Abstract
The study aimed to investigate the spelling performance and the semantic understanding of compound words by 103 Greek primary school children (first through sixth grade). The experimental group comprised of 25 children with spelling difficulties and compared with a control group of 78 children of typical development. Children were asked to spell and define 20 concrete and abstract compounds. They were also asked to spell 20 different compounds after providing their definitions in terms of their morphological constituents. Main results indicated that concrete compounds were spelled and defined better than abstract ones, but the experimental group performed significantly lower than the control group on both word types. Children with spelling disabilities were able to use less etymological information in defining compound words than their typical classmates, suggesting that they understand less the internal structure of morphologically complex words. These results are compatible with the experimental literature and are discussed in terms of the morphophonemic nature of Greek language as a transparent orthography with a rich morphology.
While there is now considerable evidence for the central role of phoneme awareness (Bryant & Bradley, 1985; Ehri, 1999) in spelling acquisition, less emphasis has been given to other linguistic factors affecting performance, as the awareness of morphologically complex forms especially in association with the semantic understanding of their internal structure. Longitudinal studies have demonstrated that normally achieving children acquire inflectional knowledge and simple compounds very early and in a fairly consistent developmental sequence (Berman, 2009; Clark, 1993; Nicoladis, 2006). The developmental patterns that emerge from these sources indicate that children initially start out by treating compounds as single entities, then as a means of putting together two nouns with a semantic relation between them (Berman, 2009). Cross-linguistic findings suggest that children can have some understanding about the function of simple noun-noun compounds as young as 2 years of age, at least for English-speaking children (Clark, 1993), whereas Hebrew-speaking children show comprehension of novel compound nouns at around the age of 3 (Berman, 2009).
On the other hand, frequency and productivity in children’s input seems to play an important role in how early in development and how often they produce compounds (Berman, 2009; Nicoladis, 2006). Hence, while English-speaking children produce compound nouns around 2 years (Clark, 1993), spontaneous novel compounds are used by Hebrew-speaking children around 4 years (Berman & Clark, 1989), reflecting the frequency and productivity of these forms in their own language.
Clark (1993) suggested that children’s ability to produce simple compounds reflects a general developmental strategy that she called the “simplicity of form” principle. This principle suggests that in early stages of language development, children should produce compounds constructed with bare nouns or verbs (e.g., man-rat) before they combine roots and affixes (e.g., wagon-puller). Since these principles are relative to the typology of the language being acquired, children learn to adapt to the typological characteristics of their language quite early (Nicoladis, 2006). Greek children’s awareness of nominal compounds in terms of frequency and familiarity has been also observed from 2 years and 10 months(Stephany, 1997).
Following stages in later development and once literacy is well established, children are expected to make use of compounding as an option for the enrichment of their language (Berman, 2009). In that sense, other types of compounds may become more challenging, due in part to their lower frequency and greater complexity (Fowler & Liberman, 1995). A number of studies have examined whether morphological knowledge is weaker in children with identified spelling difficulties showing that they exhibit poorer morphological knowledge than their normal age peers (Sénéchal & Kearnan, 2007). Morphology is also part of the explanation for how children learn a significant number of words without an explicit instruction through morphological problem solving (Anglin, 1993; Bowers & Kirby, 2010).
Anglin (1993) carried out a remarkable study of morphological problem solving in word definitions. Children aged 6 to 10 years were asked questions such as: “What does the word ‘unbribable’ mean?” The main focus was whether the child was using morphological problem solving to deduce the meaning of the complex word. For example, the child showed evidence of morphological problem solving if he or she explained the word unbribable as follows: “I don’t let anybody . . . or anything to take my toys and money or something away” (p. 96). Findings from this study showed that morphological problem solving was evident at all grade levels and the proportion of words for which such evidence existed increased as a function of grade, an ability that could be a valuable tool to figure out the meanings of their unknown words and increase their vocabulary.
In a recent intervention study, Bowers and Kirby (2010) aimed to teach 81 fourth- and fifth-grade school children (randomly assigned to a treatment and a nontreatment group) to discover spelling-meaning connections between morphologically related words via a problem-solving approach using word matrices and word sums to support instruction about morphological structure with concrete representations. Assessments (including identification of 30 bases of inflections, derivations, and compounds) after 20 hours of instruction showed significant effects on vocabulary scores for words that were taught directly and novel words with taught bases but not for words with untaught ones. The treatment group also made better use of pretest vocabulary knowledge in learning new vocabulary, indicating that when faced with an unfamiliar word, a more skilled morphological problem solver is better able to recognize a meaningful link to a related familiar word.
