Marco Marelli, Daniela Traficante and Cristina Burani Reading morphologically complex words: Experimental evidence and learning models Abstract: The study of complex word processing has been centered on the notion of morpheme as a processing unit. Evidence from psycholinguistics and cognitive neuropsychology has been taken as suggestive of symbolic morphemic representations at the lexical level, on a par with words. However, several phenomena observed in morphological processing suggest a more complex picture. The crucial role played in reading by the distributional properties of both the complex word and its morphemic constituents (e.g., family size, morphological entropy, orthography-semantics consistency) highlights the limits of the ‘morpheme-asunit’ assumption. Moreover, results from the developmental literature show that morphology is an age-related emergent aspect of written word processing, exploited to overcome reading challenges for both typically developing readers and children with dyslexia. A unitary account for this complex scenario may be offered by learning models that focus on form-to-meaning mapping. Keywords: lexical morphology, word processing, reading acquisition, learning models 1 Introduction In its most traditional definition, morphemes are characterized as the minimal meaning-associated units in a language (Bloomfield 1933). This definition, although maybe simplistic and descriptive (Blevins 2016), makes it immediately clear why the topic has attracted so much attention in the psycholinguistic community. As information-carrying elements, morphemes are potentially very helpful in language processing, providing useful cues about the meaning of a given word. As a consequence, the question as to whether and how morphological information plays a role at the cognitive level has been central in psycholinguistic research, especially the one focusing on word processing. In this Marco Marelli, Department of Psychology, University of Milano-Bicocca, (Italy) Daniela Traficante, Department of Psychology, Catholic University of Milan, (Italy) Cristina Burani, Institute of Cognitive Sciences and Technologies, National Research Council, Rome, (Italy); Department of Life Sciences, University of Trieste, (Italy) Open Access. © 2020 Marco Marelli et al., published by De Gruyter. This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License. https://doi.org/10.1515/9783110440577-014 554 Marco Marelli, Daniela Traficante and Cristina Burani chapter we will discuss the achievements of this research field, by stressing to what extent the theoretical view on the role of morphology in word reading has changed (and it is still changing). Our main focus will be on results from written word reading, as main source of evidence in the literature. Throughout the chapter we will consider evidence drawn from two main tasks, i.e., visual lexical decision (participants decide whether a printed stimulus is a word or not) and word naming (participants read aloud as fast as possible a given letter string, be it a word or a non-word). The two tasks tap into partially different components of the reading process, thus they may highlight different roles for morphemes and morphological processing, depending on the involved processing components. Data from adult readers, brain-damaged patients, and children with and without reading deficits will be discussed. 2 A role for morphemes Most of the studies on the morphological processing of adult readers has been guided, more or less explicitly, by the hypothesis that morphemes and/or complex words are stored as representational units within the cognitive system. Researchers mostly assumed that morphologically complex words could, in principle, activate representations in the reader’s mind (or her/his mental lexicon) that are univocally associated to some linguistic concept (the ideas of morpheme and/or word). This “representational view” characterized the morphological processing research, especially in its early history. In fact, one of the most persistent experimental questions wondered whether morphemes are accessed in visual word recognition. 2.1 Listing models and parsing models In early debates, two main views on the issue were considered. Either morphologically complex words were segmented during reading, and hence morphemes were represented and accessed in the cognitive system, or the mental lexicon was populated by whole words, and hence alleged morphological effects in word processing were to be considered a by-product of the relations between independent representation units. The former position, represented by full-parsing theories, is ideally exemplified by the model proposed by Taft and Forster (1975). In this perspective, words are accessed through the representations of the morphemes that form Reading morphologically complex words 555 them, implying that an early procedure is first applied to segment the complex word in its constituents, that are in turn used to compute the whole-word representation. In the original proposal by Taft and Forster, which dealt with the processing of prefixed stimuli, the procedure was accomplished by a prefixstripping operation that is able to parse a word on the basis of the prefix, consequently isolating the constituent morphemes (for an alternative architecture, see Taft 1994). The parsing-and-recombination procedure is traditionally considered to be costly (in terms of cognitive resources), but constituted a reasonable proposal in so far (i) said procedure is necessary to understand the meaning of novel words (we easily understand the meaning of windowless, even if we have never heard the word before) and (ii) through its explicit representations, it accounts for the morphological awareness that even uneducated speakers can manifest. The latter position opposes to full-parsing theories the full-listing proposals (or whole-word approaches; e.g., Butterworth 1983), that posit an explicit representation for any complex form, being it derived, inflected, or compound. For example, in this perspective separate units for run, running, runs, runner, homerun, etc. will be included in the mental lexicon. In this model morphology would not emerge as a direct consequence of explicit morpheme access, rather depending on reliable and stable (paradigmatic, even) relations between independent lexical units. The processing of runner will not be characterized by morphological effects because the -er affix is stripped from its stem; rather, these will depend on the orthographical, phonological, and semantic overlap between runner and run. A third position has also emerged, proposing middle-ground views between either extreme proposals described above. In the mixed or dual-route models, both morphological segmentation and whole-word access are possible, with the efficiency of either operations depending on many possible factors. Examples of this position are the Augmented Addressed Morphology (AAM) model (Burani and Caramazza 1987; Caramazza, Laudanna, and Romani 1988) and parallel dual-route models by Schreuder and Baayen (1995; Baayen, Dijkstra, and Schreuder 1997) and Grainger and Ziegler (2011). In these perspectives, word processing would proceed both in a morpheme-based fashion and by means of a direct access to the word representation. The two procedures will always be in place at the same time, although in some cases a given route will be more efficient than its counterpart (e.g., the parsing route for novel words, the direct route for high-frequency words). The mechanisms underlying the interplay between the two routes vary, with some approaches assuming a completely parallel, horse-race architecture, and other positing interacting procedures. Indeed, one of the more recent models of this family, the multi-route account by Kuperman et al. (2009), is characterized by massively interacting 556 Marco Marelli, Daniela Traficante and Cristina Burani multiple procedures that maximize the efficiency in word processing by exploiting all available cues to access word meaning. 2.2 Empirical evidence for morphological effects It must be noted that, in “representational” approaches such as the ones just described, it is difficult to completely get rid of the morpheme role. Certainly, even whole-word approaches do not completely exclude morpheme-based processes, rather relegating them to limited, infrequent contexts able to trigger these “special” operations (e.g., the first time a novel complex word is encountered). And indeed, empirical results seem to be in line with the morphological assumption. In the remaining of the present section, we will describe pieces of evidence that make morpheme representations fundamental for models of written word recognition. We will focus on (i) effects of morphological structure in word reading, (ii) morphological priming effects, (iii) letter transposition effects, and (iv) morpheme frequency effects. Morphological structure is known to influence word processing, even in nonlexical orthographic strings. In their pioneering lexical-decision study, Taft and Forster (1975) found that non-words made of an existing prefix paired with an existing stem (de+juvenate) take longer to be rejected than combinations of existing prefixes and non-existing stems (de+pertoire). That is, when asked to decide, as fast as possible, whether a given string is an existing word or not, participants find more difficult to evaluate dejuvenate than depertoire. The effect holds for inflectionally suffixed non-words, as shown by Caramazza, Laudanna, and Romani (1988) in Italian, and is confirmed when combinations of existing stems and nonexisting suffixes are compared with combinations of existing stems+derivational suffixes (vetralle vs. vetrezza; e.g., Burani et al. 1997). Even when a difference in response times does not emerge (e.g., Burani, Marcolini, and Stella 2002), an effect on the responses can still be observed, with items composed of morphemes accepted as “words” more often than items not including any morpheme. Recently, Crepaldi, Rastle, and Davis (2010; see also Crepaldi et al. 2013) have shown that the effect is not only determined by the lexical status of the constituent morphemes, but also by their position. That is, a suffix is expected to be found at the end of a word, hence shootment is slower to be rejected than shootmant, whereas no difference is observed between mentshoot and mantshoot. The evidence provided by this “morpheme interference” paradigm has reliably shown that a string having an acceptable morphological structure is taken as a possible word more than a string lacking this property. Readers are particularly sensitive to this structure, and find it harder to consider corresponding items as Reading morphologically complex words 557 non-existing. It is worth noting that the “morpheme interference” effect applies also to languages with a non-concatenated morphological structure such as Hebrew, in which root letters are interleaved with pattern letters. For Hebrew morphologically complex non-words, the magnitude of the costs due to root extraction (on both response latencies and accuracy) was found to be much stronger than in English (Yablonski and Ben-Shachar 2016), suggesting that the morphemic interference effect might be correlated to the richness in morphology of a given language more than to the linearity of its structure. Implicitly, all these results provide strong support to the hypothesis that morphemes play a role in written word processing. Certainly, how this role is expressed is not necessarily straightforward – most studies tend to adopt a lexical interpretation, in which the activation of morpheme representations associated with items like shootment makes it more difficult to reject them in a lexical decision task. However, also a semantic interpretation is possible: since legal morpheme combinations are potentially novel words (e.g., Marelli and Baroni 2015), participants may take more time rejecting them because they cannot help but computing the corresponding novel meanings. Indeed, we may find it difficult to consider windowless a nonword, even if we have never heard it before, because the associated meaning is so easy to compute that seems familiar to us. Consistently with this view, Burani et al. (1999) showed that lexical decision to pseudo-words made up of a root and a derivational suffix is affected by the semantic interpretability of the root-suffix combination (although not by its grammatical appropriateness): non-existent root-suffix combinations took longer to be rejected and resulted in more false alarms when they had been rated as highly interpretable than when they had a lower interpretability. It must be noted that computation of meaning was found to be task-dependent: in the same study, the degree of interpretability of new root-suffix combinations did not affect naming performance. However, and irrespective of differences in interpretability, pseudo-words made up of two morphemes were named faster and more accurately than pseudo-words with no morphological constituency (see also Burani, Marcolini, and Stella 2002). The high sensitivity to semantic variables of lexical decision and the insensitivity to semantics of word naming was confirmed by Baayen, Wurm, and Aycock (2007) for morphologically complex words. The latter two studies, while showing the centrality of the semantic component in the licensing process involved in lexical decision on complex stimuli, also indicate the dissociability of meaning from the activation of morphological structure in word naming, thus suggesting the possibility of morpho-lexical non-semantic word naming. Whatever the interpretation (purely lexical or semantics-oriented), the results obtained from both lexical decision and naming show the importance of morphology in visual word recognition. 