The question of the processes by which children understand that letters represent sound components of words has been analyzed within the framework of children's use of written language and the knowledge that they gradually acquire about the writing system before they begin formal education. Preschool children often make attempts at writing (invented spellings). The frequency with which they come into contact with written texts and the quality of the contexts in which they live, particularly the ways in which adults play the part of mediators between the children and written language, determine the way in which they see the written code and evolve in their conceptions (Hiebert & Raphael, Reference Hiebert and Raphael1998).
The study of the characteristics of early writing in different languages (Alves Martins, Reference Alves Martins1993; Besse, Reference Besse, Pontecorvo, Orsolini, Burge and Resnick1996; Chauveau & Rogovas-Chauveau, Reference Chauveau and Rogovas-Chauveau1989; Ferreiro, Reference Ferreiro and Sinclair1988; Ferreiro & Teberosky, Reference Ferreiro and Teberosky1979; Fijalkow, Reference Fijalkow1993; Pontecorvo & Orsolini, Reference Pontecorvo, Orsolini, Pontecorvo, Orsolini, Burge and Resnick1996; Sulzby, Reference Sulzby, Morrow and Smith1989; Tolchinsky, Reference Tolchinsky and Goodman1995; Varnava-Skouras, Reference Varnava-Skouras2005) has shown that children's hypotheses about the way in which the written code functions evolves from an initial level where writing is not yet determined by linguistic criteria to alphabetic writing.
At a first level, when they write, children make no attempt to match the oral to the written language; they seek to respond to factors such as the need for a minimum number of letters for each word and for a different combination of letters to distinguish between different words (graphoperceptive writing). At this level children acquire several principles and concepts concerning the written code, as Clay (Reference Clay2000) suggests, such as the principle of directionality or the production of words with a limited number of letters.
Later on children begin to phonetize their writing, that is, to analyze the oral and look for the letters that best represent the sounds they have identified (phonetized writing). Children often start by representing some of the sounds in the words before finally representing all of the sounds in the words (alphabetic writing).
The evolution to phonetized writing is mediated by the knowledge that children informally acquire about letter names, which helps them to detect the latter in the pronunciation of certain words (Treiman, Reference Treiman, Joshi and Aaron2004). This process is thought to favor an understanding of the alphabetic logic, inasmuch as the letters themselves serve as supports for a more systematic analysis of the sequence of the sounds in words.
In the English-language literature various authors (Adams, Reference Adams1998; Treiman, Reference Treiman, Metsala and Ehri1998) have begun to attach value to early writing. They argue that invented spelling activities with preschool children help develop phonemic awareness and make it easier for them to acquire the alphabetic principle. They have shown that invented spelling activities simultaneously develop both phonemic awareness and an understanding of the alphabetic principle, to the extent that they involve metalinguistic reflection about speech. As Bowman and Treiman (Reference Bowman and Treiman2002) state, “Spelling may encourage children to use the alphabetic principle in a way that reading does not, and may teach skills that eventually transfer to reading” (p. 31).
Byrne (Reference Byrne1997, Reference Byrne1998) argues that children may possess an adequate notion of the phonemic structure of oral speech without necessarily understanding the alphabetic nature of written language. This understanding requires that children possess not only phonemic analysis skills but also a concept of writing, which refers them to the representation of the phonemic structure of words. Within the overall context of this concept of writing it is the coordination of letters and phonemic awareness that is crucial to the discovery of the alphabetic principle.
The idea that the analytical procedures inherent in children's attempts at writing can model and develop skills that enable them to analyze speech is sustained by Alvarado (Reference Alvarado1998) and Vernon's (Reference Vernon1998) work. Alvarado (Reference Alvarado1998) used an initial-phoneme deletion task applied to words, which were presented both verbally and accompanied by a written support, to test children with different levels of writing. As expected, it was the children at the alphabetic level who achieved the highest success rates of around 50% in the verbal form and 95% when the words were accompanied by a written version. Along the same lines, Vernon (Reference Vernon1998) showed that children at a less evolved conceptual level provided fewer analytical answers in a word segmentation test than their more advanced peers.
