Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Raising the Standards Presentation to the American Speech Hearing Language Association November 14, 2003 Jill de Villiers Smith College, Northampton, MA Thomas Roeper, Barbara Pearson & Harry N. Seymour University of Massachusetts Amherst, MA Research supported by NIH contract N01-DC-8-2104 Motivations For A New Standardized Language Assessment • How best to determine what a child knows about language. • How best to determine this knowledge irrespective of the child’s dialect. Traditional Language Assessment • Standardized testing • Non-standardized Approaches – Language sampling – Language probes Limitations of Traditional Assessment • Standardized testing – – – – Limited to problem/no problem question Limited to gross measures of language Limited in explanatory power Limited to Standard English targets • Non-standardized approaches – Effective but Impractical for most clinicians Raising The Standards Beyond Traditional Testing – Six standards for assessment • • • • • • Assess Language knowledge at a deeper level Select dialect specific targets for identification Avoid dialect specific targets for assessment Assess morphology that is obligatory across dialects Avoid focusing on acquired lexical vocabulary Focus on pragmatic aspects of language essential for schooling and literacy Standard 1: Deeper Level of Language Knowledge • Universal grammar • Deeper, more abstract language knowledge – Hidden properties – Variable properties – Movement rules • Disorder is most obvious Standard 2: Select Dialect Specific Targets For Identification • Useful in identifying dialect status – African American English (AAE) vs Mainstream American English (MAE) • No penalty for non-Mainstream English patterns by African American English speakers Standard 3: Avoid Dialect Specific Targets For Assessment • Non-contrastive language patterns • Less superficial • No distinction expected between dialects Standard 4: Obligatory Morphology Across Dialects • There are morphological inflections that do not vary between dialects • Past tense “was”, possessive pronouns, presentational “it.” • Their obligatory status applies to both AAEand MAE. Standard 5: Avoid Focusing On Acquired Lexical Vocabulary • Basic processes of word learning • Lexical organization and retrieval • Logical properties and scope of the word “every” Standard 6: Focus On Pragmatic Aspects of Language Essential For Schooling and Literacy • • • • Question asking Communicative role taking Linking events into a cohesive narrative Understanding mental states The DELV Tests • DELV-Screening Test – Identifies language variation status – Identifies students at risk for a disorder • DELV-Criterion Referenced Test – Diagnose speech and language disorders • Syntax, Semantics, Pragmatic, Phonology • DELV-Norm-referenced Version – Exclusively on AA children Jill de Villiers Two illustrations: Learning a new word from context Understanding complex questions How can we do better with assessing semantics? • We need to understand the process of how a child learns words in context and from context. • Research has shown that children learn words quickly from spoken, and later written, context. • Children rarely have meanings pointed out or explained. Instead, from a young age they can “fastmap” the meaning of a word from its linguistic context. • A child may have an impoverished, or different, vocabulary, but still be capable of quick learning of a new word, when given the chance. Fast Mapping of Word Meanings from Context • Verbs are dependent on using syntactic cues to meaning type: i.e. argument structure. • This should not differ across different varieties of English. It is a fundamental process of learning. • Previous work on young children distinguishing intransitive from transitive verbs (Naigles, Fisher) E.g. She is mooping (intransitive: self-action) She is mooping her (transitive: causal action) Children choose different actions from a scene, depending on the structure. They “fast-mapped” the meaning from the sentence frame. Syntactic Bootstrapping • Johnson (1999) compared intransitive, transitive, also dative and complement-taking verbs. • Nonsense verbs were used in these frames to describe strange