558 Marco Marelli, Daniela Traficante and Cristina Burani Further evidence for the crucial role of morphemes has been provided by lexical decision experiments which adopted the priming technique. In this paradigm, typically, a prime word (e.g., cat) is presented before the key, target item (e.g., dog), on which a response is requested to the participants. If participants are faster at answering in the above setting as opposed to a corresponding, control condition (e.g., dog preceded by the prime sad), it will follow that the association between dog and cat is somehow relevant at cognitive level. This very paradigm has been applied to the morphological domain by introducing complex words (as targets or, more often, primes) in place of the monomorphemic examples above, with a priming effect reliably emerging in this condition too (pairs like follower-follow elicit faster response times than paired control conditions). The phenomenon is present both in cross-modal (auditory prime and visual target) contiguous priming (e.g., Marslen-Wilson et al. 1994) and in intra-modal contiguous and long-term visual priming (e.g., Drews and Zwitzerlood 1995; Feldman and Soltano 1999; Rueckl and Aicher 2008). However, in all these variants of the paradigm, morphological priming is most evident for semantically transparent primes (e.g., punishment-punish), and it is usually not observed when semantically opaque primes (e.g., department-depart) are considered. In other words, when the association between the complex prime and its stem target is not also sustained by semantic similarity, morphological priming effects do not typically emerge (see, however, the next section for a more thorough discussion of the issue). Nevertheless, the effect cannot be discarded as a simple by-product of semantic similarity: as Feldman (2000) demonstrated, morphological priming effects are significantly larger (and more long-lasting) than the sum of effects observed in purely semantic (pledge-vow) and orthographic (vowel-vow) conditions (see also Drews and Zwitzerlood 1995). This supports the hypothesis that morphological relatedness is distinct from the composite effects of semantic and orthographic similarity, further sustaining the idea of representational units for morphemes in the mental lexicon. Often in conjunction with a priming paradigm, the transposed letter effect (Forster et al. 1987) has been also applied to the study of morphology. This effect is observed when imprecisions in the position of word letters are tolerated so that a non-word is identified as its lexical counterpart (jugde read as judge). Christianson, Johnson, and Rayner (2005) showed that primes containing letter transpositions within morphemes (e.g., baoster-boaster) facilitate word naming as much as correctly spelled primes, whereas the same advantage is not found for primes having letter transpositions across morpheme boundaries (e.g., boasetr-boaster). This pattern of results was also found in Basque and Spanish using a lexical decision task (Duñabeitia, Perea, and Reading morphologically complex words 559 Carreiras 2007). Again, these pieces of evidence support the morpheme-asunit position: jumbling letters is not particularly disruptive when morpheme boundaries are respected, suggesting that word access proceeds on the basis of its constituent morphemes. Finally, lexical processing is influenced by the frequency of the word morphemes. Lexical frequency is a measure of how often a given lexical unit appears in a language vocabulary and it is simply estimated by counting word occurrences in a large collection of texts. The frequency effect is one of the oldest and most studied effects in psycholinguistics: the higher the frequency of a word, the faster a reader will process it (in terms of, e.g., response times or eyefixation durations; Solomon and Howes 1951). It is usually considered diagnostic of word representation: if a frequency effect can be observed for a given lexical category, the member of that category should be represented in the cognitive system (and accessed during language processing). Indeed, over and above the effect of word frequency, an effect of morpheme frequency is also observed in many languages (e.g., English: Taft 1979, Hay 2001; Dutch: Baayen, Dijkstra, and Schreuder 1997; Italian: Burani and Caramazza 1987; Burani, Salmaso, and Caramazza 1984), suggesting that morpheme representations are routinely accessed when reading a morphologically complex word. However, a reader is not only influenced by the frequency of morphemes, but also by the frequency of the complex word as a whole, indicating that the representation of the complex form is retrieved as well (and leading to the development of the dual-route models described above). The specific interplay between the involved representations is still not clear at the moment, and there is evidence indicating that the scenario is probably more complex than the one described in dual-route systems. For example, Baayen, Dijkstra, and Schreuder (2007) showed that lexical decision latencies are characterized by an interaction between stem and derived-word frequencies, with inhibitory stem effects for highfrequency words and facilitatory stem effects for low-frequency words. Burani and Thornton (2003) found an interaction between stem and suffix frequency effects, with the former crucially determining the emergence of the latter. Conversely, Ford, Davis, and Marslen-Wilson (2010) showed that stem frequency facilitated responses but only to productively suffixed derived words. In conclusion, even if some results are not completely straightforward, and possibly more complex scenarios are suggested, the evidence reviewed indicates that morphological information plays a role in word processing. The dominant view on how these morphological effects are expressed at the cognitive level has been connoted in representational terms: explicit, symbolic units for morphemes are stored in the mental lexicon, and are activated during the processing of a complex word. 560 Marco Marelli, Daniela Traficante and Cristina Burani 3 Morpheme representations: From IF to HOW Having established the importance of morphological information in word reading, more recent psycholinguistic literature has investigated how this information unfolds when a complex word is processed. In this perspective, the morphological-processing literature has progressively moved its focus from the question “Are morphemes represented in the mental lexicon?” to the question “How and when are morpheme representations accessed in word processing?”, generating one of the most heated debate in the field, that has dominated the literature during the last ten years. 3.1 The early processing of morphemes The main controversy concerned how early morpheme representations come into play during word reading, and in particular whether morphemes are accessed before or after lexical access. Initially, the literature has seen two opposing approaches, usually labelled as sub-lexical and supra-lexical (or morpheme-based vs. lexeme-based, Aronoff 1994). The former position (Rastle, Davis, and New 2004; Taft 2004) ideally follows from the traditional fullparsing models (Taft and Forster 1975), positing that complex words are automatically parsed and the resulting morpheme representations lead to lexical access (either through recombination or via spreading activation). The latter perspective (e.g., Giraudo and Grainger 2000) sees morphological access as a consequence of word activation, and dependent on an abstract representation level at which lexemes are organized in morphological families. Within a hybrid account (Diependaele, Sandra, and Grainger 2009), the parallel access to wholeword representations and morpho-orthographic units should maximize the probability of successful word recognition, at least for morphologically simple languages such as English (Beyersmann, Coltheart, and Castles 2012). The focus on early processing is still evident in the present literature – the review by Amenta and Crepaldi (2012) on morphological processing discussed early effects as crucial for model adjudication – and has led to new trends in the adopted methodology. Given the central role of timing in current research on morphology, techniques that have good temporal resolutions have become progressively more important. Results from eye-tracking studies have been used to investigate what happens when a word is fixated for the first time – namely, during the first 150–250 ms of processing (e.g., Kuperman, Bertram, and Baayen 2010). Event-Related Potentials (ERP) permit to capture how brain activation (in terms of electric signal on the scalp) unfolds when reading a Reading morphologically complex words 561 complex word, and are now quite widespread in the investigation of morphological processing (e.g., Lavric, Clapp, and Rastle 2007; Morris, Grainger, and Holcomb 2008). However, probably the most popular technique used to address the “early processing” issue has been the masked priming paradigm (Forster and Davis 1984). This technique is very similar to the traditional priming approaches, with the crucial difference being how the prime stimulus is treated. In masked priming, the prime is presented very briefly (usually less than 50ms) and squeezed between a forward mask (e.g., a string of hash marks) and the target item itself. Under these conditions, the prime is virtually invisible, ensuring that the resulting priming effect will not be influenced by the conscious appreciation of the prime-target relation. Over and above this desirable aspect, masked priming is usually taken as a way to isolate early cognitive processes. The assumption is that the masking condition limits the processing of the prime, so that any observable effect on the target would depend on information that is extracted from the prime during its very short presentation; as a consequence if, for example, a morphological priming is observed under these conditions, it will follow that morphemes are accessed during the first 50ms of processing. Although this interpretation is not granted, and different positions concerning the paradigm exist (see, e.g., Tzur and Frost 2007, for a perceptual explanation, or Norris and Kinoshita 2008, for a task-dependent theory), it is the most common one in morphological processing literature. In these terms, the very evidence from masked priming strongly supports sub-lexical models of word access, with morphological effects clearly emerging from the paradigm: prime-target pairs in which the prime is a complex word (e.g., killer-kill) elicit larger priming effects than orthographic control pairs, that are similar in form but not morphologically related (e.g., scandal-scan). This was taken as evidence that words are automatically parsed into their morphemes early, which are in turn activated before word access (Rastle et al. 2000). Interestingly, masked priming effects interpretable as morphological parsing can be observed for both semantically transparent derived words (e.g., killer) and opaque words whose morphological complexity is only apparent (e.g., corner).1 That is, larger priming effects are found for both killer-kill and 1 Most often, the word recognition literature does not distinguish between genuine morphologically complex words whose meaning cannot be fully derived compositionally (semantically opaque words like courteous) and words that are apparently complex because of a sheer orthographic chance (pseudo-complex words like corner). In fact, empirical evaluations found no difference in behavioral responses between either case, at least when masked priming is applied (e.g., Longtin, Segui, and Halle 2003). Following this tradition, in the present chapter 562 Marco Marelli, Daniela Traficante and Cristina Burani corner-corn pairs, as opposed to scandal-scan, even if there is no real relation (semantic or morphological) between corner and corn. The reliability of these results, replicated in masked priming studies in a number of languages (e.g., French: Longtin, Segui, and Halle 2003; English: Rastle, Davis, and New 2004; Russian: Kazanina et al. 2008; Italian: Marelli et al. 2013), led to the hypothesis that early morphological parsing proceeds in a semantically blind, form-based fashion. In other words, early parsing is morpho-orthographic: words that are, form-wise, morphologically complex will be automatically parsed into their constituent morphemes irrespective of any high-level consideration about their meanings. Indeed, this procedure will not even be influenced by the lexicality of the complex form: priming effects are found also when using novel derived words as primes (e.g., quickify; Meunier and Longtin 2007). 3.2 Semantic modulation of morpheme access However, positing that there is no role for semantics in early word decomposition does not necessarily mean excluding semantic influence completely. In fact, how meaning can modulate morphological effects at early levels remains the center of a heated debate (Rastle and Davis 2008). On the one hand, form-then-meaning accounts (Rastle, Davis, and New 2004) assume very early morpho-orthographic parsing, with semantics entering in the picture only at later stages (where, indeed, no priming effect can be found for semantically opaque pairs; Rueckl and Aicher 2008). On the other hand, form-with-meaning accounts (Feldman, O’Connor, and Moscoso del Prado Martín 2009) point to an early involvement of word and morpheme meanings, with semantics influencing the ease of morpheme processing. Divergent predictions concerning priming patterns are quite straightforward: whereas the latter explanation implies a significant difference in masked-priming effects for transparent as opposed to semantically opaque item pairs (that is, larger priming effects for killer-kill than corner-corn), the former does not predict such a difference. The debate is far from being settled, also considering that it is difficult to empirically sustain a null effect. And if many studies have failed in finding such semantic effect (e.g., Beyersmann et al. 2015b; Kazanina et al. 2008; Longtin, Segui, and Halle 2003; Rastle, Davis, and New 2004), there are also a number of experiments in which the semantic effect significantly emerges (e.g., Feldman et al. 2012; Feldman, O’Connor, and Moscoso we will use the terms “opaque” and “opaqueness” to define both cases like courteous and cases like corner. Reading morphologically complex words 563 del Prado Martín 2009; Järvikivi and Pyykkönen 2011; Kazanina 2011; Marelli et al. 2013). Certainly, even when statistically significant, the semantic effect seems to be small. Recent evidence suggests that both explanations may be founded on simplistic assumptions, as early morpho-semantic effects are more complex than expected. Marelli et al. (2013), for example, have shown that evidence of morpho-orthographic decomposition is crucially dependent on the lexical-decision paradigm. In fact, if, in place of performing a lexical decision on a primed target, participants are asked questions concerning the semantic category of the target (e.g., “does the word denote an animal?”), eye fixation times reveal priming effect in the transparent condition only (i.e., significant facilitation for killer-kill but not for either corner-corn or scandal-scan). Said facilitation is already evident on first-fixation durations on the word, indicating that the semantic contribution emerges early during processing. Such task-dependent effects are problematic for accounts based on obligatory morphological decomposition. Moreover, Tsang and Chen (2014), in experiments on Chinese, have found significant priming effects for pairs like butterfly-milk in masked conditions. That is, even if there is automatic morpheme-access in semantically opaque compounds, the semantic features of the morphemes