is the scientific study of speech sounds.It studies how speech sounds arearticulated, transmitted, and received. It is a pure science and examines speech sounds in general.Phonetics: The general study of the characteristics of speech sounds. Phonology（音系学/音位学）: The description of the systems and patterns of speech sounds in a language.Phonology is the study of how speech sounds function in a language. It studies the ways speech sounds are organized. It can be seen as the functional phonetics of a particular language.Morphology（形态学，词法学）: The study of the way in which morphemes are arranged to form words.Morphology is the study of the formation of words. It is a branch of linguistics which breaks words into morphemes.Syntax（句法学）: The study of those rules that govern the combination of words to form permissible sentences.Syntax deals with the combination of words into phrases, clauses and sentences. It is the grammar of sentence construction. Semantics(语义学）is a branch of linguistics which is concerned with the study of meaning in abstraction. Pragmatics can be defined as the study of language in use. It deals with how speakers use language in ways which cannot be predicted from linguistic knowledge alone, and how hearers arrive at the intended meaning of speakers. In a broad sense, pragmatics studies the principles observed by human beings when they communicate with one another.Language is a system of arbitrary vocal symbols used for human communication.This definition is widely accepted because it includes some of the important characteristics of human language. Language as system ---The key word in thedefinition is "system". Language is systematic. Otherwise we would not be able to learn or use it consistently. Each language system contains two subsystems: a system of sound and a system of meaning. Language is a system—elements in it are not arranged and combined randomly, but according to some rules and principles. Language as arbitrary ---There is no natural relationship between the sound and what it means in a certain language.The relation between sound and meaning is almost always arbitrary or conventional. The relation between sound and meaning is almost always arbitrary.A rose by any other name would smell as sweet.A rose by any other name would smell as sweet.Romeo and JulietThe relation between sound and meaning is almost always conventional《荀子·正名》：“名无固宜，约之以命，约定俗成谓之宜，异于约则谓之不宜。
语言学作业班级：姓名：Chapter 1 Invitations to LinguisticsI. Please illustrate the following terms.1. Arbitrariness:The forms of linguistic signs bear no natural relationship to their meaning.The different levels of arbitrariness:(1) Arbitrary relationship between the sound of a morpheme and its meaning, even with onomatopoeic words(2) Arbitrariness at the syntactic level: language is not arbitrary at the syntactic level.(3) The link between a linguistic sign and its meaning is a matter of convention.2. DualityThe property of having two levels of structures, such that units of the primary level are composed of elements of the secondary level and each of the two levels has its own principles of organization.3. Phatic communionPhatic communion refers to the social interaction of language.4. Synchronic linguistics:A synchronic description takes a fixed instant (usually, but not necessarily, the present) as its point of observation. Most grammars are of this kind.II. Please distinguish the following terms:1. Langue vs. ParoleLangue refers to the abstract linguistic system shared by all the members of a speech community, that is, the lexicon, grammar, and phonology implanted in each individual, and it is the linguist’s proper object;Parole refers to the realization of langue, the immediately accessible data. While parole constitutes the immediately accessible data, and it is a mass of confused facts, so it is not suitable for systematic investigation..(1) Langue is abstract, while parole is specific to the situation in which it occurs.(2) Langue is not actually spoken by anyone, while parole is always a naturally occurring event.(3) Langue is relatively stable, systematic and social, while parole is subject to personal, individual and situational constraints.(4) Langue is essential while parole is accessory and accidental.2. Descriptive vs. PrescriptiveThe distinction lies in prescribing how things ought to be and describing how things are.Traditional grammar was very strongly normative in grammarians tried to lay down rules for the correct use of language and settle the disputes over usage once and for all. That is prescriptive.These attitudes are still with us, though people realize nowadays the facts of usage count more than the authority-made “standards”. Thenature of linguistics as a science determines its preoccupation with description instead of prescription.3. Synchronic vs. DiachronicSynchronic description takes a fixed instant (usually, but not necessarily, the present) as its point of observation. Most grammars are of this kind.Actually synchrony is a fiction since any language is changing as the minutes pass.Diachronic linguistics is the study of a language through the course of its history.4. Competence vs. PerformanceAccording to Chomsky:A language user’s underlying knowledge about the system of rules is called his linguistic competence.Performance refers to the actual use of language or the actual realization of this knowledge in utterances in concrete situations.A speaker’s competence is stable while his performance is often influenced by psychological and social factors, so a speaker’s performance does not always or equal his supposed competence.He believes that linguists ought to study competence rather than performance.5. Langue vs. CompetenceAccording to Chomsky:Langue is a social product, a systematic inventory of rules of the language, a set of conventions for a speech community.Competence is defined from the psychological point of view, is deemedas a property of the mind of each individuals, or underlying competence as a system of generative processes.According to Hymes:He approaches language from a socio-cultural viewpoint with the aim of studying the varieties of ways of speaking on the part of individual and the community.He extended notion of competence, restricted by Chomsky to a knowledge of grammar, to incorporate the pragmatic ability for language use. This extended idea of competence can be called communicative competence.III. Answer the following questions in brief:1. The following are some book titles of linguistics. Can you judge thesynchronic or diachronic orientation just