Psycholinguistics –

How We Process Language

Rakhshanda Fawad
› I. Language Processing
› II. Methods of Psycholinguistic Research
› III. Linguistics and Language Processing
› IV. Syntactic Processing
I. Language Processing
› A. psycholinguistics = the study of language
processing mechanisms. Psycholinguistics like to
study how word meaning, sentence meaning, and
discourse meaning are computed and represented
in the mind.
› A. the burning question: “How is language done?”
› A. Mostly an unconscious process.
– Example: We think when we’re reading words on a page that it’s a
smooth process, but our eyes actually jerk across the page – a process
called saccadic motion.
› (The individual jerks by our eyes are called saccades.)
II. Methods of Psycholinguistic Research
› A. Naturally Occurring
› 1. Spoonerisms
› 2. tip-of-the-tongue phenomenon
1. Spoonerisms = “slips of the tongue”.

› Ex: The dear old queen = The queer old dean.

(segment switching)
› Ex: The rules of word formation = The words of rule
formation. (word switching)
› Ex: I’d forgotten about that = I’d forgot aboutten
that. (morpheme switching)

What it shows: The entire phrase must be planned in advance,

or else we couldn’t switch segments, morphemes, and words
like this. This reveals something about the manner in which
sentence (and phrase) production is planned in the mind.
 2. tip-of-the-tongue phenomenon
tip-of-the-tongue phenomenon: trying to access a word based on meaning,
spelling, initial letter, rhyme, etc.
Ex: “Oh, it’s that color that’s really bright green….and it’s also a really strong
liquor…but it sounds a bit like “loose”….it starts with a “sh”
sounds….chartreuse!
That’s it!”
What it shows: how words are organized in the mind = mental lexicon.
Access of the mental lexicon must be very quick, since word recognition takes
just 1/3 of a second.
Related Questions, regarding the mental lexicon:
1) How are entries in the mental lexicon linked to each other?
1) How are entries in the mental lexicon accessed?
1) What information is actually contained in an entry?
II. Methods of Psycholinguistic Research
› B. Experimental Techniques
– 1. lexical decision
– 2. priming paradigm
– 3. Sentence Processing Techniques
1. lexical decision

– a. Process: A word flashes on a computer screen,

and the subject indicates whether the word is a
real word or a nonsense word by pressing a
button.
– b. Things which are measured (dependent
variables):
› 1. response latency = how long it takes the subject to
decide if the word is a real word or a nonsense word
› 2. response accuracy = whether the subject is correct
or not.
1. lexical decision
› c. How this relates to the mental lexicon: in order to
decide if a word is a real word or a nonsense word,
the mental lexicon must be accessed.
– 1. real word: find the mental entry
– 2. nonsense word: realize that there is no mental entry
› d. frequency effects: more frequent words take less
time to access than less frequent words. This tells
us that some part of the lexicon is organized
byindividual frequency of the word.
2. priming paradigm
› a. Process: The word to be judged as real or nonsense (the
target word) is preceded by a stimulus word (prime word)
› Example: Judging “doctor”  precede with “nurse”
› a. Dependent variable: if the prime affects the response
latency or not.
› a. Priming effect: Semantically related primes lead to
faster response times.
› This tells us that the mental lexicon must also be
organized by semantic relatedness, since a semantically
related word preceding the target word makes the target
word easier to access.
2. priming paradigm
› a. Priming effect also found for orthographically related
primes, phonologically related primes (rhyming words), and
constituent morphemes.
› a. orthographically related: dock primes for doctor
› a. phonologically related: worse primes for nurse
› a. constituent morphemes: legal primes for illegality
› These all show us ways in which the mental lexicon is
organized – by spelling similarity, by phonological similarity,
and by constituent morphemes. Thus, there are many
different ways to prime for a single entry in the lexicon –
suggesting that the entries are linked to each other in several
different ways.
3. Sentence Processing Techniques

› a. Basis: We have to parse a sentence to understand it  we

process it by understanding the meaning and structure of its
parts
› a. Timed-reading experiments
› 1. assumption: a difficult sentence takes longer to parse.
Therefore, timing how long it takes to process the sentence
allows us to rank how “difficult” different sentences are to
process.
› 2. bar-pressing paradigm: The subject reads a sentence one
word at a time, and presses the space bar to indicate they
have processed that word. One word appears on the screen
at a time.
1. moving-window paradigm
› A sentence appears with all the words
› dashed out and the subject presses the space bar to see the first word.
› Another space bar press turns that word to dashes and reveals the
› second word, and so on.
› -------------------------------------
› Captain----------------------------
› ----------Hook---------------------
› ------------------was---------------
› ------------------------quite--------
› -------------------------------angry.
› --------------------------------------
semantic and syntactic structure of the
sentence
› 1. The pattern of how long it takes to process a
word reflects the semantic and syntactic structure
of the sentence.
› a. Content words take longer to process than
function words. (semantic)
› a. Subjects pause at the end of clause boundaries.
(syntactic)
a. Eye-movement experiments
› These experiments track the eye movements (saccades) of a subject
while the subject is reading a sentence.
› 1. Subjects tend to fixate on content words.
› 2. Subjects’ eyes move backwards in the sentence when a misparse
occurs.
› 3. Syntactically complex and semantically anomalous bits of sentences
› tend to create lots of backwards movements.
› a. syntactically complex: “The defendant examined by the
› lawyer…” = The defendant who was examined by the lawyer
› b. semantic anomalous: I like my coffee with cream and socks.
› Assumption: The eye movements reflect processing. Thus, longer
fixations and backwards movements reflect processing difficulty.
d. Brain activity: ERPs

