Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
•••
•
HC5L7JS S. SCARBOROUGH
As recently as 20 years ago, leaming toread each of which has been sharpened ehrough
was not thought to commence until formal instruction and experience over many years.
instruction was provided 111 school. Accord Figure 8.1 illustrates rhe majar "strands"
ingly, reading disabilities were largely con that are woven cogether during the coutse
now abundantly clear rhat reading acquisi in recognizing individual prinred words
tion is a process rhat begins early in the from rhose involved in comprehending the
preschool period, such rhar cluldren arrive meaning of the string of words that have
ar school ha ving acquired vasdy differing been identified, even though those two
degrees of know!edge and skill pertaining ro processes operare (and develop) interactive
literacy. Attention has chus tumed to ly rather rhan independently. (For a fuller
whether preschool differences in language review of rhis material, see the rccenr report
and literacy development are reliable prog of rhe Commitree on the Prevention of
nostic indicators, and perhaps direct causes, Reading Difficultíes in Young Children,
discuss the available evidence from longitu Most children who bave trouble learning
dinal research that has examined such is to read in the early school years stumble in
sues, with particular attention to at-risk mastering the "word recognition" strands.
with reading disability and preschoolers di ken words are governed largely by the '"al
agnosed with early language impairmenrs. phabetic principie," rhe notion rhat our
The Multífaceted Nature of Reading fu! speech elements (phonemes) that make
Skilled readers are able to derive meanmg grasping the alphabetic principie will be dif
from printed text accurately and efficiently. ficult if a child <loes not yet apprcciate that
Research has shown that in doing so, they spoken words consist of phonemes, beca use
fluidly coordinare many component skills, wirhour this "phoncmic awareness" the
97
98 STRANDS OF EARLY LITERACY DEVELOPMENT
LANGUAGE COMPREHENS!ON
BACKGROUND KNOVI/LEDGE
VOCABULARY
SK!LLED READING:
coordination of word
lANGUAGE STRUCTURES
comprehension.
VERBAL REASONING
1
únre.ence. melaphor. etc)
1
UTEAACY KNOWLEOGE
i
1 1
! 1 WORD RECOGN!TION
1
PHONOLOGJCAL AWAREJJESS .::;;c;;;y:;;::=;7j'���
(syllables, phonemes, etc}
speiling-sound correspondences)
SJGHT RECOGNITION
(of farriharwords)
FIGURE 8.1. Illustration of the many srrands that are woven rogether in skilled reading.
child cannct truly understand what letters come more complex. Even if rhe pronuncia
stand for (Liberrnan, 1973). rions of ali the lerter strings in a passage are
Recognizing printed words furrher re correctly decoded, the rext will not be well
quires that one learn and apply the many comprehended if the child (1) <loes not
correspondences berween particular Ietters know the words in their spoken form, (2)
and phonemes, so thar rhe pronunciation of cannot parse the syntactic and semantic re
a prinred word can be figured out ("decod lationships among the words, or (3) lacks
stored ínformation about spoken words in tial skills to inrerprer the text appropriarely
one's mental lexicon enables the identiry of and "read berween the lines." Note that in
logical decoding is the most reliable guide to dcficits are essentially oral language limita
of exceptions (words such as "of," "two," A daunting fact about reading (dis)abi!i
"choir;" and "yacht") whose spellings must, ties is that differcnces among schookhildren
Finally, skilled reading requires that the strong stability over time, despite remedia!
1 1
processes involved in word recognition be efforrs rhac are usually made to srrengthen
1 1 come so well pracriced that they can pro the skills of lower achievers. {For a review,
ceed extremely quickly and almost effort see Scarborough, 1998). Only about 5-10%
1
'
lessly, freeing up the reader's cognirive of children who read satisfactorily in the
resources for comprehensión processes. primary grades ever scumble later, and
Although most reading disabilities are as 65-75% of clufdren designated as reading
1
sociated with deficits in phonemic aware disabled early on continue to read poorly
ness, decoding, and sight recognition of throughout rheir school careers (and be
1
prinred words, reading skill can also be seri yond). In light of this continuity, there has
ously impeded by weaknesses in the "com been increasing interest in whether children
prebension'' srrands, particularly beyond at risk for reading disabilities mighr be iden
1 second grade when reading marerials be- tifiable at early ages, so that steps couid be
l
Connecting Early Lnnguage to l.Ater Reading (Dis)Abilities
99
whac early signs to look for in arder to iden a wide variery of predictors had been used
i'9
tify which preschoolers are most likely to by rhe researchers. Table 8.1 summarizes
IJ1I,
develop reading disabilities. That tapie is re those results for three sets of skill variables:
� Mosr- research on the predicrion of future has been Ieamed from research on the cog
reading abilities hns involved samples who nitive requiremenrs of skilled reading and
·�
were first tested just prior to the start of the acquisition of its various "strands."
� schoolillg {in-thctfnited States, usually in That is, although visual and motor skills of
the kindergarten year) and who were rhen entering studenrs have been a tradicional fo
�
followed up after having received 1 or 2 cus of readiness resting, performance on
�
-
� TABLE 8.1. Average Correlations between Kindergarten Predictor Variables and Latcr Reading Seores,
fi$
Measures of oral language proficiency
!?$
General language index: expressive and
�j)
Note. Data from Scarborough (1998).
:':.'ID
··�, '
- .·
100 STRANDS OF EARLY LITERACY DEVELOPMENT
such nonverbal tasks actually providcs Iirr!c because, unlike most kindergarten predic
progncstic information abour furure read rion research, the chiidren's progres� has
ing difficulries. On the other hand, rudi rypically been observed cver severa! years
memary skills that ne in to rhe "word prior to che start of schooling. Such longitu
among rhe best predicror measures. Like reading-related skills that may both shed
wise, early differences in the sorts of verbal light on theoretical questions and provide a
abilities thar make up che "comprehension" foundarion far dcsigning carly diagnostic
strands-mosr norably vocabulary, sen and preventive programs. Very briefly, che
reading. .
with future reading achievemenr are not Severa! dozen follow-up studies have been
nearly as strong {r s .57) as rhe correlarions conducted to look ar the short- and long-
seores and rbose earned 1 to 4 years Iarer (r diagnosed (and, in mosr cases, treated} ar
= .75}. In efforrs to irnprove predictive accu- speecb-language clinics (e.g., Aram & Hall,
racy, sorne researchers have combined 1989; Bishop & Adams, 1990; Carts,
kindergarten predictor variables to compute Fey, & Tomblin, 1997; Rescorla, 1999 ;
a multiple correlation wirh reading ourcome Storhard, Snowling, Bishop, Ch.ipchase, &
seores in their samples. When rhis has been Kaplan, 1998). Virmally every such srudy
done, the resulrs (mean R = .75) suggest that has confirmed thac pceschoolers with lan-
the predicrabiliry of future reading abiliry is guage impairments are indeed at consider-
abour as srrong from kindergarten onward able risk far developing reading disabilitics
as it is from grade to grade once formal (as well as far continued oral language diffi-
1 ordinated in learning to read are racher se· The fact that reading disabilities tend ro "run
curely in place befare formal school instruc in families" has been established for nearly a
rion begins, such rhar children who arrive ar century, with h1gher mcidence noted among
school wnh weaker verbal abilities and liter rhe relatives of affected schoolchildren than
acy knowledge are much more likcly than in che families of rheir normally achieving
learning to read during the primary grades. not previously been examined in a p:ospec
Tlus raises the next question: How far back tive way, I reasoned that having a parent or
in development can the roors of the various older sibling with a reading disability should
Predicting Reading from Infant and these could be discovered by following such
Developmenral relationships berween Jan· showed that offspring of parents wirh read
guage and literacy abilities have been stud ing problems were indeed ar much higher
ied from early ages in rhree kinds of sam risk for d1fficulry in learning to .read, and
ples: preschoolers wirh eacly language that these childcen differed on language mea
impairmenrs, offspring of adults wirh read sures from ocherwise-similar pee.rs ar ages as
nfants
i or preschoolers. These studies of 1990, 1991; Scarb�}fough. Dobrich, &
y oungec children are particulacly va!uable Hager, 1991). Recently, ourcome results for
,.