Another important factor that is considered to affect literacy performance is the level of concreteness of the words. Early research (Paivio, Yuille, & Madigan, 1968) showed that words differ in terms of the amount of semantic richness they possess. In particular, concrete words are considered to be high in terms of their richness since they have a sensory reference and they are easily imaginable. Conversely, abstract words expressing concepts, processes, and states do not have the same depth of sensory reference and corresponding imagery. Nilsen and Bourassa (2008) investigated a trial-by-trial word-learning performance of 27 kindergarten children and 19 first-grade children who have not yet developed decoding skills. Experimental items (n = 40) were categorized in four groups differing in terms of their levels of concreteness and regularity (i.e., concrete-regular, concrete-irregular, abstract-regular, abstract-irregular). Findings demonstrated that children learn concrete words more efficiently than abstract ones, suggesting that semantics plays a substantial role in early literacy acquisition.
The factor of concreteness is of particular importance given the fact that school reading materials tend to become more abstract as children become more advanced in their reading and grade level (Nagy & Anderson, 1984). Similar findings also come from spoken and written texts produced by students themselves. For instance, Ravid (2006), investigating the development of nominal words from a rich body of texts created by students across primary (9–10 years) and secondary (12–17 years) education, demonstrated that nominal density and abstractness increases dramatically in texts as a function of age and grade level.
Finally, the main focus of this study is the examination of compounds during school years when literacy has emerged. Compounding has gained particular attention the past decade (Libben & Jarema, 2006; Lieber & Stěkauer, 2009; Scalise & Vogel, 2010), and it is one of the richest sources of word formation in everyday language and scientific terminology (Ralli, 2005). Greek compounds are mainly right-headed and are typically formed by inserting a linking vowel (-o-) between the first and second constituent. Among common types are simple noun-noun (e.g., ντoματ
In general, Greek contains lots of polysyllabic and morphologically complex words, whereas a large body of these entails morphemes of ancient Greek origin (i.e., oικoσ ύστημα/ikosistima/ecosystem), creating abstract items that are mostly found in literary and scientific texts of higher school grades. These morphemes are usually phonologically opaque in terms of their spelling (e.g., κoινoβoυλϵυτικός/kinovuleftikos/parliamentary), posing particular difficulties to those with impoverished spelling abilities (Rousoulioti, 2011). Acquisition of compounds, therefore, seems to be of main importance for the development of spelling in Greek after the grasp of the phonetic principle by children.
It is acknowledged that Greek is a regular orthography in terms of phonology (Seymour, Aro, & Erskine, 2003). In spelling, however, Greek is less transparent as there are some instances with one-to-many phoneme-grapheme mappings (see Note 1), which are mainly rule governed. These are likely to be misspelled by poor spellers. However, since most of such spelling patterns are explained through reference to etymological and grammatical knowledge, spelling can be assisted by a gradual learning of the rules based on lexical information and morphology (Porpodas, 2006).
To conclude, this study aims to investigate the spelling performance of compounds along with their understanding on Greek primary school children in a phonologically transparent orthography with a rich morphology. Experimental studies of this kind are scarce in Greek. A description of main hypotheses of the study is presented in the next section.
Main Hypotheses of the Study
The hypotheses were the following: (a) The experimental group is expected to perform significantly below the control group on spelling compounds (Sénéchal & Kearnan, 2007; Tsesmeli & Seymour, 2006) while their first constituent would be spelled better than the second one (Kehayia et al., 1999; Libben, 1998). (b) Nouns and adjectives are assumed to be spelled higher than verbs (Tsapkini et al., 2002; Tsesmeli & Seymour, 2006). (c) Concrete compounds would evoke higher accuracy rates in spelling and meaning than the abstract ones (Nilsen & Bourassa, 2008). (d) Differences between spelling and meaning of concrete and abstract words will also be evaluated in an attempt to identify particular profiles for each group in every word type. Group differences will also be examined for the second and third hypotheses. (e) The study will investigate whether poor readers would understand less the morphological structure of the compounds than the able readers (Anglin, 1993).
