Lexical Morphology: Structure, Process, and Development

Home , Drift (linguistics), Language production, Lexical item, Morpheme, Morphological derivation

Top Lang Disorders Vol. 33, No. 1, pp. 57–72 Copyright c 2013 Wolters Kluwer Health | Lippincott Williams & Wilkins Lexical Morphology: Structure, Process, and Development

Linda Jarmulowicz and Valentina L. Taran

Recent work has demonstrated the importance of derivational morphology to later language development and has led to a consensus that derivation is a lexical process. In this review, derivational morphology is discussed in terms of lexical representation models from both linguistic and psycholinguistic perspectives. Input characteristics, including types of frequency (lexical, surface, affix, and relative) and transparency (semantic, phonological, and orthographic), are examined as key factors that affect processing and acquisition. We introduce the possibility that lexical prosody and syllabic characteristics are relevant to lexical representation and affix separability, and we pro- pose that derivational morphemes can emerge to different degrees in a system that is sensitive to both sound and meaning. Finally, morphological development with a focus on children’s sensitivity to input characteristics is briefly reviewed, and we conclude with a perspective of how lexical representation can be a framework for derived word study in therapeutic or educational settings. Keywords: derivational morphology, lexical representation, school-age language

HAT IS MORPHOLOGY? Based on a The morpheme-based view of morphology W quick look in a textbook for stu- has been replaced by lexeme-based mor- dents of communication sciences and disor- phology (Aronoff, 1994; Feldman, 1995) and ders, the answer is along the lines of the more recently by noncompositional theories “ . . . rules that govern the use of morphemes that view morphology as an emergent struc- inalanguage...”(Berko-Gleason&Bernstein ture (Bybee, 2006; Hay & Baayen, 2005; Ratner, 2013, p. 401). The traditional view, Seidenberg & Gonnerman, 2000). that discrete meaning-based morphemic units Morphology is not isolable from semantics, can be systematically combined to make new syntax, or phonology, and its effects extend words, like stringing beads, may be a sim- beyond rule identification and application. Be- ple way to present morphology; however, cause morphemes are units of meaning, mor- morphology is not just the study of rules. phology and the study of the mental lexicon are interconnected. Morphology’s role in syntax through inflectional paradigms is promi- nent, and until fairly recently morphosyn- Author Affiliation: School of Communication tax was the archetype of rule-based morphol- Sciences and Disorders, University of Memphis, ogy. Morphology and phonology also interact, Memphis, Tennessee (Drs Jarmulowicz and Taran); such as in allomorphic variation, lexical stress and Iris Speech Solutions, LLC, Arlington, Tennessee (Dr Taran). distinctions (e.g., record (N) and re cord This work was supported in part by the Center for Re- (V)), and derivational alternations (e.g., di- search Initiatives and Strategies for the Communica- vine/divinity). Morphology is also repre- tively Impaired at the University of Memphis. sented in orthography. So, if morphology is The authors have disclosed that they have no signif- peppered throughout language, perhaps the icant relationships with, or financial interest in, any question should be: What isn’t morphology? commercial companies pertaining to this article. In the first part of this paper, we review Corresponding Author: Linda Jarmulowicz, PhD, descriptions of morphological structure from School of Communication Sciences and Disorders, Uni- versity of Memphis, 807 Jefferson Avenue, Memphis, TN the field of linguistics and evidence of the 38105 ([email protected]). nature of lexical representation and process- DOI: 10.1097/TLD.0b013e318280f5c0 ing from psycholinguistics. In addition, we 57

discuss key factors that affect processing and 2003). Inflection alters words to fit in differ- acquisition, such as distributional character- ent grammatical contexts, creating new word istics and transparency. In the second part forms, but not new lexemes. The two pro- of the paper, we discuss developmental data cesses are ordered such that inflectional mor- on morphological derivation, with a focus on phemes can be added to either a stem (e.g., children’s sensitivity to input characteristics help + ed[past]) or a derived word (e.g., help and how the data fit with current lexical mod- + er[agentive] + s[plural]), but a derivational els. We conclude with a brief review of how morpheme cannot be added after inflection understanding of lexical representation can (e.g., *help + ed[past] + er[agentive]∗). Un- be a framework for derived word study. like inflectional morphemes, derivational suffixes vary in their productivity and tend to STRUCTURE be pickier about the types of bases to which they attach. In sum, inflection is grammatical For at least 20 years, linguists have ob- morphology and derivation is a type of word served that morpheme-based theories, in formation, or lexical morphology. which words are reducible to morphemes Another area of linguistic theory that that combine by rules, are insufficient to ac- touches on morphology is lexical phonol- count for all of the phenomena encompassed ogy (Plag, 2003). Despite its name, lexical by morphology (Aronoff, 1994; Beard, 1995; phonology is as much about derivational Plag, 2003). Current linguistic accounts fre- morphology as it is about phonology. The quently posit two types of representation: the literature on lexical phonology describes lexeme and the morpheme. The lexeme is two classes of English suffixes that exhibit represented in the lexicon and is much like different kinds of phonological behavior (see a bare stem or a citation form (i.e., dictionary Lieber, 2010, for overview). For historical entry). The morpheme includes affixes, redu- reasons, English has Latin-based suffixes plication, and by some accounts modifications borrowed primarily through French (e.g., of phonological representation that change -al, -ive, -ic, -tion, -ity), suffixes borrowed a lexeme (Beard, 1995). The lexeme unites through Greek (e.g., -graph, -ology, -ist), grammatical and phonological forms (Stump, and native Germanic-based suffixes (e.g., -er, 2005). For example, the verbs am, was,and -ful, -hood, -less, -ness). These two classes been have distinct phonological and grammat- of suffixes (borrowed and native) behave ical forms, but all are related to the abstract differently. First, native suffixes can easily lexeme BE, which links these forms together. attach to borrowed suffixes (e.g., nativeness, Morphemes traditionally divide into in- signifier), but the reverse is generally not true flectional and derivational types, based on (e.g., *sorrowfulity,*boredomic). The Latin- their function in a language. Empirical data based borrowed suffixes have no trouble support the distinction in adult and pedi- attaching to other Latin-based suffixes (e.g., atric populations (Badecker & Caramazza, sensitive + ity, exception + al). Second, 1989; Clahsen, Sonnenstuhl, & Blevins, 2003); borrowed suffixes can attach to bound bases however, there are parallels in terms of (e.g., nutri + tion), whereas native suffixes how the two morphological systems operate attach only to free bases (e.g., *nutri + ful). (Deacon, Campbell, Tamminga, & Kirby, Finally, borrowed suffixes may alter the 2010; Hay & Baayen, 2005; Stump, 2005). segmental phonology or stress pattern of the An incontrovertible distinction between in- stem (e.g., sacrifice/sacrificial and active/ flection and derivation is that the latter is a type of word formation that creates new lexemes in the lexicon typically through the addition of affixes, which include prefixes, ∗We will use the linguistic convention of placing * before suffixes, and infixes (Aronoff, 1994; Plag, ungrammatical or incorrect examples.

