Interestingly, in order to learn morphological regularities across the spelling‐meaning mapping, distributed‐connectionist models must be provided with an input representation that allows them to recognize orthographic patterns that occur repeatedly in semantically similar words (Rastle & Davis, 2008). This is nontrivial because these orthographic patterns occur in different positions within words: for example, dislike , unlike , likely, likelihood. Thus far, this problem has been dealt with by providing distributed‐connectionist models (e.g., Plaut & Gonnerman, 2000; Rueckl & Raveh, 1999) with a representation that is pre‐segmented. Essentially, in providing these models with an input representation that is already segmented into morphemes, these modelers have assumed that morpho‐orthographic segmentation has occurred (see Rastle & Davis, 2008, for discussion). It seems possible that learned orthographic representations themselves might be structured morphologically, and Rastle and Davis (2008) described three potential mechanisms that might drive this form of orthographic learning. However, computational work is needed to evaluate these proposals.
More recently, researchers have begun to use alternative computational approaches to understand morphological processing. Naïve discriminative learning models simulate the relationship between orthographic patterns and aspects of meaning (e.g., grammatical class) and propose that these relationships may explain some morphological effects (Baayen et al., 2011). Similarly, distributional semantic models suggest that the functions of morphemes and the constraints that govern their combination with stems may be captured through large‐scale analysis of text (Marelli & Baroni, 2015). Like distributed‐connectionist models, these approaches to morphological processing eschew the notion of explicit morpheme representations, and instead ascribe morphemic effects to overlap in orthographic and meaning representations. Thus, these models may also struggle to account for morpho‐orthographic segmentation effects. However, proponents of these models emphasize that there is considerable heterogeneity amongst morphologically structured words that are not clearly related to their stems. Some of the words used in the relevant experiments (e.g., Rastle et al., 2004) have a subtle relationship to their stems; for example, an initiative that is fruitless might be described as one that didn’t bear any fruit. Likewise, while the word cryptic is unrelated in meaning to crypt, the suffix ‐ic functions in the appropriate manner grammatically (i.e., forming adjectives from nouns). However, evidence that genuinely pseudo‐morphological words such as corner are segmented (e.g., Longtin et al., 2003) would seem to pose a challenge for these models.
Mechanisms for Acquiring Morphological Knowledge
Evidence from skilled adult readers suggests that they have acquired morphological knowledge that is applied to any morphologically structured stimulus, irrespective of its lexical status. One question that arises immediately is why readers should acquire knowledge that gives rise to misleading segmentations for stimuli like corner ‐> {corn} + {‐er}. The answer goes back to the nature of the writing system, at least in the case of English spelling. Specifically, the analyses by Ulicheva et al. (2020) revealed that words like corner that appear erroneously to be morphologically complex are very rare in English spelling; typically, these words would be spelled in a way that does not make use of the suffix ‐er (e.g., martyr, sulphur, fibre). It is therefore unsurprising that readers should learn about such systematicity (i.e., ‐er as an affix), and capitalize on it to enable rapid, skilled reading (Rastle, 2019b).
Research has only just begun to understand the mechanisms that underpin the acquisition of morphemic knowledge in written language. Affix learning presents a particularly interesting challenge because affixes do not occur in their own right, and affixes are rarely taught explicitly as children learn to read. Therefore, the function of these letter groups must be discovered through experience with words that contain particular affixes. Through experience with words such as builder, banker, and teacher, one might learn that ‐er is agentive: someone who does {verb}. Knowledge of ‐er would permit the learner to generalize; for example, understanding that a tweeter is someone who tweets. The challenge may be more difficult in cases where affixes are used more frequently in idiosyncratic ways (e.g., hearth has nothing to do with hear), or where the nature of the transformation is less systematic (e.g., cyclist, racist, and Baptist all refer to an agent but in different ways). Perhaps the challenge of affix learning explains why morpheme effects in online reading tasks are typically observed only as children approach and move through adolescence (e.g., Beyersmann et al., 2012; Dawson et al., 2018).
Recent research has conceptualized learning about affixes as a statistical learning problem (Lelonkiewicz, Ktori, & Crepaldi, 2020; Ulicheva et al., 2020). Lelonkiewicz et al. (2020) familiarized participants with a lexicon of novel words printed in a novel orthography. Each novel word contained an affix character at the beginning of the word or at the end. Following familiarization, participants were more likely to attribute a previously unseen word to the lexicon if it had an affix, and if the affix occurred in the correct position. These findings show that readers capture statistically salient chunks in language input and use this information in their analysis of new words. Ulicheva et al. (2020) selected nonwords with English suffixes that varied continuously in the consistency with which they reflect grammatical class. Performance in tests of reading, spelling, and meaning judgment showed a high degree of sensitivity to this meaningful information; and further, that this sensitivity was graded as a function of the consistency of the relationship between suffix spelling and grammatical category. Participants were more likely to judge a nonword like sedgeness as a noun than an adjective; and their eyes were more likely to regress back onto the nonword when it occurred in an adjective context than a noun context. Likewise, participants were more likely to spell a spoken word like /sEdZnƏs/ using ‐ness when it occurred in a noun context than a verb context (see also Treiman, Wolter, & Kessler, 2020 for similar findings). Together, these results suggest that morphological knowledge may increasingly mirror the writing system as readers gradually accumulate reading experience.
Thus, it seems that morphemes are salient statistical patterns relating spelling to meaning, and that knowledge of these patterns is acquired through reading experience. Tamminen and colleagues (2015) investigated how this learning might arise in a series of artificial learning experiments. They trained adults on novel words with morphological structure (e.g., sleepnule, teachnule, buildnule), and then assessed participants’ ability to generalize knowledge of the novel affix (e.g., applying knowledge of ‐nule to understand the untrained novel word sailnule). Tamminen et al. (2015) reported that knowledge of the novel affixes depended upon two aspects of the training set: 1) novel affixes needed to occur frequently in combination with multiple stems; and 2) novel affixes needed to occur with a consistent meaningful function. These findings suggest that those affixes that occur in a semantically consistent manner, and in combination with multiple stems, may be those most likely to be represented in a context‐independent manner, and segmented from their stems in visual word recognition. These conclusions are consistent with findings from skilled readers suggesting that morphological effects are largest in these conditions (e.g., Ford, Davis, & Marslen‐Wilson, 2010).
Together, such recent findings suggest that we learn morphological statistical regularities through experience (Lelonkiewicz et al., 2020; Tamminen et al., 2015), and that the morphemic knowledge of skilled readers is graded in a manner that reflects the strength of these learned statistical regularities (Ulicheva et al., 2020). However, Treiman et al. (2020) raised an intriguing challenge. They tested adults’ sensitivity to the relationship between suffix spellings and grammatical category in a nonword spelling task, using spoken nonwords containing word‐final /Əs/ and /Ik/. They found that adults were more likely to use spellings ‐ous