Psycholinguistics

Dr. Nesreen I. Nawwab

Prof. Ola Hafez
Lec. Yusra Yassin
2019-2020
First semester
How is speech
planned and
produced?
*Talking is an activity that occupies a great deal of our
time.
*On rare occasions we tend to say one word instead of
another (speech error OR slips of the tongue).
*Speech fluency is used to refer to smooth speech with
no errors.
*Disfluencies are utterances in which there is a
reduction in the normal flow of speech.

*Introductory remarks
*Psycholinguists take into consideration
how speakers are planning and
implementing an utterance from the point
of view of selecting the utterance’s
intention to evaluating whether the
articulated utterance had the intended
effect on the listener.

*Continued…
 outline
➢ Naturalistic studies of speech production
➢ Laboratory studies of speech production
➢ Models of speech production
➢ Handwriting and other forms of
language production
How can we know
how language is
produced?
* Naturalistic Studies of
Speech Production
*A great deal of what is known about how speech is
produced has been discovered through the collection and
analysis of mistakes during speaking, known as speech
errors or slips of the tongue.
*Researchers keep systematic and careful records of the
errors they observe, using a methodology known as
naturalistic observation.
*Errors affect only 2 percent of our utterances or less,
thus, researchers must devote a great deal of time to
record enough examples for their analysis.

*Continued…
*A speech error can catch the attention nation
wide, e.g., President Alexander Lukashenko of
Belarus, who made a speech error while
addressing his nation in 2016:
Citizens should undress themselves instead of
develop themselves

*Speech Errors
*Freudian slips: any error in which the listener
could interpret the error as revealing the mind of the
speaker without the speaker wanting it revealed,
e.g.,
❑He is a great eater instead of he is a great leader

*This type of error is the rarest type.

*Types of speech errors

*Spoonerisms: those in which the speaker
transposes the initial sounds or letters of two or
more words named after Reverend William
Archibald Spooner (1844-1930) who worked at
the prestigious Oxford University.
❑Read the examples and the analysis on pages
89-90.

*Continued…
*Victoria Fromkin (1923-2000) is a pioneer in the scientific field of
speech error.
*She pointed out that speech errors have been studied as far back as
the 8th century: in Errors of Populace, al-Kissa’i, an Arab linguist,
described his efforts to collect and analyze speech errors.
*Her model of speech planning served as a starting point in the 20th
century for understanding how speech is planned and produced,
whether produced with errors or error free.
❑Read and analyze the TYPES of errors collected by Fromkin, table
4.2 Examples of Speech Errors Categorized by Type, p. 91.
❑Are they similar to the ones Spooner made?
❑What are morpheme stranding errors?

*Continued…
*Errors observed in languages with tones, e.g.,
Bantu languages of Africa; dialects of Chinese,
including Mandarin, Thai; and Norwegian.

❑Read and discuss the speech errors involving

tones in Mandarin in the language spotlight
box, p 92.

*Continued…
Based on Meringer’s (1908) German Corpus,
Cohen’s (1966) corpus of Dutch (Nooteboom,
1973), and in Schelvis’s (1985) corpus of Dutch
(Nooteboom, 2005)
1. Anticipation(61%, 78%, and 60% respectively)
2. Preservation (28%, 15%, and 22%)
3. Exchange (11%, 7%, and 18%)

* Frequency of occurrence of the different

types of errors collected by Fromkin
*Since Fromkin’s (1971) work on speech errors, there have
been numerous studies of speech errors in different
populations of participants.
*The research is important because it is valuable to know
whether the types of errors that occur do so with the same
frequencies across different types of speakers.
❑Read and discuss Jaeger’s (1992, 2005) research
comparing children’s speech errors to those made by
adults in the Classic Research box, p. 94.

* Research in the area of speech errors

*Slips of the hand
*Slips of the finger
❑Read the summary paragraph
(silent reading) page 93.