Method
Participants
Participants in this study were 103 Greek students (49 males, 54 females) following Grades 1 to 6 of a mainstream primary school from the island of Crete in Greece. They were selected so as to form two groups:
Children with spelling difficulties (n = 25) (9 males, 16 females) with a mean chronological age of 8.65 years (range: 6.05–11.11 years). They were chosen as having a rate equal or below to the 30th centile on the standardized Test of Spelling Ability (Mouzaki, Protopappas, Sideridis, & Simos, 2007), indicating a significant deviation from the mean spelling performance of typical spellers, F(1, 84) = 43.485, p < .001. Their mean performance on reading single words, F(1, 84) = 22.653, p < .001, and nonwords, F(1, 84) = 41.361, p < .001, was also differing significantly from that of the control group of typical spellers (see “Psychometric Assessment” section and Table 1).
Children with typical development (n = 78) (40 males, 38 females): Their mean chronological age was 8.87 years (range: 6.04–11.11 years), and their spelling performance was above the 30th centile on the standardized Test of Spelling Ability (Mouzaki et al., 2007). There was no report of reading and spelling difficulties by their teachers and they did not follow any learning support department in school.
Psychometric Characteristics of Students With and Without Spelling Difficulties: Mean Accuracy Rates (%) (Standard Deviations in Parentheses).
All participants from both groups were Greek monolinguals of medium socioeconomic background and they did not have any mental, sight, hearing, or serious health problems.
Psychometric Assessment
All participants were given individually a number of psychometric assessments (for results, see Table 1) in order to evaluate their reading and spelling performance and to confirm the presence of literacy difficulties in the group of the disabled spellers, as follows:
Reading performance was evaluated in terms of reading words and nonwords by the standardized Test of Reading Performance (TORP; Padeliadu & Sideridis, 2000), which is a comprehensive evaluation of reading ability in Greek. In Word-reading, the student had to read single real words (N = 40 items) taken from school reading books and ordered by ascending phonological difficulty and word frequency. In Non-word-reading, the student had to decode pseudowords (N = 19 items) of ascending phonological difficulty. Tasks were given individually to students by the second author.
Spelling performance was assessed by the standardized Test of Spelling Ability by Mouzaki et al. (2007), which consists of 60 single words taken from school books and ordered in terms of ascending phonographic difficulty and word length. Items were dictated in sentences to students by their teachers.
Experimental Assessments
Four experimental assessments were developed by authors for the purpose of the study. A full description of items in terms of word frequency is given in the appendix. Experimental stimuli were taken from children’s school books (Grades 1 to 6) and selected from a corpus of 205 compounds considered to reflect a wide variety of morphological patterns of these words found in school texts (see Note 2). All spelling tasks were dictated to students in the class by their teachers, while the meaning task was given individually to them by the second author. These are described in detail as follows.
Spelling of compounds and their composites
Spelling of compounds and knowledge of their composites as well as their spellings were evaluated by a list of 20 triplets composed of a compound (e.g., ϵξώπoρτα/exoporta/external door) and their composites, namely, the composite A (e.g., έξω/exo/external) and the composite B (e.g., πόρτα/porta/door). A post hoc investigation of the psycholinguistic features of the composites that have been shown to influence performance (Yap & Balota, 2009) showed that there were no differences (p > .14) in letter length (means for composites, A: 4.7, B: 5.6), but the differences in terms of word frequency (means for composites, A: 12509, B: 885) were significant, F(1, 38) = 5.693, p < .05.
The teacher dictated to students only the compound and asked them to write down the composites on their own. Children had to spell the compound word along with their composites. Every accurately spelled compound was given 1 point. Since production and spelling of the compound elements constitute two different abilities, only the composite found and spelled correctly by the child was given 1 point.
Spelling of compounds in terms of part of speech
Spelling of compound words in terms of part of speech was evaluated by a spelling task consisted of 20 single compounds allocated in three main categories as indexed by their distribution in the corpus taken from the school books (see Note 2). These were nouns (N = 8) (e.g., χαρτoπόλϵμoς/chartopolemos/paperwar), adjectives (N = 7) (e.g., ηλιoκαμμένoς/iliokamenos/sunburnt), and verbs (N = 5) (e.g., ηλϵκτρoδoτώ/ilektroðoto/give electricity). The mean letter length of nouns, adjectives, and verbs was 10.50, 12.14, and 12.40 accordingly, while the mean word frequency of each category was 160.37, 155.42, and 0.80 as appropriately. Differences among categories in terms of letter length were not significant (> .065), but there were significant differences between nouns or adjectives and verbs (< .05) in terms of word frequency.
