Teaching

cognitive skills develop independently of language, and so

cannot be transferred to the L2. Rather, readers of L2
who comprehend well are able to do this because they
phonology for have reached a point where they can access, from the L2,
their already existing skill in building mental structures. I

reading have demonstrated this in Walter (2004), where lower-

intermediate L2 learners had no difficulty either with

comprehension decoding individual sentences in a text, or with the ideas in

the text, but had difficulty in building mental structures
corresponding to the whole text when they read it in L2.
(An interesting corroboration of this work is the functional
Catherine Walter magnetic resonance imaging (fMRI) study of Robertson,
Gernsbacher, Guidotti, Robertson, Irwin, Mock and
My research has shown that intermediate second Campana (2000), where sentence-level comprehension
language (L2) learners can understand each was associated with increased activity in the left frontal
sentence of a text without being able to build a lobe and text-level comprehension was associated with
mental representation of the overall text. This increased activity in the right frontal lobe.)
failure is linked to the development of L2
processing abilities. My latest study (2008) Recognising main points is a by-product
shows that a factor in poor L2 processing of
written text is, counter-intuitively, an unreliable Gernsbacher’s (1997) Structure Building Framework has
L2 phonological inventory. This suggests been extensively elaborated and tested over the past
strongly that teaching phonology will help L2 twenty-five years. In common with most accounts of
learners read better. comprehension, the Structure Building Framework
involves the building of a mental representation (structure)
corresponding to the heard/read/seen narrative or
The received wisdom about L2 exposition. Good comprehenders build highly hierarchical
reading comprehension is wrong structures, where nodes corresponding to new ideas link
to already-existing nodes. Whenever a new node is added
Some things that second language (L2) teachers are told to the structure, it activates the more fundamental nodes in
about reading comprehension are: (1) reading the part of the structure that it links to. This means that
comprehension skills are transferred from the first the most fundamental nodes in a good hierarchical
language (L1) to L2; (2) learners need to be taught to structure are activated the most often. Because of this,
recognise the main points in L2 texts; (3) readers process they retain a higher base activation, and are easier to
written texts visually in order to understand what is in the recall. These most fundamental nodes correspond to the
text and to integrate it with their world knowledge. main ideas in the narrative or exposition. It is the
spreading activation in the mental structure, and not any
None of these three statements is true. consciously learnt skill, that makes the main ideas easily
accessible to good comprehenders.
We don’t transfer comprehension skills
In Walter (2007) I gave a group of lower-intermediate
learners stories to read in L1 (French) and in L2 (English).
Morton Ann Gernsbacher and her colleagues (MacDonald
The stories were initially written in English, carefully tuned
and MacWhinney, 1990; Gernsbacher, 1990, 1997;
to the proficiency level of the lower-intermediate learners
Gernsbacher, Varner and Faust, 1990) have provided
both for grammar and for vocabulary, and then translated
evidence that comprehension – the building of a mental
into French. All learners read all the stories, half in L1 and
structure corresponding to a narrative or expository
half in L2, with the language of stories counterbalanced so
account – involves the same cognitive mechanisms and
that half the learners read story A in L1 and half in L2, and
processes whether this comprehension is based on
so on. The catch was that all but one of the stories
listening, reading, looking at picture stories or watching
contained contradictions, and the learners’ task was to
silent films. Note that the last two input modes do not
detect when a statement in a story contradicted an earlier
necessarily involve language at all. In other words,
statement in the same story. The learners did well in
comprehension is not in the L1. Comprehension skills
detecting main-point contradictions in L1, and slightly less
develop at the same time as the L1, but these general

