Human Cognitive Neuropsychology (Ellis&Young, 2014)
Chapter 7: Spelling and writing
Spelling without sound
Patient EB was totally speechless by a stroke. His speech comprehension and reading
comprehension are both excellent but slow. He also appeared to have no “inner speech”/
He was unable to match rhyming words that were not spelled alike. He was also unable to
match spoken and written non-words.
His writing was highly successful with only a mild and variable agrammatism. He could write
a whole passage without assistance but could not have said any of it.
Patient MH is anomic, with fluent speech but typically anomic. She couldn’t name pictures of
objects or select rhyming pairs, but she could write the names of objects correctly. The
errors she made were not what you would expect from someone who was assembling
spellings on the basis of sounds of words, but the errors were morphologically related or
visually similar. She could repeat non-words well, but not spell them. This test shows that
the correct spellings of many real words were not generated by the process of phonic
mediation (that spelling is presumed to be mediated by the phonic sound forms of words).
Patient JS and RD were neologistic jargonaphasics. When they were asked to say and then
write the names of a set of pictures, they would say a neologism, but write down the word
correctly. If writing were based on inner pronunciation, then the spellings would surely
reproduce their mispronunciations. Yet they do not.
These cases are evidence against an obligatory phonic mediation theory of spelling.
“Phonological” dysgraphia
Patient PR suffered a left hemisphere stroke. His speech comprehension was normal
combined with fluent speech with good word choice and only occasional paraphasic errors.
His reading is rapid and effortless. His spontaneous writing was said to be labored and slow
because of motor problems and formulation problems.
Nevertheless, he wrote most common words correct, dictated to him. Errors were
morphological.
This stood in contrast with his extremely poor writing of invented non-words. He could write
individual letters when given their alphabetical names, but not when their sounds were
given. He could repeat and read the non-words aloud quite well.
When he managed to write a non-word correctly, he used a real word as a mediator.
, The graphemic output lexicon
The spellings of all these patients must have been produced by some process other than
assembly from sound. It would appear that these patients retrieved the spellings of familiair
words from some internal long-term memory store equivalent to the SOL: the graphemic
output lexicon.
The idea is that each time you learn the spelling of a new word, an entry is stored on your
memory in the GOL. Every time you write that word, the representation in the GOL will be
activated and make the spelling available to you. So the spelling need no longer to be
assembled from the sound of the word.
The representation of information in the graphemic output lexicon
Whatever is represented in and retrieved from the GOL is presumably some abstract
“graphemic” description of a letter sequence which can be outputted in each of the different
ways of spelling (Capitals, lower case, cursive). Neuropsychological support can be found for
the proposal that what is retrieved from the GOL is an abstract graphemic code rather than a
motor program for letter execution.
Retrieval from the GOL
Patients JS and RD were mentioned as cases whose written naming was better than their
spoken naming. Nevertheless, spelling was not perfectly preserved: Both produced errors
which were interpreted as being based on partial but incomplete knowledge of the correct
spellings. The errors demonstrate that there’s still substantial knowledge of the spelling (for
example words like scissors and pyramid: knowing that there is a C or a Y even though you
cannot hear them).
The distinctness of the GOL and the SOL
One can imagine a possible model of word production in which the written and spoken
forms of words are two outputs of a single internal lexicon. If this was true, then a patient
who experiences word-finding problems in speech should also experience problems in
spelling. This is however not always the case. That’s why the GOL and SOL can be seen as
disconnected and separate components of the language system.
The nature of the input to the graphemic word production system
How are the entries in the GOL accessed? What inputs activate them?
One possibility is that they receive their activation from the corresponding units in the SOL.
This proposal is attractive for at least three reasons:
1. When writing we are usually aware of an “inner voice” saying the words as we write
them.
2. A form of involuntary slip of the pen, is unintentionally writing a word which has the
same sound or a similar sound, to the desired target word. (E.g.: scene instead of
seen, their instead of there etc.)
