Langston,
Psychology of Language, Notes 5 -- Fonts, Visual Perception, and Reading
I. Goals:
A. Evolution of written language.
B. Fonts.
C. Orthography.
D. Components of reading.
E. Models of reading.
II. Evolution of written language. Spoken language
evolved as part of the human condition. In spite of the debate
about
whether the ability to perceive speech is innate or learned in
childhood,
one thing is clear: All children who acquire language acquire
speech
with no formal instruction, no children acquire reading without formal
instruction. Written language first emerged around 5,000 years
ago,
and independent forms of written language emerged in only a few parts
of
the world.
A. Three classes of writing systems: All of these have
arbitrary symbol-sound mappings. They differ in whether or not
they
have arbitrary symbol-meaning mappings.
1. Logographic: Each sign represents an entire word.
You rarely have inflections (like plurals). Ex.: House =
<a
picture of a house>
2. Syllabic: Each sign represents a syllable. This
cuts down the number of signs needed to a limited set of
syllables.
Ex.: Japanese: car = kuruma (<symbol> = ku,
<symbol>
= ru, <symbol> = ma).
3. Alphabetic: Each symbol represents a phoneme (sort
of).
English.
B. Evolutionary progression: (1st three are all
logographic).
1. Pictograms: A picture communicates an idea. A
picture is what it stands for, no clue to pronunciation in the
symbol.
Ex: Mesopotamian <a picture of an arrow> = arrow.
2. Ideograms: Pictograms are hard to use for abstract ideas
and grammatical information. If you extend pictograms to abstract
ideas, you get ideograms. They represent ideas rather than
objects.
This is a big step because you introduce arbitrary mappings between the
symbol and the meaning. Ex: <foot symbol> = to go.
3. Logograms: Each symbol is a word. Chinese is an
example of this. It works well for China because wherever you go
you can read, even though pronunciation is radically different in
different
regions. If the language was alphabetic, and the spellings were
based
on sound, different dialects would look and sound different, and
reading
another dialect would be difficult if you didn't speak it. The
problem
with logograms is that they're memorization and symbol intensive.
You need thousands just to read a newspaper.
a. Hieroglyphics: Started as pictograms and developed into
a logographic system.
b. Cuneiform: Signs are read for sound value: <waves>
= sea or see. Use rebus sentences to say what you want.
Ex.:
<eye> (I) <waves> (see) <sun> (son) = "I see my
son." This
is moving in the direction of having symbol-sound correspondence.
4. Syllabary: Each symbol = one syllable.
Advantages:
A relatively small number of symbols to learn; you know how to say it
based
on the symbols. But, it's still a lot to memorize.
5. Alphabetic: Each symbol = one phoneme. Ideally,
one sign for one sound. Advantages: Efficient: Each
language
has only a few sounds (usually around 50), so you can get by with 20-40
symbols to memorize; Flexible: Any language can be characterized
by an alphabet. Simply list the sounds and then assign them
symbols.
a. Phoenicians: Adapted hieroglyphics into 22 signs (all
consonants). They traveled a lot, encouraging the spread of
alphabetic
systems.
b. Greeks: Adapted Phoenician alphabet and added
vowels.
First true alphabet: Each sign = one sound, every sound has a
sign.
Top
III. Font development (formal printing). Most of
the reading that you do is from printed material, so no survey is
complete
without discussing it.
A. Evaluating fonts: There are two (related) dimensions
along which fonts are evaluated:
1. Legibility: How easily can the letters in the font be
perceived? The easier it is to make them out, the more legible
they
are.
2. Readability: How fatigued do you get when you read
them?
Poorly designed fonts make you more tired than well-designed fonts.
Note that these characteristics were rarely considered as fonts
developed.
Instead, you have a history of technological improvements tempered by a
desire to preserve what was done in the past. Sometimes, this
resulted
in happy accidents that improve legibility and readability. But,
it was rarely done on purpose.