Spelling of concrete and abstract compounds
Spelling of concrete and abstract compounds was assessed by a spelling list of 20 single compounds allocated equally in concrete words (N = 10) (e.g., κακoκαιρία/kakokeria/badweather) and abstract words (N = 10) (e.g., αντιλαμβάνoμαι/andilamvanome/comprehend) matched in terms of letter length (means, concretes: 11, abstracts: 11.40, p >. 43) and word frequency (see Note 3) (means, concretes: 34.20, abstracts: 46, p >.79).
Meaning of concrete and abstract compounds
Meaning of concrete and abstract compounds was evaluated by the same list of compounds (N = 20; see “Spelling of Compounds in Terms of Part of Speech”), which was given previously for spelling. Children were asked individually whether they understand the meaning of each item (e.g., “Could you, please, tell me what does the word αλoυμινόκoυτo/aluminokuto/tinbox mean?”). Their responses were categorized in terms of their ability to explain etymologically—from a synchronic perspective—the meaning of the compound words as follows:
(Etymology+): Every accurate answer that entailed the morphological constituents of the target word (e.g., “Aλoυμινόκoυτo ϵίναι ένα κoυτί από αλoυμίνιo/aluminokuto ine ena kuti apo aluminio/a tinbox is a box from tin”).
(Semantics+): When the child showed semantic understanding of the word without any reference to its morphological constituents (e.g., “Aλoυμινόκoυτo ϵίναι ένα κoυτί πoυ πϵριέχϵι μπύρα/aluminokuto ine ena kuti pu periexi bira/a tinbox is a box that contains beer”).
(Etymology–): When the child gave a wrong answer that involved the morphological constituents of the word as an indication of its etymological understanding (e.g., “Eκτυφλωτικός ϵίναι o τυφλός/ektiflotikos ine o tiflos/overbright is the blind”).
(Semantics–): Every inaccurate response of the word without any reference to its morphological composition (e.g., “Γαλανoμάτα ϵίναι η ντoμάτα/γalanomata ine i domata/a blue-eyed lady is a tomato”).
Results
The spellings produced in every experimental assessment were classified as conventionally correct or as errors. Every accurately spelled word was assigned 1 point and every misspelled word 0 points. Mean percentage accuracy rates are used in every statistical analysis described in the following section.
Spelling of Compounds and Their Composites
Table 2 shows the spelling performance of the experimental and control group on the compounds and their composites. The significance of the difference between groups on spelling compounds was tested by a one-way analysis of variance. This verified significant effects for group, F(1, 99) = 46.015, p < .001. The significance of the difference between the composites was tested in a two-way 2 × 2 analyses of variance in which composites (composite A, composite B) was a within-participants factor and group (typical spellers, poor spellers) (see Note 4) a between-participants factor. This verified significant effects for composites, F(1, 99) = 150.069, p < .001, and group, F(1, 99) = 33.473, p < .001, confirming that the first composite of the compounds are spelled significantly better than the second one and poor readers perform significantly worse on both composites than typical spellers. The interaction Composites × Group, F(1, 99) = 3.644, p = 0.059, was not significant, implying that this effect is similar for the two groups.
Mean Accuracy Rate (%) for Spelling Compounds and Their Composites (Standard Deviations in Parentheses).
Spelling of Compounds in Terms of Part of Speech
Table 3 presents group performance on the compounds in terms of part of speech (see Note 5). The significance of the difference between the word types was tested in an analysis of variance in which part of speech (nouns, adjectives, verbs) was a within-participants factor and group (typical spellers, poor spellers) a between-participants factor. The analysis revealed significant effects for the part of speech, F(2, 202) = 38.813, p < .001, and group, F(1, 101) = 29.897, p < .001, verifying that poor spellers performed significantly below typical spellers but both groups spelled better nouns and adjectives than verbs. The interaction Part of Speech × Group, F(2, 202) = 2.730, p = .068, was nonsignificant, indicating that both groups were affected similarly by the part of speech of the words.
Mean Accuracy Rate (%) for Spelling Compounds per Part of Speech (Standard Deviations in Parentheses).