are nevertheless activated (see also Tsang and Chen 2013). In other words, the segmentation process may be semantically-blind, but the morpheme activation is semantically connoted. Further evidence in this regard has been provided by Amenta, Marelli, and Crepaldi (2015) in Italian, in natural language-processing situations: the very same semantically opaque words (e.g., gallone – ‘gallon’, lit. bigrooster) are characterized by inhibitory stem-frequency effects on reading times when presented in sentences prompting their opaque meaning (gallone as gallon), as opposed to facilitatory stem-frequency effects when embedded in sentences prompting the potential transparent meaning (gallone as big rooster). A corresponding example in English would be summer, which primarily denotes a season but, in the right context, can indicate someone who sums (“I am not good at math, but I am a good summer”). In this perspective, morphologically complex words are parsed irrespective of their semantic transparency, but morpheme meanings are accessed straight away, even if their semantic contribution is not helpful for computing the complex word meaning (as it happens to be the case for semantically opaque words; see also Marelli and Luzzatti, 2012). According to the model by Marelli and Baroni (2015), a compositional perspective would be crucial to understand these pieces of evidence and, more in general, the role of meaning in morphological processing. In this perspective, a priming effect would emerge also for semantically opaque words at early processing stages not because semantics is not important, but because an erroneous, 564 Marco Marelli, Daniela Traficante and Cristina Burani “transparent”, alternative meaning is automatically computed. That is, masked priming would limit information uptake so that no explicit lexical knowledge about the whole-word meaning could be accessed; as a consequence, morphemes would be automatically combined in a productive, synchronic fashion, generating a whole-word meaning that will be semantically related to its stem even for semantically opaque words (e.g., summer as someone who sums, irony as made of iron, etc.). In other words, Marelli and Baroni (2015) moved the research focus from the “form vs. meaning” debate to the understanding of which type of semantics influences morphological effects at different processing levels. Hopefully, this change in perspective may help solving the deadlock that seemingly has characterized recent literature on early morphological effects. 4 The contribution of cognitive neuropsychology Cognitive neuropsychology is the study of the mechanisms of the mind through the assessment of brain-damaged individuals. By examining the behavior of such individuals, in principle, it is possible to establish relations between cognitive functions and brain structure, as well as to individuate specific cognitive modules (as in the Fodor’s definition; Fodor 1983). For example, if a certain patient exhibits the behavior A and not the behavior B, and in a different patient the behavior B is spared whereas the behavior A is impaired, this double dissociation will indicate that the two behaviors are the expressions of specific and separate mind components. Not surprisingly, methods from cognitive neuropsychology have been also applied to the investigation of morphological processing, with the aims of isolating the processing of morphologically complex words, or finding evidence for the representations of morphemes in the mental lexicon. Most results in favor of a morphological level of analysis in the language system come from the assessment of people with acquired reading disorders. Following a brain injury, patients can manifest an impairment when processing written materials, a problem that is evident in their language productions. For example, a patient could read the word thing as think (visual error), the word bottle as cup (semantic error), or the word speak as speaker (morphological error). This error pattern (along with the inability of reading unfamiliar and novel words) is the typical manifestation of deep dyslexia (Coltheart, Patterson, and Marshall 1980). The morphological manifestation of the disorder has been considered indicative of a representation level that is specific to lexical morphology (Patterson 1980; Job and Sartori 1984). In morphological errors, morphemes (stems or affixes) can be deleted (speaker read as speak), inserted Reading morphologically complex words 565 (speak read as speaker), or substituted (speaker read as speaking). This phenomenon has been taken as evidence of word processing being morphemebased: morphemes are the building blocks of word processing, and are hence the most affected by the lexical impairment characterizing deep dyslexia. Although these errors may be at times difficult to distinguish from visual and semantic ones (since morphological relatives are also visually and semantically related, Badecker and Caramazza 1987; Funnell 1987), more recent evidence indicates that morphological manifestations in neuropsychological patients cannot be reduced to a by-product of either semantic or orthographic similarity. For example, in the production of DE, the patient described by Rastle, Tyler, and Marslen-Wilson (2006), morphological errors were more often observed in actual derived words (killer) as opposed to pseudo-suffixed (irony) or noncomplex words with embedded lexical strings (cornea), thus ruling out an explanation in terms of visual errors. Also, a semantic explanation is very unlikely – errors were mostly novel words generated by the substitution of the affix, whereas semantic errors can occur only between words with a proper semantic representation (see also Castles et al. 1996; Marelli et al. 2011). Moreover, Badecker (1997) has described the case of a patient, FM, committing morphological errors also when presented with irregular verbs (began read as begin, as well as passed read as pass). This evidence suggests that morphological representations would be situated (also) at the morpho-syntactic level, representing functional relations between lexical elements irrespective of their morphological parsability. Further research has provided a better understanding of the role of lexical frequency in morphological processing, and how patients’ performance can be modulated by the frequency of the complex word and its constituent morphemes. For example, Luzzatti, Mondini, and Semenza (2001) showed that the patient MB was better at reading singular forms as opposed to plural forms, but the effect disappeared when considering plural-dominant words (i.e., words whose plural form is more frequent than its singular one: stars, eyes, etc.). That is, the patient’s errors affected more less-marked forms (in line with Badecker 1997), but his performance was also modulated by the frequency of the specific inflectional alternatives considered. The authors considered this evidence as supporting dual-route models of word processing (Schreuder and Baayen 1995), in which representations of complex words can also be accessed directly, in a whole-word manner, especially in case of very familiar (i.e., very frequent) forms. MB’s impairment affected the morpheme-based route more seriously, leaving the whole-word procedure relatively spared (see also Biedermann et al. 2012). Further evidence in these regards emerges from the study of neglect dyslexia, a peripheral reading disorder specifically affecting the leftmost part of 566 Marco Marelli, Daniela Traficante and Cristina Burani the presented materials: patients suffering from neglect dyslexia may read wedding as ding. Neglect dyslexia was employed as experimental model to study the processing of stems in suffixed words. Arduino, Burani, and Vallar (2002) found that patients affected by neglect dyslexia are better at reading derived words with high-frequency stems than derived words with low-frequency stems. This was taken as an indication that morpheme plays a role in word reading, with more familiar morphemes being more salient, and hence more capable to pierce through the patients’ disorder and activate the corresponding representations. Moreover, a better performance was observed for existing derived words, as opposed to novel derived words, thus speaking for the parallel role of whole-word representations. In conclusion, neuropsychological studies closely follow the psycholinguistic tradition in supporting the hypothesis that morphemes are represented in the mental lexicon, and play an important role in word processing (e.g., Job and Sartori 1984). In line with results on unimpaired participants, these studies do not exclude a parallel role of a whole-word procedure, mostly determined by the familiarity of the complex form (e.g., Luzzatti, Mondini, and Semenza 2001). The neuropsychological tradition has also highlighted an aspect that was mostly overlooked by the previously discussed studies, namely, the importance of a morphosyntactic level of analysis. Results as the ones reported by Badecker (1997) indicate that morphemes are also linked at more abstracts level of representations, where paradigmatic relations are prominent (see also Marelli et al., 2012). 5 Outside the morphological-representation comfort zone The literature reviewed so far is mostly in line with a morphemes-as-units view on the role of morphology in word processing. This is the case at both the level of empirical evidence, and the level of theoretical assumptions. On the one hand, results of experiments are quite consistent with the idea that morphemes play an important role in word reading. On the other hand, this very premise has been taken as a central assumption of most cognitive models, and was conveniently adopted as a working hypothesis when running empirical research. As a consequence, one may claim that psycholinguistic studies have found evidence for morphological units because they were designed to look for that evidence. Indeed, studies that have attempted to move outside the “comfort zone” of the morpheme-representation assumption, have provided results pointing to more complex scenarios. Reading morphologically complex words 567 5.1 Unexpected results in masked priming We have already discussed results that do not conveniently fit the dominant assumption of morphemes as form representations, devoid of meaning and function, that serve as intermediate step to whole-word lexical and/or semantic representations. The studies by Marelli et al. (2013), Tsang and Chen (2014), and Amenta, Marelli, and Crepaldi (2015) have shown that morpheme access is not purely form-based, rather reflecting complex semantic operations. Moreover, the results by Badecker (1997) indicate that morphological information is (also) represented at a higher level, providing evidence of functional links between morphologically-related elements that are not morpheme-mediated (e.g., bringbrought; see also Kelliher and Henderson 1990; Smolka, Zwitserlood, and Rösler 2007, for German). The limits of the hypothesis of a purely form-based early decomposition are also highlighted by other works employing the masked priming paradigm. Consistently with Badecker’s (1997) results, Crepaldi et al. (2010) found, in a masked priming paradigm, significant priming effects for pairs including an irregular past verb and its base form (e.g., fell-fall), whereas the same priming effect did not emerge for orthographic control pairs (e.g., bell-ball). In other words, the priming effect is elicited by a paradigmatic relation and does not require a parsing procedure of the inflected form (as in the case of regular inflection: played as play plus -ed), and cannot be explained in terms of form similarity. This effect is difficult to account for in the framework of a simple morpho-orthographic decomposition, even if this process is restricted to early processing stages: paradigmatic relations, although not morpheme-mediated, emerge very early as an important factor in word processing. Other studies have shown that early morphological effects are not strictly dependent on the full parsability of the complex word – that is, the prime does not need to be clearly decomposable in constituent morphemes to elicit a masked priming effect. In fact, the results by McCormick, Rastle, and Davis (2008) indicate that morphological priming in masked conditions is robust to orthographic alterations frequently found in complex words (e.g., missing final e: adorable–adore; shared final e: lover–love; duplicated final consonant: dropper–drop). Priming effects emerge irrespective of the semantic relations between prime and target (e.g., significant priming effect for badger-badge; McCormick, Rastle, and Davis 2008) and the lexicality of the prime (e.g., significant priming effect for adorlyadore; McCormick, Rastle, and Davis 2009). In conclusion, studies on masked priming indicate that early morphological processing is more flexible than predicted by a strict morpho-orthographic assumption, both on the syntactic (Crepaldi et al. 2010) and the orthographic 568 Marco Marelli, Daniela Traficante and Cristina Burani side (McCormick, Rastle, and Davis 2008). Relations between different morphological forms play an early role in word access irrespective of their surface properties. It is worth noting that the effects discussed (i.e., priming irrespective of full-parsability in morphemes of the complex prime word, and irrespective of a semantic relation between prime and target) are usually not found in developing readers. The emergence of purely morphological effects devoid of orthographic and semantic similarity between prime and target is the result of literacy acquisition. These issues will be discussed in the next section. 