Silva and Alves Martins (Reference Silva and Alves-Martins2002, Reference Silva and Alves Martins2003) and Alves Martins and Silva (Reference Alves Martins and Silva2006a, Reference Alves Martins and Silva2006b) confirmed the relationship that was found in the above-mentioned studies. Using writing programs, they demonstrated the existence of an explicit causal relationship between the evolution of conceptualizations about writing and the performance in phonemic tasks.
These authors conducted various experimental studies in which they compared the effect of programs designed to lead the quality of preschool children's writing to evolve. After writing a few words, the children were shown the writing of children who were more evolved. They were asked to analyze the word orally, think about the two forms of writing, choose one of them, and justify their choice. In other words, they induced the children to think about speech and print and the relationships between them. This procedure led to both a clear evolution in the quality of the children's writing and progress in their phonemic awareness.
Evolution toward phonetized writing can also be influenced by various types of linguistic factors, such as the characteristics of the phonemic sequences in the structure of a word. Some words possess a phonological structure that makes it easier to mobilize letters; the articulatory properties of the initial phonemes influence the individualized perception of both the phonemes themselves and the characteristics of the letters that need to be mobilized.
Alves Martins and Silva (Reference Alves Martins and Silva2001), Mann (Reference Mann1993), Quintero (Reference Quintero1994), and Treiman and Cassar (Reference Treiman, Cassar, Perfetti, Rieben and Fayol1997), among others, have shown that writing conventional letters to represent sounds appears to be related to the phonological characteristics of the words in question. Preschool children more often write appropriate letters to represent the sounds they have identified when they are dealing with words that contain phonetic sequences that represent letter names (the effect letter name).
Thus, when children are asked to write words whose initial syllable coincides with the name of a letter with which they are familiar (e.g. pera [perα] (pear), the letter p [pe] is mobilized more often than if we ask them to write pano [pαnu] (cloth) or parede [pαredə] (wall), in which the initial phonetic sequences do not match the name of the letter p. When children are familiar with letter names, they thus find it easier to detect those letters in the pronunciation of this type of word. This in turn facilitates the understanding of the sound notation function that the letters play in the alphabetic code.
Some studies, particularly those by Treiman (Reference Treiman1993, Reference Treiman1994), say that the effect of letter names is more accentuated for some letters than for others.
Young Portuguese-speaking children have been reported to produce more vowel- and syllable-oriented spellings than English speakers. Pollo, Kessler, and Treiman (Reference Pollo, Kessler and Treiman2005) investigated the sources of such differences and found that Portuguese words have more vowel letter names and a higher vowel/consonant ratio than English words.
According to these authors, because Portuguese speakers see more vowel letters in texts, they may write more vowel letters in their spellings. Alternatively, because Portuguese speakers hear more vowel letter names in words, they may be more encouraged to use letter name spelling strategies when spelling vowel sounds than are young speakers of English.
The nature of vowel letter sound correspondences in Portuguese may also play an important role. Even though the Portuguese language has nine oral vowels, five nasal vowels, and only five corresponding letters (a, e, i, o, u), the i and u letter names correspond to the sounds they represent. Syllables frequently contain open vowels a, e, and o that match letter names, which makes it easier to correctly represent the vowels than the consonants.
Vernon and Ferreiro (Reference Vernon and Ferreiro1999) and De Abreu and Cardoso-Martins (Reference De Abreu and Cardoso-Martins1998) also showed that in languages like Brazilian Spanish or Portuguese, the vowels are more often used in early spellings than the consonants because they have a salient phonological pattern that is attributable to the stability of the relationships between the names of the vowels and the phonemes they represent.
