Content uploaded by Lynne G. Duncan
Author content
All content in this area was uploaded by Lynne G. Duncan on Mar 18, 2015
Content may be subject to copyright.
Differing sequences of metaphonological
development in French and English*
LYNNE G. DUNCAN
University of Dundee,United Kingdom
PASCALE COLE
´
Universite´ de Savoie,Laboratoire de Psychologie et de Neurocognition,
UMR 5105, CNRS,France
P H I L I P H. K. S E Y M O U R
University of Dundee,United Kingdom
AND
ANNIE MAGNAN
Universite´ Lumie
`re,Lyon 2, Laboratoire d’Etude des Me´canismes Cognitifs/
Laboratoire Dynamique du Langage (DDL), UMR 5596, CNRS,France
(Received 30 August 2004. Revised 8August 2005)
ABSTRACT
Phonological awareness is thought to become increasingly analytic
during early childhood. This study examines whether the proposed
developmental sequence (syllable ponset-rime pphoneme) varies
according to the characteristics of a child’s native language. Experiment 1
compares the phonological segmentation skills of English speakers aged
4;11 (N=10), 5;3 (N=21), and 6;5 (N=23) and French speakers aged
5;6 (N=35), and 6 ; 8 (N=34). Experiment 2 assesses performance in
the common unit task using English speakers aged 4 ;7 (N=22), 5;7
(N=23), and 6;11 (N=22), and French speakers aged 4 ; 7 (N=20),
5;6 (N=35), and 6 ; 7 (N=33). The experiments reveal crosslinguistic
differences in the processing of syllables prior to school entry with
[*] This research was supported in the UK by a project grant from the Economic and Social
Research Council (R000222321). The authors would like to express their gratitude to
the children and schools who participated in the study, and to Elena Lieven and two
anonymous reviewers for their valuable comments. Address for correspondence: Lynne
G. Duncan, Department of Psychology, University of Dundee, Dundee DD1 4HN,
Scotland, UK. tel : +44 (0)1382 344629; fax : +44 (0)1382 229993; e-mail : l.g.duncan@
dundee.ac.uk
J. Child Lang. 33 (2006), 369–399. f2006 Cambridge University Press
doi:10.1017/S030500090600732X Printed in the United Kingdom
369
French speakers exhibiting a greater consistency in manipulating
syllables. Phoneme awareness emerges in both languages once reading
instruction is introduced and rime awareness appears to follow rather
than precede this event. Thus, the emergence of phonological awareness
did not show a universal pattern but rather was subject to the influence
of both native language and literacy.
INTRODUCTION
Linguistic development is shaped by the sound structure of a child’s native
language. Although children are born with the capacity to discriminate the
full range of possible phonetic contrasts, this ability soon diminishes as
the particular contrasts of the native language assume precedence (Werker,
Gilbert, Humphrey & Tees, 1981). By the end of the first year, infants have
amassed a functional knowledge of their native language which encompasses
the set of phonetic features, phonotactic constraints on ordering, and the
prosody of familiar words (Jusczyk, Friederici, Wessels, Svenkerud &
Jusczyk, 1993).
This early accommodation to the language environment is relevant to
comparisons of metalinguistic development in different languages. One
implication is that preschool children will be highly adapted to the processing
of their native language and its particular characteristics. Although this
strongly suggests that metaphonological development is likely to differ
between languages, existing models have tended to promote a universal
developmental pathway that can be characterised as a LARGE-TO-SMALL unit
(syllable ponset-rime pphoneme) progression in awareness of sound
(Treiman, 1987; Metsala & Walley, 1998).
Metsala & Walley (1998) argue that phonological awareness reflects
the changing status of lexical representations. These are initially holistic
but later become increasingly segmental as vocabulary expands. This
restructuring results from the mounting pressure to discriminate between
phonologically similar words and occurs, not in an all-or-none manner, but
for specific words as the need arises. Consequently, the phonological skills
of manipulating and isolating segments of sound are thought to emerge only
when a level of segmental organization of the linguistic system is sufficiently
well-established. This development is held to occur in the fixed sequence
of syllables, then onsets and rimes, and finally, phonemes.
We take the view that the generality of this large-to-small sequence has
been overstated and that the course of metaphonological development may
vary as a consequence of linguistic and educational influences. This view is
based on a model proposed by Gombert (1992) which allows for alternative
pathways in phonological development. The acquisition of FIRST LINGUISTIC
SKILLS shapes subsequent development through the influence of native
DUNCAN ET AL.
370
language on the early organization of the linguistic system. This plasticity is
maintained in three subsequent phases. The first is the EPILINGUISTIC phase
during which language processes are internally organized in an implicit
format that is inaccessible to consciousness. Gombert (1992) follows
Karmiloff-Smith (1986) in proposing that development during the
epilinguistic phase is a consequence of receptive and expressive interaction
with the linguistic environment. Karmiloff-Smith refers to a process of
REPRESENTATIONAL REDESCRIPTION whereby existing pieces of independently
stored information acquired during the earlier phase are subject to an
internally-driven process of organization in order to improve accessibility
and to increase generalization within the linguistic system. Organizational
adjustments continue in the latter phases of the model, the METALINGUISTIC
phase and the AUTOMATIZATION phase, although, by this point, change is
wholly driven by the child’s external environment. One very important
influence of direct relevance to the present study is the external demand
imposed by learning to read.
Several previous studies of phonological development in English support
Gombert’s (1992) suggestion that this succession of levels of awareness
occurs separately for different sounds (syllables, rimes, phonemes) and
does not necessarily follow a large-to-small sequence (Seymour & Evans,
1994; Duncan, Seymour & Hill, 1997, 2000). In the study by Duncan
et al. (1997), Primary 1 children who had displayed excellent rhyming
skills in an epilinguistic rime oddity task, nonetheless performed much
better with small units (phonemes) than with larger units (rimes) in the
common unit task, a task which requires the production of the sound shared
by two words, and, hence, tests metalinguistic skills. While incompatible
with a simple large-to-small progression, this corresponds with Gombert’s
proposal that phonological awareness takes epi-and metalinguistic forms
and can progress at different rates for each unit of sound according to the
external demands of the linguistic environment. Even though language
development in English seems to foster an epilinguistic sensitivity to rime,
this does not in itself bring about the ability to manipulate rimes in meta-
linguistic tasks. Progression to the metaphonological level of awareness
requires the influence of an external demand. In the case of Duncan et al.’s
findings, the external factor that increased the children’s ability to
manipulate phonemes explicitly appeared to be phonics-based literacy
instruction.
Findings consistent with Gombert’s (1992) model have also been reported
from crosslinguistic comparisons of metaphonological development in
English versus Czech (Caravolas & Bruck, 1993) and Turkish (Durgunog
˘lu
&O
¨ney, 1999) where variation in skill can be traced to contrasts in the
nature of the spoken languages such as differences in the complexity of
syllable structure or in the importance of vowel harmony.
METAPHONOLOGY IN FRENCH AND ENGLISH
371
The present research focuses on metaphonological development in
English and French with the aim of relating phonological skills to the
linguistic and literacy-related contexts which are held to shape development
in Gombert’s (1992) model. Special attention will be given to early aware-
ness of syllables because existing models make contrasting predictions
about the course of phonological development for these units of sound.
According to large-to-small theories, syllables are universally the most
salient sounds for young children, developing prior to formal schooling
and before awareness of smaller units such as phonemes (Liberman,
Shankweiler, Fischer & Carter, 1974; Alegria, Pignot & Morais, 1982).
However, linguistic differences exist between English and French in
relation to spoken syllables, which, according to Gombert (1992), might be
expected to promote contrasting levels of syllable awareness. The nature of
these differences will be discussed in the next section.
Syllables in English and French
Early speech perception appears sensitive to rhythmic differences between
French and English (Nazzi, Bertoncini & Mehler, 1998). In explanation,
French has often been described as a SYLLABLE-TIMED language in which
syllables are isochronous, and English as a STRESS-TIMED language in which
the intervals between stressed syllables are isochronous (Abercrombie,
1967). However, the basis of this proposition that there are regularities
in the timing of the basic units of perception for each rhythm class has
not received experimental support (Roach, 1982). More recent work has
suggested instead that speech rhythm is associated with a number of
phonological processes that together determine the rhythmic characteristics
of any language (Ramus, Nespor & Mehler, 1999). In what follows, the
contrast in rhythm between French and English is discussed in relation to
three important parameters:
(a)Syllable structure. French and English differ in terms of the number
and distribution of syllable types. In English, there are 16 syllable structures
(Abercrombie, 1967), whereas only 8 syllable types occur frequently in
French (Wioland, 1985). CVC and CV structures occur in approximately
the same proportion in English although closed syllables predominate
overall (56%, Dauer (1983)). In French, syllables tend to be open (80%,
Wioland (1985)) and linguistic phenomena such as LIAISON and
ENCHAI
ˆNEMENT serve to enhance CV structure.