actions in ambiguous contexts. The child then answers questions about the verb and its subjects and/or objects. • This research showed that children acquire a verb’s meaning in part through the argument frames in which it appears. This phenomenon is often called syntactic bootstrapping. Extension of Johnson’s research • Our testing was modeled closely on Johnson’s design. • Our subjects were the children tested in the try-out phase of the DELV assessment test. • They included AAE and MAE speakers, typically developing and those receiving speech-language services. • In all, there were 1014 subjects aged 4 to 9 years. Argument structures: real verbs • Intransitive: one argument E.g. the dog is barking • Transitive: two arguments E.g. The boy poured the drink • Dative: three arguments E.g. The mailman handed the letter to the boy • Complement: three arguments E.g. The policeman asked the woman to stop the car Procedure • The child saw a picture that contained at least two events. S/he heard a sentence about it containing either a REAL or a NOVEL verb. • The child had to answer a set of questions about the picture that are designed to test which action s/he has associated with the verb. Now for a novel verb: transitive The woman is temming the boy. • Which one was the temmer? • Which one was temming? • Which one was got temmed? • Which one was temmable? Intransitive/Transitive Copyright 2000 The Psychological Corporation Now for novel verbs! Try a dative: The boy is meeping the flowers to the girl. Here are the things in the pictures. I want you to show me: Which one was the meeper? Which one was meepable? Which one was wearing a green dress? ©. The Psychological Corporation The boy is meeping the flowers to the girl. Which one got meeped? Which one was meeping? ©. The Psychological Corporation Now try a complement form: The lady is ganning the waiter to send the coffee. ©. The Psychological Corporation Here are the things in the pictures. I want you to show me: Which one was the ganner? Which one did the lady gan the waiter to send? Which one was wearing a dress? ©. The Psychological Corporation The lady is ganning the waiter to send the coffee. Which one did the lady gan to send the coffee? Which one was ganning? ©. The Psychological Corporation Question types • • • • • • • • ING e.g Which one is ganning? (agent) ER e.g. Which one is the ganner? (agent) Got-ED e.g. Which one got ganned? (patient) ABLE e.g. Which one is gannable? (patient) Subj-comp e.g. Which one did the woman gan (e) to send the coffee? Obj-comp e.g. Which one did the woman gan the waiter to send (e)? Overall pattern by age: MAE and AAE speaking children. Syntactic Bootstrapping / Fast Mapping Average Score /21 20 15 AAE MAE 10 5 0 4 5 6 7 Age 8 9 Overall Pattern by age on Fast-Mapping task: TRY versus DIS subjects Syntactic Bootstrapping / Fast Mapping Average Score /21 20 15 IMPAIRED TYPICAL 10 5 0 4 5 6 7 Age 8 9 Error patterns reveal where problems are. • The disordered children get only about half of the questions correct that the normal AAE-speakers do. • The patterns can be compared – across real versus novel verbs, – across different question-forms – and across different structures, to locate what the child might be having difficulty with. • Some errors are developmental: some items are hard across the age range. More work ahead... • On the DELV, we assess – fast mapping of verbs from context, – the organization and retrieval of verb and preposition vocabulary – The understanding of the quantifier “every”. This is a TINY part of what semantics means, though at least it gets us beyond naming objects. We need to expand our vision! Example 2: Understanding complex wh-questions Why are wh-questions of significance? For the past thirty years in Linguistics, considerable attention has been paid to What? wh-questions. By who? Chomsky, among others. How? By examining data from many languages Why? Because wh-questions reveal the deep processes of grammar of which we as speakers are only dimly aware. We know that in languages like English, the wh-word moves to the front of the sentence: What did you eat? (you ate what?) Long distance movement • But the “site” from which it moves, the “gap”, can be several clauses away: • What did you say you ate (...) ? • What did Jim say he saw you eat (…)? We can “recover” the meaning across several clauses.But there are some places the wh-word can’t come from: John said he saw a man who ate a snake John said he saw a man who ate what? (no problem) What did John say he saw a man who ate? (oops!) What’s this about? • There are significant limitations on how we can move wh-questions, that do not have to do with limits on what we seek answers to! • That is, they are not semantic/pragmatic limits, but SYNTACTIC. We don’t know that we know them, but we act in accordance with these deep principles that govern grammars of all languages. • If we don’t even know these principles, how can we teach our children? • Fortunately, typically developing children know them too. Wh-Question Comprehension: Testing Procedure • If children don’t say the wrong things, how can we test if they know the principles? These sentences are fairly rare!So we use comprehension: – The child is told a brief story about a pictured event. – They are then asked the key test question about some aspect of the event. – The pictured events and stories support several possible interpretations of the question. First question: • Can children get long distance movement of Whquestions? • Can they retrieve the place where it came from, if it is two clauses away? • Children’s ability to give LD answers (without embedded false clause) was tested in piloting and then in the DELV Tryout testing. • 90% of the children ages 4-6 and 95% of the children 7-10 gave at least one Long Distance answer, so simple Long Distance items do not appear on the DELV. But we did find one item discriminating. This mother snuck out one night when her little girl was asleep and bought a surprise birthday cake. The next day the little girl saw the bag from the store and asked, “What did you buy?” The mom wanted to keep the surprise until later so she said, “ Just some paper towels.” -- What did the mom say she bought? Copyrighted picture omitted. Copyrighted picture omitted. ©. The Psychological Corporation Typical Answers to “False Clause” questions • LONG DISTANCE (LD) TWO CLAUSE responses – Ex. She said she bought paper towels. • What can children get wrong? There isn’t really an answer just to “what did she say (…)” • What they do is fail to take both verbs into account: • ONE CLAUSE responses (Incorrect) – Ex. (She bought) a birthday cake. • OTHER – “a surprise” “a bag” “I don’t know.” LD False Clause Response Types by Age and Language Status Long Distance Movement Complement with False Clause Average Correct/ of 1 1 0.8 0.6 Impaired Typical 0.4 0.2 0 4 5 6 7 Age 8 9 Do children know Barriers to Long Distance Movement? • Once we know children can do long distance movement, we can ask: do they show the same limits as adults as to where a wh-word can come from inside a sentence? • Remember that you cannot get a wh out from inside a relative clause: • Who did John say he saw a man who ate (..)? • How can we ask if children allow that? • Try this: These two boys went to the circus. A clown tickled the little boy on the nose with a feather. He sneezed so hard he blew the clown's wig off! After the circus, they were very thirsty and went to buy some milk. The little boy drank his milk through a straw, but the big boy drank his milk straight from the carton. How did the boy who sneezed drink the milk? Copyrighted picture omitted ©.The Psychological Corporation WH Barrier Response Types by Age and Language Status Comprehension of WH Barriers Average Correct/ of 5 5 4 3 Impaired Typical 2 1 0 4 5 6 7 Age 8 9 WH Barrier Responses by Age and Dialect Comprehension of WH Barriers Average correct/ of 5 5 4 3 AAE MAE 2 1 0 4 5 6 7 Age 8 9 Raising the standard • Children’s subtle knowledge of the conditions on wh-movement has been explored in many research studies, and we tap only a portion of that research to use on the DELV. • Questions prove to be a refined way to see what the child’s grammar contains or lacks, and children with disorders may reveal even more possibilities. Inference and guessing can’t get you far enough without grammar! Raising the Standard (con’t) Barbara Zurer Pearson Example 3. Restrictions on Articles Example 4. Understanding