Another aspect of the relationship between the writing of conventional letters and the sounds to which they apply is the way in which the characteristics of the letter names facilitate access to the sounds they represent. Treiman, Tincoff, Rodriguez, Mouzaki, and Francis (Reference Treiman, Tincoff, Rodriguez, Mouzaki and Francis1998) showed that the sound of letters in English is easier to learn (more accessible) when the phoneme concerned is at the beginning of the letter name (e.g., [p] in the letter “p” [pi]) than when it is at the end (e.g., [s] in the letter “s” [ϵsə]). This effect seems to be more important than the nature of the sounds that are represented by the letters (constrictive sounds or occlusive ones). Treiman and Kessler (Reference Treiman, Kessler and Kail2003) argue that these results reflect the greater obviousness of onsets compared to syllable codas.
The articulatory properties of the phonemes in the words that they have to write can also influence the quality of children's writing, because some phonemes are likely to be easier to isolate within the flow of speech than others. For example, according to Liberman, Shankweiler, Fischer, and Carter (Reference Liberman, Shankweiler, Fischer and Carter1974), children become aware of vowels more easily than consonants and find it easier to identify fricative consonants than occlusive ones. Treiman (Reference Treiman, Metsala and Ehri1998) and Byrne and Fielding-Barnsley's (Reference Byrne and Fielding-Barnsley1991, Reference Byrne and Fielding-Barnsley1993) work shows that it is easier to train children in relation to the phonetic identity of fricatives than that of occlusives, because it is easier to produce these sounds in isolation. At the same time, in the written form it is easier to confuse phonemes that only differ from one another in their voicing than those that are only different in their articulation (Treiman, Broderick, Tincoff, & Rodriguez, Reference Treiman, Broderick, Tincoff and Rodriguez1998). These linguistic factors undoubtedly have consequences in the ease or difficulty with which preschool children mobilize conventional letters in their attempts to spell.
This is the context in which we proposed to analyze the impact of some of these linguistic variables on the framework of programs that seek to work on children's writing. Our goal was to analyze the impact of the characteristics of the words that are used in writing programs, particularly the effect that the articulatory properties of the initial phoneme in a word (occlusive/fricative) have on the evolution of the processes involved in the phonetization of children's writing. We also sought to analyze the extent to which there were generalization effects on the writing of phonemes that were not worked on during the programs.
We posed the following hypothesis: graphoperceptive children who underwent writing programs that induce metalinguistic reflection about the relationships between the initial phoneme of different words and the letters they represent will use conventional letters to spell the initial sound of the words and will generalize the phonetization procedures to graphophonetic correspondences that were not worked on during the programs.
We also formulated two research questions. Will the two writing programs (use of occlusive vs. fricative initial phonemes) produce equivalent phonetized writing in the posttest? Will children from the two groups generalize the phonetization procedures to graphophonetic correspondences that were not worked on during the programs?
METHOD
Participants
The participants were 39 5-year-old children from three classes at a kindergarten attended by 75 5-year-olds, where they were not given any initiation to reading or writing. In those kindergarten classes there were no regular classroom activities or instruction relating to phonological awareness or invented spelling, which happens in many kindergartens in Portugal. The teaching of reading and writing normally begins in the first year of elementary school. The only regular activities related to reading and writing were story reading, teaching letter names, and activities in which children had to write their own names (e.g., to identify their drawings and paintings).
None of the children knew how to read, which was verified by means of individual reading tests. Only children who were familiar with the five vowels and at least with consonants B, D, F, P, V, and Z were selected (46 children). Their writing was assessed by means of an initial interview where children were asked to write a set of words and to justify their writings.
Thirty-nine children were selected whose writing was graphoperceptive or presyllabic. They were randomly divided into three groups. Their age, level of intelligence, number of consonants known, and level of phonological awareness were controlled.