*Other types of errors

*The most prominent type of disfluency is
one in which an error has been detected
by the speaker and the speaker makes a
correction.
*The term repair refers to speakers’
spontaneous correction of errors during
speech.

*Speech Disfluencies
*An example from Fromkin’s (1971) corpus is the exam
will occur in class on Wednesday, I mean, Friday.
*Studies showed that speakers tend to repair:
1. immediately following the error, rather than later in
conversation.
2. following the completion of the word rather than in the
middle of the word
3. More often than not (64% of time).

*Continued…
*The sequence of repair includes:
1. Speakers noticing an error
2. Initiating an interruption
3. Planning and producing a filled pause
4. Planning and producing the intended utterance
*Not repairing occurs approximately 36% of the time
leading to communication failures in conversation.

*Continued…
*Other ways in which speech flow may be disrupted:
1. Filled pauses
2. Silence
3. Hesitations
4. Stuttering
*Studies of filled pauses can tell us a lot about how
speech is planned and produced.

*Continued…
*Studies showed that Filled pauses are used to:
1. Indicate that the speaker wishes to continue the turn
2. When speech planning hits a snag (complication or
difficulty)
3. Before unpredictable and infrequently used words
4. Before words having more than one interpretation
*Studies also showed that um is used before longer
delays, while uh is used before shorter delays

*Continued…
*Silences during speaking also provide more time to
complete the planning for the upcoming utterance.
*They last between 250 milliseconds to over 2.5
seconds.
*They occasionally occur because the word has suddenly
slipped the mind. Often the speaker will think of the
word later.
*Such occurrences are referred to as tip-of-the-tongue
(ToT) states.

*Continued…
*In such cases, speakers have partial
awareness of the word that cannot be
fully produced. They are often able to:
1. Report similar sounding words
2. The number of syllables
3. The beginning phoneme

*Continued…
*ToT cases occur more often when trying to produce:
1. Infrequent words
2. Words that are dissimilar versus similar to other words
3. Proper names
*They also :
4. Occur in some speakers more than others
5. Increase with age
6. Occur more often in bilinguals than monolinguals

*Continued…
*The most noticeable disfluency is stuttering: difficulty
articulating speech smoothly.
*Common patterns for Individuals for whom stuttering is a
long-term problem are:
1. Repeating speech sounds at the beginning of words,
e.g., ki ki ki kick, buh buh buh ball.
2. Lengthening single sounds, e.g., sssssssssssick
*There are four types of stuttering.
❑Read and explain the four types on page 96.

*Continued…
❑What types of sounds and
words are more likely to be
stuttered?
Read and answer bottom of page
96

*Continued…
*Many people who stutter also clutter.
*Cluttering describes the relatively rare speaking disorder that
involves high levels of co-articulation especially in long words.
*Speakers diagnosed with cluttering may not be fully aware how
their speech sounds to others.
*Speech for them is not difficult or effortful.
*To others, speakers with cluttering appear unorganized, poor
listeners and distracted.
*Unlike stuttering, cluttering may also involve thought patterns as
well as speaking and writing patterns.
*Recent brain imaging research has shown that those who clutter
exhibit abnormal brain activity in several areas involved in the
control of speech.

*Continued…
❑Read summary paragraph on page
97.
❑Time out for review
Page 97
Key terms
Review questions
*Since the middle of the 20th century, researchers have used
laboratory settings to study speech errors specifically and speech
production more generally.
*They devise tasks for speakers to perform.
*Each task focuses on a specific aspect of speech processing.
*The earliest experiments documented how speakers vary from
each other, e.g., some are fast and some are slow, speaking rate in
relation to age (see last paragraph, page 98).
*A true experiment is an investigation in which the researcher
manipulates one or more variables in order to measure the
influence on one or more outcome variables. It is the only research
methodology that can establish a casual link between variables.

* Laboratory Studies of Speech Production

*We will learn about two categories of
experiments:
1. Those investigating the production of
words or word pairs.
2. Those investigating the production of
sentences.

