1. Introduction
Ever wonder how your brain changes the fast
sounds in speech or the shapes on a page into useful language. Chapter 4 of
David W. Carroll's Psychology of Language delves into the processes involved in
perceiving a language and it's more complex than one might think.
2. The Paradox of Speech Perception
You won't believe it, but we can process
25-30 phonetic segments (individual speech sounds) a second while listening to
speech. Tap your finger 30 times in 1 second – Impossible! But our brains can
decode speech at this speed easily. According to Carroll, we would not hear
anything if the recorded sounds were instead of a sequence of tones, clicks,
and buzzes.
This amazing ability points to something special to note, that speech
perception may be a unique mode of processing, unlike any other sound.
3. The Building Blocks of Speech
3.1 It's All About the Vocal Gymnastics
All speech sounds that we can produce are
basically gymnastics of the vocal folds. Carroll says consonants and vowels
differ in one important manner: vowels are open on the way out from the lungs,
while consonants involve some block.
Take the sounds [b] and [p], for example. Both are made by pressing together
your lips (bilabial consonants), but the main difference is that there will be
vibration of the vocal cords in [b] (voiced) but not in [p] (voiceless). Go
ahead and give it a shot: place your hand on your neck and try saying 'ba' then
'pa.' Notice the difference?
3.2 The Acoustic Puzzle
When researchers first looked at the speech
using spectrograms (visual representation of sounds), they were confronted with
an interesting puzzle. There are no boundaries along which one sound stops and
another begins in continuous speech. The sound [t] in the word tool goes right
into the sound [u]. The [u] goes right into the sound [l]. Although we hear
three separate sounds physically, they are all mixed together like a watercolor
painting.
When multiple speech sounds are encoded in the speech signal at the same time,
this phenomenon is referred to as parallel transmission. It's like trying to
unmelt the wax after you've already shaped the candle!
4. The Categorical Perception Phenomenon
Categorical perception is one of the more
interesting discoveries in speech perception research. A good way to think of
this is that there are two sounds, 'ba' and 'pa', that have all sorts of sounds
in between them where they sound like both and there is only one thing that is
different or varies between them and that is the voice onset time or VOT or the
timing of when your vocal cords start vibrating and it's been released from the
lip closure.
Researchers have the ability to design sounds with VOTs that can occur anywhere
along this continuum. It's amazing that you never hear the sound 'ba' slowly
turning into the sound 'pa.' Instead, you hear one or the other, never
something in between. It is as though our brains come with different categories
for speech sounds that incoming sounds fit into.
5. Making Sense of Continuous Speech
5.1 Context is King
Remember playing Mad Libs as a kid? Have
you ever played a game where you create silly sentences out of selected words?
Actually, far more complicated versions of this game are constantly being
played by our brains.
In an experiment by Carroll who is a journalist, researchers substituted a
sound for a cough in a phrase.
The state-level governors met at the capital city with their respective
legi[COUGH]latures.
Listeners were hearing the missing 's' in the word legislators. And this
phenomenon is referred to as phonemic restoration. Because our brains literally
fill in the blanks based on context, we hear these restored sounds as clearly
as those that actually occurred.
5.2 The TRACE Model: Everything Talks to Everything
The speech perception TRACE model, created
by McClelland and Elman, claims that the perception of speech is not a one-way
process that progresses from sound to meaning. It's more like a chat where the
different levels of processing—features, phonemes, and words—all communicate
with each other simultaneously.
That's why we might hear 'eel' as wheel, heel, peel or meal depending on
whether the sentence references an axle, a shoe, an orange or a table.
Knowledge about the meanings of words influences how we perceive individual
sounds.
6. The Written Word: A Different Beast
6.1 Eye Movements Tell the Story
When we read, our eyes do not smoothly
glide over the page. Rather, they move quickly in jerks known as saccades
(lasting for about 10-20 milliseconds), followed by brief stops called
fixations (lasting about 225 milliseconds). Saccades blind you to the world
because you extract information only when fixating.
The remarkable thing is that expert readers can process about 14-15 characters
to the right (or the left by Hebrew readers) of the current fixation. This
perceptual span provides us an insight of what is coming ahead, allowing us to
read better.
6.2 The Word Superiority Effect
It's surprising, but it seems we are better
at recognizing letters when they're in words rather than in isolation. When
shown word 'WORD' for a fraction of a second can assess D more accurately than
alone D This shows that we may not simply perceive the text letter by letter
but by the whole word.
7. Two Models Duke It Out
Carroll presents two competing theories of
how we read.
'Dual-Route Model says we have two distinct routes':
- A lexical route where people recognize words as a whole; they have seen it
before.
- A nonlexical route which uses letter-to-sound rules for unfamiliar words or
non-words.
The connectionist model proposes one system where the connection between
letters and sounds are strengthened through experience like a path worn through
the forest through repeated use.
Researchers still argue over both models that can explain aspects of reading
and reading disorder.
8. Why This Matters
We study language perception for various
reasons. This research has real-world applications.
- Understanding how reading works can guide teaching practices and help find
learning disabilities.
- Due to lack of sophisticated context processing ability as a human, the
performance of speech recognition systems is still not on par with ours.
- Clinicians can use their knowledge of normal language perception to help
diagnose and treat language disorders.
New studies still show how understanding and producing speech are related. A
2024 study by Arjmandi and Behroozmand notes that the relationship between
speech perception and speech production remains one of the great mysteries in
psycholinguistics.
9. The Beautiful Complexity
The most surprising thing about Carroll's
chapter is that something that we have done every day, and do effortlessly, is
more complicated than it seems. Interpreting language requires two forms of
processing: the 'bottom-up processing' route that analyzes a physical signal,
and the 'top-down processing' route which makes use of our knowledge and
expectations.
We do not just passively receive language; we constantly create meaning, using
whatever cues are available to us in terms of the acoustic signal as well as
the larger context. When you comprehend what someone is saying in a loud
restaurant or speedily glance at a page, know that your brain is performing an
astonishing computation.
Grasping these processes is not merely a matter of curiosity; it could make for
better teaching methods, more effective therapies for language disorders and
one day, artificial intelligence that behaves in a more human-like way.
Carroll's foray into language perception reminds us that even our most trivial
abilities, such as listening and reading, are remarkable feats worthy of
astonishment and continued study.
References:
Carroll, D. W. (2013). Psychology of language (6th ed.). Wadsworth Cengage Learning.