The Science of Reading. Группа авторов

Чтение книги онлайн.

CHAPTER FIVE Word Recognition III : Morphological Processing

      Kathleen Rastle

      The past 50 years of research on visual word recognition has been dominated by the view that the primary challenge of reading is to decode the printed word to a spoken language representation. Thousands of articles have been devoted to understanding how skilled readers compute sound‐based (phonological) representations from printed words (e.g., Coltheart, Rastle, Perry, Langdon, & Ziegler, 2001; Plaut, McClelland, Seidenberg & Patterson, 1996), how phonological decoding constrains word identification (e.g., Lukatela & Turvey, 1994), the time‐course of phonological decoding (e.g., Rastle & Brysbaert, 2006; Rayner et al., 1995), and whether it is obligatory (e.g., Frost, 1998). Likewise, research on learning to read has focused on how phonological decoding ability influences reading success (e.g., Melby‐ Lervåg, Lyster, & Hulme, 2012), how inconsistency in the relationship between spellings and sounds affects learning to read (e.g., Seymour, Aro & Erskine, 2003), and how children should be taught to relate visual symbols to sounds (see e.g., Castles, Rastle, & Nation, 2018). This body of research has demonstrated unambiguously that the computation of phonological representations plays a vital role in skilled reading and learning to read (see Brysbaert, this volume).

Far less attention has been given to the role of morphological information in skilled reading and reading acquisition. Morphemes are the smallest unit of meaning, typically comprising stems (e.g., luck) and affixes (including prefixes, e.g., un‐, and suffixes, e.g., ‐y). The vast majority of words in English and in other languages are built by combining a small number of stems with other stems (e.g., potluck) or with prefixes and suffixes (e.g., unlucky, unluckiest, luckily). It is obvious that an efficient reading system should treat these different words as similar to one another, and this was recognized in some of the earliest work in modern reading research (Taft & Forster, 1975). However, it was not until many years later that the field began to understand how morphological information is analysed in visual word recognition, why these processes are important in skilled reading, and how this information is learned. This chapter describes recent advances in our understanding of these questions and identifies questions for future research.

Morphemes as “Islands of Regularity”

Visual symbols in alphabetic writing systems are used to represent the sounds of language. Thus, printed words such as dig, dog, and dip that look similar also tend to sound similar, and yet they have very different meanings. Indeed, if we consider only the stems within an alphabetic writing system, we see that the relationship between spelling and sound is largely systematic (similar spellings yield similar sounds) and the relationship between spelling and meaning is largely arbitrary (similar spellings yield unrelated meanings).¹ This dichotomy provides a theoretical basis for research focused on how readers (learn to) transform visual symbols back into sounds during reading. However, the situation changes substantially when we consider words built from multiple morphemes (Plaut & Gonnerman, 2000), including words with inflections (e.g., spells, spelled, spelling), derivations (e.g., misspell, speller), and compounds (e.g., spellcheck, spellbound). In these cases, stems occur in words with similar meanings (e.g., lucky, un lucky, luckily), and affixes alter the meanings of words in highly predictable ways (e.g., repaint, relock, reload). Morphemes therefore provide “islands of regularity” in an otherwise arbitrary mapping between spellings and meanings (Rastle, Davis, Marslen‐Wilson, & Tyler, 2000).

Morphemes differ in the consistency with which they link spellings to meanings. Stems can occur in morphologically structured words in ways that do not convey the morphemes’ combined meanings. For example, the combinatorial meaning of the word whisker is “someone who whisks,” yet the word is almost always used to refer to a long hair growing from the face of a mammal, or infrequently, to a very tiny amount (e.g., she won the race by a whisker). Likewise, while some affixes are highly consistent across all words in which they occur (e.g., ‐ness virtually always transforms an adjective to a noun), others are not (e.g., ‐age in vicarage, postage, and bandage). Even where morphologically complex words are consistently related to the meanings of their stems, the specific way that affixes transform the meanings of stems can vary. For example, while the words snowman, patrolman, and milkman are all men (of sorts), the ‐man component functions quite differently in each word: a snowman is a man made of snow, but a milkman is not made of milk; a patrolman patrols, but a snowman does not snow.

      These examples suggest that the regularity that morphology brings to the spelling‐meaning mapping is graded rather than all‐or‐none. However, research has only recently begun to quantify the nature of this regularity. Emerging research has revealed a striking relationship between English suffixes and grammatical category, with most suffixes being highly diagnostic of this aspect of meaning (Ulicheva et al., 2020). For example, the suffix ‐ous virtually always denotes adjective status (e.g., nervous, envious, glamorous), and adjectives ending in the sound sequence /Əs/ must be spelled ‐ous (see also Berg & Aronoff, 2017). The sound sequence /Əs/ is virtually always spelled another way if the word is not an adjective (e.g., service, princess, haggis). This relationship means that a superficial inspection of a suffixed English word already reveals an important aspect of its meaning: whether it is an object, property, or act.

      Intriguingly, at least in English, it seems that regularity between spelling and sound is sacrificed to express these powerful regularities between spelling and meaning. If English spelling offered a perfect transcription of the sounds of words, then one might spell the words busted, snored, and kicked as bustid, snord, and kict (Rastle, 2019a). Yet, the spelling system admits spelling‐sound inconsistency (i.e., the spelling ‐ed can be pronounced in many ways) in order to transmit an important piece of information about meaning (i.e., the spelling ‐ed indicates the past). This trade‐off is ubiquitous in English spelling (Ulicheva et al., 2020), and means that meaningful morphological information is highly visible, significantly more than in spoken English (Rastle, 2019a; Ulicheva et al., 2020). Further analyses are required to quantify the strength of the relationship between spelling and meaning in other languages.

Morphology and the Spelling‐Meaning Mapping

Theoretical models propose that readers map orthographic representations onto meaning via two pathways. One pathway relates orthographic representations to semantic representations directly, while the other achieves this mapping via phonological representations (Harm & Seidenberg, 2004; see Seidenberg, this volume). Meta‐analyses of neuroimaging data are consistent with these models, revealing dorsal (spelling‐sound‐meaning; see Figure 5.1) and ventral (spelling‐meaning) pathways for reading (Taylor, Rastle, & Davis, 2013; see also Hoffman, Ralph, & Woollams, 2015; Yeatman, this volume). Though there is wide agreement that phonological decoding plays a vital role in the computation of meaning (particularly during the initial stages of learning to read), it is also recognized that skilled reading ultimately requires rapid, direct computation of meaning from a printed stimulus. The acquisition of this spelling‐meaning mapping is sometimes called orthographic learning (Castles & Nation, 2006; Castles & Nation, this volume).