(b)Prosody. Lexical stress has a contrastive function in the English
language with each word having a characteristic pattern of stress (Giegerich,
1992). French can be described as a fixed stress language in which the final
syllable of words spoken in isolation is stressed and the stress pattern of a word
can vary in connected speech without affecting meaning (Schane, 1968).
DUNCAN ET AL.
372
(c)Vowel reduction. An associated feature of the variable stress pattern
in English is the tendency to reduce vowels in unstressed syllables. This
phenomenon does not occur in French (Dauer, 1983).
Recent systems devised to describe speech rhythm have taken such factors
as their starting point (Ramus et al., 1999 ; Grabe & Low, 2002). The methods
index variation in the duration or periodicity of vowels and consonants
in order to capture the differences between syllable- and stress-timed
languages. According to each of these systems, the rhythms of French and
English are distinct. This, together with evidence from the studies reviewed
earlier, underlines the hypothesis that the rhythmic cues to word segmen-
tation, fundamental to the language acquisition process, are different for
each language.
Syllable awareness in English and French
While syllabification has often been regarded as unproblematic in French
(Wioland, 1985; Cutler, Mehler, Norris & Seguı
´, 1986), syllable boundaries
can be hard to define unambiguously in English since consonants sometimes
appear to be phonologically AMBISYLLABIC (Giegerich, 1992). Factors such
as stress, sonority and vowel length have been shown to play a part in
determining ambisyllabicity for English-speaking adults and children
(Fallows, 1981; Treiman & Zukowski, 1990 ; Treiman, Bowey & Bourassa,
2002).
Mehler, Dommergues, Frauenfelder & Seguı
´(1981) devised an auditory
detection task to investigate whether French- and English-speaking adults
differ in their use of syllables in speech processing. Response facilitation
when the probe and the first syllable of the target were congruent led
Mehler et al. to conclude that French adults divide the speech stream into
syllabic segments. Responses by English-speaking adults suggested reliance
on phonemic rather than syllabic coding. Cutler et al. (1986) argued that
these differences are related to the clarity of syllable boundaries. The fact
that French syllables are clearly bounded promotes the use of these
segments whereas the ambiguity surrounding English syllable boundaries
inhibits the use of syllables.
However, Content, Kearns & Frauenfelder (2001 a) have recently
questioned this conclusion, not least because of a failure to replicate the
Mehler et al. (1981) findings (Content, Meunier, Kearns & Frauenfelder,
2001b). They propose the ONSET HYPOTHESIS that French speakers syllabify
on the basis of syllable onsets rather than syllable boundaries. Content et al.
suggest that syllabifications may vary in French according to whether a task
requires the determination of the onset, the offset or a combination of the
two. Syllabic segmentation tasks require the detection of the offset of
the first syllable and the onset of the following syllable. Content, Dumay &
METAPHONOLOGY IN FRENCH AND ENGLISH
373
Frauenfelder (1999) compared this procedure with a partial repetition task in
which participants are asked to repeat either the first part (offset detection)
or the second part (onset detection) of a bisyllabic word. French children
behaved differently in these two tasks, producing canonical syllabifications
in the segmentation task but more varied syllabification in the partial
repetition task. In explanation of their results, Content et al. proposed that
syllable offset decisions are variable and more likely to be influenced by
factors such as sonority and spelling, whereas decisions about syllable
onsets, especially those in strong syllables, are more robust as these units
form the basis of the speech segmentation process.
To our knowledge this hypothesis has not been investigated cross-
linguistically with children. Existing developmental work by Bruck,
Genesee & Caravolas (1997) to compare syllable processing in French and
English produced a mixed outcome. The French-speaking children were
found to be more accurate in a syllable counting task but did not differ from
the English speakers in a syllable matching task. However, the matching
task, and, to a lesser extent, the counting task can be performed on the
basis of a global apprehension of sound and do not require the isolation
of precisely defined syllable segments. In Gombert’s (1992) terms, they
may be measures of epilinguistic (implicit) awareness rather than of
metalinguistic (explicit) awareness. To address the issue of syllable boundary
location it may be necessary to use tasks that are clearly metalinguistic in
their demand for the actual identification or manipulation of the syllabic
unit.
Thus, the present study uses two metalinguistic tasks : Experiment 1
compares French- and English-speaking children’s syllabification of
bisyllabic words, and Experiment 2 examines common unit identification in
French and English words using syllables, phonemes and rimes as shared
segments. The results will be examined to determine the sequence of
metaphonological development in each language and the outcome will be
related to the influence of both linguistic and educational factors.
EXPERIMENT 1: SYLLABIFICATION IN FRENCH
AND ENGLISH
The segmentation task is used to maximize the chances of uncovering
any discrepancies in syllable boundary location that might derive from the
differences between French and English phonology. The task requires
the segmentation of bisyllabic words into two parts with the expectation
that words will be split into their constituent syllables. The large-to-small
view of metaphonological development predicts that this task should be
equally easy for French- and English-speaking children as awareness of
syllables develops early and is present by the preschool stage. An alternative
DUNCAN ET AL.
374
prediction, based on Gombert’s (1992) account, is that the language groups
will differ in syllable awareness due to differences in the nature of syllables
in French and English.
In the first instance, the experiment will consider the extent to which
syllabifications conform to the MAXIMAL ONSET PRINCIPLE. This principle
states that, within the constraints imposed by the phonotactics of a language,
syllabifications that maximize the onset following the syllable boundary are
preferred (Pulgram, 1970). On the basis of Content et al.’s (2001a) findings,
it is predicted that French syllabification is likely to follow this principle.
English syllabification, which is more prone to ambisyllabicity than French,
should produce fewer maximal onset syllabifications.
Factors that might give rise to departures from the maximal onset principle
in either language will also be examined. Work by Treiman & Zukowski
(1990) indicated that three other principles act alongside the maximal onset
principle in the syllabification of English, namely, legality, sonority and
stress (see also the earlier work on this topic by Fallows, 1981). These
principles are outlined below together with a discussion of how well each
factor is likely to extend to French syllabification.
1. The LEGALITY PRINCIPLE states that each syllable of a word must be
a phonologically legal word of the language (e.g. Pulgram, 1970). In the
present context, this predicts that syllabification will respect the phonotactics
of word onsets and codas in each language. Treiman & Zukowski (1990) have
also suggested that the legality principle may account for the tendency in
English syllabification to close open syllables containing short vowels, since
this type of open syllable cannot be a lexical word in English (Giegerich,
1992). This constraint is not present in French ; moreover, early language
acquisition in French has been shown to be sensitive to the frequent
occurrence of this syllable type (Demuth & Johnson, 2003). Hence, a further
prediction is that the incidence of syllabifications containing open (CV)
syllables with short vowels will differ in French and English.
2. The SONORITY CYCLE PRINCIPLE holds that ‘ the preferred syllable type
shows a sonority profile that rises maximally toward the peak and falls
minimally towards the end’ (Clements, 1990: 301). In the present study,
critical intervocalic consonants will be sonorant (liquids or nasals), which
developmental and adult work suggests should maximize the tendency to
place the intervocalic consonant at the end of the first syllable to create a
gradual decrease in the sonority contour (Treiman & Zukowski, 1990 ;
Treiman et al., 2002). The prediction is that this should favour maximal
onset segmentations for CVC items and increase the tendency to ambi-
syllabicity for CV items in both languages.
3. The STRESS PRINCIPLE, which predicts that stressed vowels will attract
consonants, has received support from studies of adults (Treiman &
Zukowski, 1990) and children (Fallows, 1981). Manipulation of word
METAPHONOLOGY IN FRENCH AND ENGLISH
375
prosody will only be possible in English as French is a fixed (final) stress
language. English vowels that are unstressed are also reduced and this will
constitute a difference between iambic stress items in English and French,
which will be examined in analysing the results. Thus, the final prediction is
that onsets will tend to be maximized with iambic stress but that this
tendency will be reduced with trochaic stress (in English only).
METHOD
Participants
The English-speaking children all attended the same school in a middle-
class area of Dundee in Scotland. They spoke English as their native
language and were progressing normally for their age. The children in the
participating Nursery (N=10), Primary 1 (N=21) and Primary 2 (N=23)
classes had mean chronological ages of 4;11 (S.D.=0;2), 5;3 (S.D.=0 ; 3) and
6;5 (S.D.=0 ; 3), respectively. The British Picture Vocabulary Scale (BPVS)
had been used to establish that receptive vocabulary was normal for
chronological age at a previous testing session
1
and yielded standardised
scores of: Nursery=98 (S.D.=15), Primary 1=103 (S.D.=13), and Primary
2=107 (S.D.=12).