complex Passives Credits • Articles – Maratsos (1976), Karmiloff-Smith (1979) – Robin Schafer & de Villiers (BUCLD 24) – Robin Schafer & Roeper (BUCLD 24) • Passives – Bever (1972), Maratsos (1975) – Roeper (1987), Roeper & Pearson (2000, 2003) •Seminars in Speech and Language 2/2004 (it’s all there) ARTICLES: “a” & “the” • 2 of the smallest, most common words in the language And The TRICKIEST Basic Contrast • “a” • “the” INDEFINITE DEFINITE Read: un-DEFINed and DEFINed “a” = one of something, “a house” “the” = a particular one, “the house” ( or a particular set of things) “the houses” Examples “Polly want a cracker” not * “Polly want the cracker” • Strict conditions on when you have the right to use “THE.” Parrots don’t usually know them, but children do. Conditions for using “THE” • Not just a specific “whatever” • Must be one that the hearer knows about. Speaker has to know what the hearer knows (Can’t just use “the.” Have to calculate the hearer’s knowledge) How do you know that the hearer knows what you think is known? Can use “the” if: 1. It was mentioned in a previous sentence. 2. You define it in the present sentence. 3. It’s “common knowledge,” of the type your culture allows you to assume. 4. It’s conventionalized in the language (eg. “The Phillipines” “in the hospital” for any old hospital) 3 & 4 relevant for testing 2nd language learners. We’re focused on 1st language (or 1st dialect) acquisition, so we’ll look only at 1 and 2 (and 3 when it’s unavoidable). Mentioned in a previous sentence: Example: S1: I saw a play last night. S2: THE play was called “Annie.” Mentioned (and identified) in the same sentence: Ex. Did I ever tell you about the boy in the picture… (prepositional phrase) Ex. Did I ever tell you about --the boy who was just here? (relative clause) So, the basic division: “a” = “any X” “the” = “the specific thing” really “the specific thing that you and I both know about.” Not quite. We can use “A” for a specific thing: Compare: “I’m looking for a house.” (any house that fits my budget) vs “What are you doing here (in our gated community)?" “I’m looking for a house” (It’s number 84. We have friends on this street.) . That is: It’s a specific house. I know which house, but you don’t, so I can’t use “the.” The DELV: How test previous mention? No pictures. Give a little story. Ask a question. The answer should include an article. . Ex. “A bird and a snake were sitting on a rock. They were friends. One of them flew away. Which one?” The bird / Not a bird Also, DELV tests previous mention + common knowledge. If you mention an object, parts of the object are counted as mentioned as well. Ex. I have a cage. The door is broken. You don’t have to know the relationship: the language helps you. “I have a cage. The brinch is broken.” You know that the brinch, whatever it is, is part of the cage. . . DELV Article Items: 2 each of 2 types of Indefinites • “a” -- real indefinite: Polly wants a cracker. • “a” -- child knows it, but you don’t. Eg. What do you have hanging on the wall in your room? (“a Rolling Stones poster”) DELV Article Items (con’t): 2 each of 2 types of Definites • “the” -- previous mention – (the bird and snake story, above) • “the” -- part of a previous mention – Ex. Sally wanted to eat an ice cream cone. But something fell out and went splash. What was it? The ice cream Item Goals • Dialect Neutral • Discriminating between Typically Developing (TD) and Language Impaired (LI) • All items dialect neutral • Definite articles most discriminating Dialect-neutral Scores by Article Type (by dialect) percent correct 1 0.8 0.6 0.4 0.2 0 4 5 6 7 8 9 10 age years AAE "A" score MAE "A" score AAE "the" score MAE "the" score Diagnostic picture "A" Article Score by Clinical Status Percent correct 1 0.8 0.6 0.4 0.2 0 4 5 6 7 8 9 Age Years Impaired Typical Most diagnostic: “THE” "The" Article Score by Clinical Status Percent correct 1 0.8 0.6 0.4 0.2 0 4 5 6 7 8 Age Years Impaired Typical 9 Example 4 (switching item types) • Understanding Implicit (Unstated) Information in Complex Passives Passive Item Types • Simple (The elephant was pushed.) • Complex (The bear was being washed) • “locative by-phrase” (non-passives) (The ball was rolling by the boy.) It’s not so