from the titles1) English Examined: Two Centuries of Comment on the Mother Tongue2) Protean Shape: A Study in Eighteenth-century Vocabulary and Usage3) Pejorative Sense Development in English4) The Categories and Types of Present-Day English Word-Formation5) Language in the Inner City: Studies in the Black English Vernacular 1) diachronic 2）synchronic 3）diachronic 4）synchronic5）We can’t judge whether it is synchronic or diachronic orientation just from the titles.2. What is language What is linguisticsLanguage can be defined as a system of arbitrary vocal symbols used for human communication and interaction.Linguistics is the scientific study of human language. The aims of linguistic theory: 1) what is knowledge of language (Competence) 2) howis knowledge of language acquired (Acquisition) 3) how is knowledge of language put to use (Performance/language processing). Main branches of linguistics:Phonetics, Phonology Morphology, Syntax, Semantics, Pragmatics.3. How do you understand performative function of languageThe performative function of language is primarily to change the social status of persons or the situations of events, as in marriage ceremonies, the sentencing of criminals, the blessing of children, the naming of a ship at a launching ceremony, and the cursing of enemies.The kind of language employed in performative verbal acts is usually quite formal and even ritualized.The performative function can extend to the control of reality as on some magical or religious occasions.For example, in Chinese when someone breaks a bowl or a plate the host or the people present are likely to say sui sui ping an as a means of controlling the invisible forces which the believers feel might affect their lives adversely.IV. Discuss the following question in detail.How do you interpret the viewpoint that “arbitrariness is a matter of degree”1)Arbitrary relationship between the sound of a morpheme and its meaning, even with onomatopoeic words:The dog barks bow wow in English but “汪汪汪” in Chinese.2) Arbitrariness at the syntactic level: language is not arbitrary at the syntactic level.He came in and sat down.He sat down and came in.He sat down after he came in.3) The link between a linguistic sign and its meaning is a matter of convention.Arbitrariness of language makes it potentiallycreative.Conventionality of language makes learning a languagelaborious.Chapter 2 Speech SoundsI. Complete the following statements.1. Human language enable their users to symbolize objects, events andconcepts which are not present (in time and space) at the moment of communication. This quality is labeled as __________.2. The sound [p] can be described with “voiced, __________, stop.”3. The different members of a phoneme, sounds which are phoneticallydifferent but do not make one word different from another in meaning,, are_________.4. Both semantics and ________ investigate linguistic meaning, but theyfocus on different aspects.5. If certain linguistics tries to lay down rules for the correct useof language and settle the disputes over usage once and for all, it is ___________ linguistics.6. Phones that fall into allophones of a phoneme have to satisfy twoconditions, one is they are ___________________, and another is that they should be in _____________________.7. The vowel ________ is high front tense unrounded.8. A dog cannot tell people that its master will be home in a few days,because its language does not have the feature of ___________.9. Computational linguistics often refers to the problems of________________, information retrieval, and ______________.10. Halliday proposed a theory of metafunctions of language, that is,language has ___________, ____________ and _____________ functions. II. Define the following terms.1. Manner of articulation:2. Distinctive features:3. Intonation:4. Assimilation:III. Answer the following questions briefly.1. Specify the difference between each pair of sounds using distinctivefeatures.1) [l] [ł] 2) [p h] [p] 3) [b] [d] 4) [k] [g] 5) [I][u]2. Work out the features of the following sounds.1) [t h] ________________________________________2) [w] ________________________________________3) [v] ________________________________________4) [ð] _________________________________________5) [l] __________________________________________3. In some dialects of English the following words have different vowels,as shown by the phonetic transcription. Based on these data, answer the questions that follow.A B. Cbite [bʌit] bide [ba i d] tie [ta i]rice [rʌis] rise [ra i z] by [ba i]type [tʌip] bribe [b r aib] sigh [s a i]wife [wʌif] wives [wa i vz] die [d a i]tyke [tʌik] time [ta i m] why [wa i]1) What is the difference of the sounds that end the words in columnsA and B2) How do the words in column C differ from those in column A and B3) Are [ʌi] and [a i] in complementary distribution Give your reasons.4) What are the phonetic transcriptions of (a) life and (b) lives5) What would the phonetic transcriptions of the following words be inthe dialects of English shown in the data(a) trial (b) bike (c) lice(d) fly (e) mine6) State the rule that will relate the phonemic representations to bephonetic transcriptions of the words given above.IV. Discuss the questions in details.1. Illustrate phoneme, phone and allophone.2. To what extent is phonology related ot phonetics and how do they differ。
I am an individual with a multifaceted skill set that has been honed through persistent dedication, continuous learning, and practical application across various domains. This essay aims to provide you with a comprehensive understanding of my core competencies, which I have cultivated over the years.Firstly, in the realm of academics, I specialize in the field of Computer Science and Engineering. I hold a strong command over programming languages such as Python, Java, and C++. My expertise extends from front-end web development to back-end systems architecture, including proficiency in frameworks like ReactJS, AngularJS, Node.js, and Django. I have consistently delivered projects involving complex algorithms, data structures, machine learning models, and software engineering principles. Beyond coding, I possess a keen interest in computational theory and artificial intelligence, allowing me to approach problems with both a theoretical and practical perspective.Secondly, my forte lies in effective communication and leadership abilities. Throughout my academic and professional journey, I've had numerous opportunities to lead teams, organize events, and conduct presentations. These experiences have