Language (Y1)

Language and psychology

Linguistics - the scientific study of language and its structure, including the study of grammar, syntax and phonetics

Ascribe meaning to words - we are able to attribute language to symbols easily and automatically
Psycholinguistics - the scientific study of the psychological reality of language use including -
-> Language acquisition
-> Use of language
-> Language comprehension
-> The psychological mechanisms used to process and represent language in the mind

Language - Human systems of communication and personal expression which have been built upon symbols and representations

Representations - words that stand in for the things they are intended to represent
In turn, these cause us to create internal, mental representations of the words
Language and thinking - at least three principles characterise human language: generativity, recursion and displacement

Generativity allows a speaker to use a small number of words / grammatical structure to compose and infinite number of sentences and new ideas
--> Can create entirely new ideas from unique arrangement of words
--> Could have conveyed the same idea with different descriptions - use more concise one
Recursion - any sentence can be extended indefinitely by embedding clauses or phrases within or following it - ability to create complex sentences
Displacement - quality of language that allows one to converse about things that do not exist, exist in other places or are abstractions
--> Abstractions such as future tense, counterfactuals, beliefs, past tense and similes

Key aspects of language

Phonology - the fundamental sounds that make up spoken language -

Phoneme - basic auditory unit of spoken, natural language - English has 34-44, dependent on accent
Smallest sound that could change and change the meaning of the word
-> Dipthongs - letters placed together to make one sound
-> Semivowels
-> Consonants - stops, whisper, affricates, fricatives (voiced and unvoiced), nasals
-> Vowels
Critical to understand language
Rabin and Schafer (1978) - phonemes in American English - organisation due to articulation

Moon et al, 2013 - infants, even pre-birth, and children learn how natural sounds are divided into these meaningful phonemic units

30 hour old babies were exposed to vowels unique to their mother's language and to vowels unique to another language
Babies sucked harder on a dummy when listening to the foreign language - novel stimuli (high amplitude sucking)
Exposure effect of learning phonemes of own language in utero
Suggests babies listen to speech in the womb and learn about it

Children are better at phonemic discrimination - in English, it is easy to distinguish between a 't' and a 'd' despite words like wedding and wetting sounding similar - at 6 months of age, children can distinguish the phonemes of any language

Innate ability to acquire any language
In Thai, there are three sounds for d and t, due to the /dt/ phoneme, and so different languages have different phonemes
Acquisition -
-> 6 month olds can distinguish non-native language distinctions
-> By 12 months old, they can no longer
-> There is a decline in this ability during the first year of life as a specific function of language experience - reorganisation of perception

Semantics - the meaning of language, words, phrases and utterances

Conceptual meaning - linguistic function of a word e.g. money is used in exchange for goods and services
Associative meaning - concepts related to the word e.g. money and wealth
Issues -
-> Homonyms - semantically ambigious words where the orthographic and phonological information overlap but the semantic information does not
-> Polysemy - semantically ambigious words due to the fact they have different, similar functions

Studying semantics - semantic association task; a task in which the participant classifies stimuli based on meaning; the detection and classification of semantic relationships between words

Processing the third word is harder because we have to suppress the relevant meaning of a phonological neighbour (van Ogden, 1987)
Classifying flowers - tulip, rob, row - third is harder
Semantic relationships - the final pair is easier due to associative meaning (semantic priming) - doctor and nurse, quicker to recognise the word
Lexical decision making task - a task in which participants must classify whether a string of letters are a word or not

Syntax - principles that govern the structure of language - sentence structure governed by the rules and principles by which we consider sentences grammatical

This provides governance, but we also have freedom to structure sentences due to generativity
Governs utterances, with languages differing in their preferred word order
English is an SVO language - subject-verb-object, whereas other languages are SOV like Dutch
Rarest sentence construction - OSV and VSO
Sentence acceptability - study syntax; do they make sense?
-> While we can often easily distinguish grammatical and ungrammatical sentences, sometimes they can appear to be acceptable in a comparative illusion
-> Syntax rules are not always the best grammar judge

Pragmatics - the context within which the language is used

The study of the meaning in context and it deals with implied meaning as opposed to the mere lexical meaning expressed
How utterances are used, interpreting what speakers mean, how people make sense of each other
Gricean Maxims of pragmatics - rule of conversation for efective, cooperative communication where all interlocutors are understood
-> Maxims of quality, quantity, relation and manner (Week 1)
Just because we have many ways to say words, it does not mean they are all acceptable

