1
Title: Insights into reading processes through investigating diversity
Heather Winskel Southern Cross University, Coffs Harbour, Australia
Running head: Insights into reading Word count: 6600
Address for correspondence: Dr Heather Winskel Psychology School of Human and Health Sciences Southern Cross University Hogbin Drive Coffs Harbour NSW 2450 Australia
[email protected] tel.: +61 2 6659 3381
2
Abstract The general goal of reading is to obtain meaning from what is written on the page or screen. The numerous scripts around the world used to write different languages vary in terms of what aspects of language are encoded in the written form. The aim of the current review is to examine some recent research on lesser studied orthographies, in particular Thai, to illustrate the benefits of comparative studies in building a greater understanding of what processes are common or distinct when reading diverse writing systems. Three areas of reading are focused on where there is substantial variation across orthographies: (1) reading with and without interword spaces, (2) flexibility in letter position coding and initial letter position advantage, and (3) the role of lexical tone when reading. In order to effectively read a script, readers need to attend to the critical features of the script that interface with the particular language of the speaker. For example, in scripts with interword spaces, these salient visual cues form clear word boundaries, whereas in unspaced scripts other orthography-specific cues need to be identified and utilised. Comparative research shows that letter position encoding varies across languages, which is shaped by the characteristics of the orthography. Finally, research on Thai and Chinese indicates that tone takes a secondary role in comparison to segmental information (consonantal and vowel information) and appears to be processed at a later stage.
3
Introduction When we read our general aim is to understand, comprehend and obtain meaning from what is written on the page or screen (Perfetti & Dunlap, 2008). Skilled readers appear to achieve this goal relatively effortlessly. However, for a child learning to read it can be quite a challenge to learn how “to crack the code” or learn how their particular writing system maps onto their spoken language (Perfetti & Zhang, 1995). This fundamental process involves converting written symbols into words, morphemes and their underlying concepts (Perfetti & Dunlap, 2008). The numerous scripts around the world substantially differ in how and what aspects of language are encoded in the written form. For example, the predominant level represented might be the word or morpheme as in logographic Chinese, the syllable (or mora) as in Japanese or the phoneme as in alphabetic English, German or Finnish (Byrne, Samuelsson, & Olson, 2014). Even though a writing system may predominantly represent a
particular level of language, it also consistently represents other levels of language. For example, English and other European alphabets represent phonemes and also syllables, morphemes and words. An overriding question is to what extent reading and reading processes are similar or different when reading these quite distinct writing systems. Traditionally, reading research has focused on Roman script and a small number of European languages, in particular English. However, there has been a rapidly growing interest in investigating a broader range of languages and scripts, which is essential if we are to delineate between universal and orthography-specific processes as well as build more comprehensive and representative universal models of reading (see Frost, 2012). This comparative research may also challenge various assumptions that have been made based on a limited group of languages and scripts (Frost, 2012; Share, 2008). One of the much debated questions is whether universals in reading exist and can be specified and applied to all writing systems and orthographies. Coltheart and Crain (2012)
4
believe that there are no such universals of reading and consequently there cannot be a universal model of reading. On the other hand, Frost (2012) advocates that in order to construct a universal model or theory of reading, we need to focus on what is invariant in orthographic processing across writing systems. This is quite a challenge since there is such diversity across orthographies and writing systems. Although the general goal of reading is to form a mental representation of text from decoding visual symbols, the processes involved in attaining this goal may substantially differ due to the large variation in the characteristics of writing systems and orthographies. The aim of the current review is to examine some recent research on lesser studied orthographies, in particular Thai, to illustrate the benefits of comparative studies in building a greater understanding of reading mechanisms and processes. Thai forms a useful comparison with the much studied Roman script, as it is also alphabetic but has marked differences. In this short review, three aspects of reading where there are substantial differences in writing systems and