fine-tuned my interpersonal skills and ability to articulate ideas clearly and persuasively. I excel at interpreting technical information for non-technical stakeholders, thereby bridging the gap between technical and business realms. Moreover, my empathetic leadership style enables me to foster collaborative environments where team members feel valued and motivated.Thirdly, I pride myself on my critical thinking and problem-solving skills. Whether it's tackling intricate mathematical conundrums or resolving real-world business challenges, I apply analytical rigor to every task. I approach problems systematically, breaking them down into manageable components before devising innovative solutions. This strategic mindset is instrumental in project planning and execution, ensuring that objectives are met efficiently and effectively.In the domain of research and writing, I am adept at conducting meticulous literature reviews, synthesizing information from diverse sources, and presenting findings coherently. My written work reflects a high degree of clarity,depth, and adherence to scholarly standards. I have published articles in reputable journals, showcasing my ability to contribute meaningfully to the intellectual discourse in my field.Lastly, I am passionate about continuous personal and professional growth.I believe in lifelong learning and hence invest time in staying updated with the latest trends and technologies in my industry. My adaptability and eagerness to learn new skills have allowed me to navigate through different roles and responsibilities with ease.In terms of extracurriculars, I also boast significant accomplishments in public speaking and debating, demonstrating my confidence and poise under pressure. Additionally, I am fluent in English, Mandarin, and Spanish, enabling me to engage in cross-cultural dialogue and collaborate effectively in international settings.In conclusion, my strengths lie in a blend of technical prowess, leadership acumen, critical thinking, research aptitude, and cultural sensitivity. These attributes equip me to excel in dynamic, challenging environments where innovation and collaboration are key. I continually strive to refine these skills and expand my horizons, driven by an insatiable curiosity and passion for excellence. Each day presents a new opportunity to deepen my knowledge, sharpen my abilities, and make meaningful contributions to any endeavor I undertake.This self-portrait, while highlighting my current proficiencies, underscores my commitment to ongoing improvement and the relentless pursuit of quality and high standards in all aspects of my life and career.。
我在计算方面进步了英语作文250字In terms of computation, I have made significant progress. Initially, I was apprehensive about tackling complex calculations, but with consistent practice, I am now able to approach mathematical challenges with confidence. My understanding of core concepts like algebra, calculus, and probability has deepened, and I am adept at applying these principles to real-world scenarios.One notable achievement has been my proficiency in solving optimization problems. Through exposure to various techniques such as linear programming and dynamic programming, I can formulate and solve optimization models efficiently. This has proven invaluable in my academic coursework, where I have successfully applied these methods to optimize resource allocation, scheduling, and decision-making.Moreover, I have developed a strong foundation in data analysis and visualization. I am proficient in usingstatistical software packages like R and Python to analyze large datasets, extract meaningful insights, and communicate results through compelling visualizations. My ability to interpret data effectively has enabled me to contribute to research projects and make informed decisions based on empirical evidence.Furthermore, I am actively seeking opportunities to enhance my computational skills. I regularly participate in online coding challenges and hackathons, which provide a platform to test my abilities and learn from others. I am also exploring advanced topics such as machine learning and artificial intelligence, which have the potential to revolutionize various industries.Complementing my technical proficiency, I have also cultivated a strong work ethic and problem-solving mindset.I am persistent in tackling challenges, even when faced with setbacks. I believe that errors are valuable learning opportunities, and I strive to learn from my mistakes and refine my approach accordingly.Overall, my journey in computation has been marked by steady progress and a growing passion for the subject. I am eager to continue expanding my knowledge and sharpening my skills, as I firmly believe that computational thinking is an essential asset in today's rapidly evolving technological landscape.。
课程名称中英文对照表(1)工学ENGINEERING 课程中文名称课程英文名称高等数理方法Advanced Mathematical Method 弹塑性力学Elastic-Plastic Mechanics 板壳理论Theory of Plate and Shell 高等工程力学Advanced Engineering Mechanics 板壳非线性力学 Nonlinear Mechanics of Plate and Shell 复合材料结构力学 Structural Mechanics of Composite Material 弹性元件的理论及设计Theory and Design of Elastic Element 非线性振动 Nonlinear Vibration 高等土力学 Advanced Soil Mechanics 分析力学Analytic Mechanics 随机振动Random Vibration 数值分析Numerical Analysis 基础工程计算与分析Calculation and Analysis of Founda tion Engineering 结构动力学Structural Dynamics 实验力学 Laboratory Mechanics 损伤与断裂Damage and Fracture 小波分析 Wavelet Analysis 有限元与边界元分析方法 Analytical Method of Finite Element and Boundary Element 最优化设计方法 Optimal Design Method 弹性力学 Elastic Mechanics 高层建筑基础 Tall Building Foundation 动力学 Dynanics 土的本构关系Soil Constitutive Relation 数学建模 Mathematical Modeling 现代通信理论与技术 Emerging Communications Theory and Technology 数字信号处理 Digital Signal Processing 络理论与多媒体技术 Multi-media and Network Technology 医用电子学 Electronics for Medicine 计算微电子学Computational Microelectronics 集成电路材料和系统电子学Material and System Electronics for In tegrated Circuits 络集成与大型数据库 Computer Network Integrating Technology and Large scale Database 现代数字系统 Modern Digital System 微机应用系统设计Microputer Application Design 计算机络新技术 Modern Computer Network Technologies 络信息系统 Network Information System 图像传输与处理 Image Transmission and Processing 图像编码理论 Theory of Image Coding 遥感技术Remote Sensing Techniques 虚拟仪器系统设计Design of Virtual Instrument System 生物医学信号处理技术 Signal Processing for Biology and Medicine 光纤光学 Fiber Optics VLSI 的EDA技术 EDA Techniques for VLSI 电子系统的ASIC技术 ASIC Design Technologies VLSI技术与检测方法VLSI Techniques & Its Examination 专题阅读或专题研究 The Special Subject Study 信息论Information Theory 半导体物理学 Semiconductor