A huge number of things impact this, including semantics and pragmatics

We are very good at detecting these 'weird' language usages - acceptable sentences are all grammatical, but not all grammatical sentences are acceptable (Jurafsky, 1992)
-> Garden path sentences - grammatically correct sentences that often lead to initial incorrect parsing (analysis) due to their syntactic ambiguity - leads readers up the garden path - last word changes suspected meaning
-> Semantic ontology - some concepts are bounded (dead or alive) - when you stop being one, you are the other - you switch; unbounded concepts mean that you can be both concepts on a scale (wide and narrow)

Speech production and perception :

Speech production - phonemes decided by place of articulation (where airflow is obstructed)

Glottal stop - glottis (the opening where the vocal chords sit) is restricted so sound stops - glottis when open allows air to flow freely
Vocal chords also close and vibrate rapdily
SSS is voiceless and requires no larynx vibration, whereas ZZZ is a voiced sound
This is typically by the moving articulator - the part of the tongue - and the passive articulator which is the roof of the mouth
Manner of articulation - how the airflow is obstructed (oral and nasal)
Vocal chords (folds) in larynx, which vibrate to modulate the flow of air being expelled from the lungs

Air is generated by the lungs, flows through the vocal tract, vibrating over the vocal chord, filtered by facial muscles, released through mouth and nose

Although the vocal tract is a tube that roughly measures up to one foot, the sound actually takes shape in the latter six inches of the tube, known as the oral cavity
In this portion of the vocal tract, many combinations of anatomical activities can alter the passing air flow
Brain can signal vocal chord to resonate or vibrate, and this forms the glottal waveform by tightening the vocal chord muscles, the sounds' fundamental frequency can be increased, producing higher-pitched sounds
Furthermore, the air flow can be restricted and altered by positions of the tongue, teeth, lips and other facial muscles in the oral cavity

How is language perceived - is it harder to understand spoken or written language?

The Segmentation Problem - when do we decide a word has ended; speech is a continuous stream of sound and it is suggested we use top-down processing to alleviate this issue
We may segment words based on our knowledge of the language and word structure
Using language constraints to segment words -
-> Word legality (Norris et al, 1997) - harder to hear the word apple in fapple then wufapple - we are likely to reject a word segmentation that results in impossible words
-> Stress (Cutler and Butterfield, 1992) - first syllable is typically stressed in English, so we can use this as a cue to segment words

Phonemic restoration effect - top-down effects in perception - Warren; in initial experiments, Warren provided the sentence shown and first replaced the first 's' phoneme in legislatures with extraneous noise in the form of a cough

None of the participants noticed / recognised the missing phoneme
This indicated that in its absence, we filled it in, indicated by top-down processing - can backfill sounds when given later context

Theories of spoken recognition

Specific role of context is controversial (top-down v bottom-up processing)

Two of the most prominent are the TRACE and Cohort models

TRACE speech perception model (McClelland and Elman, 1986)

Explains the use of relationships among auditory features, phonemes and words
Explicit 'connectionist' model (neural network) and so has been programme on a computer - built to identify words and sounds how a human would
Simulations are run to see if the models predictions of behaviour are accurate of what we observe in participants
Information can flow from the bottom up, or from the top down
Word units <- phoneme units <- feature units (inhibitory and excitatory connections)
Positives of this model - TRACE explains context effects - allows higher level information
Negatives of this model - it overestimates the influence of context and predicts top-down effects that do not exist - it is not a human
Perceives the word and its rules and places in context to perceive the speech

Cohort model - Marslen-Wilson, 1984 -

As we hear a phoneme, we come up with a possible list of words, known as a cohort, which we narrow down as we hear more phonemes
Trespass - all other candidates in the cohort have been eliminated, so it is the only word left based on the sounds - uniqueness point of the word
Selection of a candidate is influenced by other factors such as - auditory presentation of the word and the semantic / syntax context
A word can be recognised before its uniqueness point if the context supports only one candidate in the cohort

