Can Phonemic Awareness Be Trained in Kindergarten
Can Phonemic Awareness Be Trained in Kindergarten
Can Phonemic Awareness Be Trained in Kindergarten
Most studies of child language have focussed on This ability is of obvious importance for reading
the development of the primary linguistic abilities acquisition. Because of the close grapheme-sound
9f speaking and listening. In the last decade, how- relation our alphabetic script may require beginning
ever, the child’s sensitivity to the formal properties readers to pay special attention to phonemic struc-
of language has attracted increasing attention, es- ture. However, the phonemic units seem to be
pecially among reading researchers (reviews are particularly difficult to handle for many beginners.
given by Ehri, 1979; Golinkoff, 1978; Lundberg, The difficulty to access phonemes is related to the
1978). Mattingly (1972) seemed to be the first to fact that “phonemes are not represented in the
point out linguistic awareness as a powerful source acoustic signal in discrete segments, but rather
of individual differences as opposed to linguistic are merged-‘encoded‘-into larger units of ap-
ability. The ability to approach language in a situa- proximately syllabic size” (Liberman et al., 1977).
tionally detached and conscious, self-reflective way Thus, getting access to these coarticulated or en-
appears to be of critical importance for acquir- coded elements is more a matter of abstraction and
ing reading and writing skills. reflection than perceptual discrimination.
Around the age of seven most children show Mattingly (1979) argues that phonemic blending
various signs of an increasing linguistic awareness and segmentation may be better characterized
expressed at the phonological, morphophonemic, simply as access to linguistic elements rather than
lexical, and syntactic-semantic levels. One of the conscious awareness of linguistic activity. While
many aspects of linguistic awareness is that a sen- recognizing the lack of conceptual clarity we stick
tence, previously perceived as a gestalt including to the term phonemic awareness in the present
its extra-linguistic context, now can be segmented report for reasons of current convention and to
into phrases, words, syllables, and phonemes, and emphasize the metalinguistic nature of the phoneme
that these linguistic units can be synthesized to processing.
create syllables, words, phrases, and sentences. A more urgent question than the terminological
ScandinnuinnJournal of Psychology, 24
36 A . Olofsson und I . Lundberg
is to what extent and in what ways the ability to berg, Olofsson & Wall (1980) also support this
handle phonemes can be stimulated by training conclusion.
at the preschool level. Is it possible to prevent The programs also included several less de-
reading difficulties in school by an early stimulation manding tasks such as analysis/synthesis of syl-
program? lables or morphemes and recognition of rhyme.
Impressive correlations have been reported be- Rhyming seems to be one of the earliest expres-
tween reading ability and the ability to isolate and sions of the child‘s metalinguistic development.
handle phonemes (Lewkowicz, 1980; Lundberg, Recognizing a rhyme is a rather intuitive act which
1981; Zifcak, 1981). However, it is not clear wheth- does not seem to require more explicit awareness of
er phonemic awareness is a prerequisite for read- the formal aspects of language. Lewkowicz (1980)
ing acquisition or develops as a result of reading comes to the conclusion that it may be advisable
instruction. As long as only correlational data are to start teaching rhyme recognition before seg-
available, the causal direction of the relationship mentation and blending tasks are introduced, al-
remains in principle obscure. An attempt to clarify though it is still somewhat unclear whether it should
the issue by applying path analysis on longitudinal be considered a prerequisite for those skills. Al-
data was presented by Lundberg, Olofsson &Wall though the central concern is phonemic awareness
(1980). That phonemic awareness is more than it seems reasonable to start with the more acces-
a mere indication of or a consequence of learning sible aspects of linguistic forms. Thus, the child
to read is suggested in several recent studies where first learns to redirect attention from content or
it is demonstrated that the skill can be taught to meaning to form and then eventually detects move
prereaders and that it may contribute to success abstract dimensions of form.
in beginning reading (Lewkowicz & Low, 1979; Among the more complex tasks that have been
Marsch & Mineo, 1977; Williams, 1980). Though used the following can be mentioned: counting
these results may seem rather encouraging, more the number of phonemes in a word (Zifcak, 1981),
direct and clear evidence from carefully designed deletion of a phoneme in a word (Bruce, 1964),
experimental studies is needed. If lack of phonemic specifying which phoneme has been deleted (Lund-
awareness skill is demonstrated to be causally berg, 1981), and phoneme substitution (Elkonin,
related to reading and spelling difficulties and if 1973). In addition to phonemic awareness these
the skill can be taught at preschool level then the tasks seem to require various degrees of memo-
instructional implications should be serious!y con- ry and attention capacity. In young children this
sidered. capacity is easily overloaded and task failure is
The present study is part of a long-term project difficult to interpret. Thus, we found it advisable to
on phonemic awareness and reading. Here we ad- avoid tasks of this complex nature in our program.