Physics 通信原理Principle of Communication 现代数理逻辑Modern Mathematical Logic 算法分析与设计 Analysis and Design of Algorithms 高级计算机络Advanced Computer Networks 高级软件工程Advanced Software Engineering 数字图像处理 Digital Image Processing 知识工程原理Principles of Knowledge Engineering 面向对象程序设计Object-Oriented Programming 形式语言与自动机 Formal Languages and Automata 人工智能程序设计 Artificial Intelligence Programming 软件质量与测试 Software Quality and Testing 大型数据库原理与高级开发技术 Principles of Large-Scale Data-Bas e and Advanced Development Technology 自然智能与人工智能 Natural Intelligence and Artificial Intelligence Unix操作系统分析 Analysis of Unix System 计算机图形学Computer Graphics Internet与Intranet技术Internet and Intranet Technology 多媒体技术 Multimedia Technology 数据仓库技术与联机分析处理 Data Warehouse and OLAP 程序设计方法学 Methodology of Programming 计算机信息保密与安全 Secrecy and Security of Computer Information 电子商务 Electronic Commerce 分布式系统与分布式处理Distributed Systems and Distributed Processing 并行处理与并行程序设计 Parallel Processing and Parallel Programming 模糊信息处理技术Fuzzy Information Processing Technology 人工神经络及应用 Artificial Intelligence and Its Applications Unix编程环境 Unix Programming Environment 计算机视觉Computer Vision 高级管理信息系统Advanced Management Information Systems 信息系统综合集成理论及方法Theory and Methodology of Information n System Integration 计算机科学研究新进展 Advances in Computer Science 离散数学 Discrete Mathematics 操作系统 Operating System 数据库原理 Principles of Database 编译原理Principles of Compiler 程序设计语言Programming Language 数据结构 Data Structure 计算机科学中的逻辑学 Logic in Computer Science 面向对象系统分析与设计 Object-Oriented System Analysis and Design 高等数值分析Advanced Numeric Analysis 人工智能技术 Artificial Intelligence Technology 软计算理论及应用 Theory and Application of Soft-Computing 逻辑程序设计与专家系统Logic Programming and Expert Systems 模式识别Pattern Recognition 软件测试技术 Software Testing Technology 高级计算机。
Computational Principles of Learning in the Neocortex and HippocampusRandall C.O’Reilly&Jerry W.RudyDepartment of PsychologyUniversity of Colorado BoulderCampus Box345Boulder,CO80309oreilly@11th October2000Published in Hippocampus,2000,vol10,p.389-397AbstractWe present an overview of our computational approach towards understanding the different contributions of the neo-cortex and hippocampus in learning and memory.The approach is based on a set of principles derived from converging biological,psychological,and computational constraints.The most central principles are that the neocortex employs a slow learning rate and overlapping distributed representations to extract the general statistical structure of the envi-ronment,while the hippocampus learns rapidly using separated representations to encode the details of specific events while suffering minimal interference.Additional principles concern the nature of learning(error-driven and Hebbian), and recall of information via pattern completion.We summarize the results of applying these principles to a wide range of phenomena in conditioning,habituation,contextual learning,recognition memory,recall,and retrograde amnesia, and point to directions of current development.2Computational Principles of LearningIntroductionThis paper presents a computational approach to-wards understanding the different contributions of the neocortex and hippocampus in learning and memory. This approach uses basic principles of computational neural network learning mechanisms to understand both what is different about the way these two neural systems learn,and why they should have these differences.Thus, the computational approach can go beyond mere descrip-tion towards understanding the deeper principles under-lying the organization of the cognitive system.These principles are based on an convergence of biological, psychological,and computational constraints,and serve to bridge between these different levels of analysis.The set of principles discussed in this paper werefirst developed in McClelland,McNaughton,and O’Reilly (1995),and have been refined several times since then (O’Reilly,Norman,&McClelland,1998;O’Reilly& Rudy,1999,in press).The computational principles have been applied to a wide range of learning and memory phenomena across several species(rats,monkeys and hu-mans).For example,they can account for impaired and preserved learning capacities with hippocampal lesions in conditioning,habituation,contextual learning,recog-nition memory,recall,and retrograde amnesia.This pa-per provides a concise summary of the previous work, and a discussion of current and future directions.The PrinciplesThere are several levels of principles that can be dis-tinguished by their degree of specificity in characteriz-ing the nature of the underlying mechanisms.We be-gin with the most basic principles and proceed towards greater specificity.Learning Rate,Overlap,and Interference The most basic set of principles can be motivated by considering how subsequent learning can interfere with prior learning.A classic example of this kind of interfer-ence can be found in the associative learning task(e.g.,Barnes&Underwood,1959).The repre-sents one set of words that are associated with two dif-ferent sets of other words,and.For example,the word window will be associated with the word reason in the list,and associated with locomotive on the list.After studying the list of associates,subjects are tested by asking them to give the appropriate associate for each of the words.Then,subjects study thelist(often over multiple iterations),and are subsequently tested on both lists for recall of the associates after eachLearning Trials on AC List255075100%CorrectLearning Trials on AC List255075100a) AB−AC List Learning in Humans b) AB−AC List Learning in ModelFigure1:Human and model data for AB-AC list learning.a)Humans show some interference for the AB list items as a function of new learning on the AC list items.b)Model shows a catastrophic level of interference.