B. Font characteristics:
| A list of some basic properties |
Best option |
a. Serif vs. sans-serif: A serif is the little
horizontal
mark on the tops and bottoms of vertical lines in fonts (the line on
the
bottom of this f). Originally, it was used when cutting letters
in
stone to prevent the stone from cracking, but it was preserved out of
tradition.
Luckily, serif is better.
b. Weight difference: Some lines in a character are thicker
than others (e
for example).
c. Bias: The fonts can be on a bias or they can be
vertical.
d. x-height: How much height is devoted to the body of
the characters (how tall the x is).
e. Spacing: Some characters are wider than others (i vs.
e). Typewriters force these to take up the same amount of space,
but it's not required in printing (proportional is when they take up
only
the required space) (piece vs.
piece).
f. Proportions: How big is the x-height relative to the
heights of ascenders and descenders (parts going above and below the
body
of the letter). There is an optimal proportion for each font. |
Serif
Difference
Biased
More is better
Proportional
Optimal
|
C. Why are we discussing this? Because the fonts can have
a huge influence on reading. We have to know what's being read
before
we can look at the process of reading. It all comes down to
constraint
satisfaction. We need to understand the constraints on fonts
(what
are the parts, what aren't) so that we can understand how they
constrain
reading.
As an example of the importance of features on letter identification,
consider an experiment by Neisser on visual search: You scan for
an X in a field of Z's and N's or a field of O's and P's. X is
easier
to see in letters with different features than letters with similar
features.
Try it:
N N Z N Z N Z N Z
Z N Z Z N Z Z N N
N N N Z N X N Z N
N N Z N Z N Z N Z
Z N Z Z N Z Z N N |
O O P O P O P O P
P O P P O P P P O
O O P P O X P O P
O O P O P O P O P
P O P P O P P P O |
The X should "pop-out" of the grid with dissimilar features.
A similar process of analysis-by-features probably takes place in
reading,
making it very important to understand the features.
Top
IV. Orthography. What are the spelling rules?
Again, it comes down to constraints. If there are rules that
govern
spelling patterns, then knowledge of these rules can help the process
of
identifying letters and words. Additionally, a lot of theories of
reading require a mapping from spelling to sound. If you
understand
how these correspondences work, then you can go a long way as a reader
towards knowing what's being said. The system for English was
worked
out by Vinetsky.
A. Some of the big rules:
1. Avoid letter doubling.
2. V C V, V C C V, V C: A vowel before a
consonant-vowel
is long, a vowel before a C-C-V is short, a vowel before a consonant is
short.
a. To override V C, add a dummy 'e' to get V C V.
Examples:
"fin", "can" vs. "fine", "cane."
b. To override V C V, you have to double. Example:
"cunning."
3. Especially avoid doubling at the beginnings and endings of
words.
a. Except for ff, ll, ss.
b. Except for 3-letter words (egg, inn, add, ebb).
Why?
18th century editors decided to reserve 2-letter words for function
words
(to, in).
B. Why does this matter?
1. It gives you hints at pronunciation even if you've never seen
the word before ("mabe", "mab", "mabing", "mabbing").
2. It helps you to know what letters to expect in a particular
situation. The interactive activation model was designed to take
advantage of this. By letting letter features and letter
identification
be influenced by regularities at the word level, you increase the
sources
of constraint and make it easier to identify the stuff on the page.
C. A bit on English spelling:
There are a lot of irregularities in English spelling. Part of
this is due to the great vowel shift where pronunciation changed and
spelling
didn't. Part is due to English's propensity for borrowing.
A lot of borrowed words contain spellings from other languages that
make
little sense in English. Here are some of the problems:
1. Some letters in a word may not represent any sound (through,
sign, and give).
2. A group of two or more letters can represent a single sound
(think).
3. A single letter can be two or more sounds (saxophone).
4. The same letter can represent different sounds in different
words (on, bone, one).
5. The same sound can be different letters in different words
(rude, loop, soup).
The long poem at the end of the notes has a lot of examples of weird
English spelling-sound correspondence.
Why is the English spelling system good (even though it violates
one-sound
one-symbol allowing "ghoti" to be an acceptable spelling of "fish")?