activity for segment and stress changes, sentation. The lemma bridges the conceptual respectively), whereas native suffixes do not. and the formal aspects of a word. The sec- In addition, certain suffixes prefer to attach ond level of lexical representation includes to others in what is called base-driven selec- the formal characteristics, such as morpho- tion (Plag, 2003). For example, if one wanted logical and phonological information, which to googlize his Web site, the likely noun are mapped to or from the lemma. The prod- from this verb would not be *googlizance uct of the lemma plus the result of form cod- or *googlizal or *googlizage;rather,tocom- ing is called the lexeme.∗ For production, the plete the sentence, I will not comment on lemma must be accessed first then mapped to the googliz____ of his website,mostEnglish the phonological representation. For recogni- speakers would add the suffix -ation,asin tion, the form (phonology for spoken, orthog- googlization. The suffix -ize appears to se- raphy for visual) representation is the entry lect -ation as the de-verbal suffix. Some of point to the system. these effects might best be explained by Note that the levels are free to connect and frequency of co-occurrence, as we discuss interact with other stored representations. So, below. awordlikecake might link to chocolate, ice cream, cookie,orcupcake at the lemma PROCESS level whereas the same word might link to take, kick, came at the word form or lexeme Although linguistic descriptions of mor- level. Phonological neighborhoods consist of phology are helpful for classification and un- lexemes (Storkel & Morrisette, 2002). This derstanding the range of cross-linguistic pos- type of organization allows language users to sibilities, psycholinguistic models contribute complete word association tasks by thinking to understanding of how lexical morphology about the semantic or phonological aspects is processed. Because affixes have meaning of a lexical representation (e.g., to list all the and form and because derivation is a lexical words in a particular semantic category or to process, it is important to understand how list all the words that begin with a particu- words are stored in long-term memory, how lar sound). Word-finding errors may occur at derived words are processed, and what fac- the semantic or phonological levels, both in tors affect processing. The lexical representa- adults (Nozari, Kittredge, Dell, & Schwartz, tion must unite several types of information 2010) and children (Faust, Dimitrovsky, & stored in long-term memory: conceptual, se- Davidi, 1997; German & Newman, 2004). mantic, syntactic, phonological, and eventu- Most, if not all, models recognize the inde- ally orthographic information. pendence of these two types of information in the lexicon; although exactly how each level Word storage Psycholinguistic models of the lexicon have at least a two-level architecture (Dell & O’Seaghdha, 1992; Levelt, 2001) in a highly in- ∗As a point of clarification, linguists and psycholinguists terconnected system of associations. An adap- use the same word lexeme to refer to slightly different tation of Levelt’s (1999, 2001) production concepts. To some linguists, the lexeme contrasts with a morpheme in that it is an established unit in the lexicon model is presented in Figure 1. Outside of to which morphemes may attach. Psycholinguists have the lexicon is the store of lexical concepts found that splitting the linguistic lexeme into two parts— and broader pragmatic knowledge. Associ- the lemma and lexeme—better captures some of the phe- ated with the concept level is the first level nomena they observe. In addition, corpus linguistics uses of lexical representation, the lemma.The the term lemma to mean the citation form, whereas the lexeme includes the various forms of the word (e.g., eat lemma is the core of meaning, which includes would be a lemma, eats, ate, eating, eater would be lex- the semantic and syntactic subcategorization emes). This latter sense seems to be closer to the levels information associated with a lexical repre- in psycholinguistic processing models.

Figure 1. Illustration of two levels of lexical representation (based on Levelt, 1999, 2001).

develops independently of others, is still a recognition literature included both morpho- challenge facing researchers (Deacon et al., logically simple and complex words (Cooper 2010; Storkel, 2009). et al., 2002; Soto-Faraco et al., 2001). Stressed syllables have been suggested as Over time and with literacy instruction, a layer of organizational structure within the orthographic information must be added formal layers of the lexical representation, al- to the lexical representation (Ehri, 2000). though there is conﬂicting evidence in the This includes orthographic patterns and adult literature as to whether this informa- mappings between phonology and seman- tion is important in lexical recognition across tics. Each of these types of representation— all languages (Cooper, Cutler, & Wales, 2002; semantic, phonological, orthographic—is as- Soto-Faraco, Sebastian-Galles, & Cutler, 2001). sociated with the others. Together, they con- Production models also recognize the im- stitute word knowledge. Furthermore, each portance of linguistic structure above the serves as a basis of organization across words phoneme, as they designate the lexeme as the within the lexicon. In the adult visual word location where syllabiﬁcation and stress as- recognition literature, orthographic process- signment would occur in production (Levelt, ing is critical, as it is the window used 1999; Lieber, 2010). Curiously, although lexi- to understand lexical organization (Marslen- cal stress has been acknowledged as a feature Wilson, Bozic, & Randall, 2008; Rastle & in producing monomorphemic words, there Davis, 2008). Although there appears to be is little work focusing exclusively on stress little controversy about whether semantic, in morphologically complex words. Only re- phonologic, and orthographic information cently have studies of lexical stress in the must be stored with a word, there is less

certainty about how derived words are stored and processed.