We carried out analyses of variance to compare their age (at the time the pretests took place, April), level of intelligence, number of consonants known (they knew all of the vowels), and results in two phonological awareness tests. The results were the following: age, F (2, 36) = 1.09, p = .346; level of intelligence, F (2, 36) = 0.14, p = .875; number of consonants, F (2, 36) = 0.14, p = .866; initial syllable classification test, F (2, 36) = 0.78, p = .465; and initial phoneme classification test, F (2, 36) = 0.01, p = .990. There were no statistically significant differences among the three groups. Table 1 shows the means and standard deviations for the three groups’ results in relation to these variables.
Means and standard deviations for age (months), intelligence, letters (consonants) known, and results in the initial syllable and initial phoneme classification tests of three groups
Note: S, initial syllable classification; Ph, initial phoneme classification; G.1, occlusive phoneme writing program; G.2, fricative phoneme writing program; G.3, control group.
The children's writing was assessed in a pretest and a posttest, in which they were asked to write 24 words beginning with B, D, F, V, P, and Z.
Between the two tests the two experimental groups took part in a writing program designed to lead them to use conventional letters to represent the initial consonant in words. Experimental Group 1 wrote words with the occlusive initial phonemes [b] and [p]. Experimental Group 2 wrote words whose initial phoneme was fricative ([f] and [v]). The control group underwent a set of exercises using logical blocks type materials. The children were asked to classify geometric shapes in accordance with criteria such as identical shape, size, or color.
Instruments and procedures
Assessment of writing to select the participants. In order to select the children with graphoperceptive writing, we asked all of the 5-year-old children attending the kindergarten who knew the vowels and at least the six consonants B, D, P, F, V, and Z (46 children) to spell their name and then to spell six words beginning with the above consonants, to the best of their ability. After spelling each word they were asked to read it. After spelling the different words they were asked why they used certain letters and not others and why they used more or fewer letters in some words than in others. The verbal utterings that frequently accompanied the act of spelling were recorded.
We used a grid that was constructed from those drawn up by Ferreiro (Reference Ferreiro and Sinclair1988) and Alves Martins (Reference Alves Martins1993) to classify the children's responses.
The words we asked them to spell had different sizes from a linguistic perspective (mono-, di-, tri-, and polysyllabic words). Some of them referred to items of different sizes (e.g., Boi [boy] (ox)/Formiga [furmigα] (ant)).
These words were chosen so that we could understand whether the reasoning that underlined the children's writing was based on the linguistic characteristics of the words or on the perceptual images children may have built up of what constitutes a word, which are strings of letters varying in their relative position or, in certain cases, on the reference items to which the words refer.
If the form of reasoning that governs the writing is linguistic, we can expect that the different words will be written according to the number of syllables or letters they contain and that children will justify their spellings by making reference to the correspondences between the oral and the written language.
If a child uses nonlinguistic criteria to govern his/her writing, it may be expected that he/she will write the different words using a fixed number of random letters and a different combination of letters to distinguish between the different words and that children will justify their spellings making no reference to the relations between the oral and the written language.
Thirty-nine children were graphoperceptive or presyllabic.
Assessment of writing in the pre- and posttests. Each child was asked to write 24 disyllabic words in which the syllables had a consonant–vowel (CV) structure, because these syllables are the most frequent in Portuguese. According to Vigário, Martins, and Frota (Reference Vigário, Martins, Frota, Duarte and Leiria2006), CV, V, and CVC are the most frequent syllabic patterns in the spontaneous talk of adults in European Portuguese at 46%, 16%, and 11%, respectively.
The Portuguese spelling system is relatively shallow, given its predictable grapheme–phoneme mappings and the stable contextual rules establishing grapheme–phoneme correspondences (Rebelo, Delgado Martins, & Prudêncio, Reference Rebelo, Delgado Martins and Prudêncio1978). It has six occlusive oral consonants [p], [b], [t], [d], [k], and [g]—four with regular correspondences to the letters p, b, t, and d (the graphophonetic correspondences of the other two are governed by contextual rules)—and six fricative ones [f], [v], [s], [z], [ʃ], and [Ʒ], the first two with regular correspondences with the letters f and v; [z] with the letters z or s; [ʃ] with the letters x, s, z, and ch; and [Ʒ] with the letters j, g, or s. The letters z and j always correspond to a unique phoneme.