.,
Connecting Early Language to Later Reading (Dis)Abilitie:,; 101
severa! similar studies have been reported 1. In these studies of younger children, as
that tend to converge with these findings in the kindergarten predicrion research sum
{Byrne et al., 1998; Elbro, 1999; Gallagher, marized earlier; nonvcrbal skills generally
Frith & Snowling, 1999); Lyytinen et al., have been unrelated to concurrent or future
1999; Pennington, Lefly, & Boada, 1999). language and literacy levels, whereas verbal
' Risk estimares depend, of course, on the cri skills have been much better predictors.
.
. teria used to diagnose reading disabilities in Even in infancy (bírth to age 2 years), pedi
adults and children; averaging across srud atric ratings of language milestones predice
' ies, approximarely 40% of offspring of af later reading achievement better rhan do
fected parents, but less than 10% of orher perceptual-motor indices {Shapiro et al.,
'
children (of otherwise similar backgrounds), 1990). Similarly, recent studies have found
want to focus on the commonalines among These predictive correlations have often
their findings. In reading these various Iiter tended to be weaker far measures of speech
atures over the years, I have been struck by than far other aspects of language produc
the fact that the relationship between early tion and for measures of receptive rather
language and literacy development and later than expressive Ianguege (e.g., Bryant et al.,
reading achievement has appeared to be 1939, 1990; Chaney, 1992; Gallagher et al.,
similar in many respects and not contradic 1999; Lonigan et al., 1998; Pennington et
tory in any major way, despite rhe differing al., 1999; Shapiro et al., 1990).
goals and sampling procedures of the three 4. When several domains of deve!oping
sorne empirical resulcs thac have been ob etc.] have been examined wirhin a sample,
served in at least two of the three kinds of the successful predictors of future reading
research samples. (There is a wealth of addi abilities usually have not been confmed to a
tional detail to be found m each individual single linguistic domain (e.g., Catts, Fey,
of that material is beyond the scope of this Scarborough, 1989, 1990 ; Walker et al.,
reading outcomes have been besr predicted Scarborough & Dobrich, 1990). For exam-
fcrenr ages wirhin longitudinal samples weaknesses in synractic, lexical, and phono-
(e.g., Gallagher, Frirh, & Snowling, 1999; logical skrlls might show a narrowcr range
Lyyrinen et a!., 1999). Figure 8.2, which of deficirs (e.g., in just one domain) a year
sample, illustrares rhis phenomenon. At the 6. Even when rheir eady language
youngest ages, syntactic and speech produc- deficits have lessened considerably in sever-
rion abilities were most deficienr, relative to ity (or have disappeared entirely) by the
those of the comparison group, in the group time of school entry, children wirh a family
period, however, the groups differed instead less remain at high risk for developing read-
in vocabulary and phono!ogical awareness ing problems at a [ater age (Fey, Catts, &
5. Similarly, when longitudinal data have ough & Dobrich, 1990; Stothard et al.,
weak early language skills, deficit "profiles 7. Despite the relationship that has been
have actually been observed to change over found berween preschool language prob-
time wirhin individuals during rhe preschool lems and school-age reading problems, ex-
years (e.g., Bishop & Edmundson, 1987; ceptions to tbis trend have been seen in
,.,
-o o
Sentence .•·
Complexily
•
1.2
º··,,,
ª
" \ .
······a·
"
e,
o
e 1.0 o
o .
·.
"
• I /f::. - - - - - 2\. - - - - - - - -,D.
]'.
Pronunciation ,
• 0.8
tn ,
•
u
Accuracy
, º1
\
e Phonolog1cal
� ,
Awareness
• 0.6 ,
,
3 ,
,
"
E ,
0.4 ,
1'
º
•
-c Expressive
�
02 Vocabulary
e
m
,.
ro
o
30 36 42 48 54 60
Age ( mos)
FIGURE 8.2. Changes over time in rhe aspecrs of language that differcnriated preschoolers who be
came disablcd readers from those who did not (Scarborough, 1990, 199la). Effect sizes for the differ
ences berween the group means are shown for sentence complexiry (Indcx of Prcducrive Symax), ex
pressive vocabulary {Bosron Naming Test), pronunciarion accuracy (pcrcenrage of consonants corrccdy
,.
•.'
j
Connecting Early l.Anguage to I..ater Reading (Dis)Abilities 103
1
every sample. That is, sorne children with undoubtedly a bost of reasons. Here, I want
�
early language deficirs did not devclop read !º focus on a fe-:v facrors tbar, in my opin
!j
ing disabilities, and sorne children who be ton, may have impeded the derivation of
carne peor readers had not appeared to be firm conclusions from the extanr data.
ii
behind in their preschool Janguage develop
�
ment.
Correlation versus Causality
¡;
Taken ali togerher, these resulrs suggest We all have been taught that the existcnce
;,
thar there is a great <leal of continuity be of a relationship between two variables
r, rween early developmenral differences and does not mean that one variable necessarily
Iater ones. On rhe orher hand, the data also causes the other. Establishing causation re
¡,
suggest that the pattern of across-age conti quires experimental research in which it is
� nuities is not entirely simple or straightfor demonstrated chat manipulating the pre
ward but, insread, presents sorne complexí sumed cause (X) does indeed lead to
· �
ties that mighr be overlooked were ir not fer changes in the presumed effect (Y). Such an
¡;¡
rhe fact that rhese phenomena have been experimental resulr, moreover, does..nauw,c_ _
¡; observed by different researchers in various out rhe possibility of reciproca!, rather than
kinds of samples. Sorne irnplicarions of just unidirectional, causation (i.e., that Y si·
19 these common fmdings for theory and prac multaneously exerts a causal influence on
® Theoretical Issues: Present and Future dents have shown that there is a reciproca!
opment is related to rhe acquisition of lirer Beck, Bell, & Hughes, 1987). For the other
1$
acy. Various researchers have recast and verbal abilities rhat are good prediccors of
tj)
narrcwed rhis broad question in different future reading achievement from kinder·
J ways, such as: What are the consequences garten age (Table 8.1) or earlier, however,
1"
of early language impairment? Are little evidence is available yet to determine
'!)
preschool language disorders and Jacer read their causal status. Sorne language skills
il, ing disabilities two manifestations (differ may indeed play a causal role in the devel
ing, perhaps, in severicy) of the same dinical opment of reading, but sorne may only be
�
condition at different ages? What are the "correlates" or "markers" that are charac·
� preschool antecedents of reading disabiliry, reristic of children who will have rroublc
and which ones play a causal role in its de· learning to read but that are not the reason
C')
velopment? What preschool developrnents those children have difficulry. In fact, there
1" are necessary and sufficient for successful are indications that preschool training that
::í, reading acquisition? And so forth. Although successfully ameliorates early speech/lan
these differences in emphasis have guided guage impairments is not effective in reduc
w the selection of research subjects by various ing such children's risk for lacer reading
� investigarors, the data from ali such studies problems, as it ought to be if those language
among findings is helpful in addressing the At present, the most widely held view as
i
orerical issues. to the cause of reading disabilitii:s is that af
!:! Given the wealrh of evidcnce now avail fecred children have a core phonological
able from longitudinal studíes of early lan deficit (often of constitutional1 usually ge·
�
guag� and literacy, one would think rhat netic, origin) that impedes che developmcnt
,:J
clear · answers to rhe quesrions of interest of phonemic awarcness and hence interferes
� would be rather easy to derive. Thrs is not wirh discovering the alphabetic principie
the case, however. Why has it been so diffi and with learning to decode (e.g., Liberman,
!!)
culr to answer rhese questions? There are Shankwei!er, & Liberman, 1989; Stanovich
.•.