Derivation in the mental lexicon Whether or not derivational affixes main- tain their own representations is an issue that is rarely addressed in production models. In contrast, numerous studies have demonstrated that derivational morphemes are available as visual processing units (Amenta & Crepaldi, 2012; Marslen-Wilson et al., 2008), although the exact nature and sequence of morphological processing continues to be debated (Amenta & Crepaldi, 2012; Frost, Grainger, & Rastle, 2005). At some point in Figure 2. Gradient structure model of emerging derivational suffixes. most models, morphemes are available. Affixes that are available as independent units could facilitate comprehension through Gradient structure is reflected in the decomposition and would be available for strength of the connections, which has di- composition as well. Thus, new adjectives, rect implications for development. Thus, the such as thinkative or sugarous could be cre- structure of a lexical system is not built from ated spontaneously to fill a semantic need, as a simple yes/present or no/absent;rather, these were by an eight-year-old child. Despite structure is represented by degree because the ultimate availability of affixes, representa- some connections between words or mor- tion and access to affixes is likely a process phemes are stronger than others. Stronger of discovery for children over time and over connections suggest stronger representations many exposures. upon which children can then reflect. We A another possibility is that both options have attempted to illustrate application of this are available in a highly interconnected, non- type of model to derived words in Figure 2. linear, and gradient network. Derivational In this gradient structure model, the bolder morphology is part of a larger emergent words and connections are represented more system (Bybee, 2006; Hay & Baayen, 2005; strongly in the system. The strength of the Gonnerman, Seidenberg, & Andersen, 2007; connections is not uniform because deriva- Seidenberg & Gonnerman, 2000). Lexical tion is quasi-regular, and the source of the structure is the product of both the input and gradient structure is the input characteristics. the connections made by the system. From Thus, affixes may not be represented initially this view it is not incongruous to have a net- as separate entities; rather, they emerge from work of interrelated words whose compo- the system as more types of stems are en- nent parts also become interrelated. Hay and countered with similar morphophonological Baayen (2005) noted that, although the whole patterns. Note that this type of model can cap- stored lexical representation, with its con- ture both phonological regularities and mor- nections between form and meaning, takes phological irregularity. For example, in addi- precedence, they also noted that “ . . . the tion to the suffixes -al and -ity, the sequence parts of complex wholes can also be ac- -ality will emerge from the system depicted in tive during production and comprehension” Figure 2. This explains what might previously (p. 344). Thus, both semantic and phonolog- have been described as base-driven selection ical connections throughout the system can by models focusing on rules and storage. form new associations from which new rep- Suffix combinations like -istic, -mental,and resentations emerge. -ation might emerge in the same way. This

might also explain for how neologistic (i.e., of the base word. Thus, a noun such as newly invented) suffixes like -aholic (e.g., vacation would be stored separately from workaholic) emerge as a “morpheme” from the verb vacate from which it is derived. a phonological sequence. Low-frequency derivations are thought to In gradient models, the system starts with require decomposition during recognition meaning and sound, and eventually adds or- but high-frequency derived words can be thography (Gonnerman et al., 2007). It does accessed directly (Meunier & Segui, 1999; not necessarily start with discrete morpho- Plag, 2003). Affix frequency is a measure of logical units. If this is the case, we should how often a particular affix occurs in the lan- expect differences in meaning and form reg- guage. Relative frequency is the relationship ularity to affect acquisition. Frequency and between the frequencies of a particular base transparency are undisputed input character- and its derived form. Derived words that are istics that come into play in both acquisition more frequent than their base words may be and in determining the strength of lexical con- subject to semantic drift from their stem and nections among derivatives and stems and the may have suffixes that are less likely to be used productivity of particular affixes. in novel productions (Hay, 2002; Plag, 2003). Family size is a count of unique morpholog- Input characteristics ical relatives of a base word. For example, the morphological family size of the adjec- Frequency tive deep would be small, including deepen, Frequency of occurrence is consistently deeper,anddepth, whereas the adjective demonstrated to be factor in how the lexicon active would have a larger family size, in- is shaped and organized. More frequent units cluding act, activate, action, activity, activa- have stronger representations and stronger tion,andactor. In some studies, family mem- associations. Frequency figures prominently bers may include either affixed derivatives or in any input-based model (Bybee, 2006; Hay compounds (Pylkkänen, Feintuch, Hopkins, & Baayen, 2005; Seidenberg & Gonnerman, & Marantz, 2004). Base frequency is a similar, 2000). Higher lexical frequency is a predic- but not identical, idea. It reflects the cumula- tor of lexical retrieval speed (Griffin & Bock, tive frequency of a particular base plus the fre- 1998) and production accuracy (Nozari et al., quencies of all the morphologically complex 2010). words that include that base. For example, With regard to derivational morphology, the cumulative frequency of the verb protect frequency comes in several varieties. Lexical would consist of a combined frequency for or surface frequency, affix frequency, and protect, protects, protected, protective, pro- relative frequency are believed to be charac- tection, protector,etc. teristics of the lexeme or form level, whereas family size and cumulative frequency are Transparency more closely associated with the lemma or The second factor that influences the stor- semantic level (Amenta & Crepaldi, 2012; de age, access, and use of derived words or Jong, Schreuder, & Baayen, 2003). Lexical fre- affixes is transparency. Transparency might quency, sometimes called surface frequency, also be thought of as consistency. Morpho- is an estimate of a particular lexical item logical transparency comes in several types— occurring in the language relative to all other semantic, phonological, and orthographic. lexical items. Individual lexeme frequency They map effectively onto the working model estimates can be obtained for a particular of the lexical representation and fit well with base or derived word. High-frequency derived the gradient structure model in Figure 2. In words have stronger lexical representations each case, transparency can be thought of as a than low-frequency derived words, and continuum of association or overlap between consequently are represented independent a derived word and its base.

Semantic transparency refers to the over- form may differ from the base word (e.g., lap in meaning or semantic relatedness be- dictate/dictation). Any phonological change tween the base and derived form. For ex- makes the relationship between the base and ample, teacher and mousey both have a se- derivation phonologically less transparent. mantically transparent relationship with their In English, all inflectional and most native bases, teach and mouse, whereas the pairs suffixes (e.g., -ness, -ful, -ing, -er) produce class and classify or author and author- phonologically transparent relationships. Suf- ity both have a less semantically transpar- fixes that do not alter the phonology of the ent relationship between base and derived stem are called phonologically neutral. Suf- forms. Thus, the extent to which the mean- fixes that change the stem internal phonology ing of a derivation is related to its base varies and make the base and derived word relation- across base/derivation pairs. This gradient na- ship less phonologically transparent are non- ture can be captured through semantic re- neutral (e.g., curious/curiosity). These are latedness judgments using a Likert scale to historically borrowed suffixes. Note that, it evaluate the semantic relatedness of a pair is neither the case that all borrowed suffixes that includes a base and derivation. In our are nonneutral nor that native suffixes are work, adults have judged pairs such as gener- all neutral. For example, immune/immunity ous/generosity as more semantically related and deep/depth are exceptions in that the for- than minor/minority or disciple/discipline mer includes the borrowed (i.e., Latin) suf- (Jarmulowicz & Taran, 2007; Windsor, 2000). fix -ity but no phonological changes to the Each word in the less related pairs is presum- base and the latter includes the native (i.e., ably stored separately in the lexicon, with rel- Germanic) suffix -th and a phonological atively weaker connections between the two change in the vowel quality between the base words at the lemma level. Furthermore, the and derivation. morpheme, -ity, may be more strongly fused In a more recent view of phonological trans- to the minority lemma than it is in a word like parency, Hay (2002) considered the base- generosity, which exhibits a more semanti- suffix juncture as important to determining cally transparent relationship with generous. whether a suffix is easily separated from its The next two types of transparency, phono- base. Under this view, vowel initial suffixes logical and orthographic, are formal aspects will be less easily separated from their bases of the lexical representation. These two as- than will consonant initial suffixes. This is be- pects tend to be conflated in the adult lit- cause a resyllabification process changes the erature because of the dominance of visual coda of the final syllable of the base word. recognition methods. They are studied more Thus, the neutral suffix -ish might be more independently in the developmental literature challenging for children than -less because, in with children who are preliterate. Phonolog- a derived word, the syllable is modified in the ical transparency refers to the degree of sim- former but not the latter (e.g., note the sylla- ilarity in the sound patterns of the base and ble boundaries in selfish /sεl.fI /andselfless derived form. Thus, the relationship between /sεlf.lIS/). deep /dip/ and deepen /dip n/ is more phono- Orthographic transparency is the degree to logically transparent than the relationship be- which the base word appears unaltered in the tween deep /dip/ and depth /dεpθ/, because spelling of the derived form. For example, the of the change in the vowel quality of the word pair punish–punishable has a greater base word. Derived words may vary in their degree of orthographic transparency than phonological transparency due to segmental permit–permission, because the base word in changes in vowel quality or in the consonant the transparent pair is visually identical in the at the affix juncture (e.g., decide/decision derived form. The English spelling system en- shows both vowel and consonant changes). codes both morphemes and phonemes, both In addition, primary stress in the derived of which map individual sounds to letters.