*Continued…
* Many of the laboratory speech experiments are designed to make
speakers produce speech errors more frequently than they typically
would.
* Baars and colleagues (see page 99).
ball dip
buzz door
bean dump
bat deck
darn bore
* The sequence created phonological bias.
* The procedure is known as SLIP technique.
* Participants’ errors are more likely to be words rather than nonsense
words in a pattern called lexical bias effect.

*Word Production
*Baars et al. (1975) concluded that the
lexical bias effect occurred because
speakers could edit their speech late
during speech planning.
*The speech editor’s job would be similar
to quality control (see page 100).

Continued…*
*Using the SLIP technique, Motley
(1979) supported Freud’s (1901)
view that slips of the tongue may
reveal one’s thoughts. (See the
experiment on page 100).

Continued…*
*Later research has shown the following:
1. Some types of words are more likely to be produced
with an error than others.
2. Some types of phonemes are more likely to be
involved in an error than others.
3. There is a word frequency effect for speech errors.
4. There is a neighborhood effect in speech errors (see
page 100).

Continued…*
*Since the 1970s, researchers have utilized a range of
tasks to investigate how speakers plan and produce
sentences.
1. First we will examine research on the phonological
aspects of speech planning and production.
2. Second we will examine the research on
morphological and syntactic aspects of speech
production.

*Sentence Production
*Some researchers used a technique known as the tongue
twister paradigm.
*Early research showed that:
1. the number of errors produced increased as did the
number of similar phonemes in the sentences.
2. Repeating word-initial consonants led to more errors
than repeated consonants in other word positions.
3. Speakers made more errors when speaking more
quickly.
4. With practice, the number of errors reduces.

*Phonological Processes
 ❖ Six students will be assigned the following
project
❑ The research DIY on pages 102 outlines an
experiment you can carry out with your friends
and family.
❑ The topic of the experiment is Which Types of
Phonemes Twist Your Tongue?

*DIY
*In his 1980 dissertation, Gary Dell hypothesized
that speakers not only make speech errors that
others can hear, but they also make errors that
only they hear in their minds. (see page 103)

*He concluded that it was unlikely that motor

planning and articulation aspects of speech are
involved in causing speech errors.

*Continued…
*Results of later research:
1. Fewer errors occurred after practice.
2. Inner speech practice reduced errors only for inner
speech conditions.
3. Producing inner speech does not involve the thorough
specification of articulatory features that occurs
during the production of audible speech.
4. Lexical bias was observed in both spoken errors and
inner speech errors.
5. Phonological bias affected only spoken errors.
*Continued…
*Wheeldon and Levelt (1995) explored the level
of specification of inner speech regarding the
extent to which its representations are actually
phonological in nature (see bottom of page
103).
*The study was particularly able to demonstrate
that participants were planning syllables in
Dutch words.

*Continued…
*An interesting question is how do speakers plan the
phonology of their utterances?

*Are utterances planned phoneme by phoneme or

syllable by syllable?

*Are utterances always planned the same way or can the

unit of speech planning (i.e., phoneme versus syllable)
vary across utterances?

*Continued…
*There is a lot of evidence that:
1. the planning of utterances involves planning whole syllables
rather than planning syllables phoneme by phoneme.
2. Syllable frequency influences performance in speech production
tasks.
3. There may be differences in the basic unit of speech planning
across languages.
4. Laboratory experiments in which participants are asked to name
pictures (picture-naming task) have revealed that semantic and
phonological information are activated at different times (see
page 104)

*Continued…
*How much of an utterance do speakers
plan?
*What is the smallest unite of planning?
*Do speakers plan an entire clause at a
time?
*Do speakers plan only a word at a time?

* Morphological and Syntactic Processes.

*The view that speakers plan their utterances
clause by clause predicts that before we begin to
say an utterance, we have active in memory all
the lexical entries for all the words in the
utterance.