__________________________________________________________________________________
1
IATEFL Pronunciation Special Interest Group Newsletter Issue 40
well in detecting subsidiary-point contradictions, just as won’t be enough left for the other tools to operate properly.
predicted by the Structure Building Framework. In L2, The technical term for this mental workbench is working
they detected contradictions much less well, and there memory. I have begun by calling it a mental workbench
was no significant difference between their detection of because working memory evokes ideas of storage,
main-point contradictions and their detection of subsidiary- whereas storage is only one – and arguably not the most
point ones. This was not because the stories were important – of the functions of working memory.
conceptually difficult: the same stories were being read in
both languages across the group. It was not because Working memory, the phonological
there was difficulty in decoding individual sentences: this
had been very carefully controlled for. The learners were
loop, and L1 reading
good comprehenders: they demonstrated this in L1. What
happened is exactly what is predicted if something was The model of working memory that I use is the Multi-
preventing them from accessing their structure building Component Model of Baddeley and his colleagues
ability. Learners don’t need to be taught to recognise main (Baddeley and Logie, 1999). In this model, working
points: what we need to do is to find out what is memory consists of a central executive which does what it
preventing them from accessing their structure building sounds like: it apportions capacity and organises the
ability when they read in L2, and to do something about it. activity of the working memory. In addition to this, there
are two slave systems: (1) the phonological loop, which
stores information in phonological form and rehearses that
Visual processing is only temporary information by unconscious vocalisation; (2) the visuo-
spatial sketchpad, which performs analogous functions for
In fact, and strange as it may seem, readers of languages visual information. As I explained above, it is in the
with alphabetic writing systems store the most recently phonological loop that readers of alphabetic languages
read material (about as much as the reader can say in two store the immediate products of their decoding of written
seconds), not visually, but phonologically (Baddeley, material. The phonological loop is widely accepted in
Eldridge and Lewis, 1981; Waters, Caplan and Hildebrandt, working memory research, even by authors whose models
1987). Note that I am not talking here about how written of working memory are very different from Baddeley and
words are decoded (visually, probably from multiple cues his colleagues’.
at letter, letter sequence, letter cluster and word level), but
about how these words are temporarily stored. The clear
evidence here is that the visual trace disappears in favour
The phonological loop and L2
of the phonological product. reading

The ‘mental workbench’ In my earlier work (2004, 2007), I found that for lower-
intermediate L2 readers who were poor comprehenders,
In order to explain what follows, I need to introduce a L2-based verbal working memory capacity correlated with
construct that is well documented in cognitive psychology. their ability to comprehend L2 texts. In the mental
Humans have a system of cognitive abilities that I invite workbench metaphor, these readers were using too much
you to think of as a mental workbench. We use this power for one of the tools on the workbench, or putting too
workbench to access, process and store the information much stuff on the workbench at once, and this affected
we need for complex cognitive tasks. Doctors use it when their ability to build reliable mental structures
they are listening to patients’ accounts of symptoms and corresponding to the texts they were reading. There was
trying to work out diagnoses; chess players use it to think no correlation between working memory capacity and
of all the possible moves based on all the possible moves comprehension ability for a counterpart upper-intermediate,
(and so on, recursively); waiters who don’t write down good comprehender group: this suggests that the good
orders use it to process orders, respond to questions, offer comprehenders had enough L2-based verbal working
suggestions and store results as they go along. memory capacity to handle the L2 texts, and more
Comprehenders use it to comprehend texts. capacity didn’t make them better comprehenders.

Think of the mental workbench as having limited capacity: I wondered if one of the problems that the poor
if you put too much stuff on it, some of the stuff will fall off. comprehenders were encountering was in phonological
It also has tools for processing, accessing and storing storage: suppose that representations of just-decoded
information; but it has only a limited amount of power, so material were unreliable, because the learners didn’t have
that if you are using too much power for one tool there a fully elaborated phonological inventory of the L2.