much the sentence that differs in the items; it’s the alternatives the child has to choose from. Dialect Neutral Passive Comprehension 12 10 8 AAE MAE 6 4 2 0 4 5 6 7 8 9 Simple Passive -Child chooses between active and passive The elephant was pushed. Copyrighted picture omitted Copyright 2000 The Psychological Corporation Simple passives (con’t) • Not the most revealing items, but give a baseline for the child on this task. Basic Passives (Moderate Discrimination) Basic Passive Scores by Clinical Status Average Score / 5 5 4 3 2 1 4 5 6 7 8 Age Impaired Typical 9 • Basic Passive is ambiguous: – Participle form could be like an adjective – “The bear is washed”/ “The bear is ready.” Show true understanding with a fuller form: “The bear is being washed.” Not a result, it is still ongoing And without saying it, the sentence also tells us that the bear is not the one doing the washing Complex Passive Child chooses between 2 passives The bear is being washed Copyright 2000 The Psychological Corporation Complex Passives (More Discrimination) Complex Passive Scores by Clinical Status Average Score / 5 5 4 3 2 1 4 5 6 7 8 Age Impaired Typical 9 Distinguishing unstated information No new content words: it’s all in the grammatical bits. • The player dropped. – Who knows how? He could be doing it, someone else could, it could be the hand of fate. • The player was dropped. – Someone did it. Who knows when? • The player was being dropped. – Someone else is doing something to him, and it’s happening before our eyes. Item type 3: Distinguishing hidden agents • The plant was dropping by the boy. • Notice: this boy is NOT the dropper. • How do you know? – Was dropPING not was dropPED. • Can only be an agent-by-phrase (a true argument of the verb) when it is a passive construction with all the “machinery” of the passive. • Otherwise, the by-phrase is what is known as an adjunct. Item Type 3: Learning about By-phrases • John must be taken by the lake by car by someone by 5 pm. 4 “by’s”--3 of them “adverbs” – John must be taken by car • Is the car taking John? – John must be taken by the lake. • Is the lake taking John? – John must be taken by 5 pm (no) Who is taking John??? • Here, “someone.” • This is not that obvious. Locative Non-Passives Passive Morphology (Locative by-phrases) Average Correct/ of 2 2 1.6 1.2 dis try 0.8 0.4 0 4 5 6 7 Age 8 9 DELV tells you the child gets the hidden information • It’s a pretty impressive achievement. DELV tells you the child doesn’t get the hidden information • What are you going to do about it? Intervention Concepts • Avoid the ambiguities that the test exploits, and • provide natural and conversational ways to help the children • fill the gaps in their skills. PASSIVE Example • Why do children interpret passives as active? • Hypothesis: They don’t recognize hidden agents. • Goal: Help them recognize the agent. Express it when it’s not expressed. Lexical Support • Instead of using the ambiguity of the test question (which uses reversibles on purpose), use “unreversible” verbs (with fixed object relationships). Ex. • The baby ate the banana • The banana was eaten. • Can the banana eat the baby? (No) Contextual Support • Rephrase the sentence with an impossible agent. “The plant was dropping by the wall.” (Did the wall drop the plant???) • Put back into the question form: – Was the plant dropping by the wall or by the boy? Conversational support Make the implicit agent explicit and then gradually make the connection to the implicit version. 1. Somebody pushed the elephant. Who did it? 2. The plant was dropped by somebody. Who did it? 3. The cat was being dressed (-). Who did it? Intervention suggestion for Articles -1 • Help child make the discourse connection in ellipsis. Ex. Show three hats (all green) and a red shirt Ask: See these hats? Is one red? Child points to red shirt. For this child, “one” is “one anything,” but it must be “one (of them).” Must learn to tie “one” to the preceding discourse. Intervention suggestions for Articles - 2 • Help child recognize the “specificity requirement” for “THE”. Ex. Use two hats and make the child choose. Johnny got a new hat. Billy took the hat. Freddy took another hat. Show me the hat Billy took. Conclusion We have identified