Learning to read and spell English requires a child to master the alphabetic properties of the language and the orthographic rules of its writing system (e.g. silent ‘e’, soft ‘c’).  In the UK, schools explicitly teach orthographic principles as rules, without formally addressing children’s implicit learning.  For example, schools teach that for ‘silent e’ (split digraphs) the final ‘e’ makes the middle vowel long (e.g. rat-rate, hop-hope) and expect children to master the concept from the rule.



    This study aimed to determine whether children use implicit, frequency-based learning in addition to explicit instruction when they learn split digraphs.  If a child does learn split digraphs via a general rule, then she should perform consistently on all –VCe patterns.  On the other hand, if she uses implicit learning, she should perform best at the most frequent –VCe patterns. 



    We tested 52 children in Year 2 (mean age: 85.9 months) and Year 3 (mean age: 94.3 months) on reading and spelling in four different categories: high-frequency –VCe words and their –VC counterparts and low-frequency –VCe words and their –VC counterparts (e.g., ‘bame’-‘nam’; ‘pode’-‘lod’ ).  We used pseudo-words to eliminate the effects of word recognition. 



    The children performed better on high- than low-frequency -VCe words, confirming a frequency-effect in reading and spelling split digraphs.  They applied the split-digraph rule to low-frequency -VCe words at above-chance levels and did not apply it to short-vowel -VC words, demonstrating that they employ the rule beyond the frequency effect.  Finally, the frequency effect correlated to the children’s exposure to text, measured by the Title Recognition Test (Cunningham & Stanovich, 1990).  Altogether these results indicate a role for implicit learning in children’s development of the split-digraph rule and provide an explanation for the link between learning reading and learning spelling proposed by Frith (1985).

    We explored how children learn the split-digraph (‘silent e’) spelling rule.  In split digraphs, the letter ‘e’ at the end of a word indicates that the medial vowel is long (e.g. ‘hop’-‘hope’, ‘rat’-‘rate’).  Beginning readers and spellers (age 6-years) generally ignore split digraphs, interpreting -CVCe patterns as –CVC, while proficient readers and spellers (age 11-years and older) recognize them (Varnhagen, McCallum, & Burstow, 1997).



    UK schools explicitly teach this rule beginning in Year 1 (age 6-years), yet we do not know whether children learn it explicitly or implicitly.  If the explicit rule–to make the middle vowel long, add ‘e’– dominates learning, children should perform consistently on all split-digraph (–VCe) patterns.  If they learn implicitly, they should perform best on the split-digraph (-VCe) patterns they see most often: i.e. there should be frequency effects.



    The research objective was to determine whether implicit learning plays a role in children’s development of the split-digraph rule.  The study tested several predictions:




METHODS:

    The Children’s Printed Word Database (Stuart, Masterson, Dixon, & Quinlan, 2003) provided frequency-counts for split-digraphs. For example, we combined the frequencies of all words ending in ‘-ake’: total 3,822.   Using medial vowels ‘a’, i’, and ‘o’, we chose two of the three highest and two of the three lowest frequency-counts for each vowel (see Table 1).

    To eliminate the effects of word-recognition, we used pseudo-words, creating one for reading and one for spelling for each split digraph (e.g. ‘pake’, ‘nake).  We also created short-vowel pseudo-words as controls (e.g. ‘nam’, ‘vap’).  The WORD test  (Rust, Golombok, & Trickey, 1993) measured general reading and spelling abilities.  The Title Recognition Test (TRT) (Cunningham & Stanovich, 1990), consisting of five well-known children’s book titles and five pseudo-titles, measured children’s general text-interaction.  The participants were 52 children in Year 2 (mean age: 85.9 months) and Year 3 (mean age: 94.3 months) in two British schools.



RESULTS:

    Three main findings confirmed all three hypotheses.  First, a series of ANOVAs and sign tests confirmed that children perform significantly better on high-frequency than low-frequency split digraphs (by-participants F(1, 50) = 35.76, p < .01, _2 = .42; by-items F(1,20) = 11.89, p < .01, eta2 = .37).  The frequent items were learned first, indicating implicit learning. Children also read split digraphs significantly better than they spelled them (by-participants F(1, 50) = 5.58, p < .05, eta2 = .10; by-items F(1,20) = 6.69, p < .05, eta2 = .25). 



    Secondly, ANOVAs showed that children do not overextend split-digraph usage, supporting the second hypothesis: they know when to apply the rule.  They recognized appropriate split-digraph usage significantly more for long vowels than for short vowels (by-participants F(1, 50) = 299.66, p < .01, eta2 = .29; by-items F(1, 7) = 530.82, p < .01, eta2 = .99).  T-tests showed that Year 3-children recognized low-frequency split digraphs, performing significantly better than chance in reading (t(27) = 5.78, p < .05) and spelling (t(23) = 3.34, p < .05), but Year 2-children did not.  This supports development from high-frequency to low-frequency split-digraph use between Years 2 and 3.



    Finally, correlations confirmed the third hypothesis: increased exposure to text correlates to increased abilities with high-frequency split digraphs.  The TRT provided interesting results.  Although it did not correlate highly with reading either high- or low-frequency pseudo-words, in spelling, its correlation with high-frequency words (r2 = .52, p < .01) was greater than with low-frequency words (r2 = .36, p < .01).  This result supports both Frith (1985), which posited causal links between reading and spelling, and Davis and Bryant (2006), which tested the links of that theory via children’s split-digraph development, finding that reading does predict spelling.  The current study extended their work by finding a mechanism by which the learning occurs.  The frequency-effects showed that children use implicit learning in split-digraph development, and the TRT correlations specifically linked children’s amount of reading experience with their performance on spelling split digraphs.



DISCUSSION:

    Overall, the results showed a frequency-effect for reading and spelling split digraphs, a decline in this effect between Years 2 and 3, and a strong correlation between spelling high-frequency split digraphs and general reading exposure.  These results have educational implications.  Currently, reading instruction follows stage theories of reading like Marsh et al. (1980) and Frith (1985).  Within this framework, help for at-risk readers focuses on early stages, teaching sound-to-letter correspondences.  While foundations of reading are crucial, some children have difficulties with later stages, including split digraphs (Labov, Baker, Bullock, Ross, & Brown, 1998). Exploring how implicit learning can help children’s mastery of the split-digraph rule in both reading and spelling provides a constructive line of research and could be extended to children’s development with other conditional spelling rules.