Spelling of Concrete and Abstract Compounds
Table 4 presents group performance in spelling of the concrete and abstract compounds. The significance of the difference between the word types was tested in an analysis of variance in which concreteness (concrete words, abstract words) was a within-participants factor and group (typical spellers, poor spellers) a between-participants factor. This showed significant effects for concreteness, F(1, 101) = 56.037, p < .001, and group, F(1, 101) = 33.273, p < .001, confirming that concrete words are spelled significantly higher than abstract words and that spellers with learning disabilities performed significantly worse on both word types. The interaction Concreteness × Group, F(1, 101) = 6.120, p < .05, was also significant, a difference attributed to the smaller difference between concrete and abstract words by the disabled group.
Mean Accuracy Rate (%) for Spelling/Meaning Concrete and Abstract Compounds (Standard Deviations in Parentheses).
Meaning of Concrete and Abstract Compounds
Table 4 also shows group performance in the meaning of the concrete and abstract compounds. The significance of the difference between the two word types was tested in an analysis of variance in which concreteness (concrete words, abstract words) was a within-participants factor and group (typical spellers, poor spellers) a between-participants factor. The analysis verified significant effects for concreteness, F(1, 89) = 606.323, p < .001, and group, F(1, 89) = 5.931, p < .05, validating that meaning performance of the concrete words is significantly advanced than the abstract ones and that the experimental group performed significantly lower than the control group. The interaction Concreteness × Group, F(1, 89) = 0.957, ns, was not significant, implying similar effects in terms of word concreteness for both groups.
Spelling and Meaning of Concrete and Abstract Compounds
Figure 1 illustrates the spelling and meaning performance of the two groups on the concrete and abstract compounds. The significance of the difference between the spelling and meaning of the concrete words was tested in an analysis of variance in which task (spelling, meaning) was a within-participants factor and group (typical spellers, poor spellers) a between-participants factor. This revealed significant effects for the task, F(1, 89) = 136.089, p < .001, and group, F(1, 89) = 15.920, p < .001, suggesting that spelling of concrete words is significantly worse than their meaning and poor spellers performed significantly lower than typical spellers. The interaction Task × Group, F(1, 89) = 23.865, p < .001, was significant since spelling of concrete words was far worse than their meaning for poor spellers.
Spelling and meaning performance on concrete and abstract compounds by typical and poor spellers.
The significance of the difference between the spelling and meaning of the abstract words was tested in an analysis of variance in which task (spelling, meaning) was a within-participants factor and group (typical spellers, poor spellers) a between-participants factor. This verified significant effects for the task, F(1, 89) = 31.932, p < .001, and group, F(1, 89) = 17.384, p < .001, confirming that spelling of abstract words is significantly better than their meaning and poor spellers performed significantly lower than typical spellers. The interaction Task × Group, F(1, 89) = 3.937, p < .05, was significant since differences between spelling and meaning of abstract words was smaller for poor spellers. These results suggest that both groups were shown to understand better than to spell the concrete compounds but to spell better than to understand the abstract ones. The significant interactions indicated that these effects are stronger for the children with spelling disabilities on these aspects.
Etymological Effects on the Semantic Understanding of Compounds
To evaluate the issue, whether poor spellers would understand less the compound structure than the able ones, children’s data from the meaning task were categorized across four categories in terms of whether they use etymological information in their definitions of these words (see “Meaning of Concrete and Abstract Compounds”). Table 5 shows this performance of the two groups on the semantic categories in terms of etymology. The significance of the difference between the meaning categories on the accurate responses of the two groups was tested in an analysis of variance in which category (Etymology+, Semantics+) was a within-participants factor and Group (typical spellers, poor spellers) a between-participants factor. This analysis showed significant effects for the category, F(1, 89) = 11.135, p < .01, and group, F(1, 89) = 5.469, p < .05. The interaction Category × Group, F(1, 89) = 1.112, ns, was not significant, implying that differences between meaning categories was similar for the two groups.
Mean Accuracy Rate (%) for the Semantic Understanding of Compounds in Terms of Group and Semantic Condition (Standard Deviations in Parentheses).
