1/62
Looks like no tags are added yet.
Children have to solve problems at the level of
Phonology
Syntx/grammar (dog bites man vs man bites dog)
Morphology- how to deeal with the verb or when you add and dont add ED at the en
segmentation
Infants have to learn words from a continuous stream of speech nogapsemphasiedwhen talking
How to children know where one word end and the next begins
Transitional probabilities between the sound and speech (using the probability from sounds that follow the same sound are probbaly the word
saffron,Aslin & Newport
Head turn technique
Head turn technique
Played a made up languag played for 2 mins then are played different elemetnts from the speech stream and we see which they prefer based on their head turns
Testing
Can the infant extract word from the speech steam,
Saffaran question
Can infants distuingish words likee gilatu and non words e.g layout
Words here
sequences where first syllable e.g go always followed by second syllable, followed by the same syllable and children could tell the difference between them
Non words
Sequences where syllables never followed eachother= transitional probabilites = 0. Infants listened longer to non words
Experiment 2
Can infants disguinuish words and past-words?
Infants where more intresed in part words over whole words
8-month-olds can discriminate 3- syllable sequences with high transitional probabilities.
Sequences they’ve never heard p=0 (non words)
Low transitional probability- 0.33
Brusini et al (2019)
Ran the same experiment as saffran et al, but instead of using head turn (newborns can’t) they used neuroscience technique
Called
FNIRS- look into newborns
WHat did flo to
Played newborns 3 mins of stream syllables, and fond that newborns showed diff patterns of brain response to words and part words
Beyond transitional Probabilities
Speech signal needs to be segmented in diff ways dependning on the languae you learn,
Englihsh structure
Trochaiic (strong emphasis on start of war ompared to end e.g DOCtor
Jusczyk, et al
Played English speaking 7 month old infants passeges contaiing 2 syllables e.g guitar
words presented sperately found
Infants listened longer to the target words, if the words had the stress pattern of their language
Children apply
Appley trochaic bias of their language
Porved
Children are senetive to the stress pattern of thei language
Gerken 1994
English speaking children/s trochaic bias can explain the kind od segmentaion errors we see in childerns early speech.
NA-na baNAna
PU-ter comPUter
gi_RAFFE
diffferent segments transitional probabilities
Infants therefore are sensetive to
Language-general cues from birth
and language specific cues from 7.5 months
Language general vs language specific
General - Universal patterns that work across every human language on Earth—specifically transitional probabilities (how likely one syllable is to follow another)
Specific- Rules unique to the specific language the baby is growing up hearing—such as word stress patterns.
2-Learning word order rules
Children have to learn the word order rules of their language
e.g SVO in english VSO in welsh (verb comes first)
Tomasello’s (1992)
Verb-Island
Tomasello kept a detailed diary record of his daughter’s early multi-word speech. Noticid that similar verbs often behaved very differently in Travis’s speech.
Tomasello’s (1992) Verb-Island Hypothesis
Based on a diary study of his daughter Travis, Tomasello proposed that early verbs act like isolated "islands" rather than following broad grammar rules. Even when Travis used complex sentences with familiar verbs ("Mummy cut bread with knife"), newly learned verbs still started out in simple phrases ("Draw", "Draw car"), showing children learn sentence patterns verb-by-verb.
Akhtar & Tomasello (1997): Method used to test the Verb-Island Hypothesis
Researchers taught young children a novel action verb ("This is wugging") without any word-order cues, then asked them to act out a standard English SVO sentence using puppets ("Make Big Bird wug Cookie Monster").
Akhtar & Tomasello (1997): Results & what
Children performed at chance (50/50)—sometimes making Big Bird do the action to Cookie Monster, and sometimes the reverse.
they show about early word order
This supports the Verb-Island Hypothesis, showing young children lack abstract, verb-general word-order knowledge and rely on verb-specific experience
preferential-looking technique (gerstner, Fisher and Eisengart (2006)
Measure looking time to the correct vs the incorrect alternative (tha weird forg things)
Gertner, Fisher, & Eisengart (2006) & Noble et al. (2011): Preferential-looking evidence on word order
Using a preferential-looking setup (with boys/girls for 21-month-olds and bunnies/ducks for 25-month-olds), Gertner et al. (2006) showed that even 21-month-olds have verb-general SVO knowledge—understanding that the first-mentioned noun is the Agent (doer) of a novel verb. Replicated using forced-choice pointing by Noble, Rowland, & Pine (2011)
Summary of early word-order knowledge: Why did Akhtar & Tomasello (1997) vs. Gertner et al. (2006) disagree?
