Investigating orthographic and 1 RUNNING HEAD: ORTHOGRAPHIC AND SEMANTIC LEARNING IN POOR COMPREHENDERS Investigating orthographic and semantic aspects of word learning in poor comprehenders

This study compared orthographic and semantic aspects of word learning in children who differed in reading comprehension skill. Poor comprehenders and controls matched for age (9-10 years), nonverbal ability and decoding skill were trained to pronounce 20 visually presented nonwords, 10 in a consistent way and 10 in an inconsistent way. They then had an opportunity to infer the meanings of the new words from story context. Orthographic learning was measured in three ways: the number of trials taken to learn to pronounce nonwords correctly, orthographic choice and spelling. Across all measures, consistent items were easier than inconsistent items and poor comprehenders did not differ from control children. Semantic learning was assessed on three occasions, using a nonword-picture matching task. While poor comprehenders showed equivalent semantic learning to controls immediately after exposure to nonword meaning, this knowledge was not well-retained over time. Results are discussed in terms of the language and reading skills of poor comprehenders and in relation to current models of reading development. Investigating orthographic and 3 Investigating orthographic and semantic aspects of word learning in poor comprehenders Poor comprehenders are children who experience difficulties with reading comprehension, despite age-appropriate reading accuracy. Many experiments have shown that poor comprehenders are poor at text-level processes such as making inferences and comprehension monitoring (for reviews see Cain & Oakhill, 2007; Nation, 2005). However, poor comprehenders’ difficulties are not restricted to the comprehension of written text. They experience relative weaknesses in comprehending orally presented sentences and discourse, and with listening comprehension more generally (Catts, Adlof, & Weismer, 2006; Nation, Clarke, Marshall, & Durand, 2004). Narrative expression is also compromised in both written (Cragg & Nation, 2006) and oral (Cain, 2003) modalities. Researchers studying poor comprehenders take care to select children whose reading accuracy is age-appropriate; nevertheless, there is some evidence that poor comprehenders show relative weaknesses when reading exception words – words that have atypical mappings between spelling and sound (e.g., break, yacht). Nation and Snowling (1998) first reported that poor comprehenders were significantly less accurate at reading exception words than skilled comprehenders, despite the two groups being matched for phonological decoding (nonword reading) and nonverbal reasoning scores. This finding has since been replicated (Ricketts, Nation, & Bishop, 2007). Exception words can only be partially decoded using the alphabetic principle (Share, 1995). This suggests that good decoding skills alone are not enough to support efficient exception word reading. Various proposals have been made about additional skills that might underpin exception word reading. One is that orthographic processing Investigating orthographic and 4 skills are important (e.g., Harm & Seidenberg, 1999; Manis, Seidenberg, Doi, McBrideChang, & Petersen, 1996). On this view, the exception word reading deficit seen in poor comprehenders may be due to weak orthographic knowledge. Typically, orthographic processing skills are assessed using tasks such as orthographic choice (e.g., which is a word: assure or ashure?) or by measuring proxy variables such as print exposure. It is worth noting that such measures are not independent of the word recognition process and therefore it is not clear whether such tasks measure word recognition itself, rather than being an index of a separable orthographic construct (for a fuller discussion of this issue see Castles & Nation, 2006; Burt, 2006). Notwithstanding the lack of consensus as to what constitutes “orthographic processing”, in our research we have found that poor and skilled comprehenders do not differ in two measures of orthographic knowledge print exposure and orthographic choice (Ricketts et al., 2007). Nation and Snowling (1998) presented an alternative explanation, suggesting that poor comprehenders’ exception word reading deficits may be a consequence of semantic weaknesses in the oral domain. Their theorizing was inspired by a connectionist model of visual word recognition and its development (Plaut, McClelland, Seidenberg, & Patterson, 1996). Simulations of this model revealed that a word recognition system trained with input from semantics was better able to compute correct pronunciations of exception words than a model trained without a semantic contribution (Harm & Seidenberg, 2004; Plaut et al., 1996) thus suggesting that semantic knowledge may play a direct role in the reading of exception words. Similarly, Keenan and Betjemann (2007) suggested that semantic knowledge knowledge of word meanings could provide compensatory support for exception word reading where mappings between spelling and Investigating orthographic and 5 sound are weak. For example, if an exception word is unknown, a child will have no choice but to attempt to decode it. This would result in a mispronunciation that is similar to the target word e.g., pronouncing yacht to rhyme with matched rather than pot. A child who has good semantic knowledge and a large vocabulary is more likely to be familiar with the exception word that they are trying to read. Also, Bowey and Rutherford (2007) suggested that a child with stronger vocabulary skills is more likely to vary the pronunciation of an unknown word to read the word successfully. Consistent with the idea that good vocabulary skills contribute to successful exception word reading, vocabulary predicts later exception word reading above and beyond decoding and several studies find oral vocabulary deficits in some poor comprehenders (Cain, Oakhill, & Lemmon, 2004; Nation & Snowling, 1998, 2004; Ricketts et al., 2007; but see Stothard & Hulme, 1992). Mirroring