_,
.. i
"
�
'
1
& Siegel, 1994). Powerfo! and parsimo weaknesses, solely or in conjunction with
nious though this theory is, rt has been chal phonological deficits (e.g., Bowers & :X7"olf,
lenged for failing to account readi!y for sev 1993; Manis et al., 1999). There is no con
eral empírica! rrends. For cxample, training sensus as ro the nature of the addirional
programs designed in accordance with the deficit{s), bowever, and empiricai support
phonological deficit hypothesis have not for the hyporhesized subrypes is fairly hmir
been completely effective in preventing and ed. Moreover, proposed qualítative differ
rrcatíng reading disabilities (e.g., Torgesen, ences have tended ro be confounded with
Wagner, & Rashotte, .l 997). .Also, sorne severiry of impairment. Although the notion
initial difficulties in leaming to decode in re severa! decades has been rather unsuccessful
ro fall behind again in reading at a later abled readers, and I am not sure rhat rhe lar
book are rhe correlarional data reviewed It is possible, however, to imagine a sin
earlier, That is, severa! Íacets of verbal abili gle-deficit model of reading disabilities rhar
ry orher than phonological awareness have incorporares tbe strengths of rhe phonolog
been shown to be equally srrong predictors ical-deflcir hypothesis and also accou11ts for
of Iarer reading, not just from kindergarten thc preschool corrdational data reviewed in
age but also ar much younger ages. Similar chis chapter. To do so requires, however,
ly, phonological awareness itself seems to be thac we Stop thinking about causaliry only
predicted as well by previous lexical and in terms of a "chain" of evencs that mflu
syntactic abilities as by phonological ones. ence each other in turn (e.g., succcssive
Findings such as rhese suggesr that rhe deficits in phonological processing, attain
phonological core deficit hyporhesis may ing phonological awareness, grasping the
nor accounc fully for the developmenc of alphabetic principie, and, finally, learning
cal d é ficit hypothesis have argued cogently manner (e.g., the disease glaucoma), orhers
language all stem from more fundamental syphilis, for instance, do not constitute a
weaknesses in rhe phonological domain.. causal chain. Insread, rhe root cause is a
That is, even though other sorts of language persisting bacteria! infecrion, which pro·
deficits are predicrive of future reading diffi duces different symptoms at differcnt stages
culries, they are jusr correlares (or secondary of the disease. Note that knowing which
symptoms) rarher than true causes of read rype of causal model accounrs fer a disor·
ing disability (Shankweiler & Crain, 1986). der has imporranr implicarions not jusr for
The developmental parteros and relauve theocy but also far irs treatment. For a
nor readily accord wnh this explanation. symptam along the way will prevem the
(Consider, fer example, rhe data in Figure emergence of ali successi,,e stages of the dis
8.2.) Also, results of a recent generic analy ease. In contrasr, for a syphilis-like disorder,
sis of the herirabiliry of phonological aware effecrive treatmenr of a symptom will not
ness, general language abilities, and reading halr rhe progression of the disorder; in
skills were inconsistent wirh rhis accounr stead, it is necessary to idenrify and rreat
pase moving from a single-defrcir to a dou model that incorporares both an underlying
ble-deficit (or; m principie, a mulriple condition ( e .g. , a genetic predisposition to
deficit) model of reading disabiliry. These have difficulty learning certain kinds of !in·
are_ subryping hypotheses, according to guistic patterns) thar is rhe root cause of a
whicb some children's reading difficulries series of different symptoms and sorne
stem from phonological déficits, whereas causal influences berween symptoms. Wirh
others' have their roots in different language regard tQ reading d1sabilines1 far instance,
Connecting Early Language to Later Reading (Dis)Abilities
105
�-u'�º�'R_s_ -- b-�
:e
''-"'"-:e'-'e:'�cl- -- -- DAWIGE
SEXUAL
CONTACT
Jt
·e-==-
_BACTERIAL
rr JNFECTION_
'--- ___)
{L
C. HYBRID MODEL
��� SYMPTO;�
,{t
I
ft
UtlOERLYING
fr'---{L
CONOffiON ___)
FIGURE 8.3. Models of possible causal relationships in rhe developmental progression of a condition
or disorder. (In variations of the hybrid modd, sorne horizontal arrows could be absent.)
suppose rhat successive "symptoms" in other srrands during larer stages of reading
elude deficits in early syntactic proficiency, acquisition, leaving the child ar risk far fu
phonological awareness, and decoding of rure difficulties despire having attained ade
�
prinr, respectively. Although there are siz quate skill in decoding. AH rhese predicrions
W..-
refi able correlations among ali three measures, from the model are consisrenr wirh the re
the syntactic deficit might have no causal in search rhar I have reviewed earlier.
�
fluence on the subsequent developmenc of Differences in severiry of rhe underlying
¡y,�
the other two deficirs, but the weakness m impairment, furthermore, would lead to dif
phonological awareness would indeed be an ferences in che number and sevenry of
!
"""'
�
imporrant ("proximal") cause
fected by the underlying condition, treat struction) are sure to play a causal role too.
menr of an early syntactic impairment Hence, anomalous cases {successful reading
would not reduce the child's risk far reading achievement by a chi!d who had previously
� disability, but training in phonological been diagnosed with a language impair
awareness would be of benefit in eliminat menr, olfld conversely, on!y subclinical
�
ing or ameliorating the child's difficulty in weaknesses in early language in a child who
learning to decode. lt would still be possi iater exhibired a reading disabiliry) can be
ble, however; that rhe underlying condition accommodated by the model, albeit not
would continue to exert its influence on wíthout costs in terms of parsimony.
r
To explain rhe changing preschool deficir skill is nonhncar; deficits in that skill will be
profi!es rhat have been observed may re rnosr readily detectable during periods
quire anorher break from rraditionally lin when normal development undergoes a
ear ways of rhinking abour developrnenral spurr (e.g., when rapid!y developing chil
disorders. Ir is fairly well established that dren are reaching the postspurt plareau,
growrh in some {perhaps ::i.11) componenrs of and growth of rhe slower developing chil
language consiscs of spurts and plateaus ar dren may jusr be srarting to accelerare).
particular rimes rarher than steady incre Spurts in particular language skills occur ar
mental advances. If so, then a delay in ac different ages, on average (e.g., the well
quis ition will ruean thar spurts and plateaus known vocabulary spurt, typically occur
will .. occurcar ¡¡......fil!!!}e,yhªr_older age than ring ar about 18 monrhs, precedes the peri
curve in Figure 6.4. In such a case (depend symax from age 2 to 4 years). Therefore, ar
ing, of course, on rhc durations of plateaus any given time, conditions for detecting in
and the degree of dela y), there may be ages dividual or group differences in a skill will
a t which rhe performance levels of delayed be besr when that skill is normally "aseen·
and nondelaved cases will be virtually iden danc." According to this ascendancy hy
rical, a phenomenon rhar Scarborough and porhesis, furthermore, rhe milder rhe lan
Dobricb ( 1 9 9 0 } rermed "illusory recovery," guage de/ay (i.e., the smaller the horizontal
Ir provides :
1 simple explananon for the oth distance between the dashed and salid
erwise puuling fact, noted earlier, rhar len curves in Figure 8.4), the more uansient
guage and reading problems often and domain-specific the pattem of observed
(re)emerge at older ages in children who deficirs will be. A severe delay, in contrast,
normal
delayed
platea u
.. - · · · · · · · · · · - · · · · · - �
.
- o
í •
• '--v--"
•
-¡¡; •
"illusory
>
ID .
__, spurt recovery"
•
•
,•
.
,
•
•
.
•
•
••
.··
---····-----------------·-
Age
FIGURE 8.4. Illumarion of how non!inear development of language skills might lead to periods of "il
lusory recovery" by children who had previous!y appeared ro be delayed. Data from Scarborough and
Dobrich (1990).
.