English-speaking children have verb-general knowledge of SVO word order by 21 months, but whether they can display it depends on task demands. Acting out puppet scenes (Akhtar & Tomasello, 1997) is too demanding and masks their competence, whereas preferential looking (Gertner et al., 2006) and forced-choice pointing (Noble et al., 2011) reveal it.
So by the time thy’re 21 month old
ENglisj speaking children have verb general knowledge of SVO ford order by 21 months
challenge 3- learning how to mark verbs
Toddlers need to learn how to maek verbs for
Agreement (e.g she play-s)
Tense (e.g, she play-ed yesterday)
Brown 1973
Studied early multi-word speech in 3 children (Adam, Eve, and Sarah). Found that early speech is "telegraphic" (including only the most important content words, like a telegram) and that grammatical verb-marking is often absent (e.g., "Adam want it", "We go to the park yesterday").
Brown (1973): Gradual acquisition of verb marking (Adam example)
Correct verb marking develops gradually rather than all at once. For example, Adam began using third-person singular "-s" (3sg -s) around age 2;6, but did not reach 90% accuracy in obligatory contexts until a full year later at age 3;6.
Developmental Language Disorder (DLD): Definition & prevalence (Norbury et al., 2016
Previously called Specific Language Impairment (SLI), DLD is a language deficit (1.5+ SDs below the mean) in the absence of any known sensory or cognitive deficits. Norbury et al. (2016) found a UK prevalence of ~7.5% (8% boys, 6% girls), equivalent to about 2 children in every classroom.
Verb-marking in DLD (Rice, Wexler, & Hershberger, 1998)
Although children with DLD can show deficits across many language areas (e.g., word learning, syntax)
English-speaking children with DLD have a particular difficulty with verb marking. Rice, Wexler, & Hershberger (1998) found significant deficits in producing third-person singular "-s" (3sg -s) and past tense "-ed" compared to both age-matched and language-matched controls.
Bloom 1990 claim
(Processing-Capacity Limit): Toddlers drop small grammatical endings (like -s or -ed in English) simply because their young brains do not have enough processing capacity to say every word and add all the grammatical endings at once.
The Problem / Counter-Evidence From Other Languages:
If toddlers only dropped endings to save effort, they wouldn't go out of their way to add an extra ending onto a verb. Yet in other European languages, toddlers do add an ending—they just attach the wrong (infinitive / "to do") ending instead of the tense-marked ending:
German: Kaffe trink-en ("Coffee to-drink")
French: Poupée dorm-ir ("Dolly to-sleep")
Spanish: Niños cant-ar ("Children to-sing")
The MOSAIC Computer Model:
esearchers built a computer simulation called MOSAIC to test whether toddlers' verb errors across English, Dutch, German, French, and Spanish can be explained purely by how children hear and remember real adult speech.
Learns from real Child-Directed Speech
MOSAIC was fed real transcripts of parents talking to young children—no innate grammar rules were programmed into it.
Learns from the "right edge" (Recency Effect)
Because young children remember the most recent words at the end of a sentence best (recency effect), MOSAIC gradually builds sentences backward from the right edge (the end of the utterance):
Adult says: "Mummy likes to drink coffee."
Child/MOSAIC first remembers just the end: "Coffee"
Next, it remembers a bit more from the right edge: "Drink coffee" (notice how "drink" is an unmarked infinitive form because it came after "likes to" at the end of the adult's sentence!).
Freudenthal, Pine, & Gobet (2010): MOSAIC model & the recency effect
Used a computer model (MOSAIC) trained on transcripts of child-directed speech to simulate verb errors in English, Dutch, German, French, and Spanish. Because of the recency effect, MOSAIC gradually learns sentences from the right edge (the end of utterances—e.g., "Mummy likes to drink coffee" → "Coffee" → "Drink coffee"), explaining why children produce unmarked/infinitive verb forms from the input they hear.