Strengths -
-> Allows for the use of higher information to limit lower level information
Weaknesses -
-> If the first phoneme is mispronounced, we would never be able to recognise a word - we catch up by resetting the cohort
-> Findings suggesting context does not eliminate words from the cohort
Cross modal priming - weaknesses
-> Listeners complete a lexical decision task while listening to a sentence
-> Auditory stimuli can be manipulated to affect language comprehension
-> Auditory stimulus - word visually presented
-> Context should only prime one word with a semantic association to the sentence / auditory stimulus
-> But, also primed to another word, and thus context does not eliminate words from the cohort

Revised cohort model (Marslen-Wilson, 1990) - solved the two problems

Context no longer allowed to influence the early states, including the initial cohort i.e. the revised model is less interactive, more bottom-up
Words are not totally eliminated from cohort, but instead their activation level decreases and can be revived later if new information comes in - this deals with mispronunciations etc

How do we read new (or non) words - in English, 90% of non-words are pronounced with the same grapheme-phoneme rules as known words

It is easier to know how to pronounce an unknown word if there are no similarily spelled but irregularly pronounced words
These general findings are actually quite specific to English

Language acquisition - history of language theory -

Knowledge of universal grammar is considered innate in human beings, and it is universal grammar that forms the internal language faculty - not accepted anymore

Children learn language effortlessly - and this is the case whether the language they are exposed to is spoken or signed
Specific genes appear to be involved in the acquisition and use of language while infants prefer the sounds of spoken language to other sorts of sounds

behaviourists such as B.F. Skinner argued that language is a behaviour like any other and is learned through reinforcement
Chomsky argued that children would never acquire the tools needed for processing an infinite number of sentences if the language acquisition mechanism was dependent on language input alone
Suggested instead that humans have an internal language faculty: human beings possess innate, specialised cognitive structures dedicated to acquiring and using language
Based on many shared characteristics of languages
Universal grammar is a concept to describe a basic internal structure or set of intrinsic rules that languages share in common

Non-human language - animal communication systems do not contain the properties of generativity, recursion and displacement and therefore are not a language

Non-human communication systems are unable to express an infinite number of new thoughts and ideas and refer to non-existent entities
They are also unable to express very many things at all
Attempts to teach chimpanzees actually to speak have always been unsuccessful due to the way the chimpanzee larynx and throat are structured
Gardeners' taught a chimpanzee to produce more than 130 signs, but this considered a communication over a language
Nim Chimpsky - trained by Herbert Terrace - claimed Nim had learned that when he made certain signs, they obtained things they wanted such as food
Likened this language learning process to being more similar to operant conditioning than actual language acquisition

Do other species of animals have language? How do language and other abilities relate?

Unique to humans -
-> Vocabulary size (no limits on words learned)
-> Recursivity
-> Grammatical complexity (German has 4 cases, 3 genders, Japanese has special forms to speak to the Emperor etc)
-> Relative sophistication of arbitrary sentence learning - we soak up statistical regularities
-> Distance in time and space from immediate content domain
-> Amazing tendency for a rich and systematic language to develop, with the help of children, from impoverished circumstances (pidgin to creole)
-> Transmission from one generation to the next

Methods for studying language -

mistakes between children and adults (tenses)
brain imaging
studying language perception and production of damaged patients (Broca and Wernicke areas)
temporarily damaging or enhancing brain areas with TMS
building machine translation systems
questionnaire experiments
computational analyses of text corpora
cross-cultural studies e.g. colour terms and perception
recording from monkeys
eye movements from reading studies
building computational models of reading
behavioural experiments

What we can learn about language from brain damaged patients - neuropsychology:

Broca's aphasics - difficulty with grammar and speech production
Wernicke's aphasics - difficulty with semantics

Sex and gender differences in language, cognitive and perceptual abilities:

Girls on average have better vocabulary - bedroom culture
This is relative to boys, at whatever age
However, variability within boys and girls is huge
Importance of joint attention and parent-child book reading
Memory tasks - every memory task favours women
-> Similar story as far as huge variation within women, men and small differences in averages
Spatial (e.g. possible object rotation) tasks - tend to favour men
Again, small average differences compared to standard deviation between men and women

Summary - at least three principles characterise human language - generativity, recursion and displacement

At least 4 key aspects of language interpretation and production - phonology, syntax, semantics and pragmatics
Recognising spoken words is difficult, but we are very good at using context to overcome these issues
Many researchers believe that the capacity for language is a special internal faculty comprising innate, specialised neural and cognitive structures and 'wiring' - Trace and cohort models
Animals communicate, but do not have language - unlike human language, animal communication systems are closed ended rather than open ended