1. Because the meaning is more transparent. We have to
work harder to learn pronunciation, but we can then figure out word
derivations
based on spelling, helping us to get the meaning. Some examples:
electric -- electricity /k/ and /s/ are both covered by 'c'
insert -- insertion /t/ and / / are both covered by 't'
right -- righteous /t/ and / / are both covered by 't'
bomb -- bombard ø and /b/ are both covered by 'b'
damn -- damnation ø and /n/ are both covered by 'n'
produce -- production / / and / / are both covered by 'u'
2. Dialects. In parts of the country, "car" would be
spelled
"ca" to stick with the one-sound-one-symbol rule. Different
speakers
would have different spelling systems.
3. Because we can distinguish homophones by different
spellings.
Some examples:
to, too, two
bare, bear
no, know
flea, flee
sore, soar
Without spelling, readers would have to figure this out on their
own.
That's not as easy because there's more room for ambiguity. Of
course,
irregular spelling is a different sort of ambiguity, but it's easier to
handle.
So, crazy English spelling is actually good. In a sense, English
orthography is not based on pronunciation. So, it's not really
trying
for one-sound-one-symbol. Instead, it's preserving the same
spelling
for all variants of a morpheme. In other words, it's a
morphological
system, more sensitive to morphological differences than phonemic
differences.
D. Examples of orthography constraints:
1. Word superiority effect: Letters are perceived better
in words than alone. This is probably due to information from the
orthography (surrounding letters). The interactive activation
model
of reading uses the extra information from the bottom (more letters)
interacting
with information from the top (knowledge of words) to arrive at an
identification
for a given word.
2. Effects on reading:
a. People with Broca's aphasia don't suffer as much disruption
if their writing is logographic because logographic systems use a
different
part of the brain. For example, Japanese Broca's patients can
have
difficulty with hiragana or katakana (syllabaries), but be OK with
kanji
(logographic).
b. Congenitally deaf people have an easier time learning
logographic
systems than alphabets or syllabaries (probably due to trouble with
sound-spelling
correspondences).
c. Learning logographic systems is more memorization intensive,
but reading is easier right from the start (for the symbols you
know).
Alphabets are hard to master. Syllabaries aren't as bad because
children
seem to be better at identifying syllables than phonemes. Even
(phonologically
naive) adults can have a hard time segmenting words into phonemes but
can
do syllables with ease. Note that a syllabary would be
impractical
for English.
Top
V. Components of reading. What are the basic
processes?
A. From eye-tracking studies: In eye-tracking experiments,
subjects sit in front of a computer monitor to read some text. A
small camera records their eye position, and this information is used
to
determine exactly where on the screen the subject is looking.
Based
on this, the amount of time spent on each word can be computed.
From
observational studies of readers in eye-tracking studies we know that
reading
consists of:
1. Gazes/fixations: These are pauses on words. This
is where information is accumulated about words. Some facts:
a. Length is a function of the reader and the material.
Some people have a different time constant than others.
Generally,
a fixation lasts around 250 ms (1/4 second).
b. Approximately 83% of content words get fixated and 38% of
function words.
c. A fixation generally contains 2-3 characters to the left and
7-12 characters to the right (for a language like English which is read
left to right).
d. In the periphery, readers are affected by the physical
characteristics
of the text about 12 characters ahead. They're affected by the
semantic
content of the text about 6 characters ahead.
2. Saccades: If you try to observe yourself reading, you'll
probably feel that your eyes travel smoothly across the page.
But,
what really happens are periods of relative stability (fixations)
separated
by rapid jumps (saccades).
a. They take around 10 ms, and generally move forward 10
characters.
The exact size will depend on the reader and the text.
b. They're not uniform. You generally jump over function
words and fixate about 2 characters into content words.
c. The saccades allow for processing and erase your sensory
registers
to stop perceptual processing of the last fixation and prepare for the
next fixation (masking).
3. Regressions: About 15% of saccades are back to some
earlier portion of the text. The number of regressions depends on
the reader and the text.
B. Additional processes/factors: We've been structuring
things as ecological survey -> theoretical explanation. The
reason
is that you can't explain something until you've described it
properly.