5.2 Paradigmatic effects in morphological processing The importance of paradigmatic aspects is also evident in the effects of morphological entropy in word reading. Literature on entropy effects has explicitly refuted the decompositional interpretation, rather redefining morphology in an information-theoretical framework (Milin et al. 2009). The basic assumption of this approach is that words are organized in paradigms, and morphology is reflected in this very organization. Entropy provides a convenient way to measure the information carried by a given paradigm, quantifying the amount of predictability of the system. The less predictable a paradigm, the more the carried information, the larger the associated entropy. Entropy increases with the number of members in the paradigm, and when the probabilities of the members are more similar (with maximal entropy in the case of uniform distributions). Baayen, Wurm, and Aycock (2007) showed that both inflectional entropy (i.e., entropy computed on the set of possible inflected forms with a given stem) and derivational entropy (i.e., entropy computed on the set of possible derived forms with a given stem) have facilitatory effects in lexical decision: the more informative the associated paradigm, the faster the responses to the target word. Milin, Filipović Đurđević, and Moscoso Del Prado Martín (2009) employed entropy measures for describing paradigmatic effects in the context of complex paradigm classes (Serbian nominal paradigms). They computed the typical probability distribution of the frequency of use among the forms of an inflectional class (e.g., for feminine nouns belonging to the 3rd class, the -e form is associated to a probability of about .40) and the distribution of the inflectional paradigm of a specific stem (e.g., the -e form of the stem knjiga, ‘book’, is associated to a probability of about .40, whereas the -e form of the stem pučina, ‘open sea’, is associated to a probability of about .20). Once these two probability distributions are obtained, the relative entropy of a stem simply captures the degree of divergence between them. And response times were shown to be positively associated with the new measure: the larger the relative entropy (i.e., the more deviant the paradigm of a Reading morphologically complex words 569 given noun), the longer the response times. Baayen et al. (2011) further showed that relative entropy modulates priming effects in masked conditions. Finally, also aphasic patients were shown to be sensitive to morphological entropy (Van Ewijk and Avrutin 2011; Marelli et al. 2012b; note however that patients were not tested on word reading in these studies). Altogether, results on the effects of entropy measures propose a different view of the morphological effects on word reading, namely one that focuses more on the idea of paradigms than that of morpheme units. Approaches positing morphemes to be represented as independent representation entries have clearly a hard time explaining paradigmatic effects. Conceptually related to these paradigm effects, family size also constitutes an important predictor of the processing of written words. Family size is defined as the type count of the morphological relatives of a given word. For example, the family size of run is the number of complex words including run as constituent morpheme (e.g., runner, running, homerun, runaway, etc). Family size is a reliable predictor of response times in word reading: the larger the family size, the faster the processing. Effects of family size are observed for both simple (Schreuder and Baayen 1997) and complex words (Bertram, Baayen, and Schreuder 2000). However, family size is believed not to be strictly morphologically-connoted, rather capturing semantic and syntactic aspects of the considered word. In fact, it provides better predictions in terms of response times if semantically opaque forms are excluded from the count (Bertram, Baayen, and Schreuder 2000; Moscoso del Prado Martín et al. 2004). Conversely, irregular relatives must be included, irrespective of the lack of morpheme-based association with the target word (De Jong, Schreuder, and Baayen 2000). Moreover, and most surprisingly, family size is predictive of monolinguals’ lexical decision latencies across unrelated languages (Moscoso Del Prado Martín et al. 2005), indicating that family size is not simply associated to morphological forms, rather capturing cross-language similarities in semantic space. Although alternative explanations for these effects are possible (e.g., De Jong, Schreuder, and Baayen 2003; Grainger and Jacobs 1996), it seems that also family size is best characterized in an information-theory framework (Moscoso del Prado Martín, Kostić, and Baayen 2004), in line with the paradigmatic effect described above. 5.3 The impact of non-morphological form-meaning associations The interpretation of morphemes as fundamental units between form and meaning is also jeopardized by the impact of non-morphological form-meaning relations to word reading. Let’s consider phonaesthemes. These are phonological/ 570 Marco Marelli, Daniela Traficante and Cristina Burani orthographic chunks that reliably connote aspects of the correspondent word meanings, without being characterized in morphological terms. For example, most English words beginning with sn- have meanings related to nose and mouth (snore, snack, sniff, snarl, sneeze, snort) and most words beginning with gl- have meanings related to light and vision (glimmer, glisten, glitter, gleam, glow, glint), even if none of these words is morphologically complex (i.e., it cannot be parsed in morpheme constituents). Crucially, these form-meaning associations have an effect on word reading (Bergen 2004): prime-target pairs having a phonaestheme in common (e.g., snort-sniff) elicit larger priming effects than semantic (e.g., cordrope) and orthographic (e.g., flour-flag) control pairs. Not only, priming effects for phonaestheme pairs are also larger than priming effects for word pairs with overlapping forms and meanings, if this association is not frequently observed throughout the whole lexicon (e.g., skipper-skiff). That is, phonaesthemes seem to play a role in word processing that goes beyond both morphology and simple form-meaning associations, speaking for the importance of distributional aspects in these effects: orthographic-semantic patterns have to be reliably present in language usage in order to play a role in language processing. In line with this interpretation, Marelli, Amenta, and Crepaldi (2015) have shown that lexical recognition is influenced by the orthography-semantics consistency (OSC) of the considered word. The study was inspired by a curious side effect in morphological priming experiments: stems from transparent sets elicit faster responses than stems from opaque sets, irrespective of the prime preceding them, and even if the item sets are carefully matched for a number of variables. This effect is explained in terms of OSC: stems from transparent sets (e.g., widow) are orthographic strings that only appear in words that are related to the correspondent meaning (e.g., widower, widowed, widowhood), whereas stems from opaque sets (e.g., whisk) are orthographic strings appearing also in words unrelated to the correspondent meaning (e.g., whisker, whiskey, whisky). In this perspective, the former words are, throughout the whole lexicon, reliable orthographic cues for their own meanings, whereas the latter provide unreliable information in these regards. In other words, the former are better symbols than the latter. OSC, computed as the frequency-weighted average semantic-similarity between a word and its orthographic relatives, provides an efficient estimate of this reliability-as-symbol, and represents a new example as to how distributionally based form-meaning associations, over and above morphological considerations, are central to word reading. Given these pieces of evidence, it is sensible to ask ourselves whether morphological effects really need dedicated modules/representations to be explained, or may be rather accounted for by more general-purpose mechanisms that capture statistically strong form-meaning patterns. Reading morphologically complex words 571 In the present section, we have presented a series of phenomena that deviate from those expected by approaches taking morphemes as fundamental processing units in word reading. This is not to say that said approaches are necessarily unable to explain the described effects. Rather, predictions in these regards do not emerge naturally in these frameworks, and corresponding models must be adjusted with ad-hoc solutions (e.g., dedicated modules, additional processing steps). Conversely, architectures in which said effects automatically follow from their basic assumptions will provide more general and epistemologically valid interpretative frameworks for the complex pattern of results we presented. 6 The emergence of morphology in reading acquisition In the present section, we will discuss evidence for the role of morphology in reading acquisition. Also in this domain the dominant view of morphemes as crucial processing units seems to fit uncomfortably with many of the obtained results, which rather depict morphology as a by-product system emerging from language learning. The relation between orthographic patterns and morphemic units seems to be grasped by children before formal teaching is provided to them. Byrne (1996), in research on the hypotheses developed by pre-literate children on the relationship between print and spoken language, noted that pre-literate children tend to grasp the grapheme-to-morpheme transcription (e.g., the plural /s/ in cats) more easily than the grapheme-to-phoneme correspondence (e.g., the phonemic /s/ in bus). These data show that children are able to map the English orthography onto the morphemes learnt during language acquisition. Children often encounter new complex words as far as they proceed through school grades, and the opportunity of recognizing known morphemes embedded in strings of letters can help them to read and understand the meaning of the derived or compound words they come across (Bertram, Laine, and Virkkala 2000). 6.1 Morphological awareness and literacy acquisition Morphological awareness can be defined as the awareness of the morphemic structure of words in oral language, and the ability to reflect on and manipulate 572 Marco Marelli, Daniela Traficante and Cristina Burani that structure (see Carlisle 1995; Deacon, Tong, and Mimeau 2019). It is typically described as a late emerging ability closely connected to orthographic skills (Ehri 2005), and it plays an important role not only in decoding skills, but also in text comprehension. The role of morphemic structure in reading acquisition has been shown by Mann and Singson (2003). These authors found that in early school grades, phonological awareness failed to reach significance level in predicting reading skills beyond third grade, whereas morphological awareness increased its role from fourth grade (when also the number of new complex words in the texts increases). In particular, children decoded derived words with a low frequency stem less correctly than derived words with a high frequency stem, and this data suggests that children refer to stems, at least by the third grade. It is also worth noting that about 20% of the errors made by children consisted in attaching the correct suffix to a wrong stem (e.g., careful was read as *creeful), while errors involving the suffix were made less frequently (about 4% of errors were of the kind methodical read as method; 5% of errors were like imaginable read as imagination). This pattern of errors suggests that the suffix is “the best (and perhaps the only) place to look for hints as to how to decode the word” (Mann and Singson 2003: 19), when a new complex word has to be read. Evidence for the role of morphological structure in word naming comes also from Carlisle and Stones (2005), who found that both lower elementary readers (second and third grade) and upper elementary students (fifth and sixth grade) were more accurate in reading derived words with transparent structure than simple words, matched by length and frequency. However, only younger readers, in front of a derived word, were faster too. The ability to parse words into morphemes has been studied in relation to morphological awareness also by Deacon, Kieffer, and Laroche (2014). In a longitudinal study through third to fourth grades, they found both a direct effect of morphological awareness on reading comprehension and an indirect effect via word reading skills. These data are consistent with other studies on the early years of primary school (Jarmulowics et al. 2008), but are different from the results of studies carried out with children in the sixth grade, where only a direct effect of morphological awareness on reading comprehension has been found (Kieffer and Lesaux 2012), irrespective of decoding skills. This pattern of results suggests a developmental shift, from early literacy to a mastery phase: at the beginning, morphological awareness has an important role in gaining accuracy and fluency in decoding and, as a consequence, in driving reading comprehension. In skilled readers morphological awareness can help in connecting different parts of the text and in making inferences, but does not influence accuracy and fluency of decoding anymore. In children with dyslexia, there is Reading morphologically complex words 573 evidence that morphological and semantic processing might offer compensation mechanisms of phonological deficits (Elbrö and Arnbak 1996; Casalis, Colé, and Sopo 2004; Catts, Adlof, and Ellis Weismer 2006). However, further research is required to better explain the reliability and the direction of the effects involving morphological awareness, comprehension and decoding skills. In fact, there is also evidence for a reciprocal influence of reading comprehension and morphological awareness: third and fourth grade children with good skills in text comprehension might apply the understanding of the meaning to detect morphemic parts of novel words (Deacon, Kieffer, and Laroche 2014). 