Chapter 7: Spelling and writing
Spelling without sound
Patient EB was totally speechless by a stroke. His speech comprehension and reading
comprehension are both excellent but slow. He also appeared to have no “inner speech”/
He was unable to match rhyming words that were not spelled alike. He was also unable to
match spoken and written non-words.
His writing was highly successful with only a mild and variable agrammatism. He could write
a whole passage without assistance but could not have said any of it.
Patient MH is anomic, with fluent speech but typically anomic. She couldn’t name pictures of
objects or select rhyming pairs, but she could write the names of objects correctly. The
errors she made were not what you would expect from someone who was assembling
spellings on the basis of sounds of words, but the errors were morphologically related or
visually similar. She could repeat non-words well, but not spell them. This test shows that
the correct spellings of many real words were not generated by the process of phonic
mediation (that spelling is presumed to be mediated by the phonic sound forms of words).
Patient JS and RD were neologistic jargonaphasics. When they were asked to say and then
write the names of a set of pictures, they would say a neologism, but write down the word
correctly. If writing were based on inner pronunciation, then the spellings would surely
reproduce their mispronunciations. Yet they do not.
These cases are evidence against an obligatory phonic mediation theory of spelling.
“Phonological” dysgraphia
Patient PR suffered a left hemisphere stroke. His speech comprehension was normal
combined with fluent speech with good word choice and only occasional paraphasic errors.
His reading is rapid and effortless. His spontaneous writing was said to be labored and slow
because of motor problems and formulation problems.
Nevertheless, he wrote most common words correct, dictated to him. Errors were
morphological.
This stood in contrast with his extremely poor writing of invented non-words. He could write
individual letters when given their alphabetical names, but not when their sounds were
given. He could repeat and read the non-words aloud quite well.
When he managed to write a non-word correctly, he used a real word as a mediator.
, The graphemic output lexicon
The spellings of all these patients must have been produced by some process other than
assembly from sound. It would appear that these patients retrieved the spellings of familiair
words from some internal long-term memory store equivalent to the SOL: the graphemic
output lexicon.
The idea is that each time you learn the spelling of a new word, an entry is stored on your
memory in the GOL. Every time you write that word, the representation in the GOL will be
activated and make the spelling available to you. So the spelling need no longer to be
assembled from the sound of the word.
The representation of information in the graphemic output lexicon
Whatever is represented in and retrieved from the GOL is presumably some abstract
“graphemic” description of a letter sequence which can be outputted in each of the different
ways of spelling (Capitals, lower case, cursive). Neuropsychological support can be found for
the proposal that what is retrieved from the GOL is an abstract graphemic code rather than a
motor program for letter execution.
Retrieval from the GOL
Patients JS and RD were mentioned as cases whose written naming was better than their
spoken naming. Nevertheless, spelling was not perfectly preserved: Both produced errors
which were interpreted as being based on partial but incomplete knowledge of the correct
spellings. The errors demonstrate that there’s still substantial knowledge of the spelling (for
example words like scissors and pyramid: knowing that there is a C or a Y even though you
cannot hear them).
The distinctness of the GOL and the SOL
One can imagine a possible model of word production in which the written and spoken
forms of words are two outputs of a single internal lexicon. If this was true, then a patient
who experiences word-finding problems in speech should also experience problems in
spelling. This is however not always the case. That’s why the GOL and SOL can be seen as
disconnected and separate components of the language system.
The nature of the input to the graphemic word production system
How are the entries in the GOL accessed? What inputs activate them?
One possibility is that they receive their activation from the corresponding units in the SOL.
This proposal is attractive for at least three reasons:
1. When writing we are usually aware of an “inner voice” saying the words as we write
them.
2. A form of involuntary slip of the pen, is unintentionally writing a word which has the
same sound or a similar sound, to the desired target word. (E.g.: scene instead of
seen, their instead of there etc.)