Because we wanted to compare occlusive and fricative phonemes, we chose words beginning with consonants corresponding to a single occlusive or fricative phoneme.
Twelve words began with the consonants B, D, and P (4 words for each consonant), which correspond to the occlusive phonemes [b], [d], and [p]. The other 12 words began with the consonants F, V, and Z (4 words for each consonant), which correspond to the fricative phonemes [f], [v] and [z].
The second letter of all of the words was a vowel (A, I, O, or U) that always corresponded to the same four phonemes [a], [i], [o], and [u].
Alves Martins and Silva (Reference Alves Martins and Silva2006b) demonstrated that 5-year-old children's ability to analyze the phonemic components of words was linked to its size and complexity. Treiman, Berch, Tincoff, and Weatherston (Reference Treiman, Berch, Tincoff and Weatherston1993) also showed that preschoolers have more difficulty in isolating the initial consonant of a long word than the initial consonant of a short one. The complexity of the phonological structure of words can have the same kind of outcome because some sequences of phonemes are more difficult to segment than others. These authors proved that children have more difficulty in segmenting and spelling syllables of words that contain clusters.
Our aim was to analyze the way in which children represented the initial consonant of the words. We controlled the characteristics of the initial syllable of the different words, their length and complexity, and their frequency. According to Nascimento et al. (Reference Nascimento, Casteleiro, Marques, Barreto, Amaro and Veloso2000), all of the words were very frequent, except those begun with [z], which are less frequent in Portuguese.
Children were evaluated in the posttest using the dictation of the same set of words. We analyzed if the children correctly represented the first consonant in the different words. To compare the number of words that each of the two groups phonetized, we allocated 1 point for each word in which the first letter was spelled correctly.
Assessment of phonological awareness
In order to assess the children's phonological awareness, they were given an initial syllable classification test and an initial phoneme classification test, which were both taken from Silva's (Reference Silva2002) battery of phonological tests. We sought to take into account both the size of the units (syllables and phonemes) and the phonological properties of the initial phonemes in the words.
Each test was made up of 14 items, preceded by 2 training items. In these tests the children were given four words in figurative form, two of which began with the same syllable or the same phoneme, and they were asked to identify the words. The children had to categorize two target words out of four, using a syllabic or phonemic criterion. One point was given for each correct answer.
Assessment of intelligence
The level of the children's intelligence was assessed using the colored version of Raven's Progressive Matrices (Raven, Raven, & Court, Reference Raven, Raven and Court1998) because it is not dependent on verbal aspects.
Writing programs
The writing programs lasted for five sessions of around 15 min each, and they were designed to lead the children to use conventional letters to represent the initial consonant in each word. The words used in the programs were always different from those in the pre- and posttests. Experimental Group 1 wrote words with the occlusive initial phoneme ([p] or [b]), and experimental Group 2 wrote words with the fricative initial phoneme ([f] or [v]).
In each session each child was asked to write a word as best he/she could and was then shown the same word written by a hypothetical child from another class, who had used the correct initial consonant to represent the first syllable in the word. He/she was asked to think about his/her writing and that of the other child and to try to think which was the better way to write the word and why. The child's attention was drawn to the first letter in the word. Ten words were used in each session. In the first 2 words the initial syllable matched the name of the letter. In the other 8 the initial letter was followed by the vowel a, o, i, or u. For example, in the first session of the program followed by experimental Group 1, the children were asked to write the words Pena [penα] (Feather) and Pera [perα] (Pear), in which the first syllable matches the name of the letter P, and then the words Papa [papα] (Porridge), Pala [palα] (Awning), Povo [povu] (People), Popa [popα] (Stern), Pico [piku] (Peak), Pino [pinu] (Top), Pulo [pulu] (Jump), and Punho [pu
u] (Fist).