6.2 Morphological effects in different languages The attitude of children to use morphemic units in processing complex words has been observed in English (e.g., Deacon, Whalen, and Kirby 2011), and in several other languages (see, e.g., Verhoeven and Perfetti 2011). Overall, the opportunity of recognizing morphemic units increases accuracy and/or reduces latency in reading. However, Casalis, Quémart and Duncan (2015), through a direct comparison of English and French, proposed that morphological effects may vary according to the consistency of grapheme-to-phoneme correspondence and morphological richness of the language. These authors found that, in a lexical-decision task, the recognition of a stem within a complex word led French fourth-grade children to faster responses in comparison to simple words. However, the presence of a stem in the complex word tended to inhibit whole-word recognition in English speaking children matched by grade with their French peers, a finding easily interpretable as the result of lexical competition. In English morphologically complex words, the stem corresponds to the base word (e.g., farm-er), which might cause lexical interference rather than a facilitation effect. However, in both languages, children were slower and less accurate in rejecting pseudo-words in which there was a stem and/or a suffix, thus replicating for opaque orthographies the morpheme interference effect obtained on the lexical-decision accuracy of third and fifth grade Italian children (Burani, Marcolini, and Stella 2002), similarly to adults of different languages. Interestingly, when naming was used as an experimental paradigm in transparent orthographies, the morphemic effect resulted in a strong facilitation, not inhibition, of response to pseudo-words containing real morphemes. Italian children of different ages have been repeatedly shown to gain advantage in both latency and accuracy from the presence of a stem when naming pseudo-words (Burani et al. 2008; Burani, Marcolini, and Stella 2002; Traficante et al. 2011), irrespective of its being combined with either a suffix or a non-suffix (e.g., *bagn-ezza, 574 Marco Marelli, Daniela Traficante and Cristina Burani ‘*bath-ness’; *bagn-ezzo, ‘*bath-noss’; Traficante et al. 2011). However, the presence of a suffix (e.g., *bagn-ezza, ‘*bath-ness’; *bogn-ezza, ‘*bith-ness’) had a positive effect on accuracy, as compared to matched pseudo-words that did not include any morpheme (e.g., *bogn-ezzo, ‘*bith-noss’; Traficante et al. 2011). Even more interestingly, the morphological benefit on reading performance was found to be particularly strong in children with dyslexia (Burani et al. 2008; Traficante et al. 2011). The advantage due to stem-suffix composition in word naming has been reported also for low-frequency words: these were read faster and more accurately than simple words matched by frequency (Marcolini et al. 2011; see also Deacon and Whalen 2011). However, it is worth noting that children with dyslexia showed such an advantage also for high frequency words (Marcolini et al. 2011). This evidence suggests that the use of reading units (morphemes), larger than the single grapheme but shorter than the word, can be particularly useful for children who are struggling in processing the word as a whole-unit (De Luca et al. 2010), irrespective of word frequency. Similar results were found by Suárez-Coalla and Cuetos (2013) for Spanish children with dyslexia, who showed shorter latencies in reading both words and pseudo-words composed of morphemes than simple stimuli. All these cross-linguistic data suggest that, during literacy acquisition, children learn to detect and exploit frequent and stable chunks of letters corresponding to morphemes, shared by several words. In this way they optimize fluency and accuracy in decoding new and unfamiliar words, but the gain they get from morphemic structure may vary according to the characteristics of their language, the frequency of morphologically complex stimuli, and their reading ability (for reviews, see Burani 2010; Deacon, Tong, and Mimeau 2019). The orthographies of languages like Italian, Spanish and German present quite consistent grapheme-to-phoneme correspondences, that allow children to reach a good level of accuracy using small grain-size units in decoding new words (as suggested by the grain-size theory by Ziegler and Goswami 2005). However, the opportunity of detecting morphemic units (stems and affixes) that are larger than single graphemes, can allow them avoiding the time-consuming grapheme-to-phoneme reading procedure, gaining in fluency, and this morphemebased reading behavior is particularly useful for children with dyslexia (Burani et al. 2008; Marcolini et al. 2011; Suárez-Coalla and Cuetos 2013; Traficante 2012). Moreover, readers of morphologically rich languages are used to encounter long complex words, whose stem can be combined with several different affixes (both inflectional and derivational), and the higher the number of words that share the same stem is, the higher the probability for that stem to become a useful chunk for decoding (Traficante and Burani 2003; Traficante et al. 2014). Thus, for children exposed to a language with rich morphology the Reading morphologically complex words 575 presence of a stem may improve both fluency and accuracy in word recognition and naming. In English, the opacity of the orthography does not allow the reader to rely on small grain-size units, thus children must memorize the association between whole-word orthographic and phonological representations as soon as possible to reach a good level of accuracy. It is worth noting that most English words are simple and quite short (mono- or disyllabic), the affixes have a key-role in stress assignment and are recognized as useful chunks early in learning to read (Mann and Singson 2003); finally, in the case of semantically transparent derived words, the stem corresponds to the base word. For all these reasons, in processing a long complex word, a good strategy for English-speaking children should be to strip affixes away to isolate the stem. In this way the probability of a correct pronunciation, if the stem is a known word, increases in comparison to the strategy of reading through smaller units. 6.3 What drives morphemic parsing in young readers? A final issue is when, during development, morphological relationships are established in the mental lexicon as autonomous from orthographic and semantic similarity. In the preceding sections, we have presented evidence for morphological priming effects on adult word recognition irrespective of full parsability in morphemes of the complex prime word, and irrespective of a semantic relation between prime and target. As anticipated, such effects are usually not found in developing readers. A first set of studies showed that children do not always gain advantage of the morphemic structure of complex words, but only when they can easily detect the stem embedded in the complex word, without orthographic modifications. In Hebrew, despite its rich morphology, Schiff, Raveh, and Kahta (2008) reported that third and seventh graders, in a masked priming paradigm, showed morphological priming only when the root prime had the same three- or two- consonantal letter root representation as the target (e.g., NGŠ primed mtNGŠ, ‘bump into’; GŠ primed hGŠh, ‘handing over’), but not when prime and target had allomorphic root representations (NGŠ did not prime hGŠh, ‘handing over’). This finding is not consistent with adults’ performance (McCormick, Rastle and Davis 2008; Schiff, Raveh, and Kahta 2008; Velan et al. 2005) to conclude that in the mental lexicon of Hebrew children both types of root representations (three- and two- letters roots) are present, but the allomorphic forms are not connected yet. These results are consistent with those of Quémart and Casalis (2014) on French, who showed that children’s visual word recognition was a function of 576 Marco Marelli, Daniela Traficante and Cristina Burani the phonological and orthographic relationships between derived words and their stems. Quèmart and Casalis (2014) submitted third, fourth and fifth grade typically developing children and adults to a lexical decision task in which a base word was preceded by a derived word prime in different conditions of orthographic-phonological prime-target relationship. At the shortest (60ms) prime duration the authors reported significant priming effects in children only when prime and target shared a morphological relationship without any form shift in the stem (as in nuageux-NUAGE, ‘cloudy-cloud’). With longer (250 ms) prime exposure, significant priming effects were found in all conditions of morphological relatedness, irrespective of orthographic-phonological alterations (i.e., priming effects were found also when the morphological relationship involved a phonological although not orthographic modification of the stem, as in bergerie-BERGER, ‘sheepfold-shepherd’, and when the relationship between stem and derived word involved an alteration that was both phonological and orthographic, as in soigneux-SOIN, ‘careful-care’). In contrast, adults showed significant priming for all the morphological conditions at all prime exposure times. Thus allomorphic variations may prevent lexical activation of stems at the fastest word exposures in children but not in adults. According to the authors, children need more time to activate a stem in the case of a phonological or an orthographic shift because activation of the allomorphic stems is not automatized yet. The results obtained using the priming technique find a correspondence with other results drawn from unprimed lexical decision and naming experiments. Carlisle, Stone and Katz (2001) studied two groups of English-speaking young readers, with and without reading difficulties (with ages ranging from 10.75 years to 15.75 years) and a group of adults who performed both tasks. Participants were presented with stable words (in which no phonological alteration occurred between base and derived words, as in cultural), shift words (in which a phonological alteration occurred, as in majority) and foils. Results showed an effect of phonological transparency in all groups, confirming that it takes longer to respond to complex words with a phonological alteration in the base than to complex words without a phonological alteration. Finally, Lázaro, García, and Burani (2015) submitted Spanish third and fourth graders, with and without reading deficits, to a lexical decision task. They found that, irrespective of reading ability, all children gained more advantage from stem frequency in words in which the stem had no orthographic alteration with respect to the base word (e.g., colorista from color, ‘colorful’ from ‘color’) than in words with modified stem (e.g., dentista from diente, ‘dentist’ from ‘tooth’). All these studies, carried out in several languages with different orthographies show that, contrary to what was found by McCormick, Rastle, and Davis (2008) in adults, Reading morphologically complex words 577 orthographic-phonological alterations of the stem make it more difficult for children to benefit from access to the base word in performing visual tasks on derived words. 6.4 Children’s sensitivity to form-meaning association in accessing morphemic structure As mentioned in the preceding sections, the masked priming paradigm in lexical decision has been adopted in adults to assess whether the detection of morphemic units is driven by semantic information (Longtin, Segui, and Halle 2003; Rastle, Davis, and New 2004). Recently, this paradigm has been used with children and a complex pattern of results emerged, with age of children and language features influencing the size of priming effects (Beyersmann, Castles, and Coltheart 2012; Quémart and Casalis 2015; Quémart, Casalis, and Colé 2011; Shiff, Raveh, and Fighel 2012). Beyersmann, Castles, and Coltheart (2012) presented English-speaking third and fifth graders with morphologically (e.g., golden-GOLD), pseudo-morphologically (e.g., mother-MOTH) related pairs and control condition. With 50ms prime exposure, a priming effect was found only with truly suffixed primes. These data have been interpreted as evidence that the ability of children in using the morphemic structure of complex words is based on activation of the meaning shared by different words with the same stem. A different pattern of results came from French. Quémart Casalis, and Colé (2011) presented French third, fifth and seventh graders with a masked primed paradigm, using three different degrees of stimulus-onset-asynchrony (SOA) (60ms, 250ms, 800ms) between prime and target, to assess the relation between level of processing of the letter string and type of information used in word recognition. The true morphological relationship (e.g., tablette-TABLE, ‘little table-table’) was associated to reliable priming at any SOA, while the pseudo-derivation condition (e.g., baguette-BAGUE, ‘French stick-ring’) produced a priming similar to the morphological one at 60ms, but lower than that condition at 250ms prime exposure. With the longest SOA (800ms) the pseudoderivation priming effect disappeared, suggesting that with long prime duration morphemic parsing is based only on the activation of semantic properties of morphemes. In order to interpret this inconsistent pattern of results with children reading different languages, the authors proposed that the richness of French morphology might lead young readers to be more competent in detecting morpho-orthographic units than their English-speaking peers. As for the inconsistency between English-speaking children’s and adults’ data, a developmental trend has been hypothesized, suggesting that in early 578 Marco Marelli, Daniela Traficante and Cristina Burani literacy levels the form-meaning relationship would be the main dimension that drives morphemic parsing, while in skilled reading the morpho-orthographic dimension would be prevalent. In other words, purely morpho-orthographic decomposition would be a late-occurring milestone in reading acquisition (Beyersmann, Castles, and Coltheart 2012). To assess the relationship between reading proficiency and morpho-orthographic decomposition ability, Beyersmann et al. (2015a) presented French primary school children through second to fifth grade with a masked primed lexical decision task, and took into account not only performance to the task, but also children’s literacy skills. It was expected that effects of morpho-orthographic priming should increase with increasing reading proficiency, as suggested by the difference between English speaking children and adults. To avoid any idiosyncratic relationship between stem and suffix in the pseudo-derived condition (see criticisms made by Baayen et al. 2011 to the results by Rastle, Davis, and New 2004), the experimental primes other than in true morphological relation (e.g., tristesse-TRISTE, ‘sadness-sad’), were all pseudowords, with different relations with targets. There were a suffixed non-word prime condition (e.g., *tristerie-TRISTE, ‘*sadery-sad’), a non-suffixed non-word condition (e.g., *tristald-TRISTE, ‘*sadald-sad’), and an unrelated prime control condition (e.g., direction-TRISTE, ‘direction-sad’). Results showed a reliable priming effect for suffixed words, larger than the priming effects of other suffixed and nonsuffixed non-word conditions, confirming that the segmentation of the string of letter in stem + suffix units is driven by semantic interpretability of the combination. The two non-word conditions (suffixed and non-suffixed) produced priming of a similar size, suggesting that French children (irrespective of grade) are able to detect and use the stem unit. The advantage gained from the stem embedded in a non-word prime was positively correlated with literacy skill: the higher the skill, the larger the obtained priming effect. In other words, skilled readers are more likely to exploit a known stem in a string of letters, than low-proficiency readers. These results are consistent with those of a previous study (Quémart and Casalis 2015), in which French children with dyslexia showed a reliable priming effect for morphological condition (e.g., tablette-TABLE, ‘little table-table’), but not for pseudo-derived condition (e.g., baguette-BAGUE, ‘French stick-ring’) and orthographic control (e.g., abricot-ABRI, ‘apricot-shelter’). In Hebrew, Schiff, Raveh and Fighel (2012), studying the role of semantic consistency in parsing derived words, found a