orthographic coding will be examined. The three areas that are focused on are: (1) reading with and without interword spaces, (2) flexibility in letter position coding and initial letter position advantage, and (3) the role of lexical tone when reading. First, it is important to understand some of the distinctive characteristics of Thai prior to examining research in these three areas. Thai, in contrast to Roman script, has a nonlinear configuration as it has vowels that can be written above, below, or to either side of the consonant as full letters or diacritics. Several vowels precede the consonant in writing but phonologically follow it in speech (e.g., แบน ‘flat’ is spoken as /bɛ:n/), whereas other vowels are spoken in the order that they are written, as occurs in English (e.g., บาท ‘Baht’ is spoken as /ba:t/). This nonalignment of vowels is a characteristic shared by other Brahmi-derived scripts (e.g., Devanagari, Kannada, Sinhala and Burmese). Thai is also a tonal language, which is a characteristic it shares with other regional neighbours (e.g., Chinese, Burmese, Lao and
5
Vietnamese) (for further discussion refer to Winskel, 2014; Winskel, Padakannaya, & Pandey, 2014). Thai has five tones conceptualised as high, mid, falling, rising, and low and four tone markers that orthographically occur above the initial consonant in the syllable, for example ขาว /khã:w4/ (white) rising tone, ข่าว /khà:w/ (news) low tone, and ข้าว /khâ:w2/ falling tone (rice). The tone determination of a syllable is influenced by a combination of the class of initial consonant, the type of syllable (open or closed), the tone marker, and the length of the vowel (for further detail refer to Winskel, 2014; Winskel & Iemwanthong, 2010). Thai has a high degree of consistency in mapping between phonemes and graphemes but there are multigrapheme to phoneme correspondences, which mean that there are alternative spellings for some phonemes (e.g., ซ ศ ษ ส /s/). Thai also does not normally have interword spaces.
(1) Reading with and without interword spaces Thai similar to Chinese, Japanese, Lao, Khmer, Balinese, Sundanese, Tibetan and Myanmar does not have salient interword spaces that demarcate word boundaries. Interword spaces are highly salient perceptual features that delineate between the beginning and end of a word, as illustrated by the current text. The lack of these salient visual word segmentation cues implies that during normal reading, there is a degree of ambiguity in relation to which word a given letter belongs to (an example to illustrate this difficulty: คุณพ่อของฉันชอบรับประทาน อาหารที่มีรสจัด). Even some European languages such as Finnish and German have long unspaced compound words. Interword spaces form clear parafoveal word segmentation cues so that saccades can be readily targeted close to the centre or Optimal Viewing Position (OVP) (O’Reagan, 1990) of the next word to be read. In scripts that do not have these salient visual cues, other strategies need to be utilised by the reader to identify word boundaries and demarcate where a word begins and ends.
6
Not surprisingly in skilled Roman script readers, who have had a lifetime of reading spaced text, when these salient boundary cues are removed eye movement control and word identification are substantially disrupted (Morris, Rayner, & Pollatsek, 1990; Rayner, Fischer, & Pollatsek, 1998; Spragins, Lefton, & Fischer, 1976). In English, reading is typically slowed down by 30-50%, disrupting both the way the eyes move through the text and the word identification process. The masking or removal of interword spaces has been found to be more deleterious to the reading of (relatively unfamiliar) low-frequency words than when reading length-matched (relatively familiar) high-frequency words (Rayner et al., 1998). These results are interpreted as indicating that removal of spaces interferes with word identification. Visuomotor control was also disrupted as indicated by substantial changes in the spatial distribution of incoming saccades or landing site distributions over target words. Readers tended to land a bit to the left of the middle of the word, whereas when spaces were removed they tended to land closer to the beginning of the word. But what happens in readers who are accustomed to reading a script without these perceptually salient interword spaces, for example Chinese, Japanese or Thai? One possibility is that insertion of these salient perceptual cues may facilitate reading, or alternatively, as these readers have had a lifetime of experience reading text without interword spaces, adding spaces may have a deleterious effect on their reading. With the aim of addressing this question, the eye movements of Thai-English bilinguals were examined when reading both Thai and English with and without interword spaces, and their eye movements were compared when reading English to those of English monolinguals (Winskel, Radach, & Luksaneeyanawin, 2009). The frequency of critical target words in the sentences was manipulated so that information could be gained about the effect of spacing on word identification (Rayner, 1998; Radach & Kennedy, 2004).