:
- .f
J
,,
, Connecting Early Language to Later Reading (Dis)Abilities 107
9
and across-the-board deficir profi!e. {Note Practical Implications:
iJ
rhar what looks like a qualicative difference, Present and Future
¡j or subtype, would really be a quantitative
severiry difference.) If this ascendancy hy With regard to diagnosis, che risk factors
g
pothesis correctly captures the rneasure thar have been idennfied by che correlation
3 ment situation far early language skills, al research on preschoolers and kindergarr
t, chen che notion rhar a single underlying lan ners provide the best current guidelines far
guage disorder could manifest itself as a se designing screening barrenes to identify
tJ
ries of deficirs in different aspects of lan those young children who are most likely
tJ guage, each correlared with che next, is to develop reading disabilities. As noted
precisely what would be expected far mild earlier, researchers who have assessed
j
' '
- to-moderare severity levels. kindergartners on various subsets of such
;;; In sum, I believe that grearer" pOWec"and variables havr- attained high multiple corre
,
flexibility in theorizing about che relation [ations wirh subsequent reading seores. Fig
-----shi-ps-between-language and literacy devel ure 8.5 shows how well such screening bat
"
:j opment can be obtained b-)-;-conSídering ar: teries have succeeded in rhe typical study of
ternatives to causal chains and linear this sort, in which 89% predicrion accuracy
�
growth assumptions. I have not tried to has been obtained in samples of about 200
g
construct a full theory bue, rather, to illus children. For most purposes, chis is a rea
;'! trate how sorne interesting phenomena seen sonably satisfactory leve! of success. Note,
in the available literature can perhaps be however, that virtually every study has ob
é1 explained more sarisfacrorily than ar pre rained few "miss" errors (i.e., children not
..
"
;,)
,,
"'
-
-
17% 18 16 34 were predicted
v �
O) "'
rn at- (false to become RO;
<l)
;
¡] el
e
-,;
risk alarms) 53% of them did.
¡'.'!
,) '2
� [31·75%)
e o
9
.. el
<l)
-u
el
e•
-"'
e
<l)
�
ü,
83% 6 160 166 were predicted
,j
1
"'
ro
"' m
�
'
�
' 24 RO: 176 N R O :
1
75% were 9 1 % were
1
"'
t
1
predicted. predicted.
1
,
, [56-93%) [80-95%)
"
"Sensitivity" "Speclficity"
�
1
·]
FIGURE 8.5. Typical results obtained by combining kindergarren-age measures to predict later reading
in samples of about 100 children. The ranges of values across studies are shown in brackets. RD, read
.
1 ing disabled; NRD, not reading dísab!ed. Based on data from Scarborough {1998, Table A-7) .
,. 1
'
'
•'
'
1
became disabled readers) bur a sizable pro should be a focus. The besr candidares far
fred by rhe screen as being ar risk, but who program are those suggcsted by the correla
later achieved adequarely on rhc reading rional research, namely, print concepts,
measure). If early intervention is rargered at retention of verbal material, and oral lan
all children designated as at risk, rhis means guage skills (cspecially expressive vocabu
thar about half of rhose receiving ir might lary). Alrhough there is no guaranree that
not acrually be in need of ir. At present1 chis training in these skills will facilitare reading
is rhe most serious conceru associared wirh acquisition, rhis imporrant causal question
using such screens, uor jusr because rhe can be investigated rhrough follow-up srud
1 costs of intervention are substantially raised ies of rhe efficacy of inrervention programs.
bur also because the possible negarive edu Finally, especially wirh regard to younger
1,
cacional and psychologica! consequences of �t-risk preschoolers (especially rhose with a
mislabding "false alarms" are not known. diagnosis of language impairment and those
If this issue is handled senstbly, though, I with a family histocy of reading disabil1ry),
rhink a good case can be made thar early inrerventions based on an accurate causal
tdentification and intervcntion are warrant model are likely to be mosr effective in re
The available data also indicare rhat diag ply addressing these children's current
ages. The observations rhar language deficir language rherapy apparently does not re
profiles change over time within individual duce such risk, probably beca use weakness�
preschoolers (as well as berween groups, as es in speech and language do not causally
in Figure 8.2) means rhat assessmenr at a impedc reading acquisirion, at least over the
single rime poim may be misleading as to shorr rerm. Hence, as Fey (1999) has urged,
mighr be. In the future, I believe thar diag causal factors (such as phonological aware
nosric improvements can be achieved by ness) may be required. If and when evidence
giving increased consideranon to rhe possi accrues for rhe existence of an underlying
bility of nonlinear growrh in skills, to rhe "root" cause of reading abilirr differences,
youngsrers. Firsr, although eguacing corre Supporr for the preparation of this rev1ew was pro
[ation wirh causaliry is a false inference, rhe vided bv Grant No. HD-01994 to Haskins Labora·
opposite-that a lack of correlation imphes torles. i wou!d ,1Jso Jikt to thank Susan Brady and
ment.
Anrhony, J. L., Lonigan, C. J., Dyer, S. J\·L, &
Second, among rbe stronger predictors,
Bloomfield, B. (1997, April). The dcvefopment of
only phonological awareness has ver been pbonofogical processing in preschool-aged chil·
demonstrared to play a causal role '¡n learn dren: Prelimhrary ev1dence (rom confimratory
ing to read. A successful inrcrvention pro factor a11alys1s. Paper presenred at the meeting of
�
®,f
-
-·aV
9
•••
� •
lJi,
USHA GOSWAMI
I · •
'
-
' '
;:@
mi
;'!.@
of spoken language and the child's subse argue that we need to understand more
�
quent development of literacy are becoming about the relative weight rhat needs to be
� the child's phonological development-the syllable, rhyme, and phcnerne and their
progression in representing in the brain the connecrions with sequences of letters in dif
�
speech units that make up different words ferent orthographies. For example, analo
'llíl is now recognized to play a causal role in gies based on rhymes may be particularly
the acquisition of literacy. In this chapter, 1 important for reading acquisition in Eng
�
describe the development of children's abili lish. Further; reading insrrucuon can make
""
e@ phonological units in spoken words,
the phonemic
developmenr,
leve!. Using
particularly
rhe theoretical
at
SÍ!í) rization skills are thought to be acquired in consider possible contributions of varied
and nursery rhymes in helping rhe child to Phonological Skills and Learning
1 $
rhen review sorne of the evidence thar the tasks that require a child to reflecr on or to
child's awareness of the phonology of his or manipulare the componcnt sounds of spo
her language is one of the most importan! ken words. A wide vanety of such tasks has
predictors of thar child's progress in learn bcen designed, including asking children to
:t,;
ing to read and to spell, noting that cross monitor and correct speech errors (e.g.,
linguistic research is increasingly showing "sie'' to "pie"), to select the "odd word
that the phonological units that are high- out" in terms of sound (e.g., which word
111
'
or "d-i-sh'' to make "dish"; see, e.g., Goswami, in prcss aj. Furrher, rhe develop
Bradley & Bryant, 1983; Chaney, 1992; mental process of making phonological
Liberman, Shankwei!er, Fischer, & Carrcr, knowledge expbcit seems to be much easier
1974; Mersala, 1999, Treiman & Zukows for svllables, onsets, and rimes than it is far
ki, 1991). Performance in a!l rhese phono phonemes. Phonemic awareness does not
important to point out rhar the cognitive mic awareness; e.g., Morais, Cary, Alegria,
demands made by these different tasks vary, & Bertelson, 1979). Instead, ir appear s to
so that performance will reflect not just largely depcnd on direct instruction in read
phonological awareness per se but also ex ing and spelliug (e.g., Liberman et al.,
traneous task demands (e.g., McBride 1974) or on the receipr of training ar the
Chang, 1995; Yopp, 1 9 8 8 ) . One recent way phonemic leve! (e.g., Byrne & Fielding
logical awareness, derived from work on & Bertelson, 1986). Although sorne prelir
guistic" processing (see Gombert, 1992). awareness tasks (e.g., Stuarr & Coltheart,
Gombert argues rhat one approach is to 1988; Thomas & Senechal, 1998), in gener
shared phonological segments (such tasks Phonemic development can be rapid once
require conscious awareness) or the identifi As this chapter is mainly concerned wirh
ical segments (such tasks require metalin on the processes underlying the develop
guistic processing, meaning that the child ment of segmented representations of words
knowledge explicir in arder to perform rhe cessinf:. In general, this does not include tbe
phonological segments (e.g., Butterworth, rime, and these are the phonological units
1992), which mighr form rhc basis of re highlighted in early linguistic romines such
word stimuli (see Coswarni & East, 2000, development seem to be less vulnerable to
Studies suggest that there is a develop later. For example, onser�rime correspon
mental progression from phonological dences are still available to adult acquired
awareness of "large" segments or units of dyslexics and alexics even when phonemic
phonology (syllables, onsets, and rimes; the correspondences are not (e.g., Patterson &
onset in a spoken syllable refers to rhe con Marce!, 1992; Shallice, Warrington & Mc
sonant phonemes before the vowel, and the Carthy, 1983), and syllab1c correspondences
rime is the vowel pbonemes and any subse- may still be available to adult phono!ogicai
� Early Plwnological Development
113
dyslexics even when onser, rime, and Dollaghan, 199 4 ). Although phonological
�
phoneme correspondences are not (Lesch & output typical!y remains imprecise initially
·�
Martin, 1998). (sometimes only a regular caretaker can de
� Given its imporrance for literacy, surprising the brain in a way that will distinguish them
ly litde work has been done en the linguistic from ocher words and allow the child to rec
�
and lexical factors that might determine the ognize them accurately and quickly during
all children. In thinking about what rhese year-old probably knows the words "cot,"
· �
-facrors might be, 1 have found the recent "cat" and "cut," "hor," "not"_ and "lot"
processes that are an imnnsic part of lan berween rhese similar-sounding words borh
�
guage acquisirion useful ("lexical restructur quickly and accurately, child linguists argue
t:;@ ing theory"; Metsala, 1999; Mersala & that children must begm to represent rhe se
Walley, 1998). Lexical restructuring theory quences of sounds thac constiture each
�
is based on the premise that in the normal known word in their brains. They musr rep
l'll!!
