Language 1 Flashcards
1
Q
Language
A
- Uniquely human communication system- does not mean we are the only species that are able to communicate but how we communicate is different
- Main difference is the expressive capacity of language e.g. talking about the past present and future
- Other species communicate, but this is not remotely comparable to the expressive capacities of human language
- Non-human primates: message confined to here and now
- Human language: past, future, possibility
- Uniquely human communication system
- There are many definitions, but essentially:
finite set of elements + combinatorial rules allowing us to create an infinite number of utterances
2
Q
Language building blocks 1: words
A
- Representations stored in the mental lexicon (ML)- Long term memory representation (not born knowing words we need to learn it)
- ML: mental ‘catalogue’ of words, like a mental dictionary
- The information about a word in the ML:
- Spelling
- Pronunciation
- Meaning
- Grammatical category
Average person has about 50000 words in their ML
3
Q
language building blocks 2: phonemes
A
- The sound units of language
- Allow discrimination between words: /r/, /s/, and /m/ are distinct phonemes as they allow the differentiation between rat, sat, and mat
- Depending on how you combine them it creates different meanings e.g. rat and cat
- Lots of evidence for phonemes- particularly in speech errors e.g. spoonerisms
- Over 100 phonemes across the world languages, 40 or so in English- not born with them but they develop as we grow and learn
- Infants can distinguish between most phonemes but then tune to the native language ones by the age of one (Kuhl et al, 1992)- specialised in the origin language
- Some phonemes are more equal than others
- A book for geeks
- A geek for books
- The ‘s’ is not just a phoneme but a morpheme- it carries some meaning
- ‘s’ carry meaning about plurals or number of things
Morphological stranding
4
Q
spoonerisms
A
- Exchange of sounds, pointing to the existence of phonemic units
not shuffling round entire words but just particularly phonemes- shows that there are blocks within words that can be moved
5
Q
Language building blocks 3: Morphemes
A
- The smallest units in the language that carry meaning
- Words can be morphologically simple and complex
- Complex words contain more than one morpheme
- dog + s
- build + er
- think + ing
- brave + ly
- sad + ness
- English language have short words with simple morphology
- Morphological overlap affects word identification
Morphological structure is cognitively real- encoded in our mental lexicon
6
Q
language building blocks 4: syllables
A
- Rhythmic unit of language
- One vowel, with or without surrounding consonants
- Outrageous
- Out-ra-geous
- Evidence from expletive infixation rule- inserting an expletive into a word
- Outrageous
- Out-bloody-rageous
“The insertion of expletive is only possible in words with multiple syllables, where the word has the main stress preceded by a secondary stress and preferably an unstressed syllable” - McCarthy (1982)
7
Q
language building blocks 5: stress
A
Relative emphasis to certain syllables
Can alter the meaning of a word
record record
content content
console console
object object
Some patients can correctly produce the individual phonemes but stress the wrong syllable (e.g., CV, Cappa et al, 1997)
8
Q
language processing
A
- Progressively complex
- Multiple levels of analysis
- Bottom-up and top-down
sometimes may miss the intial part of a sentence but due to the context you can have an idea of what they are saying (bottom up)
9
Q
neurobiological architecture
A
- Paul Broca (1824-1880)
- patient Tan, lesion in the left inferior frontal lobe
- impaired production, relatively intact comprehension
- Karl Wernicke (1848-1905)
- lesion in the left posterior temporal lobe
fluent but disordered production, impaired comprehension
- lesion in the left posterior temporal lobe
10
Q
the dual pathway model of language processing
A
- Commonly activated regions: left inferior frontal gyrus (Broca’s area), superior, middle and inferior temporal gyri in both hemispheres
- Also major role of the white matter tracts, especially arcuate fasciculus (dorsal stream) and the extreme capsule (ventral stream)
Partially consistent with the old neuropsychological findings, but way more complex and not strictly left-lateralised
11
Q
the language network: current views
A
- An extensive set of interconnected regions. Left-hemisphere dominant, but encompassing both hemispheres.
- The involvement of the RH depends on task difficulty and the type of stimuli involved.
- Different parts of the network engaged depending on the task (comprehension / production) and the input modality (written / spoken language).
Brain regions do not act in isolation: interactions are the rule.
12
Q
visual word recognition
A
- Properties of written language
- Components of the visual word recognition process
- Models of reading (Dual-route and Triangle)
Neural basis of written word processing
13
Q
written language
A
- Recent cultural invention
- The earliest known system: pictographs (Mesopotamia 4000 BC)
- Alphabetic scripts emerged even later (Greece 1000 BC)
Evolutionary newcomer, yet we are expert readers with fixed brain circuitry attuned to reading
14
Q
writing systems
A
- Logographic: unique symbol for each word / morpheme (Chinese)
- Syllabic: unique symbol for each syllable (hiragana, katakana)
- Alphabetic: unique (ish) symbol for each phoneme (English, Russian, etc.)