7
Several findings in this study (Winskel et al., 2009) suggested that spacing facilitated later word processing rather than word targeting or early lexical segmentation. The refixation measures (gaze duration and total fixation time) were significantly shorter in duration on the target words in the spaced than unspaced sentences, but notably first fixation durations were not different. First fixation landing positions and landing site distributions were also not influenced by the spacing manipulation. First fixation landing positions in both the spaced and unspaced condition were just left of word centre or what is termed the Optimal Viewing Position (OVP) (O'Reagan, 1990). Thus, these results gave qualified support for the view that interword spaces have a facilitative function, as word processing was found to be facilitated, but on the other hand eye guidance (word targeting and lexical segmentation) was neither facilitated nor disrupted by the insertion of interword spaces. Similar findings were found in a subsequent study, in which initial and internal letter transpositions of target words as well as interword spacing were manipulated (Winskel, Perea, & Ratitamkul, 2012). These results on Thai were similar in some respects to what has been found when reading Japanese sentences composed of mixed Hiragana-Kanji script with interword spaces inserted (Sainio, Hyönä, Bingushi, & Bertram, 2007). Japanese has three distinct scripts: Hiragana and Katakana (syllabic scripts), and Kanji (an ideographic script originating from Chinese). For the mixed Kanji-Hiragana text (ideographic with syllabic), Sainio et al. (2007) found a tendency for spaced text to be read slower than unspaced text, although this difference did not reach significance. Results on reading Hiragana-only script were similar to English, as spaces facilitated both eye guidance and word identification although the facilitatory effects of spacing were considerably smaller than in English (12% in Hiragana compared to 30-50% in English). In addition, similar to Thai, initial saccade landing positions for Kanji-Hiragana were not affected by spacing, although the Preferred Viewing Location (PVL) for the two languages was not the same. In the Japanese study, the PVL was found to
8
be at the word beginning, which is typically occupied by a perceptually salient Kanji character, whereas the PVL for Thai was observed to be just left of mid-word position. For Chinese, Bai, Yan, Liversedge, Zang and Rayner (2008) found that sentences with an unfamiliar word spaced format were as easy to read as visually familiar unspaced text. They also found that demarcating word boundaries, either through the use of spaces or highlighting neither hindered nor facilitated reading. In Chinese script, most words include only two to three characters and these characters are often perceptually quite distinct and vary in visual complexity, pointing to the possibility that pre-attentive feature parsing may play a role in perceptual segmentation (Wang, Inhoff, & Chen, 1999). From research on these scripts without interword spaces, there appears to be a balance between two opposing forces that affect the reading of spaced or segmented text; the facilitation effects on word processing and the detrimental effect that the unfamiliar visual text format has on reading (Bai et al., 2008; Sainio et al., 2007). If we consider the different demands of reading scripts without interword spaces with those that do, somewhat different processes emerge. Interword spaces serve an important function as they form clear segmentation or word boundary cues in the parafovea prior to word fixation, so that initial word processing can be readily instigated with the end goal of forming a coherent mental representation of the text. In contrast, when a writing system provides no spatial segmentation cues, determining the extent of the letter cluster or characters that forms a word becomes an intrinsic part of the initial stage of word processing. Thus, there is an additional in-built process or step involved in reading an unspaced script, which involves demarcating where words begin and end using other segmentation cues besides salient interword spaces. Skilled readers of these scripts without interword spaces such as Chinese, Japanese or Thai have presumably acquired knowledge of the language- or script-specific word segmentation patterns or cues such as the frequency of characters or
9
letter clusters occurring at word boundaries (Bertram, Pollatsek, & Hyönä, 2004). In Thai, potential language specific candidates for word or syllable segmentation are, for example, the vowels that occur prior to the consonant at the beginning of