of phonetic features with age (e.g., Fowler, ed thar segmental phonology is represenred
acquisition. When children first begin to ac phonological characteristics. Early in lan
�
quíre spoken language, in infancy, rheir spo guage development, the child needs to dis
li;®
ken vocabularies ccnsist of rathcr few criminare relatively few unique words, and
in terms of certain semantic features ("Dad logical forms will suffice (e.g., Ferguson,
il1'1ll
dy" may refer to a person of a certain sex 1986; Juscyk, 1993). However, as more and
� and size) and also m terms of certain phono more words are acquired, children are
logical features (rhe child can recognize rhat thought to begin to represent smaller seg
�
"Daddy" is a diffcrent word from "dog ments in words. From the phonological
� gy"). At this developmental time point, the awareness data discussed earlier, it seems
motor program for producing the word likely that children will first represent the
�
"Daddy" is probably quite sketchy, and the number of syllablcs in a word and the "on
� sets" and "rimes" in each syllable. This
child may say "Dada" or "Da" when in
tending ro name "Daddy." Small vocabu process may begín as early as age 1 or 2
�
lary size at this time may also lead roddlers (e.g., Swingley, Pinto & Femald, 1999).
�
to "overextend" rhe words that they do The syllable is thought to be the primary
know, and rhe child may use his or her word linguistic processing unir for English, as it is
for "Daddv" to refer to such visirors as the distinguished by a nurnber of audirory cues
milkman and the postman, and to uncles including rhythm and stress. Within the syl
and other adult males as well. lab!e, the most prominent phonological seg
Most children go through a dramatic ments are the onset and the rime. Linguisti
burst in naming activiry between rhe ages of cally, the rime is a salient phonological umt
1 and 2 years. Spoken vocabulary suddenly and seems to have an organizing function
grows exponentially (by the age of 6, the av far English phonology. Many of the lan
erage child comprehends 14,000 words; see guage games, linguistic routines, and nurs-
-,
example, popular nursery rhymes have processing from infancy onwards {e.g.,
_ -- strong rhythms that emphasize syllabifica Eimas, Siqueland, Jusczyk, & Vigonto,
tion (think of Humpty Dumpry), and many 1971) but, rather; emerges as a representa
contrast rhyming words in ways that distin tional unir via spoken language experíence.
guish the enser from rhe rime (e.g., "Twin However, an importanr aspecr of early lan
kle Twinkle Little Star" rhymes "star" with guage experience that is missing from the
"are," and "Incv \v'incy Spider" rhymes current version of lexical resrructuring rheo
Mersala and Walley (1998) argue that the cal similarity relations that characterize dif
grearer pressure for restrucruring, and con As lexical restructuring theory proposes
ousers, and rimes in many of the words in sounding words constitute the basis for the
also depends on word frequency or familiar seems logical that the namre af the phono·
'
:1
:
iry. Words that are encountered many rimes logical neighbors in the child's kxican in
:
or rhat were acquired early are more likely different languages will influence this devel
, 1
to have been restructured, beca use rhe child opmental process. The tradirional linguisric
needs ro eccess rhese words rapidly and ac similarity metric for defining a phonological
lcxical resrructuring depends on rhe num words that differ by rhe addirion, delerion
lexicon {"neighborhood densiry"). Words in cording to chis metric, rime neighbors such
"cor;' wirb manv similar-sounding neigh such as "cough," and "consonant" neigh
bors) should experience most pressure for bors such as "kit," are ali considered to be
from a large number of orher extremely "cot." However, given the psychological
similar words. \Vords in "sparse" neighbor salience of rhe rime ro young children, it
hoads, which musr only be distinguished seems possible thar many phonological
fram a small number af other extremely neighbors in English mighr be rime neigh
similar words, should experience less pres bors. Far example, a word such as "cor"
sure for restructuring. Metsala and Walley (dense neighborhood) has been estimated to
propase rhat the degree to which segmenral have 49 phono!ogical neighbors (Luce &
represenration has taken place wi!I deter Pisoni, 1 99 8 ) , 24 of which are rime neigh·
mine how easily rhe child will become bors (49%). A word such as "crib" (sparse
phonologically aware and will learn to read neighborhood) has been estimated ro have
primarily as the result of spoken vocabulary neighbars in the English phonological lexi
growth and assocíated changes in rhe famil can, rhis might help to explain rhe salience
iariry of individual Jexical irems and in and utility of onset-rime representations in
' 22.5
M 12.2 3.8 6.5 38.1 22.1 7.2 8.7 9.4 3.9 1.5 4.0
SD 10.1 7.8 3.2 4.4 5.3 6.5 3.0 4.6 2.9 2.9 1.8 2.9
Note. Neve represents all phonological neighbors that diffcr from a targct word by onc onset, uowe{, or coda
substitution, dclction, ar addition. RN, rime neighbor ; CN, consonant phoneme nclghbor; OVN, onset-vowd
ncigbbor.
. - . -------- --
(OVN), and consonant neighbors (CN) in early lexical resrructuring to the rime leve!
tively. We used two rneasures of pbonologi Because rhe similarity indices shown in
cal neighborhood; the tradicional speech Tables 9.1 and 9.2 were derived from an
processing definition (addition, deletion, ar adulr lexical database, ir is not clear how
subsriturion of one phoneme, called here N applicable these indices are to young ehil
:t. 1), and a definition based on a linguistic dren. Children's lexical neighborhoods are
analysis acccrding to which monosyllables smaller than those of adults and are con
can be coded in rerms of the phonological stantly being updated, meaning thar neigh
units onset, nucleus, coda (see Treiman, borhood statistics are much more dynamie.
1988, far review). This seeond measure, From this perspective, estimares of neigh
here Neve, was derived on the basis of rhe borhood similarity based on adult data can
has demonstrated the psychological salience {e.g., Charles-Luce & Luce, 1990, 1995;
of onsets and rimes for young children. The Dollaghan, 1994; Legan, 1992). Neverrhe
chief difference psychologically would be less, there is no reason to suppose that rime
that whereas words such as "spor" and neighbors are under-represented in the
"cor" in the Neve darabase, they wou!d not one developmenral function of language
count as rime neighbors of cot in the N :t. 1 play such as nursery rhymes may be to high
darabase. Tables 9.1 and 9.2 show our light che salience of rime neighbors in Eng
analyses (De Cara & Goswami, 1999). lish, and nursery rhymes certainly do not re
lt is clear from the tables that rime neigh- stricf their rhyming panerns to N = 1
AII monosyllabic words High N (Neve e!: 32) Low N (1 s Neve :S 13)
M 13.4 5.7 3.8 3.9 24.9 11.5 7.2 6.2 5.0 1.7 1.5 1.8
SD 8.7 4.6 3.2 3.1 7.6 4.7 3.0 4.6 3.2 1.7 1.8 2.0
Note. N :; 1 represents al! phonologica! neighbors that differ from a targer word by one pbor.eme subsrirution,
deletion, ar addaicn. RN, rime ncighbor; CN, consonant phoncmc ncighbor; OVN, onser-vcwel nc,ghbor.