This is a survey of the processes and factors involved in
reading.
Any models that we discuss will have to consider these components:
1. External factors:
a. Physical stimulation: The lighting and the text
characteristics
(that font stuff we talked about).
b. Word frequency: Some words appear in print more
frequently
than others, meaning that some words are encountered more frequently
when
reading. This frequency has an effect on reading (you read more
frequent
words faster).
c. Sentence construction:
1) Some sentence forms are read faster (active sentences are
easier to process than passive sentences).
2) Length: Generally, shorter sentences improve
comprehension.
3) Anaphors (part of the text that refers back to an earlier
part): More make it worse.
d. Density of propositions (idea units): The more you have
per sentence, the harder it is to comprehend.
e. Text construction:
1) More coherence between propositions is better (the
propositions
in consecutive sentences should be related).
2) The more it corresponds to story grammars the better (for
instance, fairy tales have a particular sequencing, following that
improves
comprehension).
3) Organization: The more headers, labels, and structure
you provide the better.
f. Content area: Some kinds of texts (newspapers) and some
content areas are easier to understand.
g. External goal set: Why are you reading? It's
different
if you're memorizing, summarizing, or just understanding.
2. Internal factors:
a. Schema knowledge and application: Most events have a
regular structure, and most articles and books follow a
structure.
The more familiar you are with an area the more of this you can
apply.
For example, me reading a paper for this class works very differently
than
you reading it.
b. Procedural knowledge: Knowledge about reading.
For example, knowing how to identify main points.
c. Decoding ability: Getting from the letters to a
representation
of meaning. This is an important individual difference
factor.
My decoding ability for Japanese is very bad, middling for Romanian,
and
good for English.
d. Viewpoint/purpose: An internal goal set for reading
(what's my motivation).
e. Cognitive resources: How much mental energy you have
to devote to the task. This is another important individual
difference.
Two points:
1) More is better. There doesn't seem to be any way to
change a person's amount of capacity.
2) Some people use it better than others. This is something
you can influence.
Top
VI. Models of reading. I'm presenting one example
of each of the two major classes of models. Obviously there are
many
more approaches.
A. Production systems: Just and Carpenter (1980):
A production system is a list of condition -> action pairs.
Usually,
it's like an if-then statement. So, "if I see a green light, then
I start to drive". The action is only carried out if the
condition
is met, and the action will be carried out automatically if the
condition
is met. They string a bunch of these together to model
reading.
Kinds of productions:
1. Get next input: A set of productions to initiate a
saccade
to get more information
2. Feature analysis: Pull out the physical features of
the words.
3. Word encoding and lexical access: Use the features from
2 to access the meaning of a word. Access can come from:
a. Perceptual information (the way the word looks).
b. Spreading activation from related words already read.
c. Serial productions: Information that's read goes into
a limited capacity working memory, and anything that's in here gets
refreshed
periodically. As word meanings become less important, they're
replaced
by other meanings, and this source stops providing activation.
This process is influenced by:
a. The number of syllables in the word.
b. Word frequency (how often it appears in print).
c. Whether or not it's a novel word.
4. Case role assignment: A case role is the role a word
plays in a sentence. For example, in "John pounded the nail",
John
is the agent, nail is the patient. Knowing the case role can help
you narrow down what the next word might be (more constraints).
For
example, if I start the sentence "The man was interrogated by the",
you're
probably expecting an agent. You should be surprised to hear
"window",
even though it is acceptable.
5. Interclause integration: You have to relate the
components
of a text via some kind of structure if you're going to understand it,
and this process builds that structure. This process is affected
by:
a. Ends of sentences.
b. Ends of paragraphs.
6. Sentence wrap-up: At the end of a sentence you have
a good opportunity to make sure that all ambiguities have been cleared
up, all meanings have been accessed, and that everything fits the
structure.
It's like clearing the table for the next course.
Some notes about the model:
1. They used eye-tracking data to test the predictions of the
model. The model accounts for 72% of the variability. This
is really good.