Case of Genie - issues developing syntax (grammar rules) which suggests that there is a critical period of language development and acquisition - it also suggests that there is an interaction between cognitive and environmental factos in forming language acquisition

Language is vital to evolution as it promotes social bonds and interactions as it is an effective way to persuade others
Speech production is conceptually driven - top-down manner
Principal use of language is through speech
Non-language vocalisations convey information and gestures can supplement or substitute for spoken language - body language and tone of voice are also important
6000 languages, but many are going extinct, such as indigenous languages - 4% of languages spoken by 96% of the population
Languages vary based on their lexicons and rules for sentence construction - but all can express complex and new ideas

Linguistic universals - Aitchison, 1977 -

have consonants and vowels
combine basic sounds into larger units
have nouns and verbs
can combine words in meaningful ways
can express who did what to whom
can express negatives and questions
structure dependent
allow recursion

Issues - sign language, some of these distinguishing categories are not clear cut, tonal languages rely more on tone for communication, syllable is also a universal unit of speech, unusual classes of consonant sounds such as the click Xhosa

Evolutionary - aided communication in hunting as it mimics natural environment sounds and would not have startled prey

Hockett's design features for human language (1960) - 16 total - these are 16 features that are unique to human languages; some animal communication share some features, but only human language has them all

Vocal-auditory communication channel
Broadcast transmission and directional reception
Rapid fading
Interchangeability
Feedback
Specialisation
Semanticity
Arbitrariness - relationship between meaning and sounds of the word
Discreteness
Displacement
Productivity - novel utterances
Cultural transmission
Duality of patterning
Prevarication
Reflexiveness
Learnability

Concepts are not independent, and only apply fully to spoken language, not sign language.

Different phones that are treated as the same phoneme within a language of allophones - phonemes are abstract representations of the phonological units of language - subjective recognition by native speakers

Ability to distinguish allophones decreases with age - critical period beyond which the adult is tuned into native language
Phonological and phonotactic rules describe what sounds go together - rules based on what sounds natural and easy to produce
Speech segmentation relies on knowledge of word boundaries using information about phonotactic probabilities in language
Minimal pairs - words that differ by one phoneme

Morphemes - meaning units, building blocks of words - morphology - special case of syntax

Free and bounded morphemes (singular and plural)
Verb endings are inlectional morphemes
Derivational morphemes - create new words with new meaning when added to a stem and change the rules of its use
Applies to adjectives and verbs
Function words - words that do the grammatical work of a sentence - do not change
Lexicon - part of semantic memory system
Words are symbols - meaningful sounds and generally have a particular referent - greetings and social conventions
Semantics - meaning of words, morphemes - relationship between objects and the words used

Discourse - multi-sentence speech and includes dialogue, conversation and narrative - social conventions that affect language processing become increasingly relevant and people rely on schemas to process language

Pragmatics - understanding the context and use of the language
Difference between linguistic competence and communicative competence
Effective discourse is based on shared understanding between those engaging in conversation
Conversations involve turn taking and cooperation, following implicit social conventions - range of cues
Universals in turn taking patterns - Gricean Maxims
Denotes - literal meaning
Connotations - implied meaning

Speech errors

Malfunctions - slip of the tongue, Freudian slips and tip of the tongue effects and aphasias

Hesitations and pauses -
-> Disfluencies - more common than actual errors, and vary with the situation and the individual; natural characteristic of fluent speech
-> Filled pauses are more frequent than silent pauses and can serve to announce a delay in speech
-> Pauses allow next words to be articulated
-> Pause use varies on context
-> Some disfluencies can facilitate comprehension and hesitations have also been implicated as a cue for deception

Slips of the tongue - Freud (parapraxes) - action slips are errors can be informative in regards to language processing

Freudian slips - semantic slips to a phonologically similar word - show the cognitive processes underlying sentence formulation, rather than unconscious motivation or conflicts

Examples of types of speech error - Fromkin (1971) -

phonological substitutions
semantic substitutions
additions
deletions / omissions
blend
transposition / exchange errors (spoonerisms)
perseveration
anticipation

Majority of speech based errors are sound based, and tend to occur at one linguistic level