'
116 STRANDS OF EARLY LITERACY DEVELOPMENT
neighbors (e.g., Hickory Dickory Dock ness of one rime compared to another,
rhymes dock wirh clock, and Jack and Jill However; if such an effect could be demon
rhymes down. with c.rown; see also Goswa srrared, it would help us to understand the
It is nota priori clear wherher more sirni phonological awareness must be a cense
[arities or differences would be expected be quence of how the brain processes language.
neighbors for rwo differcnr estimares of ear rhen effects of neighborhood density should
and Logie (1980) and rhose given by Morri even when the chi!cl is recognizing a salienr
son, Chappell, and Ellis (19--9-7).(see..Oe Cara - phonolog1cahrnit in very familiar words.
& Goswami, 1999). Table 9.3 shows rhese In recent work (De Cara & Goswami,
results far monosyliables onlv, Although 1999), we used rwo different phonological
rhese analyses did not attempt to provide es· awareness rasks to test rhis hypothcsis. One
tunares of absolure lexicón size, they illus was the oddiry rask pioneered by Brad!ey
trare that rime neighbors are not likely to be and Bryant (1983), in which children muse
underrepresenrcd in children's phonological select the "odd word out" from a triple of
The nexr step in tesring rhe proposal rhar lor, wair," and triples of words from sparse
the basis for the emergence of phono!ogical neighborhoods, snch as- "mud, thud, good."
j awareness is the child's implicit campar· The second task was the same-different
see whether there is any evidence rhat chil Zukowski ( 1 991 ) , in which children musr
dren are more accurate at processing rimes decide whether or nor rwo spoken words
actually affect the developmenr. of phono- abour pairs of words from dense neighbor
logical awareness, then children should find hoods, such as "lick, sick" and pairs of
ir easier to decide that "cot" and "pot" words from sparse neighborhoods, such as
rhynie than to decide that "thud" and "soot, loor." Both rasks were chosen be
"m ud " rhyme. Given rhat rhese are ali high ca use rhey should be measures of "epilin
seern counrerinruitive to propase rhat chil tasks, we hopedro tap the leve! of phono
dren will show better phonological aware- !ogical processing rhat míghr be expected ro
""""'
-
Early Phonological Development
""""
-
-
117
ported by Metsala (1999) using a blending process. The ways in which this rnight oper
�
rask. In her task, the child had to choose, are in different languages an1 for different
� far example, rhe picrure of a "bush'' wheñ phonologicai units (e.g., onset, rime, and
rhe experimenter said lb/-/uf-/shl. She found phoneme) create important questions Ior fu
�
---- rhat 3- and 4-year-old childrerÍ performed ture research. For example, as discussed ear
,;@
significantly better in this simple phoneme lier, the representation of phoneme-level in
blending task when the carget words were formation mighr be expected to be largely
from dense neighborhoods rarher than from dependenr on the acquisition of literacy, be
sparse neighborhoods, This version of rhe cause the feedback provided by graphemic
'
1 ,.
blending task may also be an "epilinguistic" information will help che child to represent
task, as rhe children basically had to recog segmental information at the phonemic lev
-
'
nize words spoken very slowly by che exper e! (see Morais, Alegria, & Content, 1987;
1 -
imentec The )'oung age-of the particrpacing Goswami, in press a; Goswami & Bryanc,
, o,¡¡¡¡; children makes Metsala's demonstration of 1990, for reviews]. The effects of literacy on
important.,
lexical restructuring would also be expecred
-
t'IIÍÍI
Phonological Neighborhoods and
Reading Development
man),
use only one spelling
variery of
-
>ll!!il
in our experiments, it was dear chat che ef
�
ers. Tlus is not surprising, because learning spondences in reading acquisirion (e.g.,
-
,,. ro read and spell is known to affect phono
logical
has
awareness.
demonstrated
Although
this reciprocal
most research
relation
Goswami,
Goswami,
1997).
Gornbert,
Porpodas,
& De
&
Wheelwright,
1998;
ship at the phonemic level (Goswami & chis variabiliry represents or interacts wirh
'
1 •
_
Bryant, 1990, for review), effects at the rime rhe phonological characteristics of ditferent
-
,,.
"1l!I
East, 2000). For example, sorne of the rimes
from
hoods
che
of
densest
English
phono!ogical
monosyllables
neighbor
quent rimes in English monosyllabic words, So far, we have seen chat there is sorne evi
�
but can be written as in "year," "here," dence in support of the propasa! that one
� "cheer" and "tier." As soon as children be importanr basis for rhe emergence of
gin leaming to read and to spell, rheir phonological awareness is the child's im
fected by rheir orthographic knowledge. ing words, which are a natural pare of lan
Orher very dense neighborhoods showing guage processing. Metsala and Walley
this inconsistency of rime spelling include (1998) also proposed that the degree to
the neighborhoods for /el/ (sail, whale, which segmenta! representation had raken
:
'i';ÍI gaol), /Er/ (share, hair, where, their; swear) place would determine how easily che child
and /orl (shore, for, roar, war). would learn to read and write. In a trivial
]íÍii
The possibility that che experience of sense we already know chis to be true, be
_
i'i)
learning to read and spell an alphabetic or cause there is an exrensive literature docu
play in the lexical restructuring process is phonological awareness and their reading
�
not specifically discussed by Metsala and and spelling development. We do not yet
G.,
� Walley (1998). However, ir is plausible to know whether specific predictions made by
propose that the act of becoming lirerare the !exical restrncturing hypothesis would
will m itself affect the lexical restructuring be supported for literacy acquisition, how-
:--
�
:::b
! ..
i�'
�·,i:
s.
.. 118 STRANDS OF EARLY LITERACY DEVELOPMENT
\i
l .,
cver, According ro the hyporhesis, the fac h ó o d " effecrs in borh list-reading and story
· 1,:
'•
,. tors that govern the degree of lexical re reading tasks. The less skilled readers in
¡; structuring rhar has occurred for certain rheir sample were also sensitive to rime
words should also govern early literacy ac neighborhood size when followed up 1 year
'
.,
"
,.
¡-;
cutstnon. Far examp!e, early-acquired la ter.
words should be easier ro learn toread and These data are at leasr consistenr with the
l
í
..
,. to spell, and words in dense phonological lexical resrrucruring hypotbesis in regard to
n neighborhoods should be easier to ieam to rime processing. What is required now is a
li
read and to spell. This is because these are systematic study of both phonological and
¡,
t, the words rhat are more like!y to have seg orthographic neighborhood density in arder
u
menred representations. In theory, ir should to disentangle phonological and ortho
"
..
•
be easier for children ro connecr rhcse graphic rime cffects in re:admg acquisition .
¡!,
..
words to lerters, because sorne of the sound
t.