new areas of exploration in language disorders. Now we are exploring intervention concepts, so the DELV can move seamlessly from diagnosis to intervention. Tom Roeper’s turn…. • The linguistic foundations of the DELV • Quantifier and • Double-wh items. What is the impact of dialect on complex syntax? Approach: • Avoid dialect-rich domains of grammar • Focus on what is constant in UG across dialects Deep Principles of Grammar do not vary across grammars or dialects Therefore 1. Their acquisition path should be very close 2. Disorders are common across dialects DELV research has proven this to be true Variables and UG Which of these are variables? • • • “All” “Everybody” “Who” • all the children are here all = plural • every child is here every = variable • who is here who = hidden variable (wh+every) Variables and Quantification Universal Grammar • Pragmatic variables: a. Put your finger here [on nose] Child: puts finger on own nose • Part-whole Variable b. I have that book at my house Variable Type Grammatical Variables 1. Quantifiers: Every boy has a hat =>set of boys and hats 2. Adverbial variables: John always eats ice cream =>set of situations Universal Grammar • 1. All grammars can indicate variability through terms called variables. • 2. Slight variation in how much is linked to adverbs and how much linked to nouns. Test Question: Is every girl riding a bike? (yes!) Copyright 2000 The Psychological Corp. Errors occur for children in six languages Ages: 3-12 • Is every girl riding a bike • Typical mistake: “no not this bike” Hypothesis: child moves every as if it were Like all: all the children sing => the children all sing • Is every girl riding a bike ==> • Is a girl riding every bike ==> – no, there is an extra bike Control No: Is every woman sailing a boat? Copyrighted picture omitted 4 women, 3 on boats, 1 on the beach Copyright 2000 The Psychological Corp. Deleted “Every” • Is every woman sailing a boat => • Is woman sailing a boat => “yes” Control Yes: Does every dog have a bone? “Bunny Spreading” Copyrighted picture omitted 3 dogs eating a bone, 1 rabbit eating a carrot Copyright 2000 The Psychological Corp. Event Every => “always” • Does every dog have a bone => • Is it always the case that a dog has a bone? • “No, a bunny has a carrot” Non-spreaders Non-spreaders by dialect (TD only) 1 Percent of children 0.8 0.6 AAE-TD MAE-TD 0.4 0.2 0 4 5 6 7 8 Age 9 10 11 12 Classic spreaders Classic Spreaders Only by Dialect 1 Percent of children 0.8 0.6 AAE-TD MAE -TD 0.4 0.2 0 4 5 6 7-8 9-10 11-12 Event Spreaders Control-yes "spre ading" plus test-q spreading by dialect 1 Percent of children 0.8 0.6 AAE-TD MAE-TD 0.4 0.2 0 4 5 6 7 8 Age 9 10 11 12 Wh-singletons • Who is wearing a sweater? 5/6 girls Pearson, Penner, Roeper, Schulz (2002) Singleton answers by age Among the questions we elicited were double wh-questions such as: “Who is eating what?” Or “Which person is eating which food?” Decline in singletons, subtest 2wh 1-6 Percentage singletons 30 25 disaae 20 dissae 15 tryaae 10 trysae 5 0 4 4.6 5 5.6 6 6.6 7-8 9- 1110 12 Ages “try” = Typically developing; “dis” = Disordered Wh +Control “no” Spreading Wh-exhaustivity errors: 0 1 2+ 11.2 23.3 22.4 3.9 12.9 17.6 Quantifier errors: 0 1 2 84.9 63.9 59.8 - 35% children who show one Q error, have 1 or 2 wherrors - 40% children who show two Q, errors, have 1 or 2 wherrors Theory: • Children fail with both overt every and covert every =wh [+every] Deep Principles of Grammar can be disordered and lie beyond dialect • Dialect variations still exist and may cause slight differences in the learning path Dialect affects all domains in small ways • you-system => specific/non-specific • Ask 4yr old: Can you drive from NY to Chicago? “yes” • Note: answering as non-specific “Can yawl drive from NY to Chicago” “no, I can’t drive” for 4yr old Consequence: AAE and Southern make a distinction not found in SAE Consequence: AAE and Southern should learn you-system more easily UG Features • Predict common acquisition • DELV shows that children of both dialects understand “who bought what” Conclusions 1. Dialect effect can be minimized if we raise the standard 2. A new range of important deficits-which may affect school language--can be addressed