*Research showed that participants took more

time to begin speaking when similar sounding
words appeared in the clause.

*Continued…
*Among the most interesting methods used
to study speech planning are:
1. those involving the recording of eye
movements (the visual world paradigm)
(see page 105).
2. Those investigating (verb agreement
errors) (see page 105-106).

*Continued…
Time out for review
Page 107
Key terms
Review questions
 Introductory Remarks
*When we are talking, we are usually not aware of how
much time we take between having something to say and
then producing the words.

*Most of the time, the duration is so brief that it is below

the threshold of conscious awareness.

*Researchers aim to know what cognitive processes occur

from start to finish in speech production-from idea to
articulation.
* Models of Speech Production
 Introductory Remarks
*Some psycholinguistic researchers believe that the brain
handles language processing in a modular fashion, with
different types of language information handled by
independent, non-interactive modules.

*Other psycholinguistic researchers believe that the brain

handles all processing interactively without reliance on
modules using incoming information from all sources as
soon as it is available.

*Continued…
❑Serial processing models are the early
models which proposed stages that
occurred one at a time with non-
overlapping processes.
❑Parallel processing models are the
ones in which different types of language
information can be used simultaneously.

*Continued…
 Fromkin’s (1971) model

*We start with Fromkin’s (1971) model because of its

elegance and utility in highlighting the important speech
error data that must be accounted for by any successful
speech production model.
*She produced a model in which each utterance is planned
in five stages, executed one at a time in a sequence.
*The model was inspired by her analysis of the many types
of speech errors collected in her corpus.

*Serial Processing Models

Fromkin’s Utterance Generation Model:

Source: Fromkin (1971)

Continued…*
In which stage would each of the following errors
occur?
*Steve put his bag in his book.
*Mary threw in what Bill brought out.
*Taniesha was prereading the assigned chapter instead of
rereading the assigned chapter.
*A long shory stort, Noble tons of soil, you have hissed all
my mystery lectures , my favorite fong, noman numeral.

*Continued…
*One of the model’s advantages is that it
describes the planning of prosody for
utterances, which can be affected by
errors, e.g., the spade of *kings instead
In earlier stage

of the *king of spades. The stress pattern

was planned before the ordering glitch
occurred.

*Continued…
 Merrill Garrett’ Model
*His work on speech production in the 1970s and 80s made
significant contributions to how researchers view speech
production
*His work was informed by his examinations of speech error
corpora.
*He observed that most errors involve elements (phonemes,
morphemes and words) occurring within the same clause.
*He noted that morpheme and phoneme exchanges usually
involve adjacent elements.
*His model could account for how such errors occurred,
whereas Fromkin’s (1971) model could not.

*Continued…
*According to the model, each utterance we make begins with
processing at the conceptual level where the meaning of the
message is selected.

*The speaker selects the sentence structure and the specific

words of the utterance in the next three stages.

*The last stage involves selecting the articulatory instructions

that will guide the muscles and other parts of the vocal tract
during the act of speaking.

*Continued…
 Garrett’s discussion of how different types of speech errors occur
during different stages of processing.

1. Word exchange errors (put car in my gas) stem from a glitch

happening during the functional level when speakers select the
syntactic structure of the utterance.

2. Exchanges involving phonemes occur later in planning at sound

level (luck troad instead of truck load).