(data reproduced from Mc-Closkey&Cohen,1989).iteration of learning the list.Subjects exhibit some level of interference on the initially learned asso-ciations as a result of learning the list,but they still remember a reasonable percentage(see Figure1a for rep-resentative data).Thefirst set of principles concern the effects of over-lapping representations(i.e.,shared units between two different distributed representations)and rate of learn-ing on the ability to rapidly learn new information with a level of interference characteristic of human subjects: Overlapping representations lead to interference (conversely,separated representations prevent inter-ference).A faster learning rate causes more interference(conversely,a slower learning rate causes less in-terference).The mechanistic basis for these principles within a neural network perspective is straightforward.Interference is caused when weights used to encode one association are disturbed by the encoding of another(Figure2a).Over-lapping patterns share more weights,and therefore lead to greater amounts of interference.Clearly,if entirely separate representations are used to encode two different associations,then there will be no interference whatso-ever(Figure2b).The story with learning rate is simi-larly straightforward.Faster learning rates lead to more weight change,and thus greater interference(Figure3). However,a fast learning rate is necessary for rapid learn-ing.Integration and Extracting Statistical Structure Figure3shows theflip side of the interference story, integration.If the learning rate is low,then the weights will integrate over many experiences,reflecting the un-derlying statistics of the environment(White,1989;Mc-Clelland et al.,1995).Furthermore,overlapping repre-sentations facilitate this integration process,because theO’Reilly 3ABList 1: A−>BABCList 2: A−>CBA1B CA1A2List 1: A−>B List 2: A−>C a) Remembering specifics using same representations causes interference.b) Separated (conjunctive) represen−tations and fast weight changes enable rapid learning of specifics without interference.Figure 2:Interference as a function of overlapping (same)representations versus separated representations.a)Using the same representation to encode two different associations (and )causes interference —the subsequent learningofinterferes with the prior learning of because the stimulus must have stronger weights to than to for the second association,as is reflected in the weights.b)A sepa-rated representation,where is encoded separated for the first list ()versus the second list ()prevents interference.0−0.2− 0.4− 0.6− 0.8− 1−Figure 3:Weight value learning about a single input unit thatis either active or not.The weight increases when the input is on,and decrease when it is off,in proportion to the size of the learning rate.The input has an overall probability of be-ing active of rger learning rates (.1or 1)lead to more interference on prior learning,resulting in a weight value that bounces around substantially with each training example.In the extreme case of a learning rate of 1,the weight only reflects what happened on the previous trial,retaining no memory for prior events at all.As the learning rate gets smaller (.005),the weight smoothly averages over individual events and reflects the overall statistical probability of the input being active.same weights need to be reused across many different experiences to enable the integration produced by a slow learning rate.This leads to the next principle:Integration across experiences to extract underlying statistical structure requires a slow learning rate and overlapping representations.Episodic Memory and Generalization:Incom-patible FunctionsThus,focusing only on pattern overlap for the mo-ment,we can see that networks can be optimized for two different,and incompatible,functions:avoiding interfer-ence or integrating across experiences to extract general-ities.Avoiding interference requires separated represen-tations,while integration requires overlapping represen-tations.These two functions each have clear functional advantages,leading to a further set of principles:Interference avoidance is essential for episodic memory,which requires learning about the specifics of individual events and keeping them separate from other events.Integration is essential for encoding the general statistical structure of the environment,abstracted away from the specifics of individual events,which enables generalization to novel situations.The incompatibility between these functions is fur-ther evident in these descriptions (i.e.,encoding specifics versus abstracting away from them).Also,episodic memory requires relatively rapid learning —an event must be encoded as it happens,and does not typically re-peat itself for further learning opportunities.This com-pletes a pattern of opposition between these functions:episodic learning requires rapid learning while integra-tion and generalization requires slow learning.This is summarized in the following principle:Episodic memory and extracting generalities are in opposition.Episodic memory requires rapid learn-ing and separated patterns,while extracting gener-alities requires slow learning and overlapping pat-terns.Applying the Principles to the Hippocampus and NeocortexArmed with these principles,the finding that neural network models that have highly overlapping represen-tations exhibit catastrophic levels of interference (Mc-Closkey &Cohen,1989,Figure 1b)should not be sur-prising.A number of researchers showed that this in-terference can be reduced by introducing various factors4Computational Principles of Learningthat result in less pattern overlap(e.g.,Kortge,1993;French,1992;Sloman&Rumelhart,1992;McRae& Hetherington,1993).Thus,instead of concluding that all neural networks are fundamentallyflawed,as Mc-Closkey and Cohen(1989)argued(and a number of oth-ers have uncritically accepted),McClelland et al.(1995)argued that this catastrophic failure serves as an impor-tant clue into the structure of the human brain.Specifically,we argued that because of the funda-mental incompatibility between episodic memory and extracting generalities,the brain should employ two sep-arate systems that each optimize these two objectives in-dividually,instead of having a single system that triesto strike an inferior compromise.This line of reasoning provides a strikingly goodfit to the known properties of the hippocampus and neocortex,respectively.The de-tails of thisfit in various contexts is the substance of the remainder of the paper,but the general idea is that: The hippocampus rapidly binds together informa-tion using pattern-separated representations to min-imize interference.The neocortex slowly learns about the general sta-tistical structure of the environment using overlap-ping distributed representations.