The spelling of morphological segments is than -ity. First, derived words with -ity tend preserved regardless of the pronunciation. to have higher frequencies than those with Thus, orthographic transparency refers not to -ness, suggesting -ity derivations have stronger whether the sounds are represented by ac- independent representations. When derived ceptable letters, as the -o- in curiosity cer- words have strong independent representa- tainly is, but whether the base word spelling tions, they are candidates for semantic drift, is preserved. in which the derived form takes on mean- The three dimensions of transparency are ings not necessarily present in the base word somewhat separable in English. Thus, it is pos- (Plag, 2003). This directly influences seman- sible that semantically transparent word pairs, tic transparency. Thus, a word like curiosity such as deep/depth are neither phonologi- maintains a close semantic relationship with cally nor orthographically transparent. Like- curious,whereasminority has drifted some- wise, phonological and orthographic trans- what from the narrower meaning in minor. parency can vary orthogonally such that it The flipside of frequent derivations having in- is possible to have words that are both dependent representations is that productive phonologically and orthographically transpar- suffixes can make words with very low fre- ent (e.g., enjoy/enjoyment), only phonolog- quency, because they can be used to generate ically transparent (e.g., happy/happiness), words that no one has ever made before (e.g., only orthographically transparent (e.g., ogreness, a novel word created by a child). magic/magician), or neither phonologi- Thus, not finding an understandable derived cally or orthographically transparent (e.g., word in a dictionary may actually be evidence chaos/chaotic). These are the factors that of affix productivity. make the derivational system quasi-regular. A second difference that makes -ness more productive than -ity is that -ness can only at- Productivity tach to free morphemes whereas -ity can be Frequency and transparency are interwo- found on bound roots (e.g., paternity, trin- ven into the notion of productivity. This is ity). It may be less difficult to identify the essentially the likelihood that an affix will base word or morphological family for words be used in a new form. In this sense, regu- with -ness than with -ity. A third difference lar inflectional suffixes are very productive, is that words suffixed with -ness are phono- but derivational suffixes vary in their produc- logically transparent; that is, the base word tivity. Frequent and transparent suffixes are undergoes no sound changes. The suffix - more available for use in new words than ity is notoriously nonneutral, often affecting either less frequent or less transparent suf- base segments (e.g., sane/sanity)orstress fixes. Furthermore, high-frequency suffixes patterns (e.g., active/activity), or both (e.g., that are transparent along more dimensions elastic/elasticity). Furthermore, from a per- (i.e., semantic, phonological, orthographic) ceptual standpoint, the vowel initial status of - are the best candidates for being produc- ity may make it more difficult to segment from tive affixes because they have more speci- its base word than the consonant initial -ness fied, higher quality, interrelated representa- (Hay, 2002). tions (see Reichle & Perfetti, 2003 for further Finally, although the two suffixes are or- discussion). thographically consistent across words, their The two nominalizing affixes -ness and -ity base words may be altered at the base- are good examples. They have similar mean- suffix juncture (e.g., happy/happiness, ac- ings: state of being X,whereXisanad- tive/activity). Thus, the frequency and trans- jective. The suffixes themselves have simi- parency factors conspire to affect how these lar frequencies. However, there are important affixes are stored and processed and also how differences that make -ness more productive they develop.

STRUCTURE AND PROCESS IN LEXICAL a determining factor in which process or af- MORPHOLOGY DEVELOPMENT fix children would acquire first. Thus, the agentive -er is frequently used on numerous Adult research focuses on modeling the ma- verbs to mean “one who does X” and is used ture system, not on the process or stages spontaneously by children before the less fre- of achieving that system. However, there is quent but synonymous -ist (Clark & Cohen, some overlap in how the adult representa- 1984). Windsor and her colleagues (Lewis & tion and children’s lexical development are Windsor, 1996; Windsor & Hwang, 1999) framed. Input frequency and transparency in- demonstrated that school-aged children’s suf- fluences derivational development as much fix comprehension and production was, in as they influence processing. Development of part, explained by suffix frequency. and interaction among the types of lexical In our own work examining effects of stress representation—meaning (lemma) and form in derived words, we have found several fre- (phonology and orthography)—are a focus in quency effects. Jarmulowicz (2002) found an language and literacy research. In this section, effect of suffix frequency in which 7- and we discuss areas of overlap and future direc- 9-year-old children’s judgments of accurate tions of inquiry. stress placement were better on both derived words and nonwords with higher suffix fre- Frequency quency than with lower suffix frequency. In Frequency effects in language development a study of stress production accuracy in high- are present in work that emphasizes input- frequency, low-frequency, and nonsense de- driven systems. For example, Marchman and rived words by third graders, Jarmulowicz, Bates (1994) contended that the emergence Taran, and Hay (2008) found a suffix effect of morphosyntax is the result of children ac- in which children performed better on words quiring a critical mass of verb knowledge with higher frequency suffixes than on those from which they extract the inflectional pat- with lower frequency suffixes. Second, chil- terns. Others have noted that frequency plays dren performed better on higher frequency a role in the emergence of early content words derived words than on lower frequency de- (Goodman, Dale, & Li, 2008) and in produc- rived words, and they did the least well on tion accuracy (Munson, 2001). In addition, nonsense words (i.e., nonword stems with models of phonological neighborhood den- real English suffixes). Finally, the relative fre- sity rely heavily on the recurring patterns quency between the derived words and their in the input to shape the lexicon (Garlock, stems was examined. The rationale was that Walley, & Metsala, 2001; Storkel, 2009; derived words that were higher in frequency Storkel & Morrisette, 2002). relative to their base words might be more The same factors and mechanisms that likely to be stored as independent units and shape monomorphemic words in the lexi- might affect children’s performance. There con appear in the literature examining com- was no correlation between stress accuracy plex word development (Reichle & Perfetti, and either stem or derived word frequency; 2003). Warlaumont and Jarmulowicz (2011) however, a significant negative correlation found that frequency of inflectional endings (r = –.51) was obtained for the relationship in the input was related to the order of mor- between stress production accuracy and rela- phosyntactic development. Evidence for fre- tive frequency, suggesting that children pro- quency factors in derivational morphology de- duced stress better on derived words that had velopment comes from a range of sources. lower frequency stems relative to the derived Based on data from preschool children, Clark word. (1993) considered the frequency with which Accuracy of derived word reading for a word-formation process was used as a key fourth and sixth graders is also affected by indicator of productivity, and by extension, derived word frequency; however, high-base