The French children were from two urban schools located in the South
of France (Toulon and Nice). All were predominantly middle-class, spoke
French as their native language and were progressing normally for their age.
As schooling starts a year later in France than in Britain, the French
Nursery group (N=35) was aged 5;6 (S.D.=0;5) and the French Primary 1
group (N=34) was aged 6;8 (S.D.=0;7) at the time of testing. Vocabulary
performances were assessed using the WPPSI-R for the Nursery children
(Mean=10.4, S.D.=1.3) and the WISC-R for the Primary 1 children
(Mean=9.3, S.D.=1.5).
Materials and procedure
In the French investigation there were three conditions that differed
according to the phonological structure of the bisyllabic words (see Appendix
A for all stimuli). The conditions are labelled according to the maximal
onset definition of the first syllable: (1) CV. – a consonant and a vowel (e.g.
ca/rie); (2) CVn. – a consonant plus a nasalized vowel (e.g. can/tine); and (3)
CVC. – a consonant, vowel, consonant sequence (e.g. car/ton).
2
While the
[1] Approximately 4 months before for the Nursery group and 6 months before for the
Primary 1 and 2 groups.
[2] The maximal onset principle is used to label the conditions in order to subject this
hypothesis to examination rather than to imply our adherence to this principle of
syllabification.
DUNCAN ET AL.
376
CVC. items contained multiple intervocalic consonants, the CV. and CVn.
syllable types contained a single intervocalic consonant which was a liquid
in all but one of the CV. syllables and a stop consonant in all of the CVn.
syllables. The first three phonemes of the French words were matched across
the CV. and CVC. conditions and the vowel was short for all but one of the
eight items in each condition.
Differences between the French and English language prompted some
alterations to the experimental design for the English investigation. The
investigation was limited to CV. and CVC. first syllables and stress pattern
was included as a nested factor (trochaic stress, iambic stress). Thus, there
were four experimental conditions each with 8 items: (1) CV. trochaic
stress (e.g. ca/mel); (2) CVC. trochaic stress (e.g. cam/ping); (3) CV. iambic
stress (e.g. ca/reer); and (4) CVC. iambic stress (e.g. car/toon).
The first three phonemes of the English items were matched across the
CV. and CVC. conditions. The vowel was always short and the final C in
this sequence was sonorant (nasal or liquid) with the exception of one item
in each iambic stress condition which contained a fricative in this position.
There was a single intervocalic consonant in CV. items while CVC. items
contained multiple intervocalic consonants.
In each country the task was explained to the children using demon-
stration and practice items. The experimenter read out the words to be
segmented into two parts. The words were bisyllabic and had been selected
from age-appropriate spoken vocabularies in each language. The children
were asked to speak like a robot by saying the words in ‘two little bits’. In
France, the children tapped out the two sounds at the same time as they
articulated them. In Scotland, they pointed to two counters in turn. The
responses given by each child were transcribed in a phonetic code by a
native speaker and recorded for later verification.
The French study was conducted in March when the children had been
in their classes for about six months. The British investigation of the
Nursery children also took place in March, and the Primary school children
were tested 3 months earlier, in November.
RESULTS
Crosslinguistic Comparison
A total accuracy score was calculated which reflected the ability of each
language group to satisfy the instruction to segment the bisyllabic stimuli
into two parts (see Table 1). British scores for the crosslinguistic comparison
were based on iambic stress items in order to equalize stress pattern and the
number of stimuli per condition between the two languages. Accuracy was
generally high with all groups scoring over 85 %, indicating that the task
itself did not present particular difficulties for speakers of either language.
METAPHONOLOGY IN FRENCH AND ENGLISH
377
A comparison of these total accuracy scores for the five- and six-year-olds
revealed that the French children were significantly more accurate despite
having received a year less of schooling than the English-speaking children
(Mann-Whitney U=934, p<0.001).
French language analyses
The only error made by the Nursery group was an inaccurate segmentation
(carafe p/ka/-/re/-/a/-/fe/). Analysis of the small number of errors made by
Primary 1 children showed that they mainly occurred with CVC. bisyllables
and involved the addition of a schwa to produce three open CV syllables
instead of an initial closed CVC syllable. For example, the word marquis
was segmented as /ma/-/r[/-/ki/.
The most striking feature of the French results, however, was the
consistency of their segmentations, which corresponded exclusively to
syllabic units as defined by the maximal onset principle (see Table 2, where
item structure and response structure can be seen to overlap exactly). The
TABLE 1. Mean percentage of segmentations in Experiment 1that comply
with task instructions (i.e. segmentations into any two component parts)
(a) French
Age
Item type
CV. CVC. CVn.
Nursery 5 years Mean 99.63 100.00 100.00
S.D.(2.13) (0.00) (0.00)
Primary 1 6 years Mean 98.13 91.88 97.75
S.D.(4.50) (14.13) (4.75)
(b) English
Age
Item type
Trochaic stress Iambic stress
CV. CVC. CV. CVC.
Nursery 4 years Mean 91.25 91.25 85.00 76.25
S.D.(11.88) (10.25) (11.50) (18.13)
Primary 1 5 years Mean 82.75 86.25 87.50 88.13
S.D.(28.63) (29.88) (22.00) (22.88)
Primary 2 6 years Mean 89.63 88.00 90.75 89.63
S.D.(21.88) (26.00) (12.63) (23.38)
DUNCAN ET AL.
378
TABLE 2. Range of two-part segmentations by French-speaking children in Experiment 1tabulated by item structure
and response structure (mean percentage of total responses for each item type with standard deviations in parentheses)
a
Item structure CV. items CVC. items CVn. items
Response structure C. CV. CVC. CV[C] C. CV. CVC. CVC[C] C. CVn. CVnC. CVn[C]
Five-year-olds 0.00 99.63 0.00 0.00 0.00 0.00 100.00 0.00 0.00 100.00 0.00 0.00
(0.00) (2.13) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00)
Six-year-olds 0.00 98.13 0.00 0.00 0.00 0.00 91.88 0.00 0.00 97.75 0.00 0.00
(0.00) (4.50) (0.00) (0.00) (0.00) (0.00) (14.13) (0.00) (0.00) (4.75) (0.00) (0.00)
a
Responses consistent with the maximal onset principle are highlighted. Note that the figures for each item type do not always sum to 100%
because only accurate two-part segmentations are tabulated.
METAPHONOLOGY IN FRENCH AND ENGLISH
379
pattern of syllabifications was the same in the CV. and CVn. conditions in
spite of the differences in the sonority of the intervocalic consonants in these
conditions. Thus, syllabifications that maximized the onset following the
syllable boundary were preferred subject to the constraints of French
phonotactics (e.g. tarif p/ta/-/rif/ ; tardif p/tar/-/dif/).
English language analyses
In contrast, the British results were marked by the diverse pattern of
segmentations given by the English-speaking children. Table 3 contains a
complete listing of the different types of correct (i.e. 2-part) responses.
Responses were classified according to where the division was placed: (1)
C. – after the first consonant of the word (e.g. /k/-/æmel/) ; (2) CV. – after
the first CV unit (e.g. /kæ/-/mel/) ; (3) CVC. – after the first CVC unit
(e.g. /kæm/-/el/); or (4) CV[C]. – an ambisyllabic response (e.g. /kæm/-
/mel/). An additional category applied to items with a CVC structure only :
(5) CVCC. – division such that the medial consonants were placed in the
first syllable (e.g. /gærd/-/en/).
Due to the presence of zero cells a complete analysis of response structure
could not be undertaken. Instead, the incidence of maximal onset division
was compared for CV. and CVC. items (the shaded columns in Table 3)
using a three-way ANOVA: chronological age (4 years, 5 years, 6 years)r
item structure (CV., CVC.)ritem stress (trochaic, iambic). The main
effects of item stress and item structure were significant (F(1, 51)=21.85,
p<0.001 and F(1, 51)=31.94, p<0.001, respectively) as was the interaction,
item stress by item structure (F(1, 51)=18.01, p<0.001). None of the
effects involving chronological age were significant indicating that response
pattern was stable across age levels.
A Tukey HSD test (a=0.05) revealed that CVC. items were significantly
more likely to receive a maximal onset division than CV. items, and that this
tendency was unaffected by lexical stress. For CV. items, however, stress
did have a significant effect, with trochaic stress resulting in significantly
fewer maximal onset divisions than iambic stress. Inspection of Table 3
suggests that this result reflects the greater variation in response to trochaic
stress CV. items including many more CVC. and CV[C]. (ambisyllabic)
responses than in the iambic stress condition.
