recently, research has sought to connect imaging results with other important issues in comprehension research – reading narrative versus expository texts (Aboud et al., 2019), local versus global comprehension (Egidi & Caramazza, 2013), and responses to differences in text coherence (Helder et al., 2017). Beyond identifying brain regions are proposals for how other brain networks interact with language areas during comprehension (Hagoort, 2019).
One reason for the limited contribution of imaging studies to cognitive explanations of comprehension is the low temporal resolution of the fMRI BOLD signal. In our streams metaphor, this signal shows the slower currents (and down‐stream results of the fast currents). Because it reflects the flow of oxygenated blood to brain areas, the BOLD signal develops slowly, over seconds, whereas incremental comprehension processes occur over milliseconds. Further, fMRI images provide only the strength of the correlation between the expected and observed ratios of oxygenated blood during a reading task. These factors limit the interpretations of the underlying processes of comprehension. Although this is also true for imaging studies of word identification, the results there show enough convergence to be connected to theory and behavioral results. A promising approach for comprehension is to connect the brain areas identified in fMRI to the interpretation of ERP components that reflect meaning and integration processes (Brouwer & Hoeks, 2013)
Disruptions in the reading comprehension system.
Text comprehension sometimes fails. Disruptions in comprehension can arise from processes within the comprehension system (see Figure 1.3) and the linguistic and conceptual knowledge systems (Figure 1.1) they depend on. Pressure points within the comprehension system emerge with specific demands. For example, syntactic complexity and ambiguity threaten sentence comprehension; a failure to make a required inference threatens sentence comprehension and text coherence; and a text requiring conceptual knowledge not accessible to the reader threatens global text comprehension. However, the disruptions are not so conveniently localized as these observations imply. Disruptions to word‐identification processes do not end there, but can spread to result in disruptions to sentence and global text comprehension. The lexicon is a particularly important pressure point in the system, sending its output – multiple levels of information about the word being read (context‐relevant meaning and form class, potential for argument role filling and for referential specification, and more) – to the comprehension system. A disruption at this point has consequences “downstream.” This much is a description of what can go wrong for any reader for a specific text. A skilled reader can repair comprehension failures.
The research field, however, has been concerned with individual differences in comprehension failures – “poor comprehenders” (e.g., Cain & Oakhill, 2007; Hulme & Snowling, 2011; Oakhill & Yuill, 1996). Rather than address them here, we instead emphasize that, aside from language impairments – which can affect multiple linguistic knowledge sources and processes – the ordinary range of reading skill does not include individuals who have dysfunctional processing subsystems of comprehension. Rather than faulty processes such as inference making or a lack of comprehension monitoring, activating relevant knowledge may be the main issue. Knowledge of word forms (orthography), word meanings (vocabulary knowledge), knowledge of language machinery (syntax, morphology), and conceptual knowledge combine to support successful comprehension. Readers for whom the word‐identification system works efficiently, but nevertheless consistently fail in comprehension may lack sufficient critical knowledge or fail to have the knowledge activated strongly or quickly enough to engage in inference making and comprehension monitoring when reading the text (Nation, 2005). One of the points of progress in the study of individual differences has been an increased recognition of the need to assess reading components (fluent word identification, vocabulary, relevant knowledge) in order to identify some other targeted components of the reading comprehension system (see Cain, this volume).
Advance 3. Toward a More Universal Science of Reading
The advances discussed so far come largely from research on reading in alphabetic writing systems, mainly English. Indeed, the two routes of the DRC model were intended to capture an orthographic property of English – its inconsistent mappings between letters and phonemes.2 Reading science needed to address reading more broadly and a step in that direction came from the comparative analysis of orthographies by Katz and Frost (1992). “Orthographic depth” orders orthographies according to the tradeoff they make between coding speech components and meaning. Thus, among alphabetic writing systems, Welsh and Finnish are shallow (consistent mappings to phonemes), Czech and Italian only slightly less so, with English at the deep end. Moving beyond alphabetic writing toward a more universal perspective, orthographic depth was extended to nonalphabetic writing, for example, the consonant‐based Abjad system and morpho‐syllabic Chinese.
The single scale of orthographic depth, however, fails to reflect the design principles that separate other systems from alphabets. Explicit attention to these principles was the basis of the Universal Phonological Principle (Perfetti et al., 1992; Perfetti, 2003) that reading words universally involved phonology at the lowest level allowed by the writing system and the psycholinguistic grain size hypothesis (Ziegler & Goswami, 2005), which focused on where the writing system makes its connection within the phonological hierarchy. Where this connection is made – phoneme, syllable, word – has consequences for reading development. Despite increasing recognition of writing system differences, Share (2008) correctly argued that the dominant role of English in reading research had resulted in research questions and models of reading that might not apply to other systems.
More recent progress from research across languages and writing systems was the focus of two volumes on learning to read (Verhoeven & Perfetti, 2017a) and dyslexia (Verhoeven et al., 2019) The conclusions include universals across 17 languages in learning to read, along with specific features of languages, writing systems, and instruction (see chapters by Caravolas, McBride et al., and Nag, this volume).
Table 1.2 Examples of adaptations of writing systems to language features
| Language | Adaptations of the writing system to features of the language |
|---|---|
| Chinese | Small number of syllables with tones. Extensive syllable homophony makes alphabets and syllabaries less adaptive. Characters map onto syllable morphemes and can distinguish between homophones. |
| Japanese | Agglutinative language. Many multisyllabic words and small number of syllables with open structure. Japanese syllabaries (Kana) are adaptive to these factors, but historical borrowing of Chinese supports dominant Kanji character system. |
| Finnish | Relatively small number of phonemes and long words of several syllables. Complex inflectional morphology. Highly consistent alphabetic orthography supports decoding of multisyllabic, multimorpheme words |
| English | Phonological complexity and many syllables make an alphabet efficient. Simple inflectional morphology favors morphophonemes and morpheme spellings. A mismatched letter‐to‐phoneme ratio keeps phonological consistency low. |
Cross‐language comparisons suggest that writing systems show accommodation to the properties of the language they represent (Frost, 2012; Perfetti & Harris, 2013; Seidenberg, 2011). Illustrating this possibility, Table 1.2 summarizes four of the orthographies reviewed in Verhoeven and Perfetti (2017a). Two examples of alphabetic writing suggest accommodation to phoneme inventories,