Chapter 12 Psycholinguistics Flashcards Preview

Ling 100B > Chapter 12 Psycholinguistics > Flashcards

Flashcards in Chapter 12 Psycholinguistics Deck (43):


slips of the tongue
useful for data about language processing
indicate that utterances are planned before they are articulated
provide evidence for particular linguistic units, such as phonemes and morphemes
Analyzing spoonerisms is a field technique
"You have hissed all my mystery lectures"


experimental methods to study the organization of the mental lexicon

lexical decision


lexical decision

subjects must decide whether a string of letters is or is not a word
Dependent variables: response latency (the time to respond) and response accuracy (correctness/incorrectness of the response)


Response times (lexical decision)

Higher-frequency words (e.g. free) yield faster responses than lower-frequency words (e.g. fret)

Unpronounceable nonwords (e.g. nlib) also yield slower responses than pronounceable ones (e.g. plib), indicating that phonotactic constraints play a role in processing

Non-words that sound like real words (e.g. blud, phocks) also yield slower responses than those that do not



the target item (e.g. the item to be judged in a lexical decision task) is preceded by a related prime
Response time is is faster if the prime is related: semantically (cat-dog)
orthographically (couch-touch)
phonologically (light-bite)
or morphologically (legal-illegality).



The unconscious automatic analysis of sentences.
Unlike morphemes, which are stored, sentences are generated and interpreted by means of computations


Methods of studying sentence processing

timed-reading experiments
eye-tracking experiments
event-related potential (ERP) experiments


timed-reading experiments

subjects press a bar on the keyboard to advance from one word to the next. The dependent variable is response time


what do timed-reading experiments show?

show that content words such as nouns and verbs take longer to process than function words such as determiners, conjunctions, and prepositions
show that readers pause at the end of a sentence; this is seen as the effect of integrating preceding information into the clause structure



jerky eye movements while reading a sentence, can be tracked using a video camera.usually progress forwards during reading. they are called regressive saccades when going backwards


eye-tracking experiments

look at saccades and regressive saccades
The dependent variables in eye-tracking experiments are fixation location, fixation time, and the number of regressive saccades.
• Eye fixations are typically centred on content words, and are typically longer for less frequent words.


regressive saccades

• Syntactically complex and semantically anomalous (odd) sentences are associated with a greater number of regressive saccades.
• Regressive saccades also occur more with less proficient readers.


Event-related potential (ERP) experiments

measure voltage fluctuations on the scalp, resulting from neural activity in the brain.
Random fluctuations are eliminated by averaging over many presentations of a given stimulus type
What remains is the electrical activity related to a particular type of stimulus event


well-known ERP component

the N400, a negative electrical potential (voltage) about 400 milliseconds (ms) after a word is presented
This response is stronger for words that are unexpected in their semantic context
The pizza was too hot to eat. The pizza was too hot to drink. The pizza was too hot to cry.


what do ERP experiments show

suggests that the interpretation of a sentence proceeds incrementally, not waiting until the sentence is complete


Language corpora

can be analyzed for information about words, such as:
• frequency
• age of acquisition
• syntactic contexts
• morphological family size
• semantics of neighbouring words


English Lexicon Project

provides databases of stimulus words for psycholinguistic experiments, and response times from multiple lexical decision experiments


Google Ngram

graphs the relative frequency of words or phrases (as found in Google Books) over time


bottom-up processing

• As listeners hear an utterance, they perform a phonetic analysis to isolate phonemes and word boundaries, and relate these to items in the mental lexicon.


top-down processing

isteners also develop a partial representation of what they hear, and form expectations to guide phonetic processing and word recognition


slips of the tongue evidence

phonological features play a role in language processing, since they can be involved in reversal or spreading errors


cohort model of speech processing

evidence that listeners analyze each word incrementally from beginning to end, gradually reducing the number of compatible word choices until only one is left (plus any homophones).
The unit of analysis appears to be the phoneme


cohort model of speech processing example

g: gap, get, ghost, girl, glass, glee, goat, gull... gl: glad, glass, glazier, glee, glimpse, gloat... gla: glad, glamorous, gland, glass, glazier... glas: glasnost, glass...



evidence that the syllable plays a role in speech processing
subjects identify target syllables more quickly than target non- syllables


internal structure of syllable

appears to affect language processing

When subjects blend two one-syllable words to form a single one, subjects tend to take the onset from one word and the rhyme from the other
bug + cat = bat



the components of a multimorphemic word prime their semantic associates, independent of the whole word
• because the components are activated by morphological parsing before the complete item is processed (pre-lexical decomposition), and
• because a complete lexical item activates its components (post-lexical decomposition).


pre-lexical parsing

BarkingV primes dog—but triggers negative priming for tree, suggesting that bark(N) is activated pre-lexically, then suppressed post- lexically.

pseudoaffixed word like corner primes corn, but a nonaffixed word like scandal does not prime scan


Selectional restrictions

processing times are longer for nonwords that violate these
re-: attaches to V to create V
-able: attaches to V to create A
-ize: attaches to N to create V -ity: attaches to A to create N
[re-birm]-able re-[birm-ize] *re-birm-ity



involves operations (or transformations) that change the structural relations among words and phrases—for example, wh-movement

the number of these in a given sentence does not predict its processing time.



a relatively autonomous, or independent, processing unit.
sentence processing is controlled by a module separate from the syntax (the parser)


garden- path sentences

evidence for a separation between the parser and the grammar
sentences which are grammatical (and fairly simple) but still difficult to understand
The horse raced past the barn fell
Since Jay always walks a mile seems like a short distance to him.


two principles of parsing: (garden-path sentence)

Minimal Attachment
Late Closure


Minimal Attachment

do not postulate unnecessary syntactic nodes


Late Closure

attach new words to the clause currently being processed


metaphors for language processing

A psycholinguistic model incorporates research results into a proposal about how processing takes place.
they summarize research findings and generate specific hypotheses


Serial processing models

postulate that processing proceeds step by step.
better at capturing bottom- up procedures such as phonetic perception


Parallel processing models

postulate that several processes occur simultaneously.better at capturing interactions between top-down and bottom- up processes, as in sentence comprehension


Single-route models

claim that representations are accessed in only one way.


Dual-route models

claim that representations are accessed in two competing ways.


Psycholinguistic modelling

Example: accessing frequent, infrequent, and novel compounds
• blackboard: whole-word recognition
• breadboard: morphological decomposition
• blueboard: morphological decomposition


Symbolic models

make reference to rules and representations involving symbols such as phonemes, words, syntactic categories (N, V, etc.), and so on.


Connectionist models

make reference to associations of simple units (nodes) which have no discrete boundaries


Language processing seems to involve both:

• massive storage of associated representations, and
• obligatory computations that decompose linguistic input into its syntactic, morphological, and phonological constituents.