Diverse, but sharing multiple visual features: limited number of recurring shapes, contrasting contours, an average of three strokes per character (Dehaene, 2009)
15
Q
encoding sound meaning mapping
A
- An interplay between print, sound and meaning
- The same quest, regardless of the script used – divergence is deceptive
Local differences due to specificities of individual orthographies
16
Q
illustration: the role of regularity
A
- In alphabetic languages, lots of variation in the amount of correspondence between phonemes and letters:
- English, Hebrew : letters or groups of letters represent different sounds in different contexts (deep orthography) e.g. ou in cough, through, dough and four
Finnish, Spanish: consistent correspondence between letters and phonemes (shallow orthography)
17
Q
the role of regularity
A
- Balance between an accurate representation of sound and the fast transmission of meaning
- English: 26 letters, 44 phonemes. Many short words that could not be distinguished if written phonetically (maid-made, muscles-mussels, eye-I) hence encoded through complex spelling. Once learnt, it is efficient in getting straight to the meaning
Finnish: long words, rich morphology. Can afford to go 1-on-1 on sound representation
18
Q
basic findings on reading
A
- Real words are read faster than nonwords
- Regular words (mint) are read faster than irregular words (pint)
- Frequent words (rage) are read faster than infrequent words (ire)
Regularity x Frequency interaction
19
Q
what are the components of reading process?
A
- Extracting the information from visual input (sampling, eye movements)
- Letter recognition
- Access to the orthographic lexicon
- and/or
- Grapheme to phoneme conversion
Retrieval of word meaning
20
Q
eye movements in reading
A
- Extracting the information from visual input (sampling, eye movements)
- Letter recognition
- Access to the orthographic lexicon
- and/or
- Grapheme to phoneme conversion
Retrieval of word meaning
21
Q
letter recognition
A
- Two stages:
- recognition of the letter’s visual characteristics
recognition of the letter’s identity (e.g., a = A)
- recognition of the letter’s visual characteristics
22
Q
letter recognition in the brain
A
- Alexic patients: lesions in the L posterior regions close to the fusiform gyrus
- Neuroimaging: letter processing activates the L fusiform gyrus, especially the Visual Word Form Area (VWFA)
VWFA extracts the identity of the letter string, regardless of size, shape and position
23
Q
measuring letter recognition
A
- Visual priming experiment
- Classification task on target words
Primes masked and briefly presented
24
Q
the orthographic lexicon
A
- The orthographic lexicon stores representations of spelling for thousands of familiar words
- They are activated when we read a known, familiar word
This is then followed by obtaining the meaning of the word from the semantic system
25
grapheme to phoneme conversion
- We can also read new words or pseudowords (i.e., non-words):
- mave* churp* wardonl*
- In order to do that, we must be able to assemble the pronunciation from the letters (in fact, we have not encountered these sequences before)
This suggests that word processing might rely on partially distinct mechanisms for regular words, irregular words and pseudowords
26
dual route cascaded model (Coltheart et al., 2001)
- ‘Race’ between lexical and non-lexical routes
- Lexical route faster for words
- Lexical route is frequency weighted; regular < irregular
- Non-lexical route faster for pseudowords
- Lesioned computational model:
- Damage to non-lexical route –> lexicalisation problems (‘mave’ becomes ‘cave’)
Damage to lexical route –> slow to read irregular words
27
triangle models of reading (Harm & Seidenberg, 2004)
- Learning of correspondences between spelling and sound in a semi-regular system (English - systematic but with many irregularities)
- Fully interactive
- Interconnected semantic, orthographic and phonological units
- Single mechanism for reading regular, irregular and pseudowords
Words represented as patterns of semantic, orthographic and phonological
28
triangle models and pseudowords
- Phonological units and their connections to orthographic units are of particular relevance for pseudowords. Pseudowords have no meaning, so the contribution from semantic units is limited.
Reading of pseudowords is particularly affected following damage to phonological units.
29
triangle models and irregular words
- Orthographic and semantic units and the connections between them are critical for the recognition of irregular words.
Reading of irregular words is particularly affected following damage to orthographic units.
30
triangle models and experimental data
- Can successfully simulate the performance of healthy readers
better reading of frequent words, better reading of regular words, frequency by regularity interaction
31
discriminating between models
- Both types of models can account for basic experimental findings in normal reading
- Both models can account for neuropsychological data
- Both focus on the English writing system
Neither really specifies how they are implemented in the brain
32
neuronal recycling hypothesis (Dehaene, 2009)
- The ‘neuronal recycling’ hypothesis: visual word recognition is a result of recycling cortical structures whose initial functions were for object recognition.
- In accord with evidence showing positive correlation between complexity of lines in writing symbols and that found in image fragments (natural scenes).
This could be the key factor that determined the visual appearance of written symbols.
33
bilingualism
- “The regular use of two or more languages (or dialects), and bilinguals are those people who use two or more languages (or dialects) in their everyday lives”. François Grosjean (2008).
- Different from being a native speaker of two languages!
- Bilingualism is everywhere:
- ~ 6000 languages for ~ 190 independent countries.
- EU (2012): 54% general population and 78% students can speak another language.
- Acquisition contexts: simultaneous or sequential.
Variation in the AoA, practice, proficiency, exposure, usage, motivation, etc.
34
separate or shared lexicons?