frequency can also influence reading ac- and that children with language disorders curacy, despite differences in phonologi- performed particularly poorly on words with cal transparency (Deacon, Whalen, & Kirby, phonological irregularities. 2011). In addition, both morphological fam- In our work on stress accuracy, we focused ily size and family (i.e., base word) frequency on production as a reflection of lexical and af- have also been isolated as important factors fix knowledge. We deliberately avoided writ- related to reading (Carlisle & Katz, 2006). ten support in order to limit access to the orthographic representation, and we tried to Transparency limit the influence of semantics by using a A significant amount of developmental single word production task with no seman- work has focused on transparency issues; tic context or nonsense words. A further is- however, it remains challenging to iso- sue we attempted to address was consistency late the semantic, phonological, and or- across the phonological features that differed thographic factors. Semantic, phonological, in the base and derived word. Thus, instead of and orthographic transparency appears to examining phonological transparency broadly affect acquisition of derivational morphol- defined as any phonological modification— ogy at different times in development vowel, consonant, or stress—we focused (Carlisle, 1988; Carlisle & Stone, 2005). In on only affixes that result in a predictable oral language, early-acquired suffixes are stress pattern. In Jarmulowicz (2006), we generally semantically and phonologically reported that children mastered stress pro- consistent ones. Semantic transparency af- duction of neutral words at least by age fects how easily a base word can be seven; whereas, stress production accuracy recognized within a derived form (Clark, for nonneutral words continued to increase 1993; Derwing, Smith, & Wiebe, 1995). In ad- significantly each year between the ages of dition, children spell words with semantically seven and nine. There was further evidence transparent relationships more accurately (Jarmulowicz, 2006; Jarmulowicz, Taran, & than those with less apparent semantic rela- Hay, 2007) that degree of phonological tionships, even when controlled for phono- change might make a difference in perfor- logical transparency (Deacon & Bryant, 2005). mance. For example, children seemed to Transparency also affects derivational perform least well when more phonological composition and decomposition accuracy, changes were present (e.g., vowel, conso- with implications for developmental se- nant, and stress). Clin, Wade-Woolley, and quences; however, again elements of lexical Heggie (2009) examined this idea further representation are frequently conflated. In by using a composition task (i.e., sentence an evaluation of spelling, Carlisle (1988) completion with a derived word) with third, found that fourth, sixth, and eighth grade fifth, and seventh graders. They held ortho- students produced fewer errors on phono- graphic transparency constant, but varied the logically and orthographically transparent type of phonological change—either stress derived words (e.g., enjoy/enjoyment)and shift (e.g., accident/accidental) or phonemic those with only orthographic changes (e.g., change (e.g., discuss/discussion). They found happy/happiness) than on words with phon- that derived words with stress shift were the ological changes (e.g., magic/magician) most challenging. or words with both orthographic and phonological changes (e.g., chaos/chaotic). Future directions Windsor (2000) eliminated the orthographic Unlike the research with adults, much of element and found that children were less the research on derivational morphology with accurate on both identifying suffixes and children is framed in the language of aware- producing base words from derived words ness. Thus, the focus is on demonstrating ex- that were not phonologically transparent plicit knowledge, without much discussion

of tacit or implicit knowledge and processes. But what is it exactly that children are In contrast, much of the adult data come aware of? We contend that morphological from measures of implicit knowledge, based awareness is a reflection of the organization on processing speed as a metric of effi- of the lexical system, and for children, aware- ciency, spreading activation, or inhibition. ness reflects lexical organization at a particu- Adult-processing researchers rely on precon- lar point in development. In a well-organized scious associations to better understand how system, affixes that have reached a particu- the brain processes language. lar threshold or strength of representation The limited research on children’s implicit are candidates for awareness. But this can- morphological knowledge is focused on spon- not be the entire story. If lexical represen- taneous productions and processing mea- tation includes gradient relationships that are sures. Spontaneous productions demonstrate built over time and with exposure, and that some independent representation of the af- consist of semantic, phonological, and ortho- fix and compositional capacity of the devel- graphic information, there may be multiple oping system. This has been documented as thresholds, at least initially, each of which is early as preschool with children producing potentially a type of awareness. novel derivations (e.g., toothachey and fly- As noted by others, vocabulary acquisition able) to fill lexical gaps (Clark, 1982). It is is not an all or none phenomenon (Bloom, not necessarily the case that children are 2002; Ehri, 2000; Wagovich, Pak, & Miller, aware of these suffixes as independent units, 2012). A child can know a little about a lot but clearly they have a system that allows of words, or a lot about some affixes and less these suffixes to be productive. More re- about others. A child might be able to use cently, processing measures with low task a whole derived word, but not recognize its demands have been used with school-aged parts—and vice versa. The absence of con- children (Carlisle & Stone, 2005; Deacon et scious awareness does not mean a lack of al., 2010; Deacon et al., 2011). These mea- knowledge. It might mean partial knowledge sures of processing accuracy and speed, typi- or that implicit knowledge has not consoli- cally through priming or word association re- dated to the quality required for awareness. It action time tasks, are more closely aligned might mean the task demands were too high. with lexical organizational factors than One of the problems with the awareness liter- with awareness. Deacon and her colleagues ature at present is that the diversity of tasks, (Deacon et al., 2010, 2011) have begun to variety of modalities, and plethora of variables emphasize processing measures as a way to make a general interpretation challenging. begin to tease apart the influence of seman- As the research in lexical morphology ma- tics, phonology, and orthography from mor- tures, it becomes important to begin to fine phological processing and to begin to under- tune understanding of what aspects of the stand the developing representational system, lexical representation, lexical organization, or rather than just the ability to manipulate it. underlying processes are the source of dif- The developmental literature on morpho- ficulty for children who struggle with oral logical awareness and recognition of the im- vocabulary and its written extensions. Much portance of morphological awareness has of the awareness literature in lexical mor- blossomed in the last decade. Conscious phology has focused on meaning, reading, awareness of morphemes helps children in- or spelling, leaving phonology to be the fer the meaning of unfamiliar derived words. messenger between the two. Even current Through what Anglin (1993) termed morpho- voice-triggered reaction time studies acknowl- logical problem solving, children can work edge that they overlook phonological accu- out the meaning of a novel derivation by rec- racy (Carlisle & Stone, 2005; Deacon et al. ognizing the component morphemes. 2011).