2. This is between the stage approach and parallel models.
Each production is stage-like in the sense that it only does one task,
but the productions can fire in any order, and in parallel.
3. This model makes predictions about poor readers. They
should have two problems:
a. They should have less working memory capacity. (Software
to measure capacity is on my software page.)
b. They should have trouble suppressing inappropriate
information.
Why's that bad? You don't have much room to begin with.
Keeping
stuff you don't need makes the problem worse. Also, if you're
poor
at suppressing irrelevant information, it can cause you to shift your
representation
more often than necessary, making it hard to follow the text.
Gernsbacher, Varner, and Faust (1990) looked at suppression skills
of poor comprehenders. They presented sentences like the
following:
He dug with the spade.
"Spade" is ambiguous because it could be the card suit (ace of spades)
or a kind of shovel. The shovel meaning is appropriate in this
context,
the card suit is not. Right after reading an ambiguous word, all
readers show evidence of activating both the appropriate meaning and
the
inappropriate meaning. It's as if activating all meanings and
then
suppressing the wrong one is easier than trying to search for just one
meaning to begin with. How is this assessed? Immediately
after
the sentence, present a word about which people make a judgment (is
this
word related to the sentence). For example:
"ace"
The correct response is "no." Compare that to deciding about
"ace" after:
He dug with the shovel.
"Ace" should never be activated if you read "shovel," so it should
be easy to say "no." If your read "spade," it should be hard to
say
"no." 100 ms after the sentence was presented, good and poor
comprehenders
(measured using a comprehension battery) were both slower in the
"spade"
version than the "shovel" version.
The real question is how do they do later? If good comprehenders
suppress information better, then they should get rid of the
inappropriate
meaning faster than poor comprehenders. So, the critical
condition
was to wait until 300 ms after "spade" and then test with "ace."
Good comprehenders had no problem with that. Poor comprehenders
were
still thinking about "ace." So, it looks like poor comprehenders
are slow to get rid of irrelevant information.
B. Interactive activation: McClelland and Rumelhart
(1981):
(We discussed this before, but here's something a little more
formal.)
A mostly parallel model. You have two components:
1. Nodes: Each node holds a piece of information.
You have feature nodes (that detect things like / - ) ( | ), letter
nodes
(that detect things like a f e o p), and word nodes (that detect things
like words). The nodes are arranged into layers (a feature layer,
a letter layer, and a word layer).
2. Connections: They connect nodes. They're like
dendrites and axons on neurons. The information that the model
learns
is stored in the connections. The connections can be either
excitatory
or inhibitory (make a node more or less likely to fire).
How it works: The feature level receives inputs from
(theoretically)
your eyes. Each node detects a particular feature. If the
feature
is present, any letter with that feature receives an excitatory input
along
the connection. If a feature is absent, inhibitory inputs go up
to
letters. As letters become more excited, they can feed down to
excite
some features even more. Letters also inhibit each other.
Eventually,
this part stabilizes, and some features and some letters are really
excited,
the others are suppressed. At this point, it's mostly driven by
the
physical characteristics of the word (the features).
The letters also feed up to words. So, any word with a particular
letter that's excited will also be excited. Words can feed down
to
excite other letters. Words can also inhibit each other.
All of this trading excitement and inhibition will rapidly settle on
a particular word. Then you can move your eyes to get a new
word.
Some notes:
1. This accounts for word superiority because of the additional
information coming from other letters in the word. If the word is
"work", the features excite letters, which excite the word, which in
turn
excites the letters even more. This activation coming down from
the
word isn't present when the letter is presented alone, and that's why
the
word makes letter identification easier.
2. Think about scaling this up. Once you get past physical
words you still need meaning. Then you need sentence nodes, text
nodes, etc. Where is the understanding?
Top
That concludes the formal part of the unit. Here's a bonus poem
that highlights the irregularities of the English orthographic
system.
Try to read it fast, out loud.
ENGLISH IS TOUGH STUFF
Dearest creature in creation,
Study English pronunciation.
I will teach you in my verse
Sounds like corpse, corps, horse, and worse.
I will keep you, Suzy, busy,
Make your head with heat grow dizzy.