An analysis of the English trochaic stress CV. items examined the effect of
sonority on the segmentation of these items. Half of these items contained
an intervocalic liquid (r) and the other half contained an intervocalic nasal
(n or m). Although both types of consonants are sonorous, liquids are more
sonorous than nasals, creating the expectation that the liquid intervocalic
consonants might be segmented as part of the initial syllable (with or
without ambisyllabicity) more often than the nasals. In the event, this
DUNCAN ET AL.
380
TABLE 3. Range of two-part segmentations by English-speaking children in Experiment 1tabulated by item structure and
stress pattern, and response structure (mean percentage of total responses for each item type with standard deviations in
parentheses)
a
Item structure CV. items CVC. items
Response structure C. CV. CVC. CV[C] C. CV. CVC. CVC[C] CVCC.
Four-year-olds
Trochaic stress 6.25 32.50 27.50 25.00 0.00 0.00 73.75 3.75 16.25
(19.75) (17.88) (29.88) (17.63) (0.00) (0.00) (18.13) (8.38) (18.63)
Iambic stress 0.00 66.25 11.25 8.75 0.00 3.75 72.50 0.00 1.25
(0.00) (17.75) (7.13) (10.25) (0.00) (6.00) (18.50) (0.00) (4.00)
Five-year-olds
Trochaic stress 3.63 42.25 24.38 12.50 4.13 0.63 79.75 0.00 1.75
(16.38) (33.88) (30.00) (22.38) (19.13) (2.75) (34.63) (0.00) (6.00)
Iambic stress 0.00 70.88 11.25 5.38 0.63 0.63 86.88 0.00 0.00
(0.00) (23.50) (7.75) (7.50) (2.75) (2.75) (23.88) (0.00) (0.00)
Six-year-olds
Trochaic stress 15.25 33.13 21.75 19.63 13.00 0.00 73.88 0.00 1.13
(33.13) (38.13) (31.63) (29.13) (34.38) (0.00) (38.75) (0.00) (3.63)
Iambic stress 1.13 56.00 19.63 14.13 6.50 1.63 81.00 0.00 0.50
(3.63) (35.13) (23.75) (21.75) (17.63) (4.25) (31.00) (0.00) (2.63)
a
Responses consistent with the maximal onset principle are highlighted. Note that the figures for each item type do not sum to 100% because
only accurate two-part segmentations are tabulated.
METAPHONOLOGY IN FRENCH AND ENGLISH
381
analysis failed to reveal any interaction between sonority and segmentation
response type at any age level (F<1).
DISCUSSION
Experiment 1 tested the possibility that differences in the salience of
syllables in English and French would influence the course of early
metaphonological development. This was approached using the task of
two-part segmentation of bisyllabic words. Our results indicated that : (1)
French-speaking children were more accurate overall than English-speaking
children, and (2) the patterns of performance differed markedly between
French and English. Whenever the French children were successful in
splitting the word into two parts, they invariably produced maximal onset
syllables. This pattern was in striking contrast to the wide range of
responses that were produced by the English-speaking children, especially
for CV. items with trochaic stress. To take just one example, the responses
/k/-/æm[l/,/kæ/-/m[l/,/kæm/-/m[l/ and /kæm/-/[l/ were all offered as
segmentations of the English word camel.
Treiman & Zukowski (1990) were able to account for departures from
the maximal onset principle in English syllabification by invoking alterna-
tive principles of legality, sonority and stress. Our crosslinguistic study
provides a context in which to evaluate how well these principles apply to
syllabification in a language other than English, which may in turn offer
new insight into the syllabification process itself.
Children in both countries were found to respect the legality principle
and only produced syllable types that were legal according to the phono-
tactics of their own language. The only exception to this was the production
of initial consonant segmentations (e.g. camel p/k/-/æmel/) by the British
children. It seems likely that these responses reflected the influence of
literacy instruction as their incidence was very low before the second year
of schooling. The French six-year-olds who were still in their first year of
schooling did not attempt this type of segmentation.
The legality principle may also explain the striking crosslinguistic
differences in the syllabification of CV. items. English speakers tended to
use the intervocalic consonant to close the first syllable of CV. items. This
occurred mainly for items with trochaic stress but was also observed with
iambic stress, and this link with prosody will be discussed below. The
French speakers, in contrast, never produced this type of response, preferring
instead to leave the first syllable of CV. items open. The legality principle
predicts this result as CV syllables with short vowels are permissible lexical
words in French but not in English.
Treiman & Zukowski (1990) offered another explanation of this finding
in terms of the sonority contour of syllables, arguing that participants may
DUNCAN ET AL.
382
be attempting to produce a gradual decline in sonority at the end of the
syllable contour. However this suggestion receives little support from our
data as the French children showed no inclination to close the first syllables
of CV. items in spite of the presence of sonorant intervocalic consonants.
Thus, the most parsimonious explanation of this crosslinguistic difference
in syllabification seems to be that it arises from the different phonotactics of
French and English.
In fact, the effects of sonority were generally weak in the present study.
Within the trochaic stress CV. condition, sonority did not seem to be
responsible for any significant variation in the incidence of ambisyllabicity
amongst these young English speakers as response patterns were the same
regardless of whether the intervocalic consonant was a nasal or a more
sonorant liquid. It is possible that the difference between nasals and liquids
may be too subtle a contrast in sonority to affect the performance of this age
group. Treiman et al. (2002) reported emerging effects of sonority amongst
eight-year-olds, which increased in strength with chronological age.
Content et al. (1999) demonstrated that the influence of sonority also
increases with age in French. While five- and nine-year-olds showed no
effects of sonority in a segmentation task, twelve-year-olds produced fewer
maximal onset segmentations for words with an intervocalic liquid (74 %)
than for those with a nasal (93%) or with a stop consonant (94%). In the
present study, the French five- and six-year-olds showed no differences
in maximal onset divisions between CV. items containing an intervocalic
liquid and CVn. items containing an intervocalic stop consonant.
The third and final principle, the stress principle, states that stressed
syllables will attract consonants. This principle explains the British
children’s tendency to segment intervocalic consonants as part of the first
syllable with trochaic stress, and as part of the second syllable with iambic
stress. Nevertheless, word prosody made no difference to their preference
for maximal onset divisions of CVC. items. Furthermore, crosslinguistic
differences were observed even on items that had been matched for (iambic)
stress pattern. Although maximal onset segmentations were dominant with
iambic stress, the incidence of other responses varied according to item
structure and to language group. A similar degree of preference for maximal
onset responses was apparent in both language groups for CVC. items but
the CV. items still elicited a greater variety of response from the English
than from the French group. It seems possible that this difference may
be related to vowel reduction in some way. One speculative account is that
the legality constraint against open syllables with short vowels also applies
to unstressed syllables in English, but only weakly because of the high
frequency precedent of monosyllabic function words like the which often
occur in an unstressed form and which end in a reduced (schwa) vowel
(Giegerich, 1992).
METAPHONOLOGY IN FRENCH AND ENGLISH
383
In summary, the results of Experiment 1 indicate that the early phase
of metaphonological development differs markedly in French and English.
There is consensus amongst the French children as to the location of
syllable boundaries in French, whereas English-speaking children syllabify
words in a variety of ways in accordance with the ambiguous nature of
syllable boundaries in English. The findings suggest that syllabification
depends strongly upon the speech rhythm and phonotactics of the two
languages with factors such as sonority having a much weaker effect in
the early stages of schooling. This is consistent with Gombert’s (1992)
emphasis on the importance of early linguistic skills in shaping the sequence
of metaphonological development rather than with the proposal that there
is a universal large-to-small unit progression.
EXPERIMENT 2: THE SEQUENCE OF
METAPHONOLOGICAL DEVELOPMENT IN
ENGLISH AND FRENCH
The results of Experiment 1 revealed a complete absence of variability
amongst the French-speaking children. However, as already noted, this
outcome may have been task dependent. In a task where five-year-old
French speakers had to repeat only the first syllable from bisyllabic stimuli,
Content et al. (1999) observed a varied pattern of response even though the
behaviour of this group in a segmentation task bore striking similarities to
our results in Experiment 1.
An implication is that the French-English contrast in syllable awareness
may be dependent on the form of the task employed. Therefore, in the
second experiment, we aim to replicate our observations of crosslinguistic
differences in syllable processing using a different task, the common unit
task (Duncan et al., 1997, 2000). This requires the identification of the
shared syllable at the beginning of two spoken words. Within the terms of
Content et al.’s (2001a) hypothesis, this task involves first syllable offset
detection and, therefore, should produce a greater variation in French
syllabifications.