(see also Sherry&Schacter,1987for a similar conclu-sion).Before discussing the details,a few more princi-ples need to be developedfirst.Conjunctive Representations and NonlinearDiscrimination LearningThe conjunctive or configural representations the-ory provides a converging line of thinking about the na-ture of hippocampal function(Sutherland&Rudy,1989; Rudy&Sutherland,1995;Wickelgren,1979;O’Reilly& Rudy,1999).A conjunctive/configural representation is one that binds together(conjoins or configures)multi-ple elements into a novel unitary representation.This is consistent with the description of hippocampal func-tion given above,based on the need to separate patterns to avoid interference.Indeed,it is clear that pattern separation and conjunctive representations are two sides of the same coin,and that both are caused by the use of sparse representations(having relatively few active neurons)that are a known property of the hippocampus (O’Reilly&McClelland,1994;O’Reilly&Rudy,1999). To summarize:Sparse hippocampal representations lead to pattern separation(to avoid interference)and conjunctive representations(to bind together features into a uni-tary representation).One important application of the conjunctive rep-resentations idea has been to nonlinear discrimination problems.These problems require conjunctive represen-tations to solve because each of the individual stimuli isambiguous(equally often rewarded and not rewarded). The negative patterning problem is a good example.Itinvolves two stimuli,and(e.g.,a light and a tone), which are associated with reward(indicated by)or not().Three different trial types are trained:, ,.Thus,the conjunction of the two stimuli ()must be treated differently from the two stim-uli separately(,).A conjunctive representa-tion that forms a novel encoding of the two stimuli to-gether can facilitate this form of learning.Therefore, the fact that hippocampal damage impairs learning the negative patterning problem(Alvarado&Rudy,1995; Rudy&Sutherland,1995;McDonald,Murphy,Guar-raci,Gortler,White,&Baker,1997)would appear to support the idea that the hippocampus employs pattern separated,conjunctive representations.However,it is now clear that a number of other nonlinear discrimina-tion learning problems are unimpaired by hippocampal damage(Rudy&Sutherland,1995).The next set of prin-ciples help to make sense of these data so that they can be reconciled with our interference-based principles of con-junctive pattern separation,as discussed in a subsequent section.Pattern Completion:Recalling a ConjunctionPattern completion is required for recalling informa-tion from conjunctive hippocampal representations,yet it conflicts with the process of pattern separation that forms these representations in thefirst place(O’Reilly& McClelland,1994).Pattern completion occurs when a partial input cue drives the hippocampus to complete to an entire previously-encoded set of features that were bound together in a conjunctive representation.For a given input pattern,a decision must be made to recog-nize it as a retrieval cue for a previous memory and per-form pattern completion,or to perform pattern separa-tion and store the input as a new memory.This deci-sion is often difficult given noisy inputs and degraded memories.The hippocampus implements this decision as the effects of a set of basic mechanisms operating on input patterns(O’Reilly&McClelland,1994;Has-selmo&Wyble,1997),and it does not always do what would seem to be the right thing to do from an omniscient perspective knowing all the relevant task factors—this can complicate the involvement of the hippocampus in nonlinear discrimination problems.O’Reilly5Learning Mechanisms:Hebbian and Error DrivenTo more fully explain the roles of the hippocampus and neocortex we need to understand how learning works in these systems(the basic principles just described do not depend on the detailed nature of the learning mech-anisms;White,1989).There are two basic mechanisms that have been discussed in the literature,Hebbian and error-driven learning(e.g.,Marr,1971;McNaughton& Morris,1987;Gluck&Myers,1993;Schmajuk&Di-Carlo,1992).Briefly,Hebbian learning(Hebb,1949) works by increasing weights between co-active neurons (and usually decreasing weights when a receiver is ac-tive and the sender is not),which is a well-established property of biological synaptic modification mechanisms (e.g.,Collingridge&Bliss,1987).Hebbian learning is useful for binding together features active at the same time(e.g.,within the same episode),and has therefore been widely suggested as a hippocampal learning mech-anism(e.g.,Marr,1971;McNaughton&Morris,1987).Error-driven learning works by adjusting weights to minimize the errors in a network’s performance,with the best example of this being the error backpropaga-tion algorithm(Rumelhart,Hinton,&Williams,1986). Error-driven learning is sensitive to task demands in a way that Hebbian learning is not,and this makes it a much more capable form of learning for actually achiev-ing a desired input/output mapping.Thus,it is natural to associate this form of learning with the kind of pro-cedural or task-driven learning that the neocortex is of-ten thought to specialize in(e.g.,because amnesics with hippocampal damage have preserved procedural learning abilities).Although the backpropagation mechanism has been widely challenged as biologically implausible(e.g., Crick,1989;Zipser&Andersen,1988),a recent analysis shows that simple biologically-based mechanisms can be used to implement this mechanism(O’Reilly,1996),so that it is quite reasonable to assume that the cortex de-pends on this kind of learning.Although the association of Hebbian learning with the hippocampus and error-driven learning with the cor-tex is appealing in some ways,it turns out that both kinds of learning play important roles in both systems (O’Reilly&Rudy,1999;O’Reilly&Munakata,2000; O’Reilly,1998).Thus,the specific learning principles adopted here are that both forms of learning operate in both systems:Hebbian learning binds together co-occurring fea-tures(in the hippocampus)and generally learns about the co-occurrence statistics in the environ-ment across many different patterns(in neocortex).Error-driven learning shapes learning according tospecific task demands(shifting the balance of pat-tern separation and completion in the hippocampus, and developing task-appropriate representations in the neocortex).It is the existence of this task-driven learning that com-plicates the picture for nonlinear discrimination learning problems.A Summary of PrinciplesThe above principles can be summarized with the fol-lowing three general statements of neocortical and hip-pocampal learning properties(O’Reilly&Rudy,1999): Learning rate.The cortical system typically learns slowly,while the hippocampal system typically learns rapidly.Conjunctive bias.The cortical system has a bias to-wards integrating over specific instances to extract generalities.The hippocampal system is biased by its intrinsic sparseness to develop conjunctive repre-sentations of specific instances of environmental