lang

• 8 functions of language

  1. communication
  2. thinking
  3. social interaction
  4. recording information
  5. expressing emotions
  6. expressing identity
  7. attempts to control the environment
  8. pretend to be animals
• Humans are strongly motivated to develop a language, so that even with minimal input they will spontaneously come up with symbols (words, signs) to denote things and – more importantly – rules (grammar, syntax) to combine the symbols into new messages. 
• It is unclear whether language acquisition is a unique, innate capacity of humans or whether it is based on general learning principles that also apply to other species (and machines). 
• language is a system of

  • symbols
  • rules (forms grammar and syntax)
• Stages of speech production • Speech production involves four levels or stages of processing: semantic (message planning) level, syntactic level, morphological level, and phonological level. The tip-of-the-tongue state occurs when semantic processing is successful but phonological processing is unsuccessful. 
• Speech planning • Speakers must plan their utterances. This happens at three levels: phrase, clause, sentence. The extent of speakers’ advance planning is relatively greater when speakers do not experience time pressure, speak slowly, when they aren’t performing an additional task at the same time, and when simple sentences are planned. Much of the flexibility in advance planning occurs because there is a trade-off between avoiding errors in speech and the cognitive load on speakers. 
• Speech errors • Speech errors occur when an incorrect item is more activated than the correct one. Speech errors generally form words rather than non-words, in part because we monitor our speech. • There are individual differences in the number of speech errors people make and the proportion of anticipatory errors. The more proficient a person is, the fewer errors they make; most errors made by proficient speakers are anticipatory. 
• Speech disorders and aphasia • About one in 15 children has difficulty with speech and understanding spoken language for no apparent biological or cultural reason. These children have a language development disorder. • Speech problems after brain damage are called aphasia. Symptoms can be semantic, phonological, and/or grammatical. Most aphasia patients do not stop thinking, indicating that thoughts can make use of a different code. 
• Speech perception is complex because of the speed of spoken

  language, the variable nature of the stimulus input, and the presence

  of background noise. Speech perception involves the activation of

  word cohorts fitting word beginnings, which are pruned as more

  information comes in and only the target word remains.
• Computational models

  • According to the TRACE model, there are processing units at the

  level of features, phonemes, and words. The model’s emphasis on

  top-down processes allows it to account for the word superiority

  effect. However, the model exaggerates the importance of top-down

  processes and ignores the role played by orthographic information in

  speech perception.
• The distributed cohort model shows that monosyllabic spoken

  words can also be processed with distributed representations, in

  which individual nodes no longer represent meaningful information,

  but in which information is encoded by activation patterns across a

  whole layer. This model can simulate human performance without

  the need of semantic top-down influences.
• Multi-modal aspects of speech perception

  • Speech perception not only involves the processing of the auditory

  stimulus. Also information from other modalities can contribute.

  • In fluent readers, the orthographic representations of written words

  interact with the speech input to improve auditory perception.

  • When we are listening to someone, we are also influenced by their

  lip movements, as can be concluded from the McGurk effect, the

  fact that lip movements influence the sounds we hear.
• Cognitive neuropsychology of speech perception

  • The speech perception system can be compromised by brain

  damage. In some patients this is the only processing affected,

  suggesting that speech perception uses processes not shared with

  other functions (reading, understanding other sounds).

  • Ellis and Young (1988) argued that in order to understand the

  variety of symptoms observed in patients with speech perception

  disorders, one must distinguish between three routes that connect

  spoken input to spoken output in healthy people.
• It’s in the eyes

  • Average reading rate is 240–260 words per minute. Readers move

  their eyes in saccades about four times per second. Information is

  extracted from the text during fixations. The next eye movement is

  programmed after only partial processing of the currently fixated

  word. Two computational models of eye movements in reading are

  E-Z Reader and OB 1-Reader.
• Sound as well as vision?

  • There is good evidence that reading involves phonology. This

  explains why sentences with tongue-twisters are harder to read,

  why participants find it hard to say that the word ROWS does

  not refer to a flower, and why words are processed faster after a

  homophonic prime. The phonological priming effect is also observed

  when the prime is not consciously visible, suggesting that phonological

  processing occurs rapidly and automatically.
• Two routes from print to sound

  • Theories must explain why we are capable of correctly naming both

  the word “yacht” (pronounced as “yot”) and the non-word “fusk.”

  This requires the existence of two routes. According to the DRCmodel,

  one route involves the activation of entries in an orthographic

  lexicon (needed to explain why we can correctly name

  words with unclear correspondences between the letters and the

  sounds) and another route in which graphemes are converted to

  phonemes (needed to name non-words).
• • According to the triangle model, a mental lexicon is not needed, if

  one assumes that known words activate their meaning in addition

  to the sounds corresponding to the letters.
• Learning to read in English

  • It is especially hard to learn to read English because there are many

  inconsistent spelling-sound correspondences. Children learn to

  read by translating written letters into sounds (phonics approach). Gradually they develop orthography-based word recognition, which

  allows them to recognise words with inconsistent letter-sound

  correspondences as rapidly as words with consistent correspondences.

  In addition to word-decoding skills, listening comprehension

  skills are also important.
• Dyslexia

  • Some individuals find it harder to learn to read and write than peers

  of the same intelligence and under the same learning conditions.

  This condition is called developmental dyslexia.
• dyslexia

  It is to a large extent a biological condition and thought to be due to suboptimal connections between the brain areas responsible for visible word recognition and spoken word representations. Students with dyslexia still have problems with reading and writing in higher education, often despite intensive efforts to remediate the problems.

  • Reading problems as a result of brain injury are called acquired

  dyslexia.
• Word recognition

  • Words are easier to produce and recognise when they are frequent,

  acquired early, short, and primed by the preceding context.
• Sentence parsing

  • To understand the relations between the words of a sentence,

  humans build a syntactic tree consisting of a hierarchy of clauses,

  phrases, and word types. The tree is built online as the sentence is

  being processed. From time to time, the construction is impeded

  because of syntactic ambiguities.
• According to the garden-path theory, only the syntactically simplest

  construction is pursued and needs to be revised if it turns out to be

  wrong. According to the constraint-based theory, all sources of

  information (word-related, syntactic, semantic, world knowledge)

  are used from the outset during sentence comprehension.
• When there is a conflict between the syntactic structure of a

  sentence and listeners’ knowledge of the world, they often favour

  semantic knowledge at the expense of syntactic structure. More

  generally, listeners (and readers) are often content with “good

  enough” representations that are inaccurate.
• Research with neural networks suggests that statistical learning

  alone already leads to good sentence and text understanding. It is

  unclear whether this implies that much language processing in

  humans is based on statistical learning as well.
• Discourse understanding

  • We use schemas when processing discourse to make inferences

  about information needed to understand the discourse but not

  present in the message. Schemas enhance the comprehension and

  retrieval of discourse. We usually do not make more inferences

  than needed for good-enough representations. People with high

  working memory capacity make more inferences than people with

  low working memory capacity.
• To understand the intended meaning (the pragmatics) of a message

  we often have to focus on the speaker’s goals rather than on the

  literal meaning of the message. This requires common ground

  between the speaker and the listener. Again, we do not always

  manage to establish common ground (leading to misunderstandings)

  and there are individual differences in the ability to do so.