We also aim to explore further the impact of literacy acquisition on the
progression of metaphonological development. In contrast to the internally-
driven, large-to-small sequence proposed by Metsala & Walley (1998),
Gombert (1992) maps out a series of transitions between an epi- and
metalinguistic level of representation that occur separately for different
sounds (syllables, rimes, phonemes). A key suggestion is that the emergence
of the metalinguistic level is dependent on external influences, the most
important of which appears to be the onset of literacy. However, the exact
nature of this influence seems likely to vary between cultures due to
DUNCAN ET AL.
384
differing orthographies, methods of teaching and the age at which literacy is
introduced.
One argument, commonly referred to as the ORTHOGRAPHIC DEPTH
HYPOTHESIS, is that a shallow orthography in which graphemes and
phonemes have a consistent relationship will favour reliance on sub-lexical
(phonemic) codes, whereas a deep orthography will promote the use of
a lexical code (Frost, Katz & Bentin, 1987). Learning to read shallower
orthographies may thus promote a greater metaphonological awareness of
phonemes (see Caravolas & Bruck (1993)). The same principles might apply
in non-alphabetic orthographies such as Japanese Kana, but with awareness
emerging first at the syllabic rather than at the phonemic level (Mann,
1986).
Teaching method is another important influence on metaphonological
development (Alegria et al., 1982). Duncan et al. (1997, 2000) established
that English-speaking children who had received phonics instruction were
better at the common unit task when the shared sounds were small
(phonemes) than when they were large (rimes). Goswami & East (2000)
replicated this result and demonstrated that subsequent rime-based
instruction could improve performance on rimes. Orthography and teaching
method have sometimes been confounded in crosslinguistic studies. In
the study by Caravolas & Bruck (1993), for example, the Czech children
were taught by a purely phonics approach whereas the English children were
exposed to a whole-language approach.
Therefore, Experiment 2 investigates whether the sequence of meta-
phonological development differs in French and English. Both language
groups would be expected to follow a syllable ponset-rime pphoneme
progression according to the large-to-small unit theory, with change
occurring as vocabulary develops because of the need to discriminate
between phonologically similar items. This gives rise to the prediction that
children ought to move through the large-to-small sequence at the same
age in each country as language skills mature. If any crosslinguistic difference
were to be envisaged, it might be that the sequence is advanced in French
because the smaller number of syllable types suggests a higher degree of
phonological similarity in French vocabulary.
Gombert’s (1992) model predicts variation in the sequence according
to language and age of schooling. The children are taught to read at
different ages, beginning at age 5 years in the UK and at 6 years in
France, although both groups experience a common approach that
emphasizes grapheme-phoneme decoding skills. This is expected to
promote awareness of phonemes in both countries once literacy instruction
commences. However, differences in the depth of the written orthography
(English is a deeper orthography than French) may enhance phoneme
awareness in French.
METAPHONOLOGY IN FRENCH AND ENGLISH
385
Finally, it is predicted that the use of the new phonological task will
reduce the extent of the crosslinguistic differences in syllable processing
that were observed in Experiment 1.
METHOD
Participants
Entire classes of Scottish children were tested at each of the Nursery
(N=22), Primary 1 (N=23), and Primary 2 (N=22) levels. The mean
chronological ages of the children were 4;7 (S.D.=0;3), 5;7 (S.D.=0; 3), and
6;11 (S.D.=0;3), respectively. The Nursery and Primary 1 groups were
tested in March and the Primary 2 group in June. The BPVS had been used
to verify normal receptive vocabulary development on a separate occasion
3
and yielded mean standardised scores of : Nursery=98 (S.D.=12), Primary
1=103 (S.D.=13), and Primary 2=107 (S.D.=12).
The French children came from three middle-class urban schools in the
South of France. Full sets of data were obtained for 20 four-year-old
Nursery children (mean chronological age=4;7; S.D.=0;3), for 35 five-
year-old Nursery children (mean chronological age=5;6; S.D.=0 ;5) and
for 33 children at the Primary 1 level (mean chronological age=6;7;
S.D.=0;3). The mean WPPSI-R or WISC-R vocabulary performances
were 8.75 (S.D.=0.78) for the four-year-old Nursery children, 10.51
(S.D.=1.31) for the five-year-old Nursery children, and 9.26 (S.D.=1.46) for
the Primary 1 children. The data were collected during March when the
children had completed 6 months of the school year.
Materials
Three types of common units were assessed: syllables, rimes, and initial
phonemes.
4
All of the common units were located in the initial syllables of
bisyllabic words (see Appendix B). The phonological structure of the first
syllable was manipulated as follows: four structures were tested in French
(CV., CVC., CCV., CVn.) and two structures in English (CV., CVC.).
5
Sonority effects were not investigated in this experiment and medial
consonants were randomly selected from the sonority hierarchy. Lexical
[3] Approximately 8 months later for the Nursery group and 12 months before for the
Primary 1 and 2 groups.
[4] Although the initial phoneme in CV, CVC and CVn syllables coincides with the syllable
onset, previous results have shown that, once reading instruction begins, English and
French performance in this condition reflects phonemic rather than onset awareness (see
Duncan et al. (1997) and Duncan (2004)).
[5] As in the previous experiment, the maximal onset principle is used to label item
structure.
DUNCAN ET AL.
386
stress was also manipulated in English giving a total of 12 conditions in
French and English with 4 experimental items per condition.
Procedure
At the start of each session the experimenter introduced a puppet character
and demonstrated the way in which the puppet liked to say the parts of
words that sound the same. The child then played the game with three
practice word pairs. Correct responses were praised and errors were pointed
out without making reference to positional concepts such as ‘start’,
‘beginning’ or ‘end’. Throughout testing, the question put to the child
was: ‘Which bit sounds the same in _? ’ The children repeated the word
pair and then identified the common sound. They were tested individually
in a quiet area adjacent to their classrooms. There were three sessions,
one for each common unit (syllable, initial phoneme, rime), each lasting
about 5–10 minutes. The order of administration of the three sessions
was counterbalanced with each session taking place on a separate (usually
successive) day.
Scoring
The responses given by each child were transcribed in a phonetic code
and recorded for later verification. Responses were scored correct if
they corresponded to the segment of sound shared by the word pair. For
example, in the syllable condition, the shared unit was the maximal onset
syllable and only this response was accepted as correct.
RESULTS
Crosslinguistic comparison
Figure 1 illustrates the pattern of metaphonological development for
each language and Table 4 contains the full set of accuracy scores. British
scores in the crosslinguistic comparison were based only on iambic stress
items in order to match the obligatory stress pattern in French and to
equalize the number of items per condition. The data were collapsed
across syllable structure to eliminate the presence of zero cells. The data
from each country were entered into a three-way ANOVA : language
(English, French)rchronological age (4 years, 5 years, 6 years)rcommon
unit (syllable, rime, initial phoneme). All main effects were significant
(language: F(1, 149)=74.45, p<0.001; chronological age: F(2, 149)=
60.27, p<0.001; common unit: F(2, 298)=84.22, p<0.001). Both two-
way interactions were also significant. As there was a 3-way interaction
between language, chronological age and common unit (F(4, 298)=12.83,
METAPHONOLOGY IN FRENCH AND ENGLISH
387
French-speaking children
Chronological age (years)
456
% Correct
0
25
50
75
100
syllables
rimes
phonemes
English-speaking children
Chronological age (years)
456
% Correct
0
25
50
75
100
syllables
rimes
phonemes
Fig. 1. Mean percentage accuracy in the common unit task for English- and French-
speaking children aged 4, 5 and 6 years. The scores for the English-speaking children are
based on iambic stress items only.
DUNCAN ET AL.