All of the words that were presented to the children as having been written by the hypothetical child from another school had two letters: a consonant and a vowel. The consonant represented the first phoneme of the word and the vowel the last phoneme, except for the word Voar, in which the vowel represented the third phoneme.
The following example of the interaction between the researcher (R) and a child named Luis (L) is a good illustration of the dynamic that occurred during the writing program sessions:
R: Luis, try to write the word Pera.
L writes “LOA.”
R: Try to read it, show me with your finger.
L: Pera (points LOA).
R: I was yesterday with a child named João from another school who wrote Pera in a different way. Do you want to see?
L: Yes.
R shows “PA.”
L: It is different.
R: What is the first letter that João wrote?
L: He wrote “P” [pe].
R: And do you think that Pera begins with P?
L: I don't know.
R: Try to say the word slowly.
L: Pe-ra.
R: How does the word begin?
L: Pe, Pe, with P, he has the good letter.
The control group program
We organized a set of exercises with the control group using logical blocks type materials. The children were asked to classify geometric shapes in accordance with criteria such as identical shape, size, or color.
The three programs involved five sessions that lasted approximately 15 min each and were individually conducted by us with the children over the course of 2 weeks.
RESULTS
In the pretest none of the children used conventional letters to represent the initial phoneme of the 24 words. They used strings of random letters to represent the different words, combining them and varying their order from word to word. As children used more vowels than consonants in their spellings, in some words (between 1 and 6) eight children from experimental Group 1 and from control group and nine children from experimental Group 2 used a vowel at the beginning of the word that corresponds to the second phoneme of the word.
Figure 1 gives an example of graphoperceptive writing in the pretest from a child named Beatriz in experimental Group 1. As we can see, Beatriz used several letters to represent each of the words we asked her to write. She did not use any conventional letter to represent the initial sound of the words. In three words (Dono [donu], Pata [patα], and Zona [zonα]) the first letter she wrote was a vowel corresponding to the second phoneme of the word.
Examples of Beatriz's graphoperceptive writing in the pretest.
In the posttest all of the children of both experimental groups used conventional letters to represent the initial sound in some or all of the words (between 8 and 24 words).
Figure 2 gives an example of the same child's writing in the posttest. As we can see, Beatriz was able to use a correct letter to represent the initial phoneme of all of the words.
Examples of Beatriz's writing with phonetization in the posttest.
None of the children in the control group used conventional letters to represent the initial phoneme of the words. As in the pretest, eight children used a vowel at the beginning of some words (between one and seven words) corresponding to the second phoneme of the word.
Table 2 provides the means and standard deviations for the number of words in which the letters B, D, F, P, V, and Z were used at the beginning of the word and for the total number of words whose initial consonant was correctly spelled by the two experimental groups.
Means and standard deviations for the number of words in which the children in the two experimental groups phonetized the initial phonemes [b], [d], [f], [p], [v], and [z] in the posttest
Note: G.1, occlusive phoneme writing program using [b], [p]; G.2, fricative phoneme writing program using [f], [v].
To assess the impact that the articulatory properties of the initial (occlusive/fricative) phoneme in words had on the evolution of the phonetization processes involved in the children's writing in the posttest, we carried out Mann–Whitney U tests using the group as the independent variable and the number of words in which the initial phoneme was correctly written and the number of words in which the letters B, D, F, P, V, and Z were used at the beginning of the word as the dependent variables.
These nonparametric tests were performed because of heterocedasticity concerning the seven dependent variables. We used the Bonferroni procedure to control the experiment error rate, establishing α = 0.05/7.
The results show that there were statistically significant differences between the two groups in the total number of words in which the initial phoneme was correctly phonetized (U = 16.00, p < .000). They also show that there were no statistically significant differences in relation to the consonants F (U = 78.00, p = .317), V (U = 78.00, p = .317), Z (U = 70.50, p = .253), and P (U = 60.50, p = .144); but there were such differences for the consonants B (U = 19.50, p < .000) and D (U = 18.50, p < .000).