reliable morphological priming effect when prime and target were morphologically and semantically related (e.g., prime: hNHGh, ‘leadership’; target: mNHiG, ‘leader’), irrespective of reading skill. However, differences between 4th graders and 7th graders were observed for semantically inconsistent prime-target couples. For younger readers, semantically inconsistent condition (e.g., prime: Reading morphologically complex words 579 NHiGh, ‘driving’; target: mNHiG, ‘leader’) did not produce any priming effect, whereas for older children the priming effect of the semantically inconsistent condition approached significance, as found in adult readers (Bentin and Feldman 1990; Frost, Forster, and Deutsch 1997). The authors interpreted these data as evidence that the higher the reading skills are, the more abstract and independent from semantic properties the morphological representations within the mental lexicon. To summarize, data from developmental studies on visual word recognition indicate that, in children, the meaning of the morphemic units is subliminally activated, irrespective of reading skills. Thus, semantics is likely to be the early dimension on which the ability of using morphemic structure develops, whereas the sensitivity to purely morpho-orthographic relations seems to be a late acquisition, as data from young skilled readers (Beyersmann et al. 2015a) and adults (Rastle and Davis 2008) show. So, it is possible to draw a developmental trajectory in the ability of detecting morphemic units in letter strings that originates from the semantic dimension of morphemes and leads to morpho-orthographic representations, supporting the view of morphology as a byproduct system that develops according to general language learning mechanisms. Overall, the reviewed studies suggest that morphological representations may become more and more abstract with increasing reading ability and word exposure. In the early stages of reading development, readers heavily rely on the orthographic and semantic consistency of a stem to successfully identify the derived word, and in later stages, with increased exposure to written language and increased reading ability, a certain level of generalization develops in the mental lexicon. 7 Data-driven computational models: A solution to the conundrum? In the present chapter, we have first presented the most popular idea underlying the study of morphological effects in word reading, namely that morphemes are represented in the cognitive system, and these representational units are activated when processing a morphologically complex word. This assumption has been profitably applied to psycholinguistic studies for nearly four decades. However, recent results seem to draw a more complex scenario in which, on the one side, form-meaning associations that are not morpheme-mediated have a role to play (e.g., Bergen 2004; Marelli, Amenta, and Crepaldi 2015) and, on the other side, paradigmatic effects modulate the influence of morphology in 580 Marco Marelli, Daniela Traficante and Cristina Burani word recognition (Milin, Filipović Đurđević and Moscoso del Prado Martín 2009; Moscoso del Prado Martín et al. 2004). Although these pieces of evidence do not necessarily invalidate models assuming morphemes as explicit representations, they do not seem to be directly predicted by such systems either. Moreover, the developmental literature converges on characterizing morphology as an emerging phenomenon, with morphological awareness becoming progressively more important in language development (Deacon, Kieffer, and Laroche 2014; Jarmulowicz et al. 2008). Indeed, morphological complexity seems to help word processing in early readers, thus suggesting that morphemes may act as distributional cues that can be efficiently exploited to facilitate reading (Burani 2010; Burani et al. 2008; Carlisle and Stone 2005; Deacon, Tong, and Mimeau 2019; Mann and Singson 2003; Traficante et al. 2011). To capture these phenomena in a modelling perspective it is crucial to consider learning-oriented processes, often lacking in traditional models of morphological processing. In conclusion, it may be the case that different views on morphology would provide a more efficient way to account for such a diverse pattern of effects. 7.1 Morphology as consistent associations within the language system In these terms, a promising approach could characterize morphology as an epiphenomenon of more general-purpose learning mechanisms that exploit consistencies in the language system. This notion is certainly not new. In embryo, it can be found in the full-listing proposals (e.g., Butterworth 1983), since they typically see morphology as a by-product of form and meaning similarity between independent representations. More formally, it has been an important theoretical assumption of models from the connectionist tradition (see Seidenberg and McClelland 1989). In these systems there are no explicit, symbolic representations for morphemes and/or words. Rather, connectionist architectures are populated by simple elements (graphemes, phonemes, semantic features), organized in different layers. Typically, a set of non-symbolic nodes (the hidden layer) is implemented between these, and morphology (as well as lexicality) naturally unfolds by means of consistent patterns of activation within the links connecting the different layers (Plaut and Gonnerman 2000; Rueckl and Raveh 1999). In other words, in these models the morphological status of -er is not captured through an explicit representation unit for the suffix, rather emerging as strong connection links between the graphemic units (e and r at word endings) and the corresponding semantics (abstract nodes indicating instrumental or agentive Reading morphologically complex words 581 traits). Crucially, connection weights are estimated through an iterative learning process that builds on examples of “correct” form-meaning associations. In other words, during the learning phase weights are continuously updated with the purpose to progressively minimize the error in the system output (through various learning rules, e.g., the widely used delta rule). By means of hidden units, the learning procedure ends up extracting a high-order structure from patterns of low-level features (i.e., graphemic and semantic nodes), which guarantees high generalization power and the capability of exhibiting morphological effects that cannot be reduced to simple orthographic and/or semantic similarity (Plaut and Gonnerman 2000). From a theoretical point of view, the connectionist models offer an ideal interpretative framework to address the sensitive aspects we discussed in previous sections. On the one hand, the connectionist approach defines morphology as a specific expression of a more general cognitive ability to capture formmeaning patterns, hence providing a natural explanation for the distributional, graded effects reviewed above. On the other hand, it considers morphology as the result of a learning procedure, in line with the evidence from reading acquisition. In this regard the connectionist approach has provided interesting simulations of how morphological effects may depend on the degree of morphological connectivity of a language, with morphologically rich languages as Italian and Hebrew more likely to show morphological effects in the absence of semantic relations (Plaut and Gonnermann 2000; see also Bentin and Feldman 1990). The approach has been very successful also on the empirical side. Connectionist simulations work well in reproducing priming patterns and their graded effects according to the degree of morphological connectivity of the language (e.g., Plaut and Gonnerman 2000), the acquisition of morphological chunks in learning (e.g., Moscoso del Prado Martin, Schreuder, and Baayen 2004), and morphological errors in neuropsychological patients (e.g., Joanisse and Seidenberg 1999; Plaut and Shallice 1993). 7.2 Evidence from large-scale models Still, most traditional connectionist networks are based on toy models for specific tasks, with input and training data de-facto hand-coded, and are thus limited in their explanatory power. More recent computational proposals have exploited large collections of texts (corpora) as sources of training data, hence basing their models on examples of natural language usage. For example, the Naïve Discriminative Reader (NDR), a model proposed by Baayen et al. (2011) to account for morphological effects in word processing, was trained on the 582 Marco Marelli, Daniela Traficante and Cristina Burani British National Corpus (http://www.natcorp.ox.ac.uk/). NDR architecture is similar to those proposed in the connectionist paradigm: an input layer (populated by orthographic unigrams and bigrams) is directly connected to a semantic layer (populated by symbolic word meanings); connections between layers are learnt by means of the Rescorla-Wagner equations (Rescorla and Wagner 1972), which are strictly related to the connectionist delta rule. Indeed, NDR is equivalent to a connectionist network without hidden layers, namely a perceptron (Rosenblatt 1958). Despite its simplicity, NDR provides a unique account for a wide range of morphological effects, ranging from family size (Schreuder and Baayen 1997), to inflectional entropy (Milin, Filipović Đurđević, and Moscoso del Prado Martín 2009), to priming (Rastle, Davis, and New 2004), again suggesting that morphology may simply reflect a cognitive sensitivity to systematic relations between forms and meanings. Indeed, distributional phenomena emerge naturally from the NDR perspective, which also predicted new surprising effects, crucially depending on word usage in context; namely, the effect of relative prepositional entropy indicates that the processing of isolated word is influenced by the way those words are paired with prepositions within the whole corpus (Baayen et al. 2011). Similarly, effects related to phonaesthemes (Bergen 2004) and, potentially, OSC (Marelli, Amenta, and Crepaldi 2015) follow naturally from the NDR premises. Over and above the larger scale (in comparison to traditional connectionist proposals), NDR also offered a novel perspective from the theoretical point of view. On the one hand, it moved the interpretative focus from association to discrimination (Ramscar et al. 2010): in the NDR framework, “learn” is equivalent to “learn to distinguish”. Morphemes are an epiphenomenon of simple discriminative cues, namely chunks of graphemic symbols that help distinguish between different meanings. On the other hand, NDR takes a psychologically plausible, biologically grounded stance on learning. In fact, Rescorla-Wagner equations (Rescorla and Wagner 1972) are inspired by classical conditioning (Pavlov 1927; Rescorla 1988), defining an approach in which the connection that is formed between graphemic symbols and semantic units is comparable to that emerging between conditioned and unconditioned stimuli. In conclusion, NDR provides a morphological model that does not rely on explicit morpheme representations, focuses on form-meaning statistical patterns that emerge in word distributions, and is centered on a psychologically plausible learning mechanism. Although it is not uncontroversial (for example in relation to masked-priming data; Beyersmann et al. 2015b), its explanatory power is certainly remarkable, and constitutes a profitable interpretative framework for the understanding of morphological effects in reading. Reading morphologically complex words 583 If NDR models morphology by assuming a relatively underspecified semantic system, other computational approaches have specifically looked at how morphology can inform word meanings. The system proposed by Marelli and Baroni (2015) models morphological derivation in distributional semantics (Turney and Pantel 2010). This theory is based on the assumption that similar words appear in similar contexts; as a consequence, a word meaning can be represented through the contexts the given word appears in (Landauer and Dumais 1997) or, in more formal terms, through a vector encoding the cooccurrence counts between the target word and other words in the lexicon. In Marelli and Baroni’s (2015) proposal, if word meanings can be approximated by vectors, affix meanings can be modelled as functions (i.e., matrices) mapping the meaning of the stem into the meaning of the derived form (FRACSS: Functional Representation for Affixes in Compositional Semantic Space). FRACSSs are learned through examples of stems and corresponding derived forms (similarly to NDR, the model is trained on a large natural-language corpus), and define a flexible system that is able to account for a range of semantic effects associated to morphology: semantic transparency effects in lexical decision and priming paradigms, explicit intuitions on complex word meanings, the immediate understanding of the meaning of novel words, the possibility of transparent readings of semantically opaque words. Although a high-level description of the FRACSS system fits a representational account of morphology (each stem and affix has its own distributed representation), the way morphemes are computationally characterized is quite different from the symbolic units so popular in the psycholinguistic tradition. On the one hand, stems are modelled as distributed patterns of activation across a set of sub-symbolic nodes. On the other hand, affixes are closer to processes than representation units: following the functional assumption, FRACSS can indeed be seen as a contextual update (cued by the affix orthographic signs) of a core meaning (expressed through the stem). To summarize, the architectures described in the present section represent a powerful tool for the investigation of morphological effects in reading. Indeed, they provide a convincing scenario to naturally frame the graded, paradigmatic, and learning effects we presented in the previous sections (which uncomfortably fit a strict representational view on morphemes). Notwithstanding the many differences, all these systems have a focus on data-driven methods that can extract statistically reliable patterns from examples of real language usage. These patterns emerged as crucial not only at the interplay between form and meaning (Baayen et al. 2011), but also within the orthographic (Moscoso del Prado Martín, Schreuder, and Baayen 2004) or semantic systems themselves (Marelli and Baroni, 2015). The success of these computational approaches 584 Marco Marelli, Daniela Traficante and Cristina Burani crucially highlights the importance of two (closely related) components when it comes to investigate morphology: on the one hand, the distributional characterization of the elements under investigation on the basis of real language usage; on the other hand, the definition of a learning system that is able to extract information from these very distributions. Future research should accord the proper weight to these aspects, if we aim to achieve a unified explanation of the many and diverse effects emerging in complex-word reading. References Amenta, Simona & Davide Crepaldi. 