Freudenthal, Pine, & Gobet (2010): MOSAIC results & limitation across languages
MOSAIC successfully simulated the cross-linguistic pattern of infinitive verb errors (Dutch > German > French > Spanish) because infinitives consistently appear at the right edge (end) of Dutch and German utterances, benefiting from the recency effect. However, MOSAIC could not explain why English children produce an even higher rate of errors (> Dutch).
Räsänen, Ambridge, & Pine (2014): Input frequency & English verb-marking errors
Proposed that English-speaking 3-year-olds replace less familiar verb forms with forms they hear frequently in child-directed speech (CDS). For example, children make many errors on "see vs. sees" because bare "see" is much more frequent in CDS, whereas they make few errors on "fit vs. fits" because "fits" is relatively frequent in CDS.
Räsänen, Ambridge, & Pine (2014): Elicitation method & scatter-plot findings
Researchers used picture prompts to elicit 3sg -s verbs (e.g., "Kate gives a card and Sam [gives a present]"). Verbs with a higher proportion of -s forms in child-directed speech (e.g., fit/fits, need) were produced with significantly higher -s accuracy than verbs where bare forms dominate input (e.g., see, find).
Räsänen, Ambridge, & Pine (2014): Two causes of high verb-marking error rates in English
The more children hear the -s vs. bare form in input, the more accurately they produce it. High error rates in English result from two combined mechanisms: 1) learning from the right edge of utterances (recency effect), and 2) defaulting to the most frequent (usually bare) form of a verb in input.
Kueser, Leonard, & Deevy (2018): Purpose & method
Investigated why children with DLD make verb-marking errors by re-running Räsänen et al.’s (2014) elicitation task with English-speaking children with DLD and a younger language-matched typically developing (TD) control group.
Kueser, Leonard, & Deevy (2018): Key results & what they show about DLD
Children with DLD made significantly more 3sg -s errors than younger language-matched TD controls, yet both groups showed the same upward frequency slope—getting -s forms correct more often when -s forms were more frequent in input. This shows children with DLD are sensitive to input frequency in the same way as TD children.
Chellange 4- learning how to generalize appropriately
Children must eventually go beyond their input to learn more general patterns. E.g
past=verb+ed
Kick-kicked
1st type of generalisation= Past tense over-generalisations
Children learn a “rule” (add -ed to make something past tense)
Errors
Rule gets over generalised to irregular verbs e.g Santa bringed me
Over generalisations
Some irregulars common enough for children to re-learn the correcr form e.g bring-brought they hear it a lot which nakes it plausible to asssume you replace which stops them saying bringer
Uncomon irregulars e.g bleed=bled NOt bleEEed
Blything, Ambridge & Lieven (2018)
elicited past tense forms of novel verbs Showed them the duckk or cartoon and said Look, hes verbING,
Children more likely to produce irregulars . Scread vs screed
Blything, Ambridge, & Lieven (2018) elicited past-tense forms of novel verbs (e.g., screed, blick) from 3–4, 5–7, and 9–10-year-olds. Children were significantly more likely to produce irregular forms (e.g., screed → scread rather than screeded) when the made-up verb sounded similar to real, existing irregular verbs (e.g., read → read), demonstrating phonological analogy.
Blything, Ambridge, & Lieven (2018): How do children learn irregular past-tense verbs?
Children learn irregular past-tense forms (such as "bled") by building "families" of similar-sounding irregulars in memory that support one another through phonological analogy (e.g., knowing "read → read" helps support "bleed → bled").
Bowerman 1982
Reported similar over-generalisation errors with sentence structures e.g the stick broke can be done but you cant put I giggle to Dont giggle me
Blything, Ambridge, & Lieven (2018): How do children learn irregular past-tense verbs?
Children learn irregular past-tense forms (such as "bled") by building "families" of similar-sounding irregulars in memory that support one another through phonological analogy (e.g., knowing "read → read" helps support "bleed → bled").
they taught 5-6, 9-10 year olds and adults novel terms for laughing disppearing in word form
Then they rated sentences like the funnt man giggle bart. and Bart giggled
FUNNY MAN LAUGHED HOMER vs homer laughed
worse, lauhj= high frequency
So 5-6 year olds know somethings wrng with the high frequency and THE FUNNY MAN TAMMED LISA, compared to LISA TAMMED
They’ve built a semantic class of expression verbs and knew you cant “expression” someone