these findings, children with selective weaknesses in exception word reading also show concomitant weaknesses in oral vocabulary (Bowey & Rutherford, 2007; Byrne, Freebody, & Gates, 1992). In this study we sought to investigate word learning in poor comprehenders. In order to build a sight vocabulary, a child needs to make links between phonological and orthographic information, and between these representations and a word’s semantic characteristics. According to the Lexical Quality Hypothesis (Perfetti & Hart, 2002) a good quality lexical representation is one where stored phonological, orthographic and semantic information about a word is well integrated. This hypothesis is underspecified in that it does not provide more explicit predictions about which aspect of the representation is most important. However, it does provide a useful framework for considering visual word learning in suggesting that failure to learn a novel word might be Investigating orthographic and 6 a consequence of weak phonological, semantic or orthographic knowledge, or of an inability to develop associations between any of these representations. Poor comprehenders do not have difficulty processing phonology (Catts et al., 2006; Nation et al., 2004), with making consistent links between phonology and orthography (i.e., regular word reading, Nation & Snowling, 1998) or with associating a new phonological form to its new referent (Nation et al., 2007). Yet, we know that at least some poor comprehenders have low vocabulary knowledge and these children also struggle to read exception words, despite performing at a similar level to controls on other tasks thought to tap orthographic knowledge such as print exposure and orthographic choice (Ricketts et al., 2007). This experiment examined the acquisition of semantic and orthographic representations for new words that have consistent versus inconsistent spelling-sound mappings. Castles and Holmes (1996) explored visual word learning in children with surface dyslexia children who show particular difficulty reading exception words. They sought to investigate exception word learning by training children to pronounce printed nonwords such as bouch. The nonwords were assigned an inconsistent pronunciation that subjects would not produce on the basis of grapheme-phoneme conversion rules bouch was pronounced to rhyme with touch rather than couch. Castles and Holmes found that children with surface dyslexia had difficulty with this task, suggesting that they may have exception word reading difficulties because they have difficulty learning exception words. Using a similar paradigm Bailey Manis, Pedersen and Seidenberg (2004) extended this research and found that children with dyslexia were poorer than agematched controls at learning both consistent and inconsistent items. Investigating orthographic and 7 Castles & Holmes (1996; also Bailey et al., 2004) examined the process of forming associations between orthography and phonology – a process we term orthographic learning. To our knowledge orthographic learning has not been investigated in poor comprehenders. However, Cain and colleagues (Cain, Oakhill, & Elbro, 2003; Cain et al., 2004) have looked at vocabulary acquisition in poor comprehenders. In their research poor and skilled comprehenders are exposed to novel words (nonwords) in a context that provides cues for their meaning to be inferred. Poor comprehenders were poor at inferring the meaning of the novel words. Nation, Snowling and Clarke (2007) also report a study showing that poor comprehenders have difficulty learning semantic information about novel objects. Thus it seems that as well as having poor existing oral vocabulary, poor comprehenders show weaker learning of semantic information for novel vocabulary – a process we refer to as semantic learning. In this study we adapted the paradigm used by Castles and Holmes (1996) and Bailey et al. (2004) to probe semantic as well as orthographic aspects of word learning. We trained poor and skilled comprehenders to pronounce 10 nonwords in a consistent way, and 10 in an inconsistent way. Given poor comprehenders’ deficit in reading exception words (Nation & Snowling, 1998, 2004; Ricketts et al., 2007), we anticipated that they might find it more difficult to learn the inconsistent nonwords, relative to control children. After training, poor and skilled comprehenders were exposed to the 20 nonwords embedded in story context (cf. Cain et al., 2003; Cain et al., 2004). Ten nonwords (five consistent and five inconsistent) were presented in an unhelpful context, which provided little information about their meaning. The remaining 10 nonwords were presented in a helpful context, which provided cues to each nonword meaning. This was Investigating orthographic and 8 done to assess the children’s ability to use context to infer the meaning of new words. Based on previous studies (Cain et al., 2003; Cain et al., 2004; Nation et al., 2007), we hypothesized that poor comprehenders would learn fewer novel word meanings. In summary, our primary aim was to investigate whether poor comprehenders have difficulty with orthographic and semantic aspects of word learning. To our knowledge this is the first time that semantic and orthographic learning have been contrasted, and the first time that orthographic learning has been investigated in poor comprehenders. We predicted that poor comprehenders would have difficulty with semantic learning as they show poor semantic knowledge (e.g., Nation & Snowling, 1998) and do not use context to learn the meaning of new words in other word learning experiments (Cain et al., 2003; Cain et al., 2004). We also predicted that like children with dyslexia, poor comprehenders might show poor orthographic learning overall, and perhaps particular difficulty learning inconsistent items (cf. Bailey et al., 2004; Castles & Holmes, 1996). However, this prediction was slightly tempered by our previous finding that poor comprehenders do not perform poorly on some orthographic tasks, despite being poor at reading exception words (Ricketts et al., 2007).