__________________________________________________________________________________
2
IATEFL Pronunciation Special Interest Group Newsletter Issue 40
Material in the phonological loop is subject to decay, and if There was no significant difference between the L1
the elements (sounds, or words, or onsets and rimes, or performances of the two groups; each performed near
whatever units the reader is using) are not linked quickly to ceiling for the dissimilar sequences, and modestly but
long-term representations (phonemes, however specified, significantly less well for the similar sequences. In L2
or exemplars, or lexemes, for example), decay may make English, the good L2 comprehender group performed
it difficult to associate the sounds with meanings. It is slightly less well, but in the same way as in L1: better for
known that this happens in L1 (Boada and Pennington, dissimilar than for similar sequences. However, the poor
2006; Elbro, 1996; Elbro and Jensen, 2005; Griffiths and L2 comprehender group performed significantly less well
Snowling, 2002; Swan and Goswami, 1997). If in L2 than the good L2 comprehender group, and this was
phonemes are not well distinguished from one another, a especially so for the similar sequences, where on average
small amount of decay in the phonological loop might they recalled fewer than five of the ten four-word
make it difficult or impossible to link the phonological forms sequences accurately.
in the loop to the words in the L2 mental lexicon. This
doesn’t mean that the L2 reader has to have a native-like A follow-up study excluded the possibility that the effects
realisation of each L2 phoneme; only that s/he has to have were due to the poor comprehenders decoding English
a distinctive mental representation for each phoneme, so grapheme-phoneme correspondences as if they were
that s/he can distinguish words from one another when French.
they differ by a phoneme. If this is not possible, basing
comprehension on these words becomes difficult. This meant that the poor comprehenders were performing
just as predicted if their phonological representations,
The study even their representations of English phonemes that
French learners do not usually confuse, were weaker or
less detailed than those of the Good L2 Comprehender
I examined whether lower-intermediate learners of L2 who
group and the L1 English group.
were poor comprehenders of L2 written texts differed from
upper-intermediate learners who were good
Here is my conclusion from Walter (2008:469):
comprehenders of L2 written text, in having poorly
distinguished representations of the L2 phonological It is clear that in the L2 recall task there is a
system. This study is described in detail in Walter (2008).
heightened effect of similarity for the Poor L2
I used a methodology that was first used by Baddeley in
Comprehender group. What is not clear is just
1966. Baddeley had L1 speakers listen to sequences of what mechanism is at work. One possibility is
words and try to reproduce them. He found that the that storage and/or recall may have depended
participants had substantial problems in remembering to some degree on phonological features,
words that were phonologically similar. rather than phonemes. […]
I worked with the same two groups of learners as in my
Another possibility would involve the process
earlier studies, with L1 French and L2 English. The task
was to read sequences of four words that were either
that Brown and Hulme (1995… call
phonologically similar (e.g., white, wine, wet, met) or ‘redintegration’, whereby portions of an
phonologically dissimilar (e.g., job, fine, yes, soup), and activated lexical representation are combined
then to try and recall each sequence in order. The with the incomplete phonological records of
participants did the recall task in L1 and in L2. (For the L2 another word. Perhaps participants recalled a
similar sequences, I chose English vowels that have near- portion of a word in a sequence, and
equivalences in French, and that are not normally
unsuccessfully attempted redintegration only
considered confusable for French learners of English.) My
hypothesis was that the poor comprehenders might
to arrive at an erroneous result.
struggle with immediate recall of words in L2; this is what I
Whatever the detailed explanation, it is clear that
would expect if their phonological repertory of the L2 was
developing a reliable L2 phonological repertoire
poorly elaborated – so that, for example, they had to
contributes significantly to the ability to comprehend L2
remember sounds as features, rather than chunked as
written texts. This has implications for the ways in which
phonemes.
teachers help learners comprehend L2 texts better.