388
TABLE 4. Mean percentage accuracy in the common unit task of
Experiment 2(standard deviations in parentheses)
Trochaic stress Iambic stress
Syllable structure CV. CVC. CV. CVC. CVn. CCV
English
Four-year-olds
Syllables 6.82 0.00 1.14 0.00 — —
(22.07) (0.00) (5.33) (0.00)
Rimes 0.00 0.00 3.41 0.00 — —
(0.00) (0.00) (8.78) (0.00)
Phonemes 9.09 12.50 15.91 14.77 — —
(25.05) (28.61) (35.81) (35.07)
Five-year-olds
Syllables 4.35 7.61 1.09 9.78 — —
(9.69) (13.97) (5.21) (19.57)
Rimes 6.52 0.00 3.26 0.00 — —
(21.61) (0.00) (15.64) (0.00)
Phonemes 65.22 63.04 64.13 57.61 — —
(38.98) (44.51) (42.52) (44.87)
Six-year-olds
Syllables 38.64 21.59 28.41 14.77 — —
(42.07) (32.09) (28.13) (27.45)
Rimes 65.91 5.68 23.86 11.36 — —
(25.05) (18.79) (19.64) (24.06)
Phonemes 90.91 87.50 86.36 82.96 — —
(21.19) (25.30) (28.59) (33.97)
French
Four-year-olds
Syllables — — 88.75 83.75 87.50 82.50
(12.76) (18.63) (17.21) (20.03)
Rimes — — 6.25 0.00 0.00 0.00
(11.11) (0.00) (0.00) (0.00)
Phonemes — — 28.75 1.25 0.00 0.00
(23.33) (5.59) (0.00) (0.00)
Five-year-olds
Syllables — — 88.57 90.72 88.57 96.43
(18.53) (18.28) (16.43) (8.88)
Rimes — — 10.72 6.43 5.71 13.57
(24.47) (21.30) (14.96) (27.35)
Phonemes — — 9.29 5.00 3.57 8.57
(24.32) (20.83) (17.30) (23.44)
Six-year-olds
Syllables — — 72.73 71.97 68.38 81.62
(36.10) (40.87) (37.09) (30.34)
Rimes — — 42.43 50.00 24.26 44.85
(39.77) (41.93) (35.61) (45.97)
Phonemes — — 79.55 54.56 65.44 69.85
(28.95) (39.75) (34.82) (37.33)
METAPHONOLOGY IN FRENCH AND ENGLISH
389
p<0.001), the results were interpreted at this level using the Tukey HSD
procedure.
The follow-up tests support the conclusions that are clearly evident
in Figure 1. At age 4 years, English-speaking children performed close
to floor level on all linguistic units while French-speaking children were
very accurate at identifying shared syllables (86 % correct). The French
syllable advantage persisted at age 5 and diminished slightly at age 6 years.
Accuracy of phoneme identification increased sharply at age 5 years in the
English-speaking sample (to 61%) but this did not occur in the French-
speaking group until age 6 years. This differential effect can be related to
the start of formal literacy instruction at 5 years of age in the UK and at
6 years in France. Performance on shared rime units was low in both
languages but improved at 6 years. In neither language did the results
conform to the expected large-to-small sequence of syllable ponset-
rime pphoneme : for English, the hierarchy was phoneme psyllable/
onset-rime ; for French, syllable pphoneme/onset-rime.
French language analyses
The data were analysed using a three-way ANOVA: chronological age
(4 years (Nursery), 5 years (Nursery), 6 years (Primary 1))rcommon unit
(syllable, rime, initial phoneme)rstructure (CV., CVC., CVn., CCV.).
There were significant main effects of chronological age (F(2, 86)=39.73,
p<0.001), common unit (F(2, 172)=238.04, p<0.001), and structure
(F(3, 258)=12.21, p<0.001), as well as of the three-way interaction,
chronological age by common unit by structure (F(12, 516)=2.62, p<0.01).
The results of a Tukey HSD test suggest that the basic pattern of the
French data – accurate performance on syllables at each age level and
emergence of phoneme retrieval at age 6 years – was present for all syllable
structures. Some minor variations were observed in the data from the
six-year-olds, for whom phoneme detection was most accurate in CV
syllables, a high frequency syllable type in French, and rime detection was
least evident in CVn syllables.
English language analyses
Further analyses were carried out in order to assess the effects of prosody
on common unit identification in English. The data were entered into a
four-way ANOVA: chronological age (4 years, 5 years, 6 years)r
stress pattern (trochaic, iambic)rcommon unit (syllable, rime, initial
phoneme)rstructure (CV., CVC.). All main effects were significant :
stress pattern (F(1, 64)=5.07, p<0.05); common unit (F(2, 128)=120.33,
p<0.001); structure (F(1, 64)=40.11, p<0.001) and chronological age
DUNCAN ET AL.
390
(F(2, 64)=58.75, p<0.001). All two-way and three-way interactions
were significant except for common unit by stress pattern and common
unit by stress pattern by chronological age. The results were interpreted at
the level of the four-way interaction, stress pattern by common unit by
structure by chronological age (F(4, 128)=6.69, p<0.001), using the Tukey
HSD procedure.
The pattern of performance described in the crosslinguistic comparison
for English iambic stress (phoneme psyllable/onset-rime) was mirrored
with trochaic stress. No effects of stress pattern were evident in the data
from the younger (four- and five-year-old) children. Among the six-
year-olds, stress and structure affected the relative levels of accessibility
of rime units. This group identified rimes less accurately than phonemes
except for the (phonemic) rimes in stressed CV syllables. The six-year-olds
also showed a tendency to be worse at syllable identification for CVC. items
because they often responded with the shared CV rather than the shared
CVC unit.
DISCUSSION
At all ages, French children were superior in processing shared syllables
in the common unit task. Prior to formal reading instruction, at age 4 years,
the French children correctly identified 86 % of the shared syllables,
whereas the English-speaking children scored only 1%. This advantage
continued at 5 years and it was only at 6 years that English-speaking
children began to improve significantly in syllable identification. This
outcome replicates the crosslinguistic difference in syllable awareness that
was observed in Experiment 1 in spite of the use of a task that should
produce more varied syllable processing amongst French speakers. One
possibility is that the common unit task differs from Content et al.’s (1999)
partial repetition task perhaps because its comparative nature reduces the
chance of finding variability in French performance.
As predicted, both language groups showed a similar leap in their ability
to identify shared initial phonemes at the outset of reading instruction.
This was not a maturational change as it occurred at 5 years in the UK and
at 6 years in France. These findings replicate previous work in English
(Duncan et al., 1997) and extend the findings to French-speaking children
who have also been taught to read by a method that emphasizes letter-
sounds.
The English-speaking children responded to rimes and syllables in an
equivalent manner. Explicit awareness of these large units was very poor
until about 6 years of age (see Duncan et al. (2000) for a similar result).
For French children, however, developments in rime awareness parallel
changes in awareness of initial phonemes. Both these subsyllabic units are
METAPHONOLOGY IN FRENCH AND ENGLISH
391
initially poorly identified but, once reading instruction is introduced, both
show a significant improvement although the overall accuracy level achieved
for phonemes was significantly higher than for rimes.
Error responses confirm this broad pattern. Prior to reading instruction
French children tended to reproduce syllables from the target pair whereas
English-speaking children reproduced whole words. In both groups, the
introduction of reading instruction was accompanied by an increase in
the number of segmental errors, especially the initial phoneme from one
of the target words.
Syllable structure influenced common unit identification more obviously
in French than in English. French children found it easier to isolate
phonemes in CV syllables, which are the most frequent syllable type in
French. In English, there was some advantage for rime units in stressed CV
syllables. In this case, however, syllable type seems to be important only in
so far as it determines the size of the rime unit. In CV syllables, the rime
unit corresponds to a single phoneme and is easier to detect. This effect was
not observed in unstressed CV syllables, suggesting that reduced vowel
phonemes are less readily identified in English.
No overall effect of structure emerged in the syllable condition despite
the greater inconsistency in the English syllabification of CV. items in
Experiment 1. If the outcome of Experiment 1 is interpreted as indicating a
tolerance of differing syllabifications amongst English speakers because the
syllable is not a salient unit, then there may be no reason to expect a special
disadvantage in Experiment 2 when the common unit in a CV. target pair
is a maximal onset syllable.
Thus, Experiment 2 confirms the superior syllable awareness of the
French-speaking sample using a different metalinguistic task. The import-
ance of literacy instruction in promoting the development of phoneme
awareness was the other major finding. This effect was independent of
age (5 or 6 years) or native language (French or English). Both of these
results are inconsistent with the large-to-small view of metaphonological
development which singles out vocabulary development as the catalyst of
change. Instead, our findings provide support for Gombert’s (1992) theory
in which language development and literacy are the key principles of
change.
GENERAL DISCUSSION
As we noted in the Introduction, there is currently support for the view that
metaphonological development is an internally driven process, universally
applicable across languages, which reflects a need to discriminate within
an expanding vocabulary and which follows the sequence : syllable ponset-
rime pphoneme (Treiman, 1987; Metsala & Walley, 1998). If this theory
DUNCAN ET AL.
392
was correct, we would expect to find that French- and English-speaking
children displayed similar patterns of metaphonological development that
were broadly consistent with the large-to-small unit segmental progression.