in-puts.However,this conjunctive bias trades-off with the countervailing process of pattern completion,so the hippocampus does not always develop new con-junctive representations(sometimes it completes to existing ones).Learning mechanisms.Both cortex and hippocampus use error-driven and Hebbian learning.The error-driven aspect responds to task demands,and will cause the network to learn to represent whatever is needed to achieve goals or ends.Thus,the cortex can overcome its bias and develop specific,conjunc-tive representations if the task demands require this.Also,error-driven learning can shift the hippocam-pus from performing pattern separation to perform-ing pattern completion,or vice-versa,as dictated by the task.Hebbian learning is constantly operating, and reinforcing the representations that are activated in the two systems.These principles are focused on distinguishing neo-cortex and hippocampus—we have also articulated a more complete set of principles that are largely common to both systems(O’Reilly,1998;O’Reilly&Munakata, 2000).Models incorporating these principles have been extensively applied to a wide range of different cortical phenomena,including perception,language,and higher-level cognition.In the next section,we highlight the ap-plication of the principles presented here to learning and memory phenomena involving both the cortex and hip-pocampus.6Computational Principles of LearningApplications of the PrinciplesThe principles just developed have been applied to a number of different domains,as summarized in the following sections.In most cases,the same neural network model developed according to these principles(O’Reilly&Rudy,1999)was used to simulate the em-pirical data,providing a compelling demonstration that the principles are sufficient to account for a wide range offindings.Conjunctions and Nonlinear Discrimination LearningWefirst apply the above principles to the puzzling pattern of hippocampal involvement in nonlinear dis-crimination learning problems.The general statement of the issue is that although we think the hippocampus is specialized for encoding conjunctive bindings of stimuli (and keeping these separated from each other to mini-mize interference),apparently direct tests of this idea in the form of nonlinear discrimination learning problems have not provided clear support.Specifically,rats with hippocampal lesions can learn a number of these non-linear discrimination problems just like intact rats.The general explanation of these results according to the full set of principles outlined above is that:The explicit task demands present in a nonlinear discrimination learning problem cause the cortex alone(with a lesioned hippocampus)to learn the task via error-driven learning.Nonlinear discrimination problems take many trials to learn even in intact animals,allowing the slow cortical learning to accumulate a solution.The absence of hippocampal learning speed advan-tages in normal rats,despite the more rapid hip-pocampal learning rate,can be explained by the fact that the hippocampus is engaging in pattern comple-tion in these problems,instead of pattern separation.We substantiated this verbal account by running com-putational neural network simulations that embodied the principles developed above(O’Reilly&Rudy,1999; Figure4).These simulations showed that in many—but not all—cases,removing the hippocampal compo-nent did not significantly impair learning performance on nonlinear discrimination learning problems,matching the empirical data.Figure5shows one specific example, where the negative patterning problem discussed previ-ously(,,)is impaired with hippocampal lesions,but performance is not impaired on a very simi-lar ambiguous feature problem,,,,,HippocampusCortexEC_inFigure4:The O’Reilly and Rudy(1999)model,showing both cortical and hippocampal components.The cortex has12dif-ferent input dimensions(sensory pathways),with4different values per dimension.These are represented separately in the elemental cortex(Elem).Higher level association cortex(As-soc)can form conjunctive representations of these elements,if demanded by the task.The interface to the hippocampus is via the entorhinal cortex,which contains a one-to-one mapping of the elemental,association,and output cortical representations. The hippocampus can reinstate a pattern of activity over the cortex via the EC.(Gallagher&Holland,1992).See O’Reilly and Rudy (1999)for a detailed discussion of the differential per-formance on these tasks.To summarize,this work showed that it is essential to go beyond a simple conjunctive story and include a more complete set of principles in understanding hippocampal and cortical function.Because this more complete set of principles,implemented in an explicit computational model,accounts for the empirical data,this data provides support for these principles.Rapid Incidental Conjunctive Learning TasksA consideration of the full set of principles suggests that another class of tasks might provide a much better measure of hippocampal learning compared to the non-linear discrimination problems suggested by Sutherland and Rudy(1989).As we just saw,the very fact that non-linear discrimination problems require conjunctive rep-resentations is what drives the cortex alone to be able to solve them via error-driven learning.Therefore,we suggest that incidental conjunctive learning tasks,where conjunctive representations are not forced by specificO’Reilly 7100200300T o t a l E r r o r s100200300T o t a l E r r o r sa)b)c)d)510152025A B − E r r o r s510152025A B − E r r o r sFigure 5:Results for the negative patterning (left column)and ambiguous feature (,,,;right column)prob-lems.The top row shows data from rats from Alvarado and Rudy (1995),and the bottom row shows data from the model.Intact isintact rats/networks,and HL is rats/networks with hippocampal lesions.N=40different random initializations for the model.The hippocampally-lesioned system is able to learn the problems,and all conditions require many trials (i.e.,large number of errors).Negative patterning is differentially impaired with a hippocampal lesion.Data from O’Reilly and Rudy (1999).task demands,may provide a much better index of hip-pocampal function (O’Reilly &Rudy,1999).Further-more,the task should only allow for a relatively brief period of learning,which will emphasize the rapid learn-ing of the hippocampus as compared to the slow learning of the cortex.Thus,we characterize these tasks as rapid,incidental conjunctive learning tasks .There are several recent studies of tasks that fit the