*Continued…
3. In an adjacent phonological or morpheme exchange errors, e.g.,
an anguage lerror, the phoneme /l/ appears one word later and the
article is realized as an rather than a, suggesting that when the
phonology of the article was planned, the following word began
with a vowel instead of a consonant. Thus, the planning stage that
resulted in the inappropriate placement of the phoneme /l/ must
have preceded the planning stage that specified the phonological
realization of the article

4. In the error the coffee never run outs in the morning, is an

example of what Garrett referred to as “accommodation” : errors
in which a moved element is phonologically realized appropriate
to the new location in the utterance rather than the source
When you make the sound of new elements appropriate
location. to the new position

*Continued…
 Bock and Levelt (1994) model of speech production
The model relies on four of the same stages of processing proposed
by Garrett (1980, 1982, 1988).
1. The message stage: the speaker has something to say.
2. The functional stage: involves the selection of words and
determining their syntactic role in the sentence.
3. The positional stage: involves planning the order of words in the
utterance, which includes specifying syntactic and morphology
structure.
4. The phonological coding stage: involves specifying the
phonological structure of the utterance, including phonemes and
prosody.

*Continued…
 Bock and Levelt’s discussion of how different types of speech errors occur
during different stages of processing.

1. Exchange errors stem from processing in the functional

level.
2. Stranding errors stem from processing in the positional level.
3. Sound errors stem from processing in the phonological
encoding level.
4. Just prior to articulation, the speaker is able to monitor the
phonetic planning of their speech and filter the utterance

*Continued…
 Differences between Garrett’s model and Bock and
Levelt’s model.

1. Bock and Levelt (1994) provided more detail about the nature of
processing occurring within the functional and positional stages.
2. Bock and Levelt showed through experiments that the positional
level of planning can be examined separately from the functional
level of processing.
3. They also showed that utterances can be produced faster if
speakers have previously produced the same syntactic structure
even when the words within the structure are different.
People can produce the word faster if they
have use the same syntactic structure

*Continued…
4. They proposed that during the functional stage,
speakers access lexical information for the words
and that lexical information about words is stored
in memory in a semantic network, as first
described by Collins and Loftus (1975) (see page
111). Lexical info

*Continued…
A. When information is accessed, the node is activated, and the
activation can spread along connections activating related
nodes.
B. There are three levels of nodes involved in lexical access for
any word.
T.o.T = the top of tongue we cannot produce
C. Their view of interconnectedness of different types of
information activated during lexical access provided
straightforward way to explain ToT state: the activation levels
spreading through the network from the message level down to
functional and positional level do not sufficiently activate the
phonological level.

*Continued…
The evidence was obtained from an
experiment in a laboratory setting
using a task that involved two
activities: naming a target when
presented with its definition, and
rating a list of words on how easily
they could be pronounced.

*Continued…
The researchers varied the relationship
between the words in the two activities.
The results showed that participants
produced more correct answers with less
ToT states in the primary activity when
words in the secondary activity where
phonologically similar to the target words.

*Continued…
In a similar experiment on speakers of
Italian (nouns’ gender is grammatical),
participants experiencing ToT state
were able to report the gender of the
target word 84% of the times.

*Continued…
❑Is the interconnectedness of
different types of lexical
information consistent or
inconsistent with the view that
language is modular? (page 112)

*Continued…
*Dell (1986) proposed a model using connectionist
model architecture.
*The stages of speech planning occur in parallel and
the processing occurring at each level had the
potential to influence processing at every other
level.
*In contrast with the unidirectional connections in
the Bock and Levelt (1994), the connections in this
model are bidirectional with equal weightings.

*Parallel Processing Models

*Dell’s model is successful in accounting for the
lexical bias effect in speech errors.

*Dell explains the effect as occurring due to backward

activation flowing from the phonological level back
to a morphological level. In contrast, in the serial
models, the effect occurs due to forward activation
flowing from the semantic nodes to the lemma level
to the phonological encoding level.

*Continued…
*The model is unique in its ability to account for one of the
rarest type of speech error-the mixed error, in which an
error is related in both sound and meaning to the intended
word, e.g., lobster instead of oyster.

*The model, however, cannot account for all research

findings in speech production, e.g., during picture naming,
the activation of semantic information occurs earlier than
the activation of phonological information. It is unclear
how the consistent pattern of semantic processing
preceding phonological process in planning could be
produced.

*Continued…
Time out for review
Page 113
Key terms
Review questions