Our data do not fit this account. French- and English-speaking children
follow different developmental pathways that reflect the contrasting
characteristics of their native languages. In neither case do the data conform
to the standard version of the large-to-small unit theory. We therefore
think that an alternative framework, such as the one outlined by Gombert
(1992), may be needed. The important contribution of this theory is the
acknowledgement that metaphonological development is likely to reflect
features of the native language and to respond to demands arising from
the external linguistic environment, especially those associated with the
acquisition of literacy.
Experiment 1 examined the consistency of syllable boundary placement
using a segmentation task. Although the French-speaking children were
significantly more accurate, the English speakers were able to generate
two segments with relative ease. However, a striking qualitative difference
became apparent between the language groups once the children’s actual
responses were analysed in detail. English-speaking children generated a
wide variety of segments and thus appeared unsure about exactly where the
syllable boundaries lay. French children, by contrast, consistently made
the two segments correspond to syllables as defined by the maximal onset
principle.
For English speakers, accuracy in tasks such as syllable matching may
have more to do with the large size of these units than with their linguistic
status. Walley, Smith & Jusczyk (1986: 227) provided supportive evidence
from a phonological similarity task, concluding that ‘ younger children’s
more global perceptions cannot be equated simply with syllable perception ’.
Treiman & Zukowski (1996) found mixed results when they compared
kindergartners’ sensitivity to syllables and rime units of equivalent size in
a word-pair comparison task. A small superiority for syllables emerged only
when they substituted nonsense stimuli for the original bisyllabic words.
Thus, it is perhaps tasks that require children to identify syllables explicitly
that provide the best opportunities to examine the status of these units in
the developing linguistic system.
In the present study, the consistent maximal onset segmentations
of the French children contrasted markedly with the varied pattern of
segmentations provided by the English-speaking children. The principles
of legality, sonority and stress that have been identified in previous studies
of English syllabification did not add any explanatory power to the maximal
onset account when applied to French syllabification. One possibility is that
these principles have a graded relevance for the syllabification process
that is related to linguistic rhythm, such that the principles are of most
METAPHONOLOGY IN FRENCH AND ENGLISH
393
relevance in languages like English and of least relevance in languages like
French.
The English-speaking children appeared to obey the legality principle by
avoiding clusters that could not occur at the beginnings or ends of words
and by tending to avoid the production of stressed CV syllables with a short
vowel. The sonority principle might also explain the British children’s
tendency to close CV syllables although this conclusion is weakened by the
failure to find any effect of the sonority contrast between liquids and nasals
in determining the extent of this tendency. Finally, the children’s responses
were generally in keeping with the stress principle for CV items. Beyond
this, however, the overwhelming impression was one of diversity in syllable
production, especially for CV items.
In Experiment 2, we examined the accessibility of syllables by means of
the common unit task. This requires retrieval of the shared syllable of
two spoken words and involves the detection of the OFFSET of the first
syllable without having to make a decision about the onset of the following
syllable. According to Content et al. (2001a), the performance of French-
speaking children might be expected to be more variable and less accurate
in this situation. Contrary to this view, the French children displayed
excellent performance when manipulating syllables in the common unit
task. The level of accuracy far exceeded the performance of the English-
speaking children whose poor performance was consistent with their
uncertainty about syllable boundaries in Experiment 1. Syllable identification
in English did not start to improve until around 6 years of age. This
contrasted with the performance of the French children who were very
accurate at identifying common syllables from nursery age onwards.
Therefore, the different formats of the phonological tasks in Experiments 1
and 2 did not appear to impact greatly on the performance of either
language group. The removal of the requirement to detect the onset of the
final syllable did not lead to uncertainty about syllable structure amongst
the French children. Nor did it alter the uncertainty evident amongst the
English-speaking children.
While the metalinguistic awareness of French and British children
differed with regard to syllables, both groups showed strikingly similar
effects of literacy instruction on the emergence of awareness of phonemes.
A marked increase in the accuracy of common phoneme identification
was observed at age 5 years in Britain and at age 6 years in France. This
age difference establishes that phonemic awareness is independent of
chronological age and directly related to the onset of alphabetic literacy.
The conclusion is consistent with previous studies of English-speaking
children (Duncan et al., 1997; Goswami & East, 2000), which have shown
that an explicit awareness of phonemes develops in response to instruction
that emphasizes letter-sounds.
DUNCAN ET AL.
394
Explicit awareness of rime units seemingly takes longer to develop. The
French-speaking children develop a good explicit awareness of phonemes
and, to a slightly lesser extent, of rimes in response to reading instruction.
It is not clear why rime awareness improves at age 6 years in both France
and Britain. In previous investigations of English (Duncan et al., 2000), it
was proposed that rime awareness emerges because children are beginning
to utilise rime units in reading. It remains to be seen whether this also
applies to French. Rime-based literacy instruction is another factor that has
been shown to improve explicit awareness of rimes in English (Goswami &
East, 2000). Although such instruction was not experienced by the present
British sample, it will be important to establish whether French children
routinely receive any formal or informal instruction that emphasizes rimes
or, indeed, syllables. One conclusion that is clear is that metaphonological
awareness of the rime does not necessarily occupy an intermediate position
between the syllable and the phoneme.
Orthographic depth did not seem to affect performance as the French
children did not show a more rapid acquisition of phoneme awareness
despite the shallower orthography of French. The early placement of the
test point in Primary 1 (after only 6 months of instruction) may have been
responsible for the lack of an effect. Future crosslinguistic studies of
English and French might explore this issue further and also consider how
the contrasting phonological profiles of the two languages impact on literacy
acquisition. For example, the French syllable awareness results can be
linked to the findings of Cole
´, Magnan & Grainger (1999). These authors
used a visual version of Mehler et al.’s (1981) segment detection task
in which French children had to detect a word-initial CV or CVC ortho-
graphic sequence which was either syllabic or non-syllabic. So, the children
had to detect PA or PAR in PAROLE (CV first syllable) or PARDON
(CVC first syllable). Early in the first school year, performance was sensitive
only to the length of the segment, suggesting involvement of grapheme–
phoneme correspondences. By the end of the year, however, the better
readers detected the segment faster when it corresponded to the first
syllable, suggesting use of a syllabic code. It is noteworthy that evidence of
the use of orthographic syllables in reading French seemed to emerge AFTER
mastery of the alphabetic principle. Of interest is the extent to which
this pattern would be replicated with English-speaking children whose
awareness of syllables appears to be rather less secure.
CONCLUSIONS
Our findings indicate that phonological awareness cannot be characterized
as a universal large-to-small sequence. Children’s phonological awareness
appears to be dependent upon the linguistic characteristics of their native
METAPHONOLOGY IN FRENCH AND ENGLISH
395
language. French is a syllable-timed language whereas English is stress-
timed. For native speakers, this aspect of spoken language seems to
constrain the formation of the phonological system with the result that
French-speaking children display a greater sensitivity to syllables than
English-speakers. Phonological awareness also appears to be influenced by
literacy acquisition. Reading instruction that emphasizes letters and their
sounds promotes awareness of phonemes in both English and French but
at different ages due to later school entry in France. Taken together our
results point to the value of a crosslinguistic perspective in investigating the
link between phonological awareness and literacy.
REFERENCES
Abercrombie, D. (1967). Elements of general phonetics. Edinburgh : Edinburgh University
Press.
Alegria, J., Pignot, E. & Morais, J. (1982). Phonetic analysis of speech and memory codes
in beginning readers. Memory & Cognition 10, 451–56.
Bruck, M., Genesee, F. & Caravolas, M. (1997). A crosslinguistic study of early literacy
acquisition. In B. Blachman (ed.), Foundations of reading acquisition and dyslexia.
Hillsdale, NJ : Erlbaum.
Caravolas, M. & Bruck, M. (1993). The effect of oral and written language input on
children’s phonological awareness : a cross linguistic study. Journal of Experimental Child
Psychology 55, 1–30.
Clements, G. N. (1990). The role of the sonority cycle in core syllabification. In J. Kingston
& M. E. Beckman (eds), Papers in laboratory phonology 1 : between the grammar and physics
of speech. Cambridge: Cambridge University Press.
Cole
´, P., Magnan, A. & Grainger, J. (1999). Syllable-sized units in visual word recognition :
evidence from skilled and beginning readers. Applied Psycholinguistics 20, 507–32.
Content, A., Dumay, N. & Frauenfelder, U. H. (1999). Segmentation syllabique chez
l’enfant. Actes des IIe
`mes Journe´es d’Etudes Linguistiques. Nantes, France.
Content, A., Kearns, R. K. & Frauenfelder, U. H. (2001 a). Boundaries versus onsets in
syllabic segmentation. Journal of Memory and Language 45, 177–99.
Content, A., Meunier, C., Kearns, R. K. & Frauenfelder, U. H. (2001b). Sequence detection
in pseudowords in French: where is the syllable effect? Language and Cognitive Processes
16, 609–36.