,,
•
As argued previously, rbe phonological seg decades ago, Bradley and Bryant ( 1978 ,
1
·
mems that should a priori be mosr affecred 1 983) demonstrated the importance of
•
by spoken vocabulary growrh and associat rhyme awareness for reading development
· lexical irems and interitem phonological oddiry rask describe.d earlier. Bradley and
similaricy relarions should be the syllable, Bryant found rhat rhyme awareness mea
the onser, and rhe rime. Although we do sure.d in preschoolers was a significant pre
nor yet know wherher, for exarnple, words dictor of later progress in readlng and
in dense phonological neighborhoods are spelling, e.ven when other factors such as IQ
easier to learn to read and ro spell, it has and memory were conrrolled in mulriple
been shown rhar beginning readers are regression equations. They also reponed
more likely ro read words with large ortho that backward readers had poorcr rhyming
graphic "rime neighborhoods" correctly skills than <lid younger children reading at
than words with moderare or small orrho the same leve] as rhem. MacLean, Bryant
1
graphic rime neighborhoods (Leslie & Cal and Bradley ( 1 9 8 7 ) found a significant con
hoon, 1995). Orthographic neighborhood nection bctween rhyming skills at ag:e 3 and
rime spellings nored earlier (e.g., "year," ple 2 years latcr, Bryant, MacLean, Bradley,
"here," "cheer," and "tier" would a!l be in and Crossland (1990) reportcd a significanr
che same phonological rime neighborhood relationship between nursery rhyme knowJ.
bur differenc onhographic rime neighbor edge ar age 3 and success in reading and
hoods). The orthographic rime neighbor spelling at ages 5 and 6, e.ven afrer factors
hood is a measure of how many cther such as social background and IQ were con
same rime. Thus a word such as "ship" has More recently, Chane.y (1992) reported
(dip, hip, chip, lip, nip, skip, rip, slip .. ere.) rhyming rasks, and she found relationships
as well as a large phonological rime nergh berv,,een this early rhyme awareness and lat
has a small orrhographic rime neighbor Chane:y's rhyme measure (rhyme produc
hood (deem, teem), even though the phono tion) was rhe best correlare of the other met
logical rime neighborhood is much larger alinguisric skills in her srudy. Burgess and
·,
{13 rime neighbors, e.g .. cream, dream, Lonigan (199 8 ) found rhat "phonological
rhcme). Leslie and Calhoon (1995) found senmiviry" measure:d in a large sample of
�SiA
·
,q-,¡.
Early Pllonological Deuelopment
119
�
onset and rime oddity task performance and work in more than ene way. For example
�
rasks of blending and segmenting com in Bryant et al.'s (1990) study, a path analy:
¡
7J/J
pound words inro words or syllables) pre sis showed a route from nursery rhymes to
� dicted performance in borh letter-name and rhyme awareness to reading and an inde
letter-sound knowledge tasks 1 year Iarer, pendent route from nursery rhymes to
�
The latter were argued to be rudimemary phoneme awareness to reading. This is
1'$jj reading skills. Cronin and Carver (1998) shown in Figure 9.1. On the basis of this
used an onset oddiry task and a rhyme kind of evidence, we have previously argued
�
matching task to measure phonological sen (Goswami & Bryant, 1990) that rhyme
!a!ij
siriviry in a group of 57 5-year-olds, and awareness might comribute ro reading de
found that phonological sensirivity signifi velopment in at leasr rwo ways. First, rhyme
�
cantly discrirninated the three different might contribute to reading because rhyme
� achievement levels used to group rhe chil _awareness is a predicror of which children
dren in terms of reading abiliry ar rhe end of will find ir easier to develop phoneme
..,.
'�
firsc grade, even when vocabulary levels awareness. In terms of lexical restructuring
� Scher; and Williams (1998) showed rhat sented the onsets and rimes in single-syllable
kindergarten nursery rhyme· knowledge was words will go on to segment rhe onsct and
-
O!@
the strongest predictor of word attack and
ed for an additicnal 11 % and 18% of the Children's awareness of rhyme might thus
�
variance, respcctively. allow them to form implicit phonobgical
•
�
This selection of studies showing a con
the subsequent acquisition of literacy or lit egories with srrings of lerrers, children could
Q
eracy-related skills demonstrares that the learn spelling sequences for onsets and
� developmental pathway to reading acquisi rimes, which are important spelling cate
tion in English critically involves rhyme. gories in English {see later). Children might
�
However; the connection with rhyme might rhus be able tO use analogies berween words
!$
il!!ll Reading
l't)
�
Phonernes
,!j
¡:$
i;:;j
Rhyme
�
via analogy
é'$
Rhyme
_,,,
;$ili
'-�----
STRANDS OF EARLY LITERACY DEVELOPMENT
120
sharing spelling patterns for rimes as one 91 %), the pronunciation of vowcls was very
Research has shown that borh children and sharing a C1 V spelling had a consistent pro
adults can use analogies to decode unfamil nunciation {e.g., bea in beak and bean), 77%
tar words (Glushko, 1979; Goswami, of CVC words sharing a VC2 spelling had a
sponraneously ln reading (after about 10 This sratistical analysis of che spelling sys
years of age, see Marsh, DesDerg, & Coop rem of-English-cshows that rhe spelling
er, 1977; Marsh, Friedman, \\7dch, & Des sound consistency of rhe wrirren language is
berg, 1981), it is now known that analogies greatest for initia l consonants (onsers), final
are used rnuch earlier. Even beginníng read consonanrs, and rimes. lt indicares rhat the
ers can make analogies between shared conrext of che final consonant ar canso·
spelling partems in words (see Goswami, nants can disambiguare che pronunciation
1999, far a recent overvíew). Younger read of rhe vowel (e.g., "a" makes a different
ers make fewer analogies than do older sound in cat, ball, car, day, saw, cake, and
readers. This was thought to be because care, bue these different phonemic corre
rhey have smallcr reading vocabularies spondences are consisrenr within rhyming
(Goswami, 1986). Analogies are also used groups: car, mar, bat ; ball, fall, wall . . .
when children are reading stories (Goswa ; care, dare, stare ; and so on). These
mi, 1988). This suggested rhat analogy is a sratisrical relarionships may help us to un
largely implicit process, driven by che derstand the developmental parhway be
child's phonological skills and the ortho rween rhyme awareness and reading.
che onhography that they are learning ro many of the alremative pronunciations for
read. The facr that an.'.'1.k•gy can be an im vowel graphemes become highly predicrable
pli clt process does not mean rhar we do not if rhe rime is considered as a unit. A child
need ro teach children to use analogies. The with good onser-rime awareness is thus in a
use of an ana!ogy stratcp· shou[d develop betrer posirion to discover the stabiliry of
One reason why orthographic analogies anorher possible reason for the link berween
may be useful in reading English is suggested rhyme awareness and phoneme awareness
Treiman, Mullennix, Bijeljac-Babic, and Far exarnple, the vowel digrapb "ou" is al
Richmond-We!ty (1995). They calculated ways used to represent one sound in the
how many times individual lerrers mapped to rimes shout and house. lt is used to repre
rhe same sounds when tbey occurred in the sent a different sound in soup and group,
same posit1ons across different words far ali but this is predictable. The sound /u/ is
the monosyllabic words of English wirh a spelled differently in rimes such as tooth
consonanr-vowel-consonant (CVC) phono and spoon but is again predictable given the
logical strncture (e.g., "c" in cat, c11p, cone, rime. In rruly "alpbaberic" languages, read
etc., "p" in cup, top, cheap, etc.). The CVC ing instruction based on teaching children a
words in this analysis included words spelled fixed sequence of grapheme-phoneme cor
with vowel digraphs, such as "rain" and respondences is commonplace (e.g., Wim
"beak," and words with "rnle of e" mer, 1993). However; far a language such as
Treiman et al. {1 9 95) found chat whereas rhe phoneme correspondences and rimes may
pronunciation of initial and final consonants be more appropriare (Goswami & Easr,
parisons between similar-sounding words consisrent wirh these possibilities. Far ex-
�
rhat are a natural part of language process- ample, the study by Burgess and Lonigan
'11lP ing, rhen children whose environment en- (1998) nored earlier found that receptive
z;¡¡t courages chcm to make such implicit com- and expressive oral language skills mea-
comes to learning toread and to spell. Simi- __ at time 1 predicted performance in the oddi
I mi, larly, children who have experieñced inrer- "ty task 'measured 1 year iater (onset and
vention designed to teach them ro compare rime versions). This is consistent with the
1·�
ar an advantage when it comesfo Iéáfiiirig --ship-6etween vocabulary sblls and the de-
1 �
"'® examine these hypotheses. Nevertheless, the was a connection between lexical develop-
studies that have been done suggest thar di- ment at age 1 (measured by the number of
'.l:!lll
rect instructional influences can have an mappings of meanings to speech units thac
�
important effect on the development of each child had at 1 year) and phonological
""'
'1!ij
are reflected in rhe orthography appears to significantly more accurately with early-
be necessary, at leasc far children at risk of acquired words than witb larer-acquired
underachievement in reading. For these chil- words. Alrhough not directly relevant to the
' �
dren, direct insrruction in sound-letter cor- hypothesis that a richer language environ-
I �
At present, we can only speculace how rhe Cupples, and Lovegrove ( 1 9 9 8 ) have shown
""
1"ll, child's environmenr might encourage the that rhyme detection measured at 4 years 11
� language processing and thereby stimulare This could reflect the fact rhat children who
the developmental pathway to phonological have rcstructured more words to rhe on
sllil
awareness. However, it seems likely that set-rime level at time 1 (consequently per
children who experience a rich lmguistic en forming better in a rhyme-detection task)
vironment in early life will acquire a larger are those same children who ate activelv ac
�
vocabulary a t a faster rate rhan do children quiring more words, and who thus have
� larger vocabularies at time 2.