There is a consensus that phonologically them, and the input factors affecting them transparent derived words are easier for chil- are useful clinically and in the classroom. dren to understand and easier to extract suf- This framework can be a foundation for word fixes from. But exactly what makes those study. words with nonneutral suffixes so difficult is Ensuring that children have concept knowl- less clear. Two phonological issues, both of edge to map to the lexical representation and which concern phonological units larger than vice versa is an important first step. For in- the phoneme, have yet to be fully explored. stance, the child who possesses the concept First, is the hypothesis forwarded by Hay of a three-dimensional round body may not (2002) that vowel-initial suffixes fuse more have a lexical representation for the word closely to the base word, which makes the suf- spherical. Conversely, a child may possess a fix more difficult to extract perceptually from lexical representation for which he does not the base word. Thus, it may be the syllable yet possess an accurate concept. structure at the base-affix boundary that partly Instruction at the lemma level can intro- explains phonological transparency. The sec- duce the syntactic function of the base word ond issue concerns linguistic prosody. Stress (e.g., sphere is a noun), derived word (e.g., perception, rhythm detection, and stress spherical is an adjective), and affix (e.g., production have been the focus of research, -al often indicates an adjective and attaches particularly as performance on these tasks ap- to nouns). Additional focus on the lemma pear to predict reading ability, independent level might include embedding this word of phoneme-level tasks (Holliman, Wood, & in a network of semantically related words Sheehy, 2010; Jarmulowicz, Taran, & Hay, (globe, ball, round, orb) and morphological 2007; Whalley & Hansen, 2006). It is possible family members (e.g., sphere/spherical). then, that a missing piece of the morpholog- Differences in meaning and use among ical awareness puzzle is children’s ability to the semantically related words might help reflect and manipulate linguistic stress. Know- establish the lemma. Further explicit instruc- ing the sounds in a word is an important part tion of morphological problem solving can of lexical knowledge. Stress, particularly in bring the client’s attention to the semantic polysyllabic words, may need to be part of lex- relationships among morphologically related ical knowledge as well. Furthermore, as illus- words (sphere/spheric/spherical,orwords trated in Figure 2, stress patterns may emerge like biosphere, atmosphere). Other words along with suffixes over repeated exposures. with similar affixes may be introduced Finally, as children progress through (theater/theatric/theatrical; cycle/cyclic/ school, new words increasingly enter the cyclical) to reinforce the meaning of the lexicon through encounters with text. Links suffix. between morphology and phonology be- At the lexeme or form level, phonological come reinforced through repeated activa- patterns can be reinforced through pronun- tion through orthography. As a result, the ciation of the target word as well as other spelling of a word may be a window into words that share the pattern. The systematic the word’s lexical representation (Deacon & phonological changes of nonneutral suffixes Bryant, 2005; Perfetti, 1992). can be explicitly taught to students. For example, in words formed with the suffixes -ic, THEORY INTO PRACTICE -ion,and-ity, primary stress placement occurs on the syllable before the suffix. For younger The models and research discussed in this children, just getting them to pay attention to paper highlight the multiple levels of rep- patterns of sound at the ends of words that oc- resentation one must consider when plan- cur in other words would tap this level of rep- ning assessment and intervention. Attend- resentation. Further attention to differences ing to these levels, the processes between in both suffix type (phonologically neutral or

Table 1. Variables to consider for lexical morphology word study

Unit of Representation Frequency Length (in Syllables)

Base High Mid Low 1 2 3 4 + Affix Prefix High Mid Low 1 2 Suffix High Mid Low 1 2 Derived word High Mid Low 2 3 4 5 + Lexical Characteristics Transparency Semantic High Mid Low Phonologic Neutral High Low Nonneutral High Low Orthographic High Low

nonneutral), suffix structure (vowel initial or The pervasive effects of frequency suggest consonant initial) may change the relative dif- that frequent encounters with a derived word, ficulty of isolating the affixes. Linking these despite the actual frequency of occurrence, sounds back to meaning would complete the will promote more detailed representations. circle. For example, to target the lexeme level, Increasing the number of encounters a child the student may be instructed that some suf- has with a target word in a variety of semantic fixes change the sound of the root word and contexts and modalities should strengthen others do not. After reading or repeating mor- the connections with other words and add phologically related word pairs with neutral detail to each type of representation of the (e.g., allow/allowable, use/useable) and nonword. Likewise, providing frequent oral rep- neutral patterns (e.g., equal/equality, reg- etitions of the word would increase connec- ular/regularity), the student may be asked tions with the phonological representation which root words changed sounds in the suf- (McKeown, Beck, Omanson, & Pople, 1985). fixed form and which stay the same. The Transparent words (e.g., act/active)maybe words could then be defined and used in sen- a starting point when planning assessment or tences to target the lemma level of the repre- intervention, but it is important that children sentation. not be denied the challenge of less trans- Likewise, calling attention to system- parent examples (e.g., explode/explosion; atic orthographic patterns may be bene- demolish/demolition, muscle/muscular). ficial. An example would be contrasting A gradient system can manage irregu- more frequent spellings of suffixes (e.g., at- larity with enough input. Intermixing tract/attraction) with less common forms degrees of frequency and transparency (e.g., permit/permission). Presentation of tar- gives the instructor a range of variables to gets in authentic texts can help anchor the manipulate. orthographic characteristics of words with Although little specific research has been the semantic and syntactic features (Beck, reported on the most appropriate instruc- McKeown, & Kucan, 2008). An example of tional sequence, we have provided a table this would be selecting books or magazines with a list of the factors discussed in this pa- with many examples of the target suffixes. In- per.OnemightuseTable1toevaluatethe struction may consist of reading the passages, probes or stimuli chosen for an activity or to pointing out the derived forms, and using the perform an item analysis. For example, for a text as a basis for discussing the meaning of child who shows little awareness of affixes, the words. one might begin with short, high-frequency