2012. Morphological processing as we know it: An analytical review of morphological effects in visual word identification. Frontiers in Psychology 3 (232). https://doi.org/10.3389/fpsyg.2012.00232. Amenta, Simona, Marco Marelli & Davide Crepaldi. 2015. The fruitless effort of growing a fruitless tree: Early morpho-orthographic and morpho-semantic effects in sentence reading. Journal of Experimental Psychology: Learning, Memory & Cognition 41 (5). 1587–1596. Arduino, Lisa S., Cristina Burani & Giuseppe Vallar. 2002. Lexical effects in left neglect dyslexia: A study in Italian patients. Cognitive Neuropsychology 19 (5). 421–444. Aronoff, Mark. 1994. Morphology by itself: Stems and inflectional classes. Cambridge, MA: MIT press. Baayen, R. Harald, Ton T. Dijkstra & Robert Schreuder. 1997. Singulars and plurals in Dutch: Evidence for a parallel dual-route model. Journal of Memory and Language 37 (1). 94–117. Baayen, R. Harald, Petar Milin, Dusica Filipović Đurđević, Peter Hendrix & Marco Marelli. 2011. An amorphous model for morphological processing in visual comprehension based on naive discriminative learning. Psychological review 118 (3). 438–481. Baayen, R. Harald, Lee H. Wurm & Joanna Aycock. 2007. Lexical dynamics for low-frequency complex words: A regression study across tasks and modalities. The Mental Lexicon 2 (3). 419–463. Badecker, William. 1997. Levels of morphological deficit: Indications from inflectional regularity. Brain and Language 60 (3). 360–380. Badecker, William & Alfonso Caramazza. 1987. The analysis of morphological errors in a case of acquired dyslexia. Brain and Language 32 (2). 278–305. Bentin, Shlomo & Laurie B. Feldman. 1990. The contribution of morphological and semantic relatedness to repetition priming at short and long lags: Evidence from Hebrew. The Quarterly Journal of Experimental Psychology 42 (4). 693–711. Bergen, Benjamin K. 2004. The psychological reality of phonaesthemes. Language 80 (2). 290–311. Bertram, Raymond, R. Harald Baayen & Robert Schreuder. 2000. Effects of family size for complex words. Journal of Memory and Language 42 (3). 390–405. Bertram, Raymond, Jukka Hyönä & Matti Laine. 2000. The role of context in morphological processing: Evidence from Finnish. Language and Cognitive Processes 15 (4–5). 367–388. Reading morphologically complex words 585 Bertram, Raymond, Matti Laine, R. Harald Baayen, Robert Schreuder & Jukka Hyönä. 2000. Affixal homonym triggers full-form storage, even with inflected words, even in a morphologically rich language. Cognition 74 (2). B13–B25. Bertram, Raymond, Matti Laine & Minna Maria Virkkala. 2000. The role of derivational morphology in vocabulary acquisition: Get by with little help from my morpheme friends. Scandinavian Journal of Psychology 41 (4). 287–296. Beyersmann, Elisabeth, Anne Castles & Max Coltheart. 2012. Morphological processing during visual word recognition in developing readers: Evidence from masked priming. The Quarterly Journal of Experimental Psychology 65 (7). 1306–1326. Beyersmann, Elisabeth, Max Coltheart & Anne Castles. 2012. Parallel processing of whole words and morphemes in visual word recognition. The Quarterly Journal of Experimental Psychology 65 (9). 1798–1819. Beyersmann, Elisabeth, Jonathan Grainger, Séverine Casalis & Johannes C. Ziegler. 2015a. Effects of reading proficiency on embedded stem priming in primary school children. Journal of Experimental Child Psychology 139 (11). 115–126. Beyersmann, Elisabeth, Johannes C. Ziegler, Anne Castles, Max Coltheart, Yvette Kezilas & Jonathan Grainger. 2015b. Morpho-orthographic segmentation without semantics. Psychonomic Bulletin & Review 23 (2). 533–539. Biedermann, Britta, Antje Lorenz, Elisabeth Beyersmann & Lyndsay Nickels. 2012. The influence of plural dominance in aphasic word production. Aphasiology 26 (8). 985–1004. Blevins, James P. 2016. The minimal sign. In Gregory T. Stump & Andrew Hippisley (eds.), The Cambridge Handbook of Morphology, 102–137. Cambridge: Cambridge University Press. Bloomfield, Leonard. 1933. Language. New York: Holt. Burani, Cristina. 2010. Word morphology enhances reading fluency in children with developmental dyslexia. Lingue e Linguaggio IX (2). 177–198. Burani, Cristina & Alfonso Caramazza.1987. Representation and processing of derived words. Language and Cognitive Processes 2 (3–4). 217–227. Burani, Cristina, Francesca M. Dovetto, Alberto Spuntarelli & Anna M. Thornton. 1999. Morpho-lexical access and naming: The semantic interpretability of new root-suffix combinations. Brain and Language 68 (1–2), 333–339. Burani, Cristina, Francesca M. Dovetto, Anna M. Thornton & Alessandro Laudanna. 1997. Accessing and naming suffixed pseudo-words. In Geert E. Booij & Jap van Marle (eds.), Yearbook of Morphology 1996, 55–72. Dordrecht: Kluwer Academic Publishers. Burani, Cristina, Stefania Marcolini, Maria De Luca & Pierluigi Zoccolotti. 2008. Morphemebased reading aloud: Evidence from dyslexic and skilled Italian readers. Cognition 108 (1). 243–262. Burani, Cristina, Stefania Marcolini & Giacomo Stella. 2002. How early does morpholexical reading develop in readers of a shallow orthography? Brain and language 81 (1–3). 568–586. Burani, Cristina, Dario Salmaso & Alfonso Caramazza. 1984. Morphological structure and lexical access. Visible Language 18 (4). 342–352. Burani, Cristina & Anna M. Thornton. 2003. The interplay of root, suffix and whole-word frequency in processing derived words. In R. Harald Baayen & Robert Schreuder (eds.), Morphological Structure in Language Processing, 157–208. Berlin: Mouton de Gruyter. Butterworth, Brian. 1983. Lexical representation. In Brian Butterworth (ed.), Language production (Vol. 2), 257–294. London: Academic Press. 586 Marco Marelli, Daniela Traficante and Cristina Burani Byrne, Brian. 1996. The learnability of the alphabetic principle: Children’s initial hypotheses about how print represents spoken language. Applied Psycholinguistics 17(4), 401–426. Caramazza, Alfonso, Alessandro Laudanna & Cristina Romani. 1988. Lexical access and inflectional morphology. Cognition 28(3). 297–332. Carlisle, Joanne F. 1995. Morphological awareness and early reading achievement. In Laurie B. Feldman (ed.), Morphological aspects of language processing, 189–209. Hillsdale, N.J: LEA. Carlisle, Joanne F. & C. Addison Stone. 2005. Exploring the role of morphemes in word reading. Reading Research Quarterly 40 (4). 428–449. Carlisle, Joanne F., C. Addison Stone & Lauren A. Katz. 2001. The effect of morphological transparency on reading complex words. Annals of Dyslexia 51 (1), 249–274. Casalis, Séverine., Pascale Colé & Delphine Sopo. 2004. Morphological awareness in developmental dyslexia. Annals of Dyslexia 54 (1). 114–138. Casalis, Séverine, Pauline Quémart & Lynne G. Duncan. 2015. How language affects children’s use of derivational morphology in visual word and pseudoword processing: Evidence from a cross-language study. Frontiers in Psychology 6 (452). https://doi.org/10.3389/fpsyg.2015.00452. Castles, Anne, Max Coltheart, Greg Savage, Andree Bates & Lauren Reid. 1996. Morphological processing and visual word recognition: Evidence from acquired dyslexia. Cognitive Neuropsychology 13 (7). 1041–1057. Catts, Hugh W., Suzanne M. Adlof & Susan Ellis Weismer. 2006. Language deficits in poor comprehenders: A case for the simple view of reading. Journal of Speech, Language & Hearing Research 49 (2). 278–293. Christianson, Kiel, Rebecca L. Johnson & Keith Rayner. 2005. Letter transpositions within and across morphemes. Journal of Experimental Psychology: Learning, Memory & Cognition 31 (6). 1327–1339. Coltheart, Max, Karalyn Patterson & John C. Marshall. 1980. Deep dyslexia. London: Routhledge & Kegan Paul. Crepaldi, Davide, Kathleen Rastle, Max Coltheart & Lyndey Nickels. 2010. ‘Fell’ primes ‘fall’, but does ‘bell’ prime ‘ball’? Masked priming with irregularly-inflected primes. Journal of Memory and Language 63 (1). 83–99. Crepaldi, Davide, Kathleen Rastle & Colin J. Davis. 2010. Morphemes in their place: Evidence for position-specific identification of suffixes. Memory & cognition 38 (3). 312–321. Crepaldi, Davide, Kathleen Rastle, Colin J. Davis & Stephen J. Lupker. 2013. Seeing stems everywhere: Position-independent identification of stem morphemes. Journal of Experimental Psychology: Human Perception and Performance 39 (2). 510–525. Deacon, S. Hélène, Michael J. Kieffer & Annie Laroche. 2014. The relation between morphological awareness and reading comprehension: Evidence from mediation and longitudinal models. Scientific Studies of Reading 18 (6). 432–451. Deacon, S. Hélène, Xiuli Tong & Catherine Mimeau. 2019. Morphological and semantic processing in developmental dyslexia across languages: A theoretical and empirical review. In Ludo Verhoeven, Charles Perfetti & Kenneth Pugh (eds.), Developmental dyslexia across languages and writing systems, 327–349. Cambridge: Cambridge University Press. Deacon, S. Hélène, Rachel Whalen, & John R. Kirby. 2011. Do children see the danger in dangerous? Grade 4, 6, and 8 children’s reading of morphologically complex words. Applied Psycholinguistics 32 (3). 467–481. Reading morphologically complex words 587 De Jong, Nivja H., Robert Schreuder & R. Harald Baayen. 2000. The morphological family size effect and morphology. Language and Cognitive Processes 15 (4–5). 329–365. De Jong, Nivja H., Robert Schreuder & R. Harald Baayen. 2003. Morphological resonance in the mental lexicon. In R. Harald Baayen & Robert Schreuder (eds.), Morphological structure in language processing, 65–88. Berlin: Mouton de Gruyter. De Luca, Maria, Cristina Burani, Despina Paizi, Donatella Spinelli & Pierluigi Zoccolotti. 2010. Letter and letter-string processing in developmental dyslexia. Cortex 46 (10). 1272–1283. Diependaele, Kevin, Dominiek Sandra & Jonathan Grainger. 2009. Semantic transparency and masked morphological priming: The case of prefixed words. Memory & cognition 37 (6). 895–908. Drews, Etta & Pienie Zwitserlood. 1995. Morphological and orthographic similarity in visual word recognition. Journal of Experimental Psychology: Human Perception and Performance 21 (5). 1098–1116. Duñabeitia, Jon Andoni, Manuel Perea & Manuel Carreiras. 2007. Do transposed-letter similarity effects occur at a morpheme level? Evidence for morpho-orthographic decomposition. Cognition 105 (3). 691–703. Ehri, Linnea C. 2005. Learning to read words: Theory, findings and issues. Scientific Studies of Reading, 9 (2), 167–188. Elbro, Carsten & Elisabeth Arnbak. 1996. The role of morpheme recognition and morphological awareness in dyslexia. Annals of Dyslexia 46 (1). 209–240. Feldman, Laurie B. 2000. Are morphological effects distinguishable from the effects of shared meaning and shared form? Journal of Experimental Psychology: Learning, Memory & Cognition 26 (6). 1431–1444. Feldman, Laurie B., Aleksandar Kostić, Vasilije Gvozdenović, Patrick A. O’Connor & Fermín Moscoso del Prado Martín. 2012. Semantic similarity influences early morphological priming in serbian: A challenge to form-then-meaning accounts of word recognition. Psychonomic bulletin & review 19 (4). 668–676. Feldman, Laurie B., Patrick A. O’Connor & Fermín Moscoso del Prado Martín. 2009. Early morphological processing is morphosemantic and not simply morpho-orthographic: A violation of form-then-meaning accounts of word recognition. Psychonomic bulletin & review 16 (4). 684–691. Feldman, Laurie B., & Emily G. Soltano. 1999. Morphological priming: The role of prime duration, semantic transparency and affix position. Brain and Language 68 (1–2). 33–39. Fodor, Jerry A. (1983). The modularity of mind: An essay on faculty psychology. Cambridge, MA: The MIT press. Ford, Michael A., Matthew H. Davis & William Marslen-Wilson. 2010. Derivational morphology and base morpheme frequency. Journal of Memory and Language 63 (1), 117–130. Forster, Kenneth I. & Chris Davis. 1984. Repetition priming and frequency attenuation in lexical access. Journal of Experimental Psychology: Learning, Memory & Cognition 10 (4). 680–698. Forster, Kenneth I., Chris Davis, Colin Schoknecht & Ronald Carter. 1987. Masked priming with graphemically related forms: Repetition or partial activation? The Quarterly Journal of Experimental Psychology 39 (2). 211–251. Frost, Ram, Kenneth I. Forster & Avital Deutsch. 1997. What can we learn from the morphology of Hebrew? A masked-priming investigation of morphological representation. Journal of Experimental Psychology: Learning, Memory & Cognition 23 (4). 829–856. 588 Marco Marelli, Daniela Traficante and Cristina Burani Funnell, Elaine. 1987. Morphological errors in acquired dyslexia: A case of mistaken identity. The Quarterly Journal of Experimental Psychology A: Human Experimental Psychology 39 (3). 497–539. Giraudo, Hélène, & Jonathan Grainger. 2000. Effects of prime word frequency and cumulative root frequency in masked morphological priming. Language and Cognitive Processes 15 (4–5). 421–444. Grainger, Jonathan & Arthur M. Jacobs. 1996. Orthographic processing in visual word recognition: a multiple read-out model. Psychological Review 103 (3), 518–565. Grainger, Jonathan & Johannes C. Ziegler. 2011. A dual-route approach to orthographic processing. Frontiers in Psychology 2 (54). https://doi.org/10.3389/fpsyg.2011.00054 Hay, Jennifer. 2001. Lexical frequency in morphology: Is everything relative? Linguistics 39 (6). 1041–1070. Howes, Davis H. & Richard L. Solomon. 1951. Visual duration threshold as a function of wordprobability. Journal of Experimental Psychology 41 (6). 401–410. Jarmulowicz, Linda, Sarah E. Hay, Valentina L. Taran & Corinna A. Ethington. 2008. Fitting English derivational morphophonology into a developmental model of reading. Reading and Writing: An Interdisciplinary Journal 21 (3). 275–297. Järvikivi, Juhani & Pyykkönen Pirita. 2011. Sub-and supralexical information in early phases of lexical access. Frontiers in Psychology 2 (282). https://doi.org/10.3389/fpsyg.2011. 00282. Joanisse, Marc F. & Mark S. Seidenberg. 1999. Impairments in verb morphology after brain injury: A connectionist model. Proceedings of the National Academy of Sciences 96 (13). 7592–7597. Job, Remo & Giuseppe Sartori. 1984. Morphological decomposition: Evidence from crossed phonological dyslexia. The Quarterly Journal of Experimental Psychology 36 (3). 435–458. Kazanina, Nina. 2011. Decomposition of prefixed words in Russian. Journal of Experimental Psychology: Learning, Memory & Cognition 37 (6). 1371–1390. Kazanina, Nina, Galina Dukova-Zheleva, Dana Geber, Viktor Kharlamov & Keren Tonciulescu. 