› 1. ERP = event-related potentials

› 1. We can measure activity in the brain when a
subject is reading a sentence.
ERP: Negative voltage change
› 1. N400 = negative
voltage change
approximately 400ms
after a word is read
which is semantically
odd.
› Ex: “The pizza was too
hot to cry.” = N400
(instead of eat)
ERP: positive voltage change
› 1. P600 = positive
voltage change
approximately 600ms
after a word is read
which is syntactically
odd.
› Ex: Sarah’s belief in
fairies vs. Sarah’s in
5. What this means: processing of sentences is
belief fairies
immediate and “online” – happens as each word is read,
rather than waiting until the end of a
sentence/clause/phrase to put things together.
III. Linguistics and Language Processing
› A. Top-down vs. bottom-up processing
› B. Cohort Model of word recognition
› C. Evidence that syllables are also fundamental
units of (auditory) perception
› D. Morpheme Activation
A. Top-down vs. bottom-up processing
› 1. top-down: We begin interpretation of a sentence
spontaneously and automatically based on what
information is available to us. For instance, we do
not have to wait until we have analyzed all the
phonemes in a sentence in order to understand it.
› 1. bottom-up: Do analysis to isolate phonemes,
word boundaries, and relate these things to the
mental lexicon. Can happen only piece by piece –
no forward projection, no prediction.
1. Comparing top-down and bottom-up
processing:
› In a lexical priming study, suppose a word is lexically
ambiguous and so has two meanings. However, suppose that
only one of those meanings is appropriate, given the syntactic
structure of the sentence the word is in. A person using very
strong top-down processing would only be primed for the
meaning which is appropriate, given
› the syntactic structure. A person using very strong bottom-up
processing would be primed for both meanings, despite the
fact that only one meaning is appropriate.
› Ex: Hoggle fell gracelessly to the ground.
› Top-down processing: prime only for soil
› Bottom-up processing: prime for soil and grind
B. Cohort Model of word recognition

› 1. William Marslen-Wilson proposed that in word

comprehension, words are recognized from beginning to end.
› 2. ex: hearing “crystal”. = [krIstal]. First, we process the [k],
and initially consider all the words that begin with [k]. All the
words considered are called the cohort. Then we process the
[r], and consider all the words that begin with [kr]. (The
cohort is reduced from all the words that begin with [k] to all
the words that begin with [kr]). And so on, until we process
all the segments of “crystal”.
› 3. Shown with experiments that this is the case.
› 4. Suggests that the segment is a fundamental unit of
auditory perception.
C. Evidence that syllables are also
fundamental units of (auditory)
perception
› 1. Syllables used successfully as primes in lexical
decision tasks.
› 2. Word-blending tasks: subjects unconsciously split
words at natural points in the syllable  the onset
vs. the rhyme.
› 3. Example: “bark” + “meow” = “beow” (rather
than “baow”)
› 4. Subjects prefer to create word blends according
to the syllable structure of their language.
D. Morpheme Activation
› 1. Individual morphological components play a role
in processing.
› 2. Individual morphemes in compound words
automatically activated during word recognition.
› 3. Evidence: crowbar primes bird.
IV. Syntactic Processing
› A. syntactic parser: (together with the grammar) guides
the order of which elements of a sentence are processed
and the manner in which syntactic structure is built up.
› B. Some grammatically complex sentences are easy to
parse and some grammatically easy sentences are hard
to parse.
› Ex: (complex sentence but easy parsing)
› Sarah saw the goblin who displeased Jareth the other day.
› Ex: (easy sentence but hard parsing)
› The horse raced past the barn fell.
C. garden-path sentences
› C. garden-path sentences: sentences which are easy for the
grammar to produce, but hard for the parser.
› A. The parser is “led down the garden path” to the wrong
structure.
› Ex: The horse raced past the barn fell.
› Parser’s garden path: “The horse raced past the barn” = simple
sentence.
› “The horse raced past the barn fell” = Ack!
› Correct Parse: “The horse raced past the barn” = “The horse which
was raced past the barn”  “raced past the barn” modifies “the
horse”.
› “The horse raced past the barn fell” = The horse (modifier) fell.
C. garden-path sentences
› B. Garden path sentences reveal the
preferences of the parser.
› Ex: Given “The horse raced past the barn”, the
parser can choose a simple Subj V PP structure, or a
Subj + modifying phrase structure. The parsers
chooses a simple Subj V PP structure because it
prefers this analysis.
C. garden-path sentences
› More examples of garden-path sentences:
› 1. The prime number few.
› 2. Fat people eat accumulates.
› 3. The cotton clothing is usually made of grows in Mississippi.
› 4. The man who hunts ducks out on weekends.
› 5. Mary gave the child the dog bit a bandaid.
› 6. The girl told the story cried.
› 7. I convinced her children are noisy.
› 8. Helen is expecting tomorrow to be a bad day.
C. garden-path sentences
› 9. I know the words to that song about the queen don't
rhyme.
› 10. She told me a little white lie will come back to haunt me.
› 11. The dog that I had really loved bones.
› 12. The man who whistles tunes pianos.
› 13. The old man the boat.
› 14. The raft floated down the river sank.
› 15. We painted the wall with cracks.
› 16. The tycoon sold the offshore oil tracts for a lot of money
wanted to kill JR.
C. The key
› These sentences are ambiguous at the beginning. There are two
structures the parser can choose up until the disambiguating word is
reached. Sometimes context can help the parser decide which structure
to choose.
› Ex: There were these two horses. And they both were raced that day by
an expert jockey. He raced one of them past the barn, and one of them
past the house. The horse raced past the barn fell.
› Why does context help here?
› We have two horses in the discourse. As soon as we mention The horse,
we seem to assume that there is a unique horse we are talking about.
Therefore, we are looking for some kind of description to attach to The
horse in order to pick out which one we’re talking about. The parser
would be pushed by top-down processing to interpret raced past the barn
as a modifier to the horse so we can tell which one we’re talking about.
D. Parsing Strategies (in the absence of context)