Shows the importance of the phrase as a unit of production, as errors rarely jump across phase boundaries
Exchange errors show a lexical bias
Stressed and unstressed syllables do not change - suggests a systematic process of sentence construction in which the skeleton does not switch but details do, and that those details exchange with other parts of the language category
Spoonerisms - swapping words or first letters of words in sentences
Tip of the tongue state -
-> Temporary inability to access a word from memory
-> Brown (1991) - occurs about once a week, is universal, increases with age, affects recall of proper names, involves an available initial letter, accompanied by other words and is resolved on almost half of occasions
-> TOT experience can cause phonological or semantic errors
-> Bilinguals are more prone to this experience - due to lexicon crossover - lexical activations increase the likelihood of this state

Dabrowka (2015) - What is universal grammar, and has anyone seen it?

Explanations - universality (all human languages have the same properties), convergence (all have the same grammar despite having different input) and poverty of the stimulus (children know things they could not have learned without available input)
However, there are few true universal and most languages are distinct and so the fundamental cross linguistic fact that needs explaining is diversity, not university
Differences in adult native speakers knowledge of the grammar of their language, and so children do not all learn the same grammar
Unique in our ability to language learn
Universal grammar is therefore instead an acquisition style that is a language making capacity, rather than an innate knowledge bank of language structures that allow it to be learned

Okanda et al (2015) - Gricean Maxims in preschoolers:

Studied violations of the maxims in Japanese children aged 4-6 and adults
Understanding of the maxims was assessed - quality, quantity, relation and manner
Sensitivity to violations increased with age - at 4, it was chance, at 5 they understood some maxims (first quantity, then the others) and 6 year olds and adults understood all the maxims
Preschoolers acquired the maxim of relation first, and found quality the most difficulty to comprehend
Adults and children differed in their comprehension of the politeness maxim - development of pragmatic understanding of the maxims and implications for the construction of developmental tasks from early childhood to adulthood

Miller and Eimas (1995) - Speech Perception Review:

Map acoustic signals onto lexical signals and language comprehension in which in the early stages we analyse the continuous signal of speech onto discrete prelexical linguistic representations

Connectionism - McClelland and Cleermans (2009)

Computational models used to model aspects of human perception, cognition and behaviour, the learning processes underlying this behaviour and the storage / retrieval of this information from memory
They are representations of the connections in our cognitive networks and how we process information when cognition is occurring (models of connection)
Central features - emphasis placed on mechanisms of change - some mechanisms involved in cognitive processes should be informed by the manner in which the system changed over time as it developed and learned
Activation - how each neuron propagates and activates mechanisms of cognitions in neural networks
Representation - pattern of activation when something is being perceived and also the value that the connection weights link to processing units due to the weight of previous experiences
-> Patterns of activation formed during processing are not necessarily stored as memories, but instead leave a repeatable trace in the network through the adjustments they make to connection weights

Processing - propagation of activation signals among processing units - continuous time process subject to random variation; same input can give different real-time trajectories
-> Much of our experience is a continuous flow, with many networks settling into attractor states (attractor networks) which are a stable pattern of activity that remains until a reset occurs so the next input can be presented
Learning - adjustment of the connection weight due to the frequency of the use of that connection in cognition e.g. frequently used heuristics and categories cause stronger weighting
Knowledge is always implicit whereas symbolic systems are explicit - connectionism relies on knowledge stored in connections, and this leads to consciousness
Some parts of the connectionist modelling are used to control the states of activation in other parts of the network
-> Difference between weight-based processing and activation-based processing - activation based on activation, maintenance and updating, weight based on adaptation to amounts of input and output
-> Activation more flexible and faster - speed and flexibility are both salient characteristics of high-level cognition
-> Activation based processing is a central characteristic of the frontal cortex - evolved for controlled processing such as working memory, inhibition, executive control and monitoring of ongoing behaviour
-> Processing in the frontal cortex is therefore characterised by mechanisms of active maintenance through which representations can remain strongly activated for long periods of time so these representations can bias processing elsewhere

Connectionist models can be thought of as implementing a process of global constraint satisfaction whereby biased competition between neural coalitions result in the network settling onto the most likely interpretation of the current input
Strong link between attention, working memory, cognitive control and availability to conscious experience due to the connected nature of the processing between these areas - information processing occurs depending on the operation of the same computational principles