base words and affixes, with fairly transparent difficulties is still obscure. Children with lan- relationships between the base and derived guage disorders may struggle with underde- word (e.g., walk/walker, jump/jumper). velopment in any of the three parts of lex- In conclusion, models of adult lexical rep- ical representation or with building connec- resentation can inform research and prac- tions across representations. There may be tice with children. Complete lexical repre- subtypes of children who struggle with dif- sentations incorporate all three parts: seman- ferent parts of the representation, but not tic, phonological, and orthographic. These others, although subtypes have not yet been parts may not all develop at the same time identified. Finally, there are aspects of the or the same rate, leaving some words with representation that researchers are still try- strong phonological but weak semantic repre- ing to understand, such as the role of lin- sentations, or strong orthographic but weak guistic prosody. Linguistic awareness, in this phonological representations. Frequency af- case morphological awareness, is awareness fects each type of information. Morpheme of the linguistic system at a particular point units emerge from this system as a func- in development. The lexical system, includ- tion of affix productivity for typically devel- ing lexical morphology, is dynamic; thus, it oping children throughout early and mid- is imperative that researchers and practition- dle childhood. Although areas of weakness ers understand its theoretical architecture. are becoming clearer for children with pri- Hopefully, this paper can contribute to that mary language disorders, the exact source of understanding.

REFERENCES

Amenta, S., & Crepaldi, D. (2012). Morphological process- Carlisle, J. F., & Katz, L. A. (2006). Effects of word and ing as we know it: An analytical review of morpholog- morpheme familiarity on reading of derived words. ical effects in visual word identification. Frontiers in Reading and Writing, 19, 669–693. Psychology, 3, 232. doi:10.3389/fpsyg.2012.00232 Carlisle, J. F., & Stone, C. A. (2005). Exploring the role Anglin, J. M. (1993). Vocabulary development: A mor- of morphemes in word reading. Reading Research phological analysis. Monographs of the Society for Quarterly, 40, 428–229. Research in Child Development, 58(10), 1–166. Clahsen, H., Sonnenstuhl, I., & Blevins, J. P. (2003). Aronoff, M. (1994). Morphology by itself: Stems and in- Derivational morphology in the German mental lex- flectional classes. Cambridge, MA: MIT Press. icon: A dual mechanism account. In Baayen, H., & Badecker, W., & Caramazza, A. (1989). A lexical distinc- Schreuder, R. (Eds.), Morphological structure in lan- tion between inflection and derivation. Linguistic In- guage processing (pp. 125–155). Berlin: Mouton de quiry, 20(1), 108–116. Gruyter. Beard, R. (1995). Lexeme-morpheme base morphology: Clark, E. V. (1982). The young word-maker: A case study A general theory of inflection and word formation. of innovation in the child’s lexicon. In Wanner, E., SUNY Series in Linguistics. Albany, NY: State Univer- & Gleitman, L. R. (Eds.), Language acquisition: The sity of New York Press. state of the art (pp. 390–426). Cambridge: Cambridge Beck, I. L., McKeown, M. G., & Kucan, L. (2008). Creating University Press. robust vocabulary: Frequently asked questions and Clark, E. (1993). The lexicon in acquisition. New York: extended examples. New York: Guilford Press. Cambridge University Press. Berko-Gleason, J., & Bernstein Ratner, N. (2013). The de- Clark, E., & Cohen, S. R. (1984). Productivity and memory velopment of language (8th ed.). Upper Saddle River, for newly formed words. Journal of Child Language, NJ: Pearson. 11, 611–625. Bloom, P. (2002). How children learn the meanings of Clin, E., Wade-Woolley, L., & Heggie, L. (2009). Prosodic words. Cambridge, MA: The MIT Press. sensitivity and morphological awareness in children’s Bybee, J. (2006). From usage to grammar: The reading. Journal of Experimental Child Psychology, mind’s response to repetition. Language, 82(4), 104, 197–213. 711–733. Cooper, N., Cutler, A., & Wales, R. (2002). Constraints Carlisle, J. F. (1988). Knowledge of derivational morphol- of lexical stress on lexical access in English: Evidence ogy and spelling ability in fourth, sixth, and eighth from native and non-native listeners. Language and graders. Applied Linguistics, 9(3), 247–266. Speech, 45(3), 207–228.