- “Separate store models” propose separate lexicons for each language (e.g., Potter et al, 1984). à inspired by neuropsychological cases (selective language recovery in stroke patients - Pitres, A. (1895)). Strong evidence against these models comes from studies that have shown that the two languages are co-activated and compete for selection (e.g., Spivey and Marian, 1999; Costa & Caramazza, 1999; Costa et al., 2000).
“Common store models” propose a shared lexicon between languages and a shared semantic memory system. (Green, 1998; Costa et al., 2005)à The majority of current models of the bilingual language system are based on this. Largely supported by evidence from neuroimaging, psycholinguistics as well as patient studies (e.g., acquired language impairments due to stroke, traumatic brain injury and dementia).
35
evidence for separate lexicons (Kim et al., 1997)
Sentence generation task: Distinct activation in Broca’s area for covert production of two different languages.
36
evidence for shared lexicons )Costa & Caramazza, 1999)
- Spanish-English bilinguals naming pictures in Spanish (L1) with distractors in Spanish (L1) and in English (L2).
Semantic interference does not depend on the language of the distractor.
37
meta-analysis study: l1 and l2 naming (Indefrey, 2006)
- L2 > L1 in late AoA, low proficiency, low exposure bilinguals.
- Differences in BA 44, 47 L&R inferior frontal gyrus.
- Converging behavioural and neuroimaging evidence for strong effects of AoA, proficiency and usage/exposure.
- Less fluently spoken languages trigger more activity than well-mastered languages.
Differences are seen in regions that respond to processing difficulty and selection, across a range of tasks and domains.
38
the hard problem: selecting the right word in the right language (Costa et al., 2000)
Parallel activation of the two languages: example from Spanish-Catalan bilinguals naming (cognate and non-cognate) pictures in Spanish.
39
language specific selection models- conception selection account (La Heli, 2005)
- An early hypothesis was that the language a speaker wants to use at a given point in time is specified at the conceptual level as part of the preverbal message, which permits the brain to essentially activate words only in the target language. However, this model is not really supported by evidence of parallel activation and competition between languages.
- A second critical assumption of this account is that word production unfolds in a fully serial fashion thereby only the word corresponding to the selected concept is accessed at the lexical level.
- The target language is specified at the conceptual level (pre-verbal message).
No parallel activation and no cross-language competition.
40
language specific selection models- language specific selection account (costa et al., 1999)
- proposed instead that target and related concepts activate the corresponding words in both languages however,
- words in the non-target language would not interfere since lexical selection mechanisms only consider words in the target language and selection only depends on the activation levels of the words in the target language. The selection mechanism is here conceptualized as a lexicon-external monitoring device capable of restricting lexical search exclusively to the intended language while ignoring activated words in the non-intended language.
- The target language is selected at the lexical level.
Parallel activation, but no cross-language competition.
41
inhibitory control model (language non specific selection) (Green, 1998)
- The target language is selected by applying reactive inhibitory control (IC) to the non-target language (via language task schemas);
- IC is proportional to the strength of the language (L1 needs to be inhibited more strongly than the L2);
- IC is applied to all the non-target lexical representations (global inhibition);
- IC is a domain-general mechanism.
- Predictions:
* Cross-language competition proportional to the strength of the language (strong interference from L1 when using L2)
Transfer effects from language control to domain-general cognitive control (e.g., bilingual advantage).
42
evidence from language switching
- Asymmetrical switch costs are not consistently observed (see Costa et al., 2004, 2006);
Rather than reflecting persisting inhibition, switch cost asymmetries may reflect the need to strongly activate the weaker of the two languages (see Branzi et al., 2014; 2020).
43
neural mechanisms for language selection (green&abutalebi, 2013)
- Summary of functional neuroimaging investigations of switching between languages, translation and language selection.
-
- Evidence across studies for critical involvement of the cognitive control network in language selection.
- Asymmetries in neural language switch costs?
Is language selection supported by the same mechanisms as those involved in task selection?
44
asymmetries in neural switch costs?
- Activity in inferior frontal gyrus (IFG), anterior cingulate cortex (ACC), inferior parietal lobule (IPL), and caudate is observed during language selection.
Neural switch cost asymmetries have been observed in various studies (Branzi et al., 2016; Wang et al., 2007; Abutalebi et al., 2013).
45
domain-general mechanisms for language (Abutalebi et al., 2008)
- Picture naming task in German-French bilinguals.
- Three contexts:
* 1) SNc: Simple naming in L1, naming baseline.
* 2) TSc: Task selection in L1 (name image or produce related verb in L1); selection baseline.
* 3) LSc: Language selection (name image in L1 or L2).
ACC and Caudate are particularly engaged when managing conflict between languages (against the assumption of the ICM).
46
evidence from neuropsychology (Calabria et al., 2014)
- MR (January 2011)
* Multiple white matter lesions at the subcortical level around ventricles and in the corpus callosum.
* The L Caudate showed lesions both in its posterior part and in its tail, whereas the R Caudate showed lesions only in its tail.
*
Involuntary switches to Spanish in MS patient with damage to the Caudate (see also Abutalebi et al., 2000)
47