The children in experimental Group 1, who were given a writing program that worked on the phonetization of occlusive initial phonemes, generalized the phonetization processes and applied them to the fricative phonemes in a posttest situation, in which they achieved the same results as the children in experimental Group 2 when they had to phonetize [f], [v], and [z]. The children in experimental Group 2 found it more difficult to generalize the phonetization processes in relation to occlusive phonemes, and they did not do as well as the experimental Group 1 children when it came to phonetizing [b] and [d], although they did obtain similar results with [p].
When the children from both groups did not use a correct consonant to represent the initial phoneme of the words, what did they use? They generally used a correct vowel corresponding to the second phoneme of the word. Table 3 provides the means and standard deviations for the words in which a correct vowel was used instead of the correct initial consonant by the two groups.
Means and standard deviations for the number of words phonetized in the posttest
Note: The values are for the number of words in which the children in the two experimental groups phonetized the second phonemes (vowels) instead of the first phonemes (consonants) in the posttest. G.1, occlusive phoneme writing program using [b], [p]; G.2, fricative phoneme writing program using [f], [v].
Thus, it seems that when children were not able to correctly spell the first phoneme they generally spelled the second one, a vowel, which in the Portuguese language is easier.
DISCUSSION
This study confirms that conducting intervention programs that work on preschool children's writing leads to an evolution in the children's thinking about the characteristics of the written code. We found that the children in both experimental groups, whose writing was graphoperceptive at the time of the pretest, started phonetizing their writing in the posttest. It would thus seem that the tasks that were put to them—to think about their writing and compare it with the phonetized writing of a hypothetical schoolmate—made it possible to initiate metalinguistic thinking processes at the level of segments of speech and print and of the relationships between them, which in turn modeled a conceptual understanding that the code is a system for writing down sounds. This interpretation is in line with those that Treiman (Reference Treiman, Joshi and Aaron2004) posits when she says that “The spelling task may prompt children to use more systematic methods of deriving spelling from sounds” (p. 30).
Given the initial-letter hypothesis, according to which the systematic mapping between the oral and the written occurs first in relation to initial letters (Bowman & Treiman, Reference Bowman and Treiman2002), we focused in more detail on an analysis of the phonetization of the first letter in words. Our results indicate that the number of words in which the initial phonemes were correctly phonetized in the posttest situation was greater in the children in experimental Group 1, whose writing program used the occlusive initial phonemes [b] and [p], than it was for those in experimental Group 2, whose writing program used the fricative initial phonemes [f] and [v].
This advantage is because the children in experimental Group 1 generalized the phonetization procedures that they had learned over the course of the program to all of the phonemes they had not worked on: the fricative phonemes [f], [v], and [z] and the occlusive [d]. Although the children in experimental Group 2 generalized the phonetization procedures to the fricative phoneme [z] and the occlusive [p], they did not manage to do so for two of the occlusive phonemes ([b] and [d]), which they had not worked on in their program. These data, in addition to both groups phonetizing virtually 100% of the fricative phonemes, confirms that it is easier to acquire the phonetization of fricative phonemes than that of occlusive ones.