Deacon, S. H., & Bryant, P. (2005). The strength of chil- Griffin, Z. M.,, & Bock, K. (1998). Constraint, word fre- dren’s knowledge of the role of root morphemes in quency, and the relationship between lexical process- the spelling of derived words. Journal of Child Lan- ing levels in spoken word production. Journal of guage, 32, 375–389. Memory and Language, 38, 313–338. Deacon, S. H., Campbell, E., Tamminga, M., & Kirby, Hay, J. (2002). From speech perception to morphology: J. (2010). Seeing the harm in harmed and harm- Affix ordering revisited. Language, 78(3), 527–555. ful: Morphological processing by children in grades Hay, J. B., & Baayen, R. H. (2005). Shifting 4, 6, and 8. Applied Psycholinguistics, 31, 759– paradigms: Gradient structure in morphology. 775. Trends in Cognitive Sciences, 9(7), 342–348. Deacon, S. H., Whalen, R., & Kirby, J. R. (2011). Do doi:10.1016/j.tics.2005.04.002 children see the danger in dangerous? Grade 4, 6, Holliman, A. J., Wood, C., & Sheehy, K. (2010). Sensitiv- and 8 children’s reading of morphologically complex ity to speech rhythm explains individual differences in words. Applied Psycholinguistics, 32, 467–481. reading ability independently of phonological aware- Dell, G. S., & O’Seaghdha, P. G. (1992). Stages of lexical ness. British Journal of Developmental Psychology, access in language production. Cognition, 42(1–3), 26(3), 357–367. 287–314. Jarmulowicz, L. (2002). English derivational suffix fre- Derwing, B. L., & Baker, W. J. (1979). Research on the quency and children’s stress judgments. Brain and acquisition of English morphology. In Fletcher, P., & Language, 81, 192–204. Garman, M. (Eds.), Language acquisition (pp. 209– Jarmulowicz, L. (2006). School-aged children’s phonolog- 223). Cambridge, U.K.: Cambridge University Press. ical production of derived English words. Journal of de Jong, N. H., Schreuder, R., & Baayen, R. H. (2003). Speech-Language and Hearing Research, 49, 294– Morphological resonance in the mental lexicon. In 308. Baayen, R. H., & Schreuder, R. (Eds.). Morphological Jarmulowicz, L., & Taran, V. L. (2007). Exploration of structure in language processing (pp. 65–86). Berlin: lexical-semantic factors affecting stress production in Mouton de Gruyter. derived words. Language Speech and Hearing Ser- Ehri, L. C. (2000). Learning to read and learning to spell: vices in Schools, 38, 378–389. Two sides of a coin. Topics in Language Disorders, Jarmulowicz, L., Taran, V. L., & Hay, S. E. (2007). Third 20(3), 19–36. graders’ metalinguistic skills, reading skills, and stress Faust, M., Dimitrovsky, L., & Davidi, S. (1997). Naming production in derived English words. Journal of difficulties in language-disabled children: Preliminary Speech-Language and Hearing Research, 50, 1593– findings with the application of the tip-of-the-tongue 1605. paradigm. Journal of Speech Language and Hearing Jarmulowicz, L., Taran, V. L., & Hay, S. E. (2008). Lexical Research, 40(5), 1026–1036. frequency and third-graders’ stress accuracy in derived Feldman, L. B. (Ed.). (1995). Morphological aspects of English word production. Applied Psycholinguistics, language processing. Mahwah, NJ: Erlbaum. 29, 213–235. Frost, R., Grainger, J., & Rastle, K. (2005). Current is- Kirby, J. R., Deacon, S. H., Bowers, P. N., Izenberg, L., sues in morphological processing: An introduction. Wade-Woolley, L., & Parrila, R. (2012). Children’s Language and Cognitive Processes, 20(1/2), 1–5. morphological awareness and reading ability. Read- Garlock, V. M., Walley, A. C., & Metsala, J. L. (2001). Age- ing and Writing, 25, 389–410. of-acquisition, word frequency, and neighborhood Levelt, W. J. M. (1999). Models of word production. density effects on spoken word recognition by chil- Trends in Cognitive Sciences, 3(6), 223–232. dren and adults. Journal of Memory and Language, Levelt, W. J. M. (2001). Spoken word production: A the- 45, 468–492. ory of lexical access. PNAS, 98(23), 13464–13471. German, D. J., & Newman, R. S. (2004). The impact of lex- Lewis, D. J., & Windsor, J. (1996). Children’s analysis of ical factors on children’s word finding errors. Journal derivational suffix meanings. Journal of Speech and of Speech, Language and Hearing Research, 47(3), Hearing Research, 39, 209–216. 6224–636. Lieber, R. (2010). Introducing morphology. New York: Gonnerman, L. M., Seidenberg, M. S., & Andersen, E. S. Cambridge University Press. (2007). Graded semantic and phonological similarity Marchman, V. A., & Bates, E. (1994). Continuity effects in priming: Evidence for a distributed con- in lexical and morphological development: A test nectionist approach to morphology. Journal of Ex- of the critical mass hypothesis. Journal of Child perimental Psychology: General, 136(2), 323–345. Language, 21(2), 339–366. doi:http://dx.doi.org/ doi:10.1037/0096–3445.136.2.323 10.1017/S0305000900009302 Goodman, J. C., Dale, P. S., & Li, P. (2008). Marslen-Wilson, W. D., Bozic, M., & Randall, B. (2008). Does frequency count? Parental input Early decomposition in visual word recognition: and the acquisition of vocabulary. Jour- Dissociating morphology, form, and meaning. Lan- nal of Child Language, 35(3), 515–531. guage and Cognitive Processes, 23(3), 394-421. doi: doi:http://dx.doi.org/10.1017/S0305000907008641 10.1080/01690960701588004

McKeown, M. G., Beck, I. L., Omanson, R. C., & Pople, M. Soto-Faraco, S., Sebastian-Galles, N., & Cutler, A. (2001). T. (1985). Some effects of the nature and frequency Segmental and suprasegmental mismatch in lexical ac- of vocabulary instruction on the knowledge and use cess. Journal of Memory and Language, 45, 412– of words. Reading Research Quarterly, 20, 522–535. 432. Meunier, F., & Segui, J. (1999). Morphological priming Storkel, H. L. (2009). Developmental differences effect: The role of surface frequency. Brain and Lan- in the effects of phonological, lexical and guage, 68, 54–60. semantic variables on word learning by in- Munson, B. (2001). Phonological pattern frequency and fants. Journal of Child Language, 36, 291–321. speech production in adults and children. Journal doi:10.1017/S030500090800891X of Speech, Language, and Hearing Research, 44(4), Storkel, H. L., & Morrisette, M. L. (2002). The lexicon 778–792. and phonology: Interactions in language acquisition. Nozari, N., Kittredge, A. K., Dell, G. S., & Schwartz, M. Language, Speech, and Hearing Services in Schools, F. (2010). Naming and repetition in aphasia: Steps, 33, 24–37. routes, and frequency effects. Journal of Memory Stump, G. T. (2005). Word-formation and inflectional and Language, 63(4), 541–559. morphology. In P. ˇStekauer & R. Lieber (Eds.), Hand- Perfetti, C. A. (1992). The representation problem in read- book of word-formation (pp. 49–71). Dordrecht, The ingacquisition.InP.B.Gough,L.C.Ehri&R.Treiman Netherlands: Springer. (Eds.), Reading acquisition (pp. 145–174). Hillsdale, Wagovich, S. A., Pak, Y., & Miller, M. D. (2012). Ortho- NJ: Lawrence Erlbaum Associates. graphic word knowledge growth in school-age chil- Plag, I. (2003). Word-formation in English. New York: dren. American Journal of Speech-Language Pathol- Cambridge University Press. ogy, 21, 140–153. Pylkkänen, L., Feintuch, S., Hopkins, E., & Marantz, Warlaumont, A. S., & Jarmulowicz, L. (2012). Care- A. (2004). Neural correlates of the effects givers’ suffix frequencies and suffix acquisition by of morphological frequency and family size: language impaired, late talking, and typically devel- An MEG study. Cognition, 91, B35–B45. oping children. Journal of Child Language, 39(5), doi:10.1016/j.cognition.2003.09.008 1017–1042. Rastle, K., & Davis, M. H. (2008). Morphological de- Whalley, K., & Hansen, J. (2006). The role of composition based on the analysis of orthogra- prosodic sensitivity in children’s reading develop- phy. Language & Cognitive Processes, 23, 942–971. ment. Journal of Research in Reading, 29(3), doi:10.1080/01690960802069730 288–303. Reichle, E. D., & Perfetti, C. A. (2003). Morphology in Windsor, J. (2000). The role of phonological opacity in word identification: A word-experience model that reading achievement. Journal of Speech, Language, accounts for morpheme frequency effects. Scientific and Hearing Research, 43, 50–61. Studies of Reading, 7(3), 219–237. Windsor, J., & Hwang, M. (1999). Derivational suffix Seidenberg, M. S., & Gonnerman, L. M. (2000). Explaining productivity for students with and without language- derivational morphology as the convergence of codes. learning disabilities. Journal of Speech, Language, Trends in Cognitive Sciences, 4(9), 353–361. and Hearing research, 42, 220–230.