› 1. minimal attachment: no extra syntactic nodes

unless we have to.
› Ex: “I convinced her children…”
D. Parsing Strategies (in the absence of context)

› 2. late closure: prefer to attach words to the node

currently being processed.
› Ex: “The old man…”
Exercises: 1. Experimental Things

› Muffin has been dabbling in psycholinguistic experimental techniques and

has conducted an experiment
› to see if a priming effect was found for morphological roots on suffixed
past tense forms in a lexical decision task.
› a. Give an example of a morphological root for a suffixed past tense form.
› b. What would be the prime in this experiment? What would be the
target?
› c. What is the dependent variable?
› Suppose Muffin believes that there is, in fact, a priming effect found for
morphological roots of suffixed past tense forms in a lexical decision task.
› d. What sort of data would she be looking for? (How would this priming
effect manifest itself?)
Exercises: 2. And More Experimental
Things
› Muffin has been dabbling with sentence processing experimentation. She has been testing
pairs of
› sentences like the following:
› [Link] butter that had been melted in the pan started to burn.
› ii. The butter melted in the pan started to burn.
› a. What relation do these sentences have to each other? (Hint: Think sentence relations)
› b. Muffin was using a moving-window paradigm. Show what the subject would have seen
after 3 space bar presses on sentence ii. (Draw what was on the screen for press 1, press
2, and press 3.)
› c. Muffin found that subjects seem to be much slower pressing the space bar after seeing
started in sentence ii than in sentence i. Describe what this would mean, in terms of how
difficult to process sentence ii is compared to sentence i.
› d. What do you think would cause the slowdown at started in sentence ii?
› e. If Muffin conducted an eye-movement experiment using these same sets of sentences
(like i and ii above), what would you expect a subject’s eyes to do upon reading started in
sentence ii?
Exercises: 3. Yet More Experimental
Things

› Muffin just couldn’t get enough, so she tried her

claw at ERP experimentation. Using sentences like
sentence i and ii in question 2 above, Muffin noticed
a change in the ERP response of subjects at started.
› Would you expect this response to be an N400 or a
P600 response?
› Why?
Exercises: 4. Cohort Model
› According to the cohort model, how many
phonemes of each of the following words would
have to be processed before a hearer would be sure
which word had been spoken? For each word, write
the word in IPA. For each initial cluster up until the
word would have no cohort left, give one example
of a member of the cohort.
Exercises: 4. Cohort Model
› Ex: “prey” = /prej/
› /p/ = pool
› /pr/ = prefect
› /pre/ = prelude
› /prej/ = prayed
› Need to get to end of word boundary.
Exercises: 4. Cohort Model
› 1. giraffe
› 2. scram
› 3. splat
› 4. zebra
Exercises: 5. Garden Paths.
› For each of the sentences below
› a) indicate what the ambiguous region of the sentence is,
› b) indicate what the disambiguating word/region is,
› c) indicate what the initial parse would look like up until
the disambiguating word using parentheses to group
constituents,
› d) indicate what the final parse would look like using
parentheses to group constituents, and
› e) indicate what parsing strategy (or strategies)would
lead to the garden path and why.
Exercises 5. Garden Paths.
› Ex: The old man the boat.
› a) The old man
› b) the (or possibly the end of the sentence, since The old
man the boat hit is okay. is a fine sentence.)
› c) (The old man)
› d) (The old) (man) (the boat).
› e) late closure – prefer to continue processing the NP,
rather than start the VP; minimal attachment –
› prefer to not posit a VP when all the words could fit under
an NP.
Exercises 5. Garden Paths.
› 1. The man who whistles tunes pianos.
› 2. The girl told the truth cried.
› 3. She told me a little white lie would come back to
haunt me.
› 4. We painted the walls with cracks.
The End