interference effects
If readers analyse the morphemic constituents of printed words, then they might have trouble deciding that a stimulus like quickify (i.e., comprising an existing stem and affix) is not a word. In a seminal study, Taft and Forster (1975) observed that nonwords built of morphemic units (e.g., dejuvenate) are rejected more slowly in lexical decision than nonwords that do not comprise morphemic units (e.g., depertoire). In this example, juvenate is a bound stem meaning “young” (as in juvenile), while pertoire is not a bound stem. Bound stems are stems that cannot occur on their own. These findings were important because they suggested that morphemic analysis is a phenomenon that arises prior to lexical identification.
The morpheme interference effect has been replicated and extended across many studies (e.g., Burani, Dovetto, Thornton, & Laudanna, 1997; Caramazza, Laudanna, & Romani, 1988). Replicating the effect using English derived nonwords, Crepaldi et al. (2010) reported that morphemically structured nonwords such as gasful are rejected more slowly in lexical decision than nonwords without morphological structure such as gasfil. In an extension to these findings, they reported that the interference effect vanishes when the constituent morphemes are transposed (e.g., comparing fulgas to filgas), suggesting that sensitivity to affixes may be position specific (Crepaldi et al., 2010). Dawson, Rastle and Ricketts (2018) used stimuli from Crepaldi et al. (2010) to test whether the morpheme effect arises in participants of different ages: children (aged 7–9), young adolescents (aged 12–13), older adolescents (aged 16–17), and adults. They found a morpheme interference effect on accuracy in all groups, but reported an effect on response time only in adults and older adolescents. These findings provide evidence that relatively young readers are sensitive to morphological structure, but that knowledge of these important indicators of meaning builds through the process of reading acquisition (see also Castles & Nation, this volume).
Morphological priming effects
If morphologically complex words are analysed in terms of their constituents (e.g., darkness ‐> [dark] + [‐ness]), then a morphologically complex prime (e.g., darkness) should facilitate subsequent recognition of its constituent stem (e.g., DARK) relative to an unrelated control prime (e.g., darkness – DARK vs. fullness – DARK). Priming effects of this nature have been observed extensively, across different priming paradigms and across different languages (see, e.g., Amenta & Crepaldi, 2012; Rastle & Davis, 2008, for reviews).
Morphologically complex words are morphologically related to their stems (e.g., cleanly, clean), but they are also semantically and orthographically related to their stems. Early studies thus considered the possibility that morphological priming effects might simply reflect a combination of orthographic and semantic priming (Devlin, Jamison, Matthews, & Gonnerman, 2004). However, Rastle and colleagues (2000) provided evidence against this possibility in a series of masked priming experiments. Masked priming is an experimental paradigm in which primes are presented so briefly that they cannot be perceived consciously (Forster & Davis, 1984). These researchers showed that masked priming effects for morphologically related items (e.g., darkness – DARK) were significantly greater than those for semantic relatedness (e.g., cello – VIOLIN), orthographic relatedness (e.g., tapestry – APE), or combined semantic and orthographic relatedness (e.g., ghost – GHOUL). These findings indicate that there is a special relationship between morphologically related words that goes beyond their combined semantic and orthographic relatedness.
The question that has provoked most interest over the past 20 years of morphological priming research concerns the role of semantic transparency in morphological decomposition. Marslen‐Wilson and colleagues (1994) reported a series of cross‐modal priming experiments (spoken primes, visual targets) showing that morphologically complex words prime recognition of their stems, but only when there is a semantic relationship between them (e.g., departure – DEPART, but not department – DEPART). These findings led them to conclude that morphologically complex words are represented in terms of their constituents only when the meaning of the whole word can be computed from the meanings of the parts. Visual priming studies using fully visible primes also observed this pattern (Rastle et al., 2000). However, visual masked priming studies in French (Longtin, Segui, & Hallé, 2003) and English (Rastle, Davis, & New, 2004) yielded different results. These studies showed that briefly presented, morphologically structured primes facilitate recognition of their stems, irrespective of whether there is a meaningful relationship between the words. This latter class of item includes morphologically related pairs that have lost their connection over time (e.g., witness – WIT) and pseudo‐morphological pairs that never had a meaningful connection (e.g., corner – CORN). Critically, priming effects for these items were shown to be greater than those for orthographically related items without an apparent morphological structure (e.g., brothel – BROTH).
These initial findings were interpreted as evidence for a form of morpho‐orthographic decomposition based on the mere appearance of morphological structure (Rastle et al., 2004). They went on to be replicated across a number of studies in different Indo‐European languages (see Rastle & Davis, 2008, for review), bringing them into line with discoveries concerning masked morphological priming in Semitic languages (Boudelaa & Marslen‐Wilson, 2001; Frost, Forster, & Deutsch, 1997). Further research established that this form of decomposition arises rapidly and prior to lexical identification. Longtin and Meunier (2005) reported that the masked priming effects yielded by morphologically structured nonwords (e.g., darkism – DARK) were of the same magnitude as those observed for semantically transparent derived words (e.g., darkness – DARK). This was the case irrespective of whether the nonword primes were syntactically legal (e.g., quickify – QUICK) or illegal (e.g., sportation – SPORT). Finally, research indicates that this form of decomposition is robust to regular orthographic alterations that characterize morphologically complex words (McCormick, Rastle, & Davis, 2008). Masked morphological priming effects are observed when primes cannot be parsed perfectly into constituents because of a missing e (e.g., adorable – ADORE), a shared e (e.g., writer – WRITE), or a reduplicated consonant (e.g., metallic – METAL). These effects are of a similar magnitude to those using primes that can be parsed perfectly into morphemic constituents, and also arise when primes do not share a meaningful relationship with stems (e.g., badger – BADGE; McCormick et al., 2008).
Research investigating morphological priming effects in children of different ages has suggested that morpho‐orthographic segmentation may reflect a form of reading expertise. Beyersmann and colleagues reported that English speaking (Beyersmann, Castles, & Coltheart, 2012) and French speaking (Beyersmann, Grainger, Casalis, & Ziegler, 2015) primary school children show robust masked morphological priming effects, but only when morphological primes have a semantic relationship with targets. Beyersmann et al. (2012) found no evidence of priming based on the appearance of morphological structure (e.g., corner – CORN) in children between the ages of 8 and 10. Similar findings were observed for Hebrew primary school children between the ages of 9 and 12: Robust masked morphological priming when primes were semantically related to targets, but weak or null priming when they were not (Schiff, Raveh, & Fighel, 2012). These findings suggest that perhaps morpho‐orthographic segmentation is a form of analysis that is acquired only after extensive reading experience. This conclusion is consistent with work by Andrews and Lo (2013) investigating individual differences in masked priming amongst university students. They found that the morpho‐orthographic pattern is modulated by vocabulary and spelling ability, with pseudo‐morphological priming being stronger in people with good spelling skills.
Theoretical Accounts of Morphological Processing
The body of data previously described is consistent with the notion of a hierarchically organized
