introduction to psycholinguistics debates methods
TRANSCRIPT
Introduction to psycholinguistics
DebatesMethods
Important questions: the ultimate Question
• Innateness, species specificity, domain specificity– Nature vs. nurture– Are the skills involved in knowledge and use of language specific
to humans?– Are the skills involved in knowledge and use of language specific
to language?
– i.e. Does language distinguish us from all other life forms??
Important questions:the organisation of linguistic knowledge• The organisation of knowledge of language
– Are processes or levels within language independent of each other? Do they work in cooperation?
– How are different processes of language related (e.g. reading and speech)?
– What are the subprocesses involved in language production and comprehension?
– How does the study of language relate to everyday language use?
Modularity of language• Big modularity:
– How independent is language from the rest of cognition?• Little modularity:
– How independent are specific levels of language processing of each other?
Modularity (Fodor, 1983)
• A module is a specialized, encapsulated cognitive system that has evolved to handle specific information types of enormous relevance to the species.
• Input levels of processing output
• What is the relationship between different levels of processing?
A modular view
Modularity (Fodor, 1983)• Modules are
– Encapsulated (it is impossible to interfere with the inner workings of a module)– Unconscious (it is difficult or impossible to think about or reflect upon the operations of a
module)– Fast– Have shallow outputs (no information about the intervening steps that led to that output– Obligatory firing (operate reflexively, providing pre-determined outputs for
predetermined inputs regardless of context)– Ontogenetically universal (develop in a characteristic sequence)– Localized (have dedicated neural systems)– Pathologically universal (break down in characteristic fashion following some insult to
the system)– Domain specific (deal exclusively with a single information type)– It is assumed that learned systems do not display all of these characteristics.
Interactionism
• Modules are not encapsulated• They interact and cooperate• Bottom-up and top-down processes
Important questions:Critical periods in language acquisition?
• Lenneberg (1967) critical period hypothesis– Certain biological events associated with language can only
happen in an early critical period (hemispheric specialization).– Certain language events have to happen in this period for
language acquisition to follow a typical route– Language is acquired most effectively in this critical period.
Evidence• The two hemispheres are not fully lateralized at birth. An early left
hemisphere damage or hemidecortication does not only result in permanent language impairment.
– Maturational hypothesis: the two hemispheres of the brain are equipotential at birth. Language lateralization is strongest between 2-5 years, and closes by adolescence.
– Invariance hypothesis: the innate organization of the left hemisphere makes it more ready for language (given up only for good reasons like brain damage to the left hemisphere)
Evidence• Second language learning.• ‘Feral children’: Genie
• A weaker version of the critical period hypothesis: sensitive period, restricted to more complex aspects of syntactic processing
Important questions:Language universals
• Chomsky: there are certain universal constraints on rules and categories
• These constraints are biological and so innate• Language acquisition device, Universal Grammar, that
accounts for all possible human languages• Principles and parameters
• BUT see also: Evans & Levinson 2009
Types of universals• Substantive universals
– Categories necessary for analysing or constructing languages: noun, vowel, subject, etc.
• Formal universals– Constraints on the types of linguistic rules
• Implicational universals (Greenberg 1963): Of the form if x, then y– If a language marks gender on nouns, it will also mark it on
pronouns. – If a language is predominantly VSO in its word order, then the
adjective will most often follow the noun.
Important questions:Does the language system use rules?
• linguistics: explicit rules– E.g. the plural form of nouns in English is formed by adding the –s morpheme
to the stem• For a long time, rules were proposed to operate in both
comprehension and production. This view is more and more often challenged.
• E.g. in connectionist modelling– Builds on a plausible brain metaphor (?): processing is implemented by a
densely connected network of neuronlike units. The model has to be very explicit.
– Rethinking of linguistic representations. Connectionist models do not contain rules explicitly: these emerge as statistical generalisations over the data.
Psychological mechanisms
serial and parallel processing
bottom-up and top-down processing
automatic and controlled processing
Serial vs. parallel processing
Planning of clause structure
Planning of clause structure
Retrieving lexical items
Retrieving lexical items
Retrievingphonological
representations
Retrievingphonological
representations
A serial model
Intention to convey an idea
Intention to convey an idea
A parallel model
Planning of clause structure
Planning of clause structure
Retrieving lexical items
Retrieving lexical items
Retrievingphonological
representations
Retrievingphonological
representations
Bottom-up and top-down processing
HAT
stimulus
TH A
features
letters
HAT CAT
N
word
Bottom-up and top-down processing
HAT stimulus
TH A
features
letters
HAT CAT
N
word
Bottom-up vs top-down processing
• Phoneme restoration effect (Warren & Warren, 1970)• up top [peel, feel,wheel, heel]
• ?• bottom down */ee/l• Inability to report a disguised phoneme
Context can disambiguate it:“It was found that the *eel was on the orange”“It was found that the *eel was on the shoe” [peel and heel, respectively]In general, how to make sense of speech in noisy rooms?