2008. Decomposition into multiple morphemes during lexical access: A masked priming study of Russian nouns. Language and Cognitive Processes 23 (6). 800–823. Kelliher, Susan & Leslie Henderson. 1990. Morphologically based frequency effects in the recognition of irregularly inflected verbs. British Journal of Psychology 81 (4). 527–539. Kieffer, Michael J. & Nonie K. Lesaux. 2012. Direct and indirect roles of morphological awareness in the english reading comprehension of native English, Spanish, Filipino & Vietnamese speakers. Language Learning 62 (4). 1170–1204, Kuperman, Victor, Raymond Bertram & R. Harald Baayen. 2008. Morphological dynamics in compound processing. Language and Cognitive Processes 23 (7–8). 1089–1132. Kuperman, Victor, Raymond Bertram & R. Harald Baayen. 2010. Processing trade-offs in the reading of dutch derived words. Journal of Memory and Language 62 (2). 83–97. Kuperman, Victor, Robert Schreuder, Raymond Bertram & R. Harald Baayen. 2009. Reading polymorphemic Dutch compounds: toward a multiple route model of lexical processing. Journal of Experimental Psychology: Human Perception and Performance 35 (3). 876–895. Landauer, Thomas K. & Susan T. Dumais. 1997. A solution to Plato’s problem: The latent semantic analysis theory of acquisition, induction & representation of knowledge. Psychological Review 104 (2). 211–240. Reading morphologically complex words 589 Lavric, Aureliu, Amanda Clapp & Kathleen Rastle. 2007. ERP evidence of morphological analysis from orthography: A masked priming study. Journal of cognitive neuroscience 19 (5). 866–877. Lázaro, Miguel, Laura García & Cristina Burani. 2015. How orthographic transparency affects morphological processing in young readers with and without reading disability. Scandinavian Journal of Psychology 56 (5). 498–507. Longtin, Catherine-Marie, Juan Segui & Pierre A. Halle. 2003. Morphological priming without morphological relationship. Language and Cognitive Processes 18 (3). 313–334. Luzzatti, Claudio, Sara Mondini & Carlo Semenza. 2001. Lexical representation and processing of morphologically complex words: Evidence from the reading performance of an italian agrammatic patient. Brain and language 79 (3). 345–359. Mann, Virginia A. & Maria Singson. 2003. Linking morphological knowledge to English decoding ability: large effects of little suffixes. In Egbert M.H. Assink & Dominiek Sandra (eds.), Reading complex words: Cross-language studies, 1–25. Amsterdam: Kluwer Academic. Marcolini, Stefania, Daniela Traficante, Pierluigi Zoccolotti & Cristina Burani. 2011. Word frequency modulates morpheme-based reading in poor and skilled italian readers. Applied Psycholinguistics 32 (3). 513–532. Marelli, Marco, Silvia Aggujaro, Franco Molteni & Claudio Luzzatti. 2012a. The multiple-lemma representation of Italian compound nouns: A single case study of deep dyslexia. Neuropsychologia 50 (5). 852–861. Marelli, Marco, Simona Amenta & Davide Crepaldi. 2015. Semantic transparency in free stems: The effect of Orthography-Semantics Consistency on word recognition. The Quarterly Journal of Experimental Psychology 68 (8). 1571–1583. Marelli, Marco, Simona Amenta, Elena A. Morone & Davide Crepaldi. 2013. Meaning is in the beholder’s eye: Morpho-semantic effects in masked priming. Psychonomic bulletin & review 20 (3). 534–541. Marelli, Marco, Elena Barbieri, Giusy Zonca & Claudio Luzzatti (2012b). Grammatical Class, Inflectional Entropy and Imageability Effects in Picture Naming: A Multiple Single-case Study on Italian Aphasic Patients. Procedia-Social and Behavioral Sciences 61. 214–215. Marelli, Marco & Marco Baroni. 2015. Affixation in semantic space: Modeling morpheme meanings with compositional distributional semantics. Psychological Review 122 (3). 485–515. Marelli, Marco & Claudio Luzzatti. 2012. Frequency effects in the processing of Italian nominal compounds: Modulation of headedness and semantic transparency. Journal of Memory and Language 66 (4). 644–664. Marelli, Marco, Daniela Traficante, Silvia Aggujaro, Franco Molteni & Claudio Luzzatti. 2011. Grammatical and semantic effects in reading derived nouns: A study of deep dyslexia. Procedia: Social and Behavioral Sciences 23. 69–70. Marslen-Wilson, William, Lorraine K. Tyler, Rachelle Waksler & Lianne Older. 1994. Morphology and meaning in the English mental lexicon. Psychological Review 101 (1). 3–33. McCormick, Samantha F., Kathleen Rastle & Matthew H. Davis. 2008. Is there a ‘fete’ in ‘fetish’? Effects of orthographic opacity on morpho-orthographic segmentation in visual word recognition. Journal of Memory and Language 58 (2). 307–326. McCormick, Samantha F., Kathleen Rastle & Matthew H. Davis. 2009. Adore-able not adorable? orthographic underspecification studied with masked repetition priming. European Journal of Cognitive Psychology 21 (6). 813–836. 590 Marco Marelli, Daniela Traficante and Cristina Burani Meunier, Fanny, & Catherine-Marie Longtin. 2007. Morphological decomposition and semantic integration in word processing. Journal of Memory and Language 56 (4), 457–471. Milin, Petar, Dusica Filipović Đurđević & Fermín Moscoso del Prado Martín. 2009. The simultaneous effects of inflectional paradigms and classes on lexical recognition: Evidence from Serbian. Journal of Memory and Language 60 (1). 50–64. Milin, Petar, Victor Kuperman, Aleksander Kostić & R. Harald Baayen. 2009. Paradigms bit by bit: An information theoretic approach to the processing of paradigmatic structure in inflection and derivation. In: James P. Blevins & Juliette Blevins (eds.), Analogy in Grammar: Form and Acquisition, 214–252. Oxford: Oxford University Press. Morris, Joanna, Jonathan Grainger & Philip J. Holcomb. 2008. An electrophysiological investigation of early effects of masked morphological priming. Language and Cognitive Processes 23 (7–8). 1021–1056. Moscoso del Prado Martín, Fermín, Raymond Bertram, Tuomo Häikiö, Robert Schreuder, & R. Harald Baayen. 2004. Morphological family size in a morphologically rich language: the case of Finnish compared with Dutch and Hebrew. Journal of Experimental Psychology: Learning, Memory & Cognition 30 (6). 1271–1278. Moscoso del Prado Martín, Fermín, Avital Deutsch, Ram Frost, Robert Schreuder, Nivja H. De Jong & R. Harald Baayen. 2005. Changing places: A cross-language perspective on frequency and family size in Dutch and Hebrew. Journal of Memory and Language 53 (4). 496–512. Moscoso del Prado Martín, Fermín, Aleksander Kostić & R. Harald Baayen. 2004. Putting the bits together: An information theoretical perspective on morphological processing. Cognition 94 (1). 1–18. Moscoso del Prado Martín, Fermín, Robert Schreuder & R. Harald Baayen. 2004. Using the structure found in time: Building real-scale orthographic and phonetic representations by accumulation of expectations. In Howard Bowman & Christophe Labiouse(eds.), Connectionist models of cognition, perception and emotion: Proceedings of the Eighth Neural Computation and Psychology Workshop, 263–272. Singapore: World Scientific. Norris, Dennis & Sachiko Kinoshita. 2008. Perception as evidence accumulation and Bayesian inference: insights from masked priming. Journal of Experimental Psychology: General 137 (3). 434–455. Patterson, Karalyn. 1980. Derivational errors. In Max Coltheart, Karalyn Patterson, & John C. Marshall (eds.), Deep Dyslexia, 286–306. London: Routledge & Kegan. Pavlov, Ivan P. 1927. Conditional Reflexes. New York: Dover Publications. Plaut, David C. & Laura M. Gonnerman. 2000. Are non-semantic morphological effects incompatible with a distributed connectionist approach to lexical processing? Language and Cognitive Processes 15 (4–5). 445–485. Plaut, David C. & Tim Shallice. 1993. Deep dyslexia: A case study of connectionist neuropsychology. Cognitive Neuropsychology 10 (5). 377–500. Quémart, Pauline & Séverine Casalis. 2014. Effects of phonological and orthographic shifts on children’s processing of written morphology: A time-course study. Scientific Studies of Reading 18 (5). 363–382. Quémart, Pauline & Séverine Casalis. 2015. Visual processing of derivational morphology in children with developmental dyslexia: Insights from masked priming. Applied Psycholinguistics 36 (2). 345–376. Reading morphologically complex words 591 Quémart, Pauline, Séverine Casalis & Pascale Colé. 2011. The role of form and meaning in the processing of written morphology: A priming study in french developing readers. Journal of experimental child psychology 109 (4). 478–496. Ramscar, Michael, Daniel Yarlett, Melody Dye, Katie Denny & Kirsten Thorpe. 2010. The effects of feature‐label‐order and their implications for symbolic learning. Cognitive Science 34 (6). 909–957. Rastle, Kathleen & Matthew H. Davis. 2008. Morphological decomposition based on the analysis of orthography. Language and Cognitive Processes 23 (7–8). 942–971. Rastle, Kathleen, Matt H. Davis, William Marslen-Wilson & Lorraine K. Tyler. 2000. Morphological and semantic effects in visual word recognition: A time-course study. Language and Cognitive Processes 15(4–5). 507–537. Rastle, Kathleen, Matthew H. Davis & Boris New. 2004. The broth in my brother’s brothel: Morpho-orthographic segmentation in visual word recognition. Psychonomic bulletin & review 11 (6). 1090–1098. Rastle, Kathleen, Lorraine K. Tyler & William Marslen-Wilson. 2006. New evidence for morphological errors in deep dyslexia. Brain and language 97 (2). 189–199. Rescorla, Robert A. 1988. Pavlovian conditioning it’s not what you think it is. American Psychologist 43 (3). 151–160. Rescorla, Robert A., & Allan R. Wagner. 1972. A theory of Pavlovian conditioning: Variations in the effectiveness of reinforcement and nonreinforcement. In Abraham H. Black & William F. Prokasy (eds.), Classical conditioning: Current research and theory II, 64–99. New York: Appleton-Century-Crofts. Rosenblatt, Frank. 1958. The perceptron: A theory of statistical separability in cognitive systems (Project Para). Buffalo, New.York: Cornell Aeronautical Laboratory. Rueckl, Jay G. & Karen A. Aicher. 2008. Are CORNER and BROTHER morphologically complex? Not in the long term. Language and Cognitive Processes 23 (7–8). 972–1001. Rueckl, Jay G. & Michael Raveh. 1999. The influence of morphological regularities on the dynamics of a connectionist network. Brain and Language 68 (1). 110–117. Schiff, Rachel, Michal Raveh & Avital Fighel. 2012. The development of the hebrew mental lexicon: When morphological representations become devoid of their meaning. Scientific Studies of Reading 16(5). 383–403. Schiff, Rachel, Michal Raveh & Shani Kahta. 2008. The developing mental lexicon: Evidence from morphological priming of irregular hebrew forms. Reading and Writing: An Interdisciplinary Journal 21 (7). 719–743. Schreuder, Robert & R. Harald Baayen. 1995. Modeling morphological processing. In Laurie B. Feldman (ed.), Morphological Aspects of Language Processing, 131–154. Hillsdale, New Jersey: Lawrence Erlbaum. Schreuder, Robert & R. Harald Baayen. 1997. How complex simplex words can be. Journal of Memory and Language 37 (1). 118–139. Seidenberg, Mark S. & James L. McClelland. 1989. A distributed, developmental model of word recognition and naming. Psychological Review 96 (4). 523–568. Smolka, Eva, Pienie Zwitserlood & Frank Rösler. 2007. Stem access in regular and irregular inflection: Evidence from German participles. Journal of Memory and Language 57 (3), 325–347. Solomon, Richard and Davis, H. Howes. 1951. Word frequency, personal values, and visual duration thresholds. Psychological Review 58 (4). 256. 592 Marco Marelli, Daniela Traficante and Cristina Burani Suárez-Coalla, Paz & Fernando Cuetos. 2013. The role of morphology in reading in Spanishspeaking children with dyslexia. Spanish Journal of Psychology 16 (e51). 1–7. Taft, Marcus. 1979. Recognition of affixed words and the word frequency effect. Memory & Cognition 7 (4). 263–272. Taft, Marcus. 1994. Interactive-activation as a framework for understanding morphological processing. Language and Cognitive Processes 9(3). 271–294. Taft, Marcus. 2004. Morphological decomposition and the reverse base frequency effect. The Quarterly Journal of Experimental Psychology A: Human Experimental Psychology 57A (4). 745–765. Taft, Marcus & Kenneth I. Forster. 1975. Lexical storage and retrieval of prefixed words. Journal of Verbal Learning and Verbal Behavior 14 (6). 638–647. Traficante, Daniela. 2012. From graphemes to morphemes: An alternative way to improve skills in children with dyslexia. Revista de investigación en Logopedia 2 (2). 163–185. Traficante, Daniela & Cristina Burani. 2003. Visual processing of Italian verbs and adjectives: The role of inflectional family size. In R. Harald Baayen & Robert Schreuder (eds.), Morphological Structure in Language Processing, 45–64. Berlin: Mouton de Gruyter. Traficante, Daniela, Stefania Marcolini, Alessandra Luci, Pierluigi Zoccolotti & Cristina Burani. 2011. How do roots and suffixes influence reading of pseudowords: A study of young Italian readers with and without dyslexia. Language and Cognitive Processes 26 (4). 777–793. Traficante, Daniela, Marco Marelli, Claudio Luzzatti & Cristina Burani. 2014. Influence of verb and noun bases on reading aloud derived nouns: Evidence from children with good and poor reading skills. Reading and Writing 27 (7). 1303–1326. Tsang, Yiu-Kei & Hsuan-Chih Chen. 2013. Early morphological processing is sensitive to morphemic meanings: Evidence from processing ambiguous morphemes. Journal of Memory and Language 68 (3). 223–239. Tsang, Yiu-Kei & Hsuan-Chih Chen. 2014. Activation of morphemic meanings in processing opaque words. Psychonomic bulletin & review 21 (5). 1281–1286. Turney, Peter D. & Patrick Pantel. 2010. From frequency to meaning: Vector space models of semantics. Journal of Artificial Intelligence Research 37 (1). 141–188. Tzur, Boaz & Ram Frost. 2007. SOA does not reveal the absolute time course of cognitive processing in fast priming experiments. Journal of Memory and Language 56 (3). 321–335. Van Ewijk, Lizet & Sergey Avrutin. 2011. Verb retrieval in non-fluent aphasia: An informationtheoretic approach. Procedia: Social and Behavioral Sciences 23. 104–105. Velan, Hadas, Ram Frost, Avital Deutsch & David C. Plaut. 2005. The processing of root morphemes in Hebrew: Contrasting localist and distributed accounts. Language and Cognitive Processes 20 (1–2). 169–206. Verhoeven, Ludo & Charles A. Perfetti. 2011. Morphological processing in reading acquisition: A cross-linguistic perspective. Applied Psycholinguistics 32 (3). 457–466. Yablonski, Maya & Michal Ben-Shachar. 2016. The morpheme-interference effect in Hebrew: A generalization across the verbal and nominal domains. The Mental Lexicon 11 (2). 277–307. Ziegler, Johannes C. & Usha Goswami. 2005. Reading acquisition, developmental dyslexia and skilled reading across languages: A psycholinguistic grain size theory. Psychological Bulletin 131 (1), 3–29.