There is a set of data that supports the idea that the properties of the words that served as the basis for the work with experimental Group 1 were more complex from a linguistic analysis point of view. Various studies show that children find it easier to identify fricative consonants than occlusive ones in the oral flow of words (Byrne & Fielding-Barnsley, Reference Byrne and Fielding-Barnsley1991, Reference Byrne and Fielding-Barnsley1993; Liberman et al., Reference Liberman, Shankweiler, Fischer and Carter1974; Treiman, Reference Treiman, Metsala and Ehri1998). This conclusion is linked to the considerable difference in the acoustic realization of occlusives, depending on the vowels that follow them, which makes the task of identifying these particular sounds as separate units more complex (Byrne & Fielding-Barnsley, Reference Byrne and Fielding-Barnsley1991, Reference Byrne and Fielding-Barnsley1993). It is therefore more difficult to associate them with the corresponding letters. Thus, it seems to us that because the children in experimental Group 1 were led to think about the relationships between speech and writing under linguistically more complex conditions, it made it easier for them to generalize phonetization procedures to the writing of sounds that were not worked on in any way. The more widespread mobilization of vowels to represent the sounds in the first syllable by the children in experimental Group 2 when compared to the children in experimental Group 1 corroborates this interpretation. A study by Pollo, Kessler, and Treiman (Reference Pollo, Kessler and Treiman2005), which analyzed the differences in phonological spelling with Brazilian and US preschool children, indicated that the Brazilian children were three times more likely to start a word with a vowel letter than the US children; this could be explained by the different properties of the writing system to which children have been exposed. Thus, the intervention program of experimental Group 1 is more effective in the sense that it counteracts this natural tendency in the Portuguese language and successfully promotes the phonetic identity of consonants.
Regarding the phonetization of the phoneme [p], in which the two groups’ performance was quite similar, our results show that the children from experimental Group 2 were able to generalize the phonetization procedures they had learned over the course of the program with fricative phonemes to the occlusive phoneme [p]. These results were quite unexpected because there was no evidence concerning either the frequency or the familiarity that can justify the generalization to this occlusive consonant instead of the other occlusive consonants [b] and [d]. This issue should be investigated in future research.
The existence of a generalization of phonetization procedures in both groups suggests that this type of program leads children to acquire the notion that the phonemes they identify in words should be represented by letters that contain the applicable sound. It also suggests that they simultaneously acquire phonemic identification skills, that is, the ability to perceive phonemes as stable identities within different words, which they then apply when they analyze new words.
The Anglo-Saxon literature often discusses the acquisition of the alphabetic principle as depending on phonemic awareness and on the facilitating role that letter–sound correspondences play as a support in the process of becoming aware of phonemic entities (Byrne, Reference Byrne1997, Reference Byrne1998). For instance, Castles and Coltheart (Reference Castles and Coltheart2004) assume that letter–sound knowledge enables children to perform phonemic tasks because this allows them to manipulate phonemes by manipulating orthographic images of words. Other authors, such as Hulme, Caravolas, Málkova, and Brigstocke (Reference Hulme, Caravolas, Málková and Brigstocke2005), consider that the ability of phonemic manipulation can be developed in the absence of knowledge about the corresponding letters. Our results concerning the generalization across phonemes in writing reveals the possibility of the existence of a complex interaction between the processes involved in becoming aware of the oral units in speech and the understanding of the way in which the written code works that children achieve that are attributable to their invented spelling activities and their reflection about the nature of the written code.
However, this understanding appears to be mediated by the difficulties posed to the children by the characteristics of the phonological structure of words and the articulatory properties of the sounds that make them up.
We point out two limitations of this study: the first one concerns the nature of the activities developed by the control group and the second one concerns the way children's knowledge of letters was evaluated. For the first point, it would be relevant in future research to use a control group with other linguistic training that is useful for writing skills, rather than using a control group that only classified geometrical shapes. For the second point, it would be relevant to evaluate not only the knowledge of the names of the letters but also the knowledge of their phonetic counterpart.
CONCLUSIONS
We believe that programs that work on the level of writing and that lead children to think about the relationships between speech and writing under linguistically more complex conditions (i.e., those involving the mobilization of occlusive phonemes) seem to be more effective in the generalization of phonetization procedures than programs that mobilize phonemes that are easier to isolate within the acoustic flow, such as fricative phonemes.
To the extent that the properties of both words and the sounds that compose them appear to mediate a child's path to the understanding of the alphabetic principle and its application to writing tasks, it would be useful to extend this type of study to the analysis of a larger number of linguistic variables. By this we mean not only the study of different types of contrast between phonemes but also the cross-referencing of these variables with the frequency with which the various letters appear in written Portuguese.
ACKNOWLEDGMENTS
This work was supported by FCT Grant POCI 2010.