Accordingly, the significance of the difference between the categories on the inaccurate responses of the same groups was also tested in an analysis of variance in which category (Etymology–, Semantics–) was a within-participants factor. This revealed again significant effects for the category, F(1, 89) = 176.664, p < .001, and group, F(1, 89) = 5.716, p < .05. The interaction Category × Group, F(1, 89) = 12.483, p < .01, was significant due to higher performance on the inaccurate category of nonetymological responses for poor spellers. Both analyses verified that the accurate and inaccurate etymological responses are significantly higher than the relevant nonetymological ones and that poor spellers performed significantly below than typical spellers on both word types, suggesting that the two groups use at a small degree etymological information to explain the compound word but more than without any reference to their constituents. Poor spellers, however, understand less the morphological structure of the compounds than the typical spellers.
Discussion
The study aimed to investigate the spelling performance of compounds along with their semantic understanding on two groups of Greek primary school children with differing spelling abilities, an issue that has been underexplored so far despite its importance in later stages of literacy development.
Findings showed that children with spelling disabilities are significantly deprived in spelling compounds (21%) in relation to typical spellers (60%), a percentage that is relevant with their general spelling development as indexed by the standardized measure of spelling ability containing a variety of words including complex words (see Table 1). A further investigation of their ability to recognize the compound inner structure showed that Greek children are able to identify and spell better the first morphological constituent of each compound than the second one while poor spellers perform significantly worse on both composites than typical spellers. This finding is compatible with other studies where the first constituent plays an important role in recognition and spelling of the morphologically complex word (Kehayia et al., 1999; Libben, 1998; Schreuder, Neijt, Van der Weide, & Baayen, 1998) and possibly relates with the fact that word frequency of the first constituent was much larger than that of the second one.
In relation to the second hypothesis, spelling performance is affected by the part of speech of the target word. Results of the study verified this assumption, showing that students presented better performance on nouns and adjectives than verbs and poor spellers performed significantly below typical spellers on these aspects (typical spellers: 53% vs. 31%; poor spellers: 23% vs. 11%), a relevant finding with experimental literature (Tsapkini et al., 2002; Tsesmeli & Seymour, 2006). Comparisons between nominal and verbal items in terms of letter length and word frequency showed that nouns and adjectives were considerably more frequent than verbs.
But how important is the concreteness factor in affecting the performance of the compounds? Spelling data showed that concrete compounds evoked significantly higher rates than abstract compounds despite their match in terms of letter length and word frequency, and this effect was valid for the two groups. Poor spellers performed significantly worse than typical spellers on both word types; however, their differences were larger for the able spellers than the ones with spelling disabilities. On the other hand, meaning data showed a similar but more enhanced profile with spelling data, indicating that the concreteness factor influences more markedly the meaning performance of the groups on these aspects. In particular, concrete words were semantically understood significantly better than abstract words, and the experimental group performed significantly lower than the control group. However, the group with spelling disabilities presented a particularly low meaning performance on abstract words than on concrete ones in relation to their typical classmates.
Further evaluation of the differences between spelling and meaning data on each word type separately revealed an interesting profile. Figure 1 illustrates that children understand significantly better than spell the concrete words, but they spell better than understand the abstract ones. There is evidence from longitudinal data in different languages on preschool children that they appear to understand compounds typically before they are able to produce them verbally (Berman, 2009; Nicoladis, 2006). While the evidence in English for comprehension preceding production is weak, early comprehension and late production has been observed in French and Hebrew, mainly due to the complex morphology and infrequency of compounds in these languages (Nicoladis, 2006).
In relation to the present study, support comes from the concrete compounds, where their understanding has supposedly emerged during children’s preschool years but their written production after the literacy began, consequently their comprehension scores by typical spellers is significantly better (87.50%) than their spelling (65.50%). Apparently, this is not the case for the abstract words where they semantically are being acquired by typical spellers mainly from literary and scientific texts during their schooling (Ravid, 2006) and not in their verbal inputs during everyday life. As a consequence, they are able to spell these words (41%) significantly better than to comprehend them (25.86%). Undoubtedly, this picture is being obscured in the group with learning disabilities, where there are minor differences between meaning (12.38%) and spelling scores (20%) of abstract compounds as affected by their low literacy experience, while on the concrete words there is a large difference between their comprehension (79%) and their spelling (32%).