Cutler, A., Mehler, J., Norris, D. & Seguı
´, J. (1986). The syllable’s differing role in the
segmentation of French and English. Journal of Memory and Language 25, 385–400.
Dauer, R. (1983). Stress-timing and syllable-timing reanalyzed. Journal of Phonetics 11,
51–62.
Demuth, K. & Johnson, M. (2003). Truncation to subminimal words in early French.
Canadian Journal of Linguistics 48, 211–41.
Duncan, L. G. (2004). Influence de l’apprentissage de la lecture et de la langue maternelle
sur le de
´veloppement phonologique: Une perspective inter-langues. In S. Valdois, P. Cole
´
& D. David (eds), Apprentissage de la lecture et dyslexies de´veloppementales: de la the´orie a
`
la pratique orthophonique et pe´dagogique. Marseille: SOLAL.
Duncan, L. G., Seymour, P. H. K. & Hill, S. (1997). How important are rhyme and analogy
in beginning reading? Cognition 63, 171–208.
Duncan, L. G., Seymour, P. H. K. & Hill, S. (2000). A small to large unit progression in
metaphonological awareness and reading? The Quarterly Journal of Experimental
Psychology (Section A)53, 1081–104.
Durgunog
˘lu, A. Y. & O
¨ney, B. (1999). A crosslinguistic comparison of phonological
awareness and word recognition. Reading and Writing 11, 281–99.
DUNCAN ET AL.
396
Fallows, D. (1981). Experimental evidence for English syllabification and syllable structure.
Journal of Linguistics 17, 309–17.
Frost, R., Katz, L. & Bentin, S. (1987). Strategies for visual word recognition and ortho-
graphical depth : a multilingual comparison. Journal of Experimental Psychology : Human
Perception and Performance 13, 104–15.
Giegerich, H. J. (1992). English phonology : an introduction. Cambridge : Cambridge
University Press.
Gombert, J. E. (1992). Metalinguistic development. London : Harvester Wheatsheaf.
Goswami, U. & East, M. (2000). Rhyme and analogy in beginning reading : conceptual and
methodological issues. Applied Psycholinguistics 21, 63–93.
Grabe, E. & Low, E. L. (2002). Durational variability in speech and the rhythm class
hypothesis. In C. Gussenhoven & N. Warner (eds), Papers in laboratory phonology (Vol.
7). Cambridge : Cambridge University Press.
Jusczyk, P. W., Friederici, A. D., Wessels, J., Svenkerud, V. Y. & Jusczyk, A. M. (1993).
Infants’ sensitivity to the sound patterns of native language words. Journal of Memory and
Language 32, 402–20.
Karmiloff-Smith, A. (1986). From meta-processes to conscious access : evidence from
children’s metalinguistic and repair data. Cognition 23, 95–147.
Liberman, I. Y., Shankweiler, D., Fischer, F. W. & Carter, B. (1974). Explicit syllable
and phoneme segmentation in the young child. Journal of Experimental Child Psychology
18, 201–12.
Mann, V. A. (1986). Phonological awareness : the role of reading experience. Cognition 24,
65–92.
Mehler, J., Dommergues, J., Frauenfelder, U. & Seguı
´, J. (1981). The syllable’s role in
speech segmentation. Journal of Verbal Learning and Verbal Behaviour 20, 298–305.
Metsala, J. L. & Walley, A. C. (1998). Spoken vocabulary growth and the segmental
restructuring of lexical representations: precursors to phonemic awareness and early
reading ability. In J. L. Metsala & L. C. Ehri (eds), Word recognition in beginning literacy.
Mahwah, NJ : Erlbaum.
Nazzi, T., Bertoncini, J. & Mehler, J. (1998). Language discrimination by newborns : toward
an understanding of the role of rhythm. Journal of Experimental Psychology : Human
Perception and Performance 24, 756–66.
Pulgram, E. (1970). Syllable, word, nexus, cursus. The Hague : Mouton.
Ramus, F., Nespor, M. & Mehler, J. (1999). Correlates of linguistic rhythm in the speech
signal. Cognition 73, 265–92.
Roach, P. (1982). On the distinction between ‘ stress-timed’ and ‘syllable-timed ’ languages.
In D. Crystal (ed.), Linguistic controversies. London : Arnold.
Schane, S. A. (1968). French phonology and morphology. Cambridge, MA : MIT Press.
Seymour, P. H. K. & Evans, H. M. (1994). Levels of phonological awareness and learning
to read. Reading and Writing 6, 221–50.
Treiman, R. (1987). On the relationship between phonological awareness and literacy.
Cahiers de Psychologie Cognitive 7, 524–29.
Treiman, R., Bowey, J. A. & Bourassa, D. (2002). Segmentation of spoken words into
syllables by English-speaking children as compared to adults. Journal of Experimental
Child Psychology 83, 213–38.
Treiman, R. & Zukowski, A. (1990). Towards an understanding of English syllabification.
Journal of Memory and Language 29, 66–85.
Treiman, R. & Zukowski, A. (1996). Children’s sensitivity to syllables, onsets, rimes and
phonemes. Journal of Experimental Child Psychology 61, 193–215.
Walley, A. C., Smith, L. B. & Jusczyk, P. W. (1986). The role of phonemes and syllables
in the perceived similarity of speech sounds for children. Memory & Cognition 14, 220–29.
Werker, J. F., Gilbert, J. H., Humphrey, K. & Tees, R. C. (1981). Developmental aspects of
cross-language speech perception. Child Development 52, 349–55.
Wioland, F. (1985). Les structures syllabiques du franc¸ ais. Gene
`ve : Slatkine ; Paris :
Champion.
METAPHONOLOGY IN FRENCH AND ENGLISH
397
APPENDIX A: MATERIALS FOR EXPERIMENT 1
French
CV. items CVn. items CVC. items
carie cantine carton
cirage camping circuit
marin menton martien
ballon bandit balcon
carafe langage carbone
bassine dandy bastide
mare
´e canton marquis
tarif gambade tardif
English
a
CV. items CVC. Items
Trochaic
stress
Iambic
stress
Trochaic
stress
Iambic
stress
camel career camping cartoon
garage display garden dislike
sheriff maroon sherbet marquee
panic Peru panda perhaps
carrot command carton complain
runner commence runway combine
curry connect curtain confuse
bunny commute bunches compact
a
Note that the version of English spoken in Scotland is a rhotic accent (Giegerich, 1992).
APPENDIX B: MATERIALS FOR EXPERIMENT 2
French
CV. items CVC. items CVn. items CCV. items
Syllable
ga
ˆteau-galette corbeau-cordon cinquante-simplet drapeau-drage
´e
cadeau-cahier carton-carnet conseil-contact brebis-bretelle
balai-baguette marteau-mardi janvier-jambon flocon-flotteur
midi-minuit dispute-discours langage-lenteur produit-promesse
Rime
silence-piqu
ˆre garc¸on-barquette bandit-fanto
ˆme dragon-plateau
lutin-pure
´e sortie-formule nombril-tondeuse gratin-flacon
bijou-micro marche
´-jardin jongleur-fontaine plumeau-trucage
magie-tableau tortue-sorbet cantine-bandeau crate
`re-plafond
DUNCAN ET AL.
398
APPENDIX B: (CONTINUED)
French
CV. items CVC. items CVn. items CCV. items
Initial phoneme
lapin-lec¸on berceau-balcon danger-dompteur cre
´meux-cravate
sapin-secret parcours-pistache panthe
`re-pinceau pratique-pre
´sent
rocher-requin biscuit-berger cinquie
`me-sandale gravier-groseille
matin-maison palmier-portable rentre
´e-rongeur plaque-plume
English
CV. items CVC. items
Trochaic stress Iambic stress Trochaic stress Iambic stress
Syllable
panic-paddle canal-cassette bunches-bundle confuse-control
woman-woolly reward-release window-winter dislike-dismay
button-bubble command-correct marble-market surprise-survey
carrot-camel delight-demand picnic-picture complain-compare
Rime
daisy-taking baboon-career runway-bunches forbid-torment
melon-better below-reward pancake-bandage bamboo-campaign
puppy-rubbish delight-repair garbage-harvest perhaps-herself
comic-wobble display-without reindeer-painting marquee-cartoon
Initial phoneme
runner-rabbit machine-meow garden-golfer fulfil-foresee
woman-walking Peru-police camping-curtain sardine-submit
garage-going hello-hooray dolphin-darker harpoon-himself
kettle-curry balloon-beside pencil-padlock success-segment
METAPHONOLOGY IN FRENCH AND ENGLISH
399