dress the more limited question if phonemic aware- Among the tasks which require phoneme analysis
ness can be taught at all to prereaders. A system- isolating the initial phoneme of a word seems to
atic training program is described, and its effects be the easiest one-“a clearly useful task” in Lew-
on preschool children is evaluated in an experi- kowicz’s wording. It is necessary for the child to
mental setting. The long-term effects on reading master this task as a first step in phonemic seg-
and spelling progress in school will be reported mentation. Several interesting and attractive task
in a later study. varieties can also be designed with initial phoneme
analysis.
attentional strategies or cognitive style was kept The treatment of the experimental groups varied along
t o a minimum. Gross measures of nonspecific a dimension of structure. Group 1 had a highly structured
program with well specified tasks arranged in a strict
processes such as intelligence or memory will give
sequence of half-hour sessions (3-4 times per week).
normal distributions of test results because a large Group 2 also had a well structured program but the
number of subskills ar e tapped and pooled in a teacher was not very much supervised and was mainly
haphazard way. In contrast, the operation o f single guided by the written instructions. The total amount of
skills which underlie performance on the test will time spent on the tasks turned out to be somewhat less
for this group. The third experimental group followed
yield a bimodal distribution. These skills ar e either a more spontaneous, non-scheduled variety of the pro-
possessed o r not possessed by the children. Cal- gram. When the teacher found it convenient a task was
fee (1977) is one of the few researchers who h as introduced more or less “on the fly” in the normal play
bothered t o inspect the distributions of scores any activities that occupy most of the curriculum in Swedish
closer. He found that highly selective tests (i.e. kindergarten. Thus, the order between different tasks,
as well as the selection of particular phonemes was a
phonemic segmentation) yielded bimodal distribu- more haphazardous affair under this condition. A rea-
tions. sonable expectation was now that the effect of the pro-
If, on the other hand, the main problems of ab- gram was higher the more structure that was imposed
stracting the invariant features of the acoustic signal upon it.
Control group 1 worked with a program formally fairly
(assessing the phonemes) are perceptual rather than equivalent to the experimental program. It consisted of
conceptual a more normal distribution should be a number of non-verbal tasks, such as sound localization,
expected. In such a case the difficulty of each item exercises involving discrimination of various sources of
of the test will be affected by the specific character sounds, sound signals as symbolic cues for actions like
of each individual phoneme-the extent to which running, sitting down etc., the intended objective being
to stimulate the development of auditory perception
it is coarticulated (stopcontinuant), its position (Cox & Pearson, 1974). The general format, the sequenc-
in the word (initial, medial, final) etc. ing of various tasks, the amount of time spent on the
To the extent w e will find bimodal distributions tasks, group activities etc. were all vely similar to the
of our test data the interpretation will be easy, experimental program. The teachers of this group were
also convinced that sensory training had an important
while we a t the same time have t o abandon classi- impact on the children’s general mental development.
cal statistical analyses based on the normality Control group 2 provided base line data. During the
assumption. critical period they followed a normal Swedish preschool
In the design of the tests special care has also program, which means a fairly low degree of structure
and a pronounced emphasis on social and esthetic devel-
been taken t o obtain a sensitive indicator of change opment in playful settings.
in phonemic abilities with special regard to the Immediately before and after the training period in-
disabled children. After all, the ultimate concern dividual tests of phonemic analysis and synthesis were
is the development of the slow child and w e should administered to all children in the groups. As a final
not be too worried about ceiling effects for the step in the investigation a reading test was given. Swedish
children start school comparatively late (around the
bright. This focus will be partly in conflict with age of seven). The investigation took place in ApriLMay
traditional psychometric design principles. How- and most children were expected to start school by the
ever, it is relevant here t o recognize the distinc- end of August. Despite their advanced age most of the
tion between psychometric vs. edumetric validity children were presumably quite unfamiliar with written
discussed by Carver (1974). language. Only a small minority could be expected to
show any reading ability, hopefully evenly distributed
among the groups. As a check a short screening test
for reading was given to all subjects.
METHOD In a field experiment like the present one it is, for
practical reasons, impossible to randomize individuals on
General design the different treatment and control groups. Of course,
For a period of 6-8 weeks three experimental groups this limits the conclusions that can be drawn and forces
of children in kindergarten took part in a more or less us to be careful in interpreting our data.
structured program for developing phonemic awareness.