Nilsen and Bourassa (2008) found similar effects on their investigation of word-learning processes on the two word types. According to their views, there are a number of hypotheses regarding why concrete words are learned more easily than abstract words, and they argue that the level of semantic representation influences the ease with which words are learned. Moreover, they argue that there are representational differences between concrete and abstract words within children’s mental lexicons since concrete words are characterized by a greater number of semantic and specific features (i.e., color, shape, size, etc.), allowing thus for a greater retrieval of information from previous knowledge and experience. This factor seems to play a particular role for later stages of literacy acquisition, where abstract and complex words are more often found in school texts (Nagy & Anderson, 1984; Ravid, 2006), indicating that poor readers would face more difficulties with these word types at this phase of development.
Finally, an important question refers to the degree that semantic understanding of these complex forms is affected by their morphological or etymological characteristics. A thorough inspection of responses in relation to compounds’ definitions showed that the accurate etymological responses were higher than the accurate nonetymological ones, although these percentages were low even for typical spellers (31.50% vs. 25%, accordingly). These results were corroborated by the inaccurate responses of the definitions of the compounds showing an opposite but complementary pattern. In other words, it was found for both groups that responses where children acknowledged the morphological constituents of the target word (e.g., “a tinbox is a box from tin”) were rated over those that showed semantic understanding of the word without any reference to its morphological constituents (e.g., “a tinbox is a box that contains beer”).
This may suggest that this kind of awareness is present in children’s minds and despite its small magnitude can be used as a morphological strategy to enhance children’s spelling and vocabulary performance. The assumption here is that children with better developed morphological awareness skills may have an advantage in acquiring a complex vocabulary, as shown by Bowers and Kirby (2010), who found that students’ ability to identify bases in complex words contributed a significant amount of variance to their semantic knowledge, thus predicting their general vocabulary knowledge.
However, for poor spellers, data showed that they are less able to use etymological information (24.50%) in defining compounds than their typical classmates (31.50%), indicating that they understand less the internal structure of complex words, a finding that was also reported by other studies (Nunes & Bryant, 2006; Tsesmeli & Seymour, 2006). Consequently, morphological strategies can be used effectively by spellers with disabilities, causing significant effects on their spellings as shown by relevant interventions (Nunes & Bryant, 2006; Tsesmeli, 2010; Tsesmeli & Seymour, 2009).
Hence, the findings of this study have several important implications, namely, in serving as essential cognitive prerequisites for designing interventions appropriate for students who experience serious and prolonged spelling disability. Similar interventions should promote the teaching of the morphemic structure of the words in an explicit and sequential manner from the simplest to the most complex structures (i.e., nominal vs. verbal items or concrete vs. abstract items) and spelling should be combined with the understanding of these forms, since it has been widely acknowledged that morphological training should promote a student’s reflection of the hierarchical relations between words (i.e., word families or pairs; Nunes & Bryant, 2006).
These findings are important for the spellers with disabilities in Greek, which is a regular orthography (Seymour et al., 2003) in reading, but with more inconsistent phoneme-grapheme mappings in spelling. It also presents a wide variation of word productivity through compounding containing lots of polysyllabic and morphologically complex words in which morphemes are usually phonologically and semantically nontransparent (Agathopoulou, 2003; Ralli, 2005), thus posing particular difficulties for poor spellers as verified also by present results.
Some of the limitations of this study would be the deficiency of large samples for the experimental and control groups along with a more detailed description of their psychometric features (i.e., intelligence, vocabulary), due to practical circumstances in terms of school access, to increase the generalization of the findings to the relevant populations. Additionally, future experiments would involve more tight controls for item factors affecting results (i.e., morphosemantic transparency, frequency, and family size effects of compounds/constituents), as suggested by literature (Baayen, Kuperman, & Bertram, 2010). Nonetheless, it was important first to evaluate a wide variation of morphological patterns in compounds that are mostly found in school books and affect children’s performance.
However, this study is one of the few attempts of investigating the spelling and meaning of compounds on primary school children in a transparent orthography, as Greek. The findings are consistent with the experimental literature (Bowers & Kirby, 2010; Nilsen & Bourassa, 2008; Nunes & Bryant, 2006) and important for the development of alternative approaches in the assessment and intervention for individuals with spelling difficulties (Nunes & Bryant, 2006).
Footnotes
Appendix
In the following, numbers in parentheses are compound frequencies based on The Hellenic National Corpus (Institute of Language and Speech Processing, 2000) containing about 34,000,000 Greek words. Also, f = compound frequency, fA = frequency for Composite A, fB = frequency for Composite B.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.