rapid,incidental conjunctive characterizations.In these tasks,subjects are exposed to a set of features in a partic-ular configuration,and then the features are rearranged.Subjects are then tested to determine if they can detect the rearrangement.If the test indicates that the rearrange-ment was detected,then one can infer the subject learned a conjunctive representation of the original configura-tion.The literature indicates that the incidental learning of stimulus conjunctions,unlike many nonlinear discrim-ination problems,is dependent on the hippocampus.Perhaps the simplest demonstration comes from the study of the role of the hippocampal formation in ex-ploratory behavior.Control rats and rats with damage to the dorsal hippocampus were repeatedly exposed to a set of objects that were arranged on a circular platform in a fixed configuration relative to a large and distinct visual cue (Save,Poucet,Foreman,&Buhot,1992).Af-ter the exploratory behavior of both sets of rats habitu-ated,the same objects were rearranged into a different configuration.This rearrangement reinstated exploratory behavior in the control rats but not in the rats with dam-age to the hippocampus.In a third phase of the study,a new object was introduced into the mix.This manipula-tion reinstated exploratory behavior in both sets of rats.This pattern of data suggests that both control rats and rats with damage to the hippocampus encode representa-tions of the individual objects and can discriminate them from novel objects.However,only the control rats en-coded the conjunctions necessary to represent the spatial arrangement of the objects,even though this was not in any way a requirement of the task.Several other stud-ies of this general form have found similar results in rats (Honey,Watt,&Good,1998;Honey &Good,1993;Good &Bannerman,1997;Hall &Honey,1990;Honey,Willis,&Hall,1990).In humans,the well established incidental context effects on memory (e.g.,Godden &Baddeley,1975)have been shown to be hippocampal-dependent (Mayes,MacDonald,Donlan,&Pears,1992).Other hippocampal incidental conjunctive learning ef-fects have also been demonstrated in humans (Chun &Phelps,1999).We have shown that the same neural network model constructed according to our principles and tested on the nonlinear discrimination learning problems as described above exhibits a clear hippocampal sensitivity in these rapid incidental conjunctive learning tasks (O’Reilly &Rudy,1999).8Computational Principles of LearningContextual Fear ConditioningEvidence for the involvement of the hippocampal for-mation in the incidental learning of stimulus conjunc-tions has also emerged in the contextual fear condition-ing literature.This example also provides a simple ex-ample of the widely-discussed role of the hippocampus in spatial learning(e.g.,O’Keefe&Nadel,1978;Mc-Naughton&Nadel,1990).Rats with damage to the hippocampal formation do not express fear to a con-text or place where shock occurred,but will express fear to an explicit cue(e.g.,a tone)paired with shock (Kim&Fanselow,1992;Phillips&LeDoux,1994;but see Maren,Aharonov,&Fanselow,1997).Rudy and O’Reilly(1999)recently provided specific evidence that, in intact rats,the context representations are conjunc-tive in nature,which has been widely assumed(e.g., Fanselow,1990;Kiernan&Westbrook,1993;Rudy& Sutherland,1994).For example,we compared the ef-fects of preexposure to the conditioning context with the effects of preexposure to the separate features that made up the context.Only preexposure to the intact context facilitated contextual fear conditioning,suggesting that conjunctive representations across the context features were necessary.We also showed that pattern completion of hippocampal conjunctive representations can lead to generalized fear conditioning.We have simulated the incidental learning of con-junctive context representations in fear conditioning us-ing the same principles as described above(O’Reilly& Rudy,1999).For example,Figure6shows the rat and model data for the separate versus intact context fea-tures experiment from Rudy and O’Reilly(1999),with the model providing a specific prediction regarding the effects of hippocampal lesions,which has yet to be tested empirically.Transitivity and FlexibilityWhereas the previous examples concern the learn-ing of conjunctive representations,this next example is concerned with theflexible use of learned information. Several theorists have described memories encoded by the hippocampus as beingflexible,meaning that(a)such memories can be applied inferentially in novel situations (Eichenbaum,1992;O’Keefe&Nadel,1978)or(b)that they are available to multiple response systems(Squire, 1992).Although the termflexibility provides a useful de-scription of certain behaviors,it does not provide a mech-anistic understanding of how thisflexibility arises from the properties of the hippocampus.We have shown that hippocampal pattern completion plays an important role in producing thisflexible behav-ior(O’Reilly&Rudy,1999).Specifically,we showeda)b)204060PercentFreezing0.000.050.100.150.200.250.30FearResponseFigure6:Effects of exposure to the features separately com-pared to exposure to the entire context on level of fear response in a)rats(data from Rudy and O’Reilly(1999))and the model. The immediate shock condition(Immed)is included as a con-trol condition for the model.Intact rats and the intact model show a significant effect of being exposed to the entire con-text together compared to the features separately,while the hip-pocampally lesioned model exhibits slightly more responding in the separate condition,possibly because of the greater over-all number of training trials in this case.Simulation data from O’Reilly and Rudy(1999).that the transitivity studies of Bunsey and Eichenbaum (1996)and Dusek and Eichenbaum(1997)can be sim-ulated using the same model as in all of the previous examples,with pattern completion playing a key role. Interestingly,the model shows that the training parame-ters employed in these studies interact significantly with the pattern completion mechanism to produce the ob-served transitivity effects.We are able to make a number of novel empirical predictions that are inconsistent with a simple logical-reasoning mechanism by manipulating these factors(O’Reilly&Rudy,1999).Dual-Process Memory ModelsThe dual mechanisms of neocortex and hippocampus provide a naturalfit with dual-process models of recogni-tion memory(Jacoby,Yonelinas,&Jennings,1997;Ag-gleton&Shaw,1996;Aggleton&Brown,1999;Vargha-Khadem,Gadian,Watkins,Connelly,Van Paesschen,&。