Automatic and controlled (attentional) processing
• Controlled processes: complex tasks that substantially draw on limited processing capacity
• Automatic processes do not tax limited resources
• Automatic: skilled word recognition• Controlled: comprehension of main ideas
Automatic versus controlled (attentional) processing
Ambiguity
SinceJayalwaysjogsa mileseemslikea short distanceto him.
Afterthe childhad visitedthe doctorprescribeda courseof injections.
Ambiguities• Local:
– Since Jay always jogs a mile seems like a short distance to him.– After the child had visited the doctor prescribed a course of injections.– The evidence examined by the lawyer turned out to be unreliable.
• Persistent:– The policeman saw the thief with the binoculars. – Visiting relatives can be boring.
• Sources:• 1. Lexical ambiguity in meaning and/or in syntactic category (The
evidence examined…).• 2. Ambiguous dependency (The policeman saw the thief with the
binoculars).
Ambiguity: lexical
•My Blackberry Is Not Working! - The One Ronnie, Preview - BBC One
• http://www.youtube.com/watch?v=kAG39jKi0lI&feature=player_embedded
Methods in psycholinguistics
Methods
• Question is how do we get to know something about the processes taking place in the mind of a language user?
• Methods in psycholinguistics are by and large the methods of cognitive psychology
• On the basis of experimental and observational data, researchers formulate models of what might be going on in our head
Experiments• Most often used• Online and offline methods• Online – measures some kind of activity during language
behavior: • Reaction times• Eye-movement registration• Speech monitoring• Brain imaging
• Off-line• Post-hoc task, memory involvement• Grammaticality judgments
Experimental methods• Visual Comprehension (reading): lexical decision, naming, priming,
self-paced reading, eye movements, semantic categorization, brain imaging• Auditory Comprehension (listening): lexical decision, priming,
phoneme monitoring, brain imaging.• Oral Production (speaking): picture naming paradigm, spontaneous
errors, error inducing paradigm, brain imaging.• Written Production (writing): thinking-aloud protocol, ScriptLog
(software logging key strokes)
* Advantages and disadvantages of methods: unnatural settings, unnatural tasks, more natural tasks can be rather expensive
Experimental methods
Visual lexical decision – testing word length
600 MSTEREMONE NO YES
700 MS
WORD NO YES 500 MS
E.g., after 20 items per condition, 100 ms or 5% error difference in favour of short words => word length has an effect!
PSYCHOLINGUISTIC
NO YES
Dependent variables: Response latencies and error rates
Experimental methods
Naming – testing word length
500 MS
WORD 400 MS
E.g., after 20 items per condition, 100 ms difference in favour of short words
=> word length has an effect!
PSYCHOLINGUISTIC
Dependent variable: Voice Onset Time
Reaction times• How long does it take to read out a word?• How long does it take to decide about a word
– Whether it is a word or not?– Whether it contains the sound /f/?– Whether it belongs to the category of mammals?
• What properties of words influence reaction times?– Length, frequency, complexity etc.
• This is easily measured by computers (not always that easily)
Priming
• In almost all areas of psycholinguistics• If two things (representations) are related, they engage the
same level of processing, and then processing one will have an effect on processing the other.
• This effect can be facilitatory or inhibitory. The relationship can be semantic, morphological, phonological, associative etc.
Priming
birdthrush versus mammalthrush
breadbutter versus shoebutter
punishmentpunish versus bulletinbullet
Observation, surveys and questionnaires
• Mainly in child language research• CHILDES database• Diaries of (language researcher) Mums• Important in research on language pathologies• Speech errors
General methodological problems
• Most experiments tests monolingual typical university students
• Most experiments involve reading, while for most people, speech is the central language activity
• How sensitive is performance to a specific experimental technique?
A well-controlled experiment
1. Number of items2. Matching relevant factors 3. Number of subjects4. Statistical testing5. Outcome task-specific?6. Outcome language-mode specific?7. Outcome language-specific?