who experience a poorer linguisric environ
ment. Vocabulary size and rate of vocabu It has also been shown that children from
\J�
lary acquisition are both hypothesized to be lower socioeconomic status homes tend to
important far lexical restructuring to occur. perform less well on measures of phonolog
Similarly, it seems likely that children whose ical sensitiviry than children from higher
caretakers promote linguistic activities such SES homes. The reasons far this are not
as language games and nursery rhymes will well understood. Far example, Dickinson
spend more time impiicidy comparing and and Snow {1987) measured phonological
children whose caretakers do not promote who were ali attending high-quality day-
122 STRANDS OF EARLY LITERACY DEVELOPMENT
care programs, and found that those from "hat" cou1d be changed into a word such as
high-SES backgrounds performed at a sig "rat" by discarding the onser and retaining
nificanrly higher leve! than rhose from low che nme. The orher half of che experimental
SES backgrounds. Raz and Bryant (1990) grnup conrinued to receive phonological
gav,: rwo phonological awareness tasks to training onlv, At the end of the second year
high- and low-SES English children (initlal of the study, the children in the experimen
phoneme identification and rime oddiry), tal group who had had plastic lerrers train
and found that those from lower-SES back ing were 8 months further on in readmg
grounds had lower levels of phonological than the chiidien in the semcntic control
awareness at school cntry. These children group and a year further on in spelling.
were not significantly worse than rheir Compared to children who had spenr the in
--high-SES counterparts at rhis stage, but af tervening period in an additional unseen
1
ter a year in school the gap had widened control group, they were an astonishing 2
I'
dramatically, · and the low-SES children years further on m spelling, and 12 monrhs
______ --sh·o,;;1ecf-a .significanr deficir in phonological in reading. The guins made by the children
awareness. Bowey (1995) has reported sim who had conrinued to receive phonological
ilar SES findings in a sample of Australian traming only were not significant but still
children. She compared high- and low-SES notable. This study suggests that there is a
children in early word readmg achievemenr, clear ccnnection between training children
and found significant differences by SES how the alphabet is used to represent
srarus. She also reported rhar these differ sounds and reading and spelling develop
ences in phonological awareness and not by Similar results were found in a large study
A number of srudies have used drrect inrer words and attending to the first sounds in
vcnrion to improve children's phonological the childrens' names. The aim of the pro
awareness and measured consequent effects gram was "to guide rhe children to discover
on lireracy, For example, as part of the lon and attend to the phonological structure of
gitudinal study discussed earlier; Bradley language" {p. 268). The effectiveness of the
and Bryant (1983) took rhe 60 children in program in attaming this aim was measured
rheir cohorr of 400 who had performed by comparmg the children's performance in
most poorly in rhe oddity rask ar 4 and 5 various metalinguistic tasks after training to
years of age and gave sorne of rhem 2 years that of 155 children in an unseen control
of rraining in grouping words on the basis group. The trained ch1]dren were found to
of sounds. Training was based on a picture be significantly ahead of the control chil
sorting task in which the children were dren in a variety of metalinguistic sk1lls in
taught to group words by onset, rime and ciuding rhyming, syllable manipulation, and
vowel and coda phonemes (e.g., placing pie phoneme segmentation. The long-term ef
tures of a hat, a rat, a mat, and a bat togerh fect of the training on the children's reading
er for grouping by rime). A control group and spelling progress in grades 1 and 2 was
learned to sort the same picrures by seman also assessed. The impact of the training
tic category (e.g., placing pictures of a rat, a was fow1d to be significant at both grades
pig, and a cow togerher for "farmyard ani for both reading and spelling, although ef
mals"). fects were stronger for spelling.
Half of the experimental group then spent A recent German study of the effects of
the second year of che srudy learning how providing training in phonolog1cal aware
the shared phonological segments in words ness in kindergarten found a similar pattern
such as "har," "rar," and "mar" were re of results to that reported by Lundberg et
flected in shared spelling. The children were al. (1988). Schneider, Kuespert, Roth, Vise,
given plastic letters for chis task, and were and Marx (1997) developed a 6-month
taught, for example, that a word such as metalinguistic training program covenng
L L,, -
Early Phonotogsoú Deueíopment
syllables, rhymes, and phonemes and gave bases for the emergence of phonologica!
it to a sample of 130 kindergarten children awareness. lt was argued that rhe nature of
were then monitored in grades 1 and 2. guages wdl influence this deve!opmcntal
Schneider eral. (1997) found significant ef process, and it was then shown rhat phono
�
fects of rhe metalinguistic training program logical neighborhood characteristics in Eng
unseen control group, as would be expecred units of the onset and the rime. As syllsbles,
�
from Lnndberg et aL's (1983) results. They onsers. and rimes are also emphasized in
� also found significant long-term effecrs of early linguistic routines such as nursery
metalinguistic training on reading and rhymes, ir was suggested that informa! envi
�
spelling progress, with stronger effects for ronmental experiences can promote rhe or
� spelling. Recently, the same research group ganization of the mental lexicon around rhe
a¡¡¡ . has reporred signjficanr effects of the same syllable and rhyme. Diiect training m
training program on the reading and phonological caregcries can also promore
�
spelling progress of German kindergarten such an organization. lt was argued rhar the
� children assessed as being at risk for dyslex acquisition of lireracy will influence the fur
made when rhe metalinguistic program wa s tion of existing categories (e.g., when a ixe
;;,;,¡)
combined with direct trammg m quent rime has multiple spellings, as in
"""'
¡¡¡;g findings reported by Bradley and Bryant future important directions for research on
,.,,. (1983) for an at-risk sample. Whereas chil phonology and reading acquisition include
dren who are not at risk far later lireracy research designed to help us to understand
awareness training alone (cf. Lundberg et are due to basic brain function and which
�
al., 1988; Schneider et al., 1997), those are culturally influenced and to help us to
!C'1ij who are likely to have specific problems in understand the relarive weight that needs to
1
acquiring literacy seem to need a combina be given to the different phonological units
¡¡¡¡;¡
tion of metalinguistic and letter-sound of syllable, rhyme, and phoneme and their
� training. Interestingly, rhe Schneider et al. connections with sequences of letters in d1f
2,;j) {in press) srudy included an ar-risk group ferent orthographies. A better understand
1 rhat received letter-sound training alone, ing of the factors that affect the transfer of
mlll ¡
without metalinguistic rraining. This group phonological awareness :1.cross different lan
either performed at comparable levels in guages is a rhird important goal far future
�
later reading and spelling progress to che work.
�
1 metalinguistic rraining alone group or per
This chapter has reviewed evidence for the beginning literaey (pp. 263-287). Ht!!sda!e, NJ:
Er!baum.
propasa! that vocabulary acquisition pro
Bowcy, J. A. (1995). Socioeconomic �t:1.tus differ
duces developmental prcssure for the child
ences in preschool phonological sens,tinty and
to make implicir comparisons between simi
first-grade reading achievement. Journal o{ Edu
lar-sounding words in the mental lexicon, cational Psychology, 87, .176---487.
and thar such comparisons are one of rhe Bradley, L., & Bryant, P. E. (1978). Difficulties m