__________________________________________________________________________________
3
IATEFL Pronunciation Special Interest Group Newsletter Issue 40
The implications for teaching Brown, G. D. A. & Hulme, C. (1995). Modelling item length
effects in memory span: no rehearsal needed? Journal of
Memory and Language, 34, pp. 594-621.
Good L1 comprehenders do not need to learn to
comprehend in L2; they already know how to comprehend. Elbro, C. (1996). Early linguistic abilities and reading
development: a review and a hypothesis. Reading and Writing, 8,
pp. 453-485.
Good L1 comprehenders do not need to be taught to
identify main ideas in a text. The only cases where help Elbro, C. & Jensen, M. N. (2005). Quality of phonological
will be needed in this area are where the rhetorical representations, verbal learning, and phoneme awareness in
dyslexic and normal readers. Scandinavian Journal of
conventions in the L2 genre they are reading are different Psychology, 46, pp. 375-384.
from the corresponding rhetorical conventions in L1.
Gernsbacher, M. A. (1990). Language comprehension as
My study suggests very strongly that, for alphabetic structure building. Hillsdale, NJ: Erlbaum.
languages, explicit teaching of L2 phoneme recognition
Gernsbacher, M. A. (1997). Two decades of structure building.
will help L2 learners comprehend L2 texts better. If Discourse Processes, 23, pp. 265-304.
learners get to a point where their mental repertory of L2
phonemes is well elaborated, i.e., where the different Gernsbacher, M. A., Varner, K. R. & Faust, M. (1990).
phonemes in the L2 are well distinguished from one Investigating differences in general comprehension skill. Journal
of Experimental Psychology. Learning, Memory and Cognition,
another, they will be able to decode written L2 text into
16, pp. 430-445.
well-differentiated words for temporary storage in the
phonological loop. They can then use their working Griffiths, Y. M. & Snowling, M. J. (2002) Predictors of exception
memory to build mental structures representing L2 texts, word and nonword reading in dyslexic children: the severity
similar to the structures they build based on L1 texts. hypothesis. Journal of Educational Psychology, 94, pp. 34-43.

MacDonald, M. C. & MacWhinney, B. (1990). Measuring

This implies that the sorts of activities that pronunciation- inhibition and facilitation effects from pronouns. Journal of
conscious teachers have been using for decades will be Memory and Language, 29, pp. 469-492
useful for developing L2 reading comprehension skill:
activities like minimal pair recognition activities, activities Robertson, D. A., Gernsbacher, M. A., Guidotti, S.J.,
Robertson, R. R. W., Irwin, W., Mock, B. J. & Campana, M. A.
for recognising stress patterns and dictations of sentences (2000). Functional neuroanatomy of the cognitive process of
containing confusable words. Once learners can mapping during discourse comprehension. Psychological
distinguish L2 phonemes reliably, all spoken and written Science, 11, pp. 255-260.
L2 input will help reinforce their L2 phonemic repertory.
Swan, D. & Goswami, U. (1997). Phonological awareness
Not only will they become better listeners, but they will
deficits in developmental dyslexia and the phonological
also be better comprehenders of L2 written text. representations hypothesis. Journal of Experimental Child
Psychology, 66, pp. 18-41.
Catherine Walter ([email protected]) is a Senior
Walter, C. (2004). Transfer of reading comprehension skills to
Lecturer at the Institute of Education, University of London. L2 is linked to mental representations of text and to L2 working
st
From 1 May 2009, she will be joining the Education memory. Applied Linguistics, 25, pp. 315-339.
Department at the University of Oxford as a Lecturer in
Applied Linguistics ([email protected]). Walter, C. (2007). First- to second-language reading
comprehension: not transfer, but access. International Journal of
Applied Linguistics, 17, pp. 14-37.
References
Walter, C. (2008). Phonology in second language reading: not
an optional extra. TESOL Quarterly, 42/3, pp. 455-474.
Baddeley, A. D. & Logie, R. H. (1999). Working memory: the
multiple-component model. In Models of working memory:
mechanisms of active maintenance and executive control. A. Waters, G. S., Caplan, D. & Hildebrandt, N. (1987) Working
Miyake & P. Shah (Eds.), Cambridge: Cambridge University memory and written sentence comprehension. In Attention and
Press, pp. 28-61. performance: vol. 12. The psychology of reading. M. Coltheart
(Ed.), Hove, England: Erlbaum, pp. 531-555.
Baddeley, A., Eldridge, M. & Lewis, V. (1981). The role of
subvocalisation in reading. Quarterly Journal of Experimental
Psychology A, 33, pp. 439-454.
© Catherine Walter 2009. This article may be reproduced freely,
Boada, R. & Pennington, B. F. (2006). Deficient phonological as long as the author is informed in writing and in advance of
representations in children with dyslexia. Journal of Experimental publication.
Child Psychology, 95, pp, 153-193.

__________________________________________________________________________________
4
IATEFL Pronunciation Special Interest Group Newsletter Issue 40