Tear in eye, your dress will tear.
So shall I! Oh hear my prayer.
Just compare heart, beard, and heard,
Dies and diet, lord and word,
Sword and sward, retain and Britain.
(Mind the latter, how it's written.)
Now I surely will not plague you
With such words as plaque and ague.
But be careful how you speak:
Say break and steak, but bleak and streak;
Cloven, oven, how and low,
Script, receipt, show, poem, and toe.
Hear me say, devoid of trickery,
Daughter, laughter, and Terpsichore,
Typhoid, measles, topsails, aisles,
Exiles, similes, and reviles;
Scholar, vicar, and cigar,
Solar, mica, war and far;
One, anemone, Balmoral,
Kitchen, lichen, laundry, laurel;
Gertrude, German, wind and mind,
Scene, Melpomene, mankind.
Billet does not rhyme with ballet,
Bouquet, wallet, mallet, chalet.
Blood and flood are not like food,
Nor is mould like should and would.
Viscous, viscount, load and broad,
Toward, to forward, to reward.
And your pronunciation's OK
When you correctly say croquet,
Rounded, wounded, grieve and sieve,
Friend and fiend, alive and live.
Ivy, privy, famous; clamour
And enamour rhyme with hammer.
River, rival, tomb, bomb, comb,
Doll and roll and some and home.
Stranger does not rhyme with anger,
Neither does devour with clangour.
Souls but foul, haunt but aunt,
Font, front, wont, want, grand, and grant,
Shoes, goes, does. Now first say finger,
And then singer, ginger, linger,
Real, zeal, mauve, gauze, gouge and gauge,
Marriage, foliage, mirage, and age.
Query does not rhyme with very,
Nor does fury sound like bury.
Dost, lost, post and doth, cloth, loth.
Job, nob, bosom, transom, oath.
Though the differences seem little,
We say actual but victual.
Refer does not rhyme with deafer.
Foeffer does, and zephyr, heifer.
Mint, pint, senate and sedate;
Dull, bull, and George ate late.
Scenic, Arabic, Pacific,
Science, conscience, scientific.
Liberty, library, heave and heaven,
Rachel, ache, moustache, eleven.
We say hallowed, but allowed,
People, leopard, towed, but vowed.
Mark the differences, moreover,
Between mover, cover, clover;
Leeches, breeches, wise, precise,
Chalice, but police and lice;
Camel, constable, unstable,
Principle, disciple, label.
Petal, panel, and canal,
Wait, surprise, plait, promise, pal.
Worm and storm, chaise, chaos, chair,
Senator, spectator, mayor.
Tour, but our and succour, four.
Gas, alas, and Arkansas.
Sea, idea, Korea, area,
Psalm, Maria, but malaria.
Youth, south, southern, cleanse and clean.
Doctrine, turpentine, marine.
Compare alien with Italian,
Dandelion and battalion.
Sally with ally, yea, ye,
Eye, I, ay, aye, whey, and key.
Say aver, but ever, fever,
Neither, leisure, skein, deceiver.
Heron, granary, canary.
Crevice and device and aerie.
Face, but preface, not efface.
Phlegm, phlegmatic, ass, glass, bass.
Large, but target, gin, give, verging,
Ought, out, joust and scour, scourging.
Ear, but earn and wear and tear
Do not rhyme with here but ere.
Seven is right, but so is even,
Hyphen, roughen, nephew Stephen,
Monkey, donkey, Turk and jerk,
Ask, grasp, wasp, and cork and work.
Pronunciation -- think of Psyche!
Is a paling stout and spikey?
Won't it make you lose your wits,
Writing groats and saying grits?
It's a dark abyss or tunnel:
Strewn with stones, stowed, solace, gunwale,
Islington and Isle of Wight,
Housewife, verdict and indict.
Finally, which rhymes with enough --
Though, through, plough, or dough, or cough?
Hiccough has the sound of cup.
My advice is to give up!!!
-- Author Unknown
Top
Psychology of Language Notes 5
Will Langston
Back to Langston's Psychology of Language
Page