During the corresponding period one control group was Subjects
run through a program emphazising non-verbal sound The average age of the children was 6 years 11 months
discrimination and a second control group just followed (SD= 3.5 months) with no significant differences among
the regular preschool plan. There were no reasons to the groups. All groups, except experiment group 2, were
expect variation among the teachers involved as far as selected (quite randomly) from the preschool system
training and experience was concerned. of the city of UmeA (8OOOO inh.). Experiment group 2
Table 1. Mean performance scores of preschool groups with and without training in phonemic awareness
Phonemic test
Reading
Synthesis Analysis Total test
Groups N na Pre Post Pre Post Pre Post Post
phonemic segments and present them to the Experimenter more, 1 point etc. A few children indeed managed to
on the peg-board. From these cues he was supposed read the last row fluently and thus obtained 4 points.
to figure out what the child had on his picture. This test The results of the test in kindergarten correlates between
also consisted of 8 items: 4 CVC, 2 VCC, 1 VCV, and 0.55 and 0.66 with reading and spelling tests and with
1 VVC. The post-test was changed according to the same teachers’ ratings in the first school years. The relatively
principles as in the synthesis case. None of the post- strong relation still holds in grade 3 where correlations
test items were included in the training program. The of 0.59-0.62 have been obtained (Olofsson & Lundberg,
segmentation attempts were tape recorded. in preparation).
In the synthesis tests as well as in the analysis tests
some children tried to use letter names instead of sounds.
In those cases it was explicitly pointed out that this
should be avoided and the Experimenter explained the RESULTS
difference between letter names and speech sounds. In Table 1 the main results of the investigation
The scoring was liberal, however, and minor substitu-
are summarized. The overall effect of the treat-
tions (probably perceptually rather than conceptually
determined) wire tolerated, such as confusion of voiced ment is expressed as the difference between the
and voice-less consonants. This principle was in accord values in the “total” columns. Only experiment
with Wallach et al. (1977) and March & Mineo (1977) group 1 shows a clear improvement from pre- to
who found that the problems are not to be localized on post-test with an average gain of 2.52 points. In
the perceptual level (training in allophonic variations
had no effect on phonemic competence). the base-line condition of control group 2 even
The reliability of the two tests proved to be high in a slight decrease in the test scores can be observed.
both versions (KR-200.90) Fit of the latent-trait-model of The distributions of the test scores deviates
Rasch was also found, so it could be concluded that the markedly from normality as can be seen in Fig. 1.
items formed a homogeneous group. The bimodality as well as the nonrandom assign-
The total testing time for pre- and post-test, respec-
tively amounted to about 12 minutes on the average with ment of subjects t o the different groups make any
lower and upper extremes of 3 and 24 minutes respec- attempt t o test the significance of the difference
tively. between pre- and post-test rather futile.
The fact that the scores a re distributed bimodally
Reading tests
suggests, in line with the reasoning above, that
Although no child had been subjected to any formal
reading instruction it was expected that a few of them the ability t o synthesize o r segment phonemes is
had gained some reading ability with the help of parents a single skill rather than a composite of skills.
or older siblings or in some other informal ways. The In Fig. 2’the distribution of scores on the reading
reading ability was quickly assessed with the same test test is presented. About one fifth of the children
as was used by Lundberg, Olofsson & Wall (1980). The seems to have some reading ability (score 2-4),
test consists of four, hierarchically arranged rows with
increasing level of difficulty. The child starts with the while more than half of them cannot read a t all.
first row and then proceeds until s/he fails; first very From a comparison of Figs. 1 and 2 it appears that
simple CVC-words, then a three-word-sentence, a longer several of the nonreading children show practically
sentence with a subordinate clause, and finally a syntac- complete ability t o analyze and synthesize pho-
tically still more complex sentence including three long
and complex words. nemes (the non-readers outnumber the children
Children with no sign of reading ability obtained 0 with low pre- and post-test scores). This can be
point, those who managed to read the first row, but not interpreted as support for the notion that phonemic
1
f
,il
6
2 6
0 2 4 6 8 6
NUMER OF CORRECT RESPONSES i OF CORRECT RESPONS
skills are, or at least can be, developed before plotted against the corresponding pre-test score.
reading acquisition. Improved scores are displayed as points above the
The relationships between pre- and post-test line x = y . Marked deviations from this line can
scores for the different groups are illustrated in be seen in same cases from the experimental groups.
Fig. 3. The post-test score for each individual is The high reliability of the testing is also convinc-
1L -
12 -
W
2 L 6 8 10 12 lI, 16 2 L 6 8 10 12 1L 16
PRE TEST SCORE PRE TEST SCORE
2 L 6 8 10 12 16 16
PRE TEST SCORE
(bed) was presented as I s : / + ang. The initial ex- variable with large variance (Lord, 1953). Since we
tended fricative was interpreted as the imperative here deliberately designed tests sensitive to changes
sign for silence followed by the word “ang” (mead- in the low-achieving children, ceiling effects were
ow)-i.e. the quite poetic expression “silent mead- obtained which make the interpretation of distri-
ow”. The general impression was that regardless butional forms rather difficult. Perhaps testing of
of ability level the children enjoyed the exercises younger children might clarify the issue.
and games very much and were encouraged to If the stage interpretation is correct, the transi-
approach their language in a playful and creative tion from phonemic incompetence to mastery
way that was new to most of them. seems to develop quite rapidly or even as a sudden
insight. Many signs of such insights were observed
during the training sessions. The new insight might
DISCUSSION be a consequence as well as a stimulation of more
The present study provides some additional evi- general metalinguistic discoveries within the con-
dence that phonemic awareness skills can be text of the “mental revolution” that is assumed
trained at the preschool level and outside the con- to take place between the age of 6 and 8. For the
text of reading instruction. The experimental design child the phoneme is a new and abstract category
of the study permitted an evaluation of the specific which requires advanced cognitive operations to
effect of phonemic awareness training as com- handle. We now know that the development of
pared to more general effects of a stimulation pro- thinking is highly dependent on the specific material
gram. The importance of program structure was the child is working with (e.g. Donaldson, 1978;
also assessed in the comparisons between the Papert, 1980). Phonemes might be especially power-
different experimental groups. Most of the positive ful devices for developing conscious reflection and
effects have been observed in children with low exploratory curiosity. If this is the case, phonemic
ability who also were the main concern of the study. training has a deeper motivation than preparing
The testing procedure with its pronounced ceiling the child for reading instruction in school.
effects failed to demonstrate any development of
phonemic skills in a majority of the more able REFERENCES
children. However, many of these children dis- Bachke, C. C. & Gunnestad, A. Den forsta leseunder-
played creativity and initiated constructive and visningen. Oslo: Universitetsforlaget, 1981.
playful activities far beyond the scope of the train- Bruce, D. J. An analysis of word sounds by young chil-
dren. British Journal of Educational Psychology,
ing program suggesting that they profited from the 1964,34, 158-170.
training program anyhow. Evaluating the effect of Calfee, R. C. Assessment of independent reading skills.
a program for children vastly diffferent on the In A. S. Reber & D. Scarborough (Eds.), Toward a
ability continuum is of course difficult. In fact, psychology of reading. New York: Halstead Press,
it requires that the same measuring stick is used 1977.
Carver, R. P. Two dimensions of tests: psychometric
regardless of the ability of the children. Promising and edumetric. American Psychologist, 1974, 29,
attempts to handle this problem have been reported 5 12-518.
by Rop (1977; ref. in Fischer, 1977). She found Cox, D. E. & Pearson, J. Materialsfor language stimu-
that preschool program designed to support chil- lation. The College of Speech Therapists, London,
1974.
dren with low abilities improved performance of Donaldson, M. Children’s minds. Glasgow: FontanalCol-
all children, regardless of initial level. However, lins, 1978.
the relative positions of the low achievers were Ehri, L. Linguistic insight: Threshold of reading acquisi-
maintained. tion. In T. G. Waller & G. E. MacKinnon (Eds.),
The bimodal distributions of test scores were Reading research. Advances in theory and practice
(Vol. 1). New York: Academic Press, 1979.
interpreted as reflections of an underlying single Elkonin, D. B. USSR. In J. Downing (Ed.), Comparative
skill dimension. The bimodality also suggests that reading: Crossnational studies of behavior and pro-
relatively few children are in a transitory stage cesses in reading and writing. New York: MacMillan,
where they partly manage to handle the tasks of 1973.
Fischer, G. H. Probabilistic test models and their applica-
phonemic analysis and synthesis. However, bi- tions. German Journal of Psychology, 1977, I , 298-
modal distributions might also reflect high-dis- 319.
criminating tests and a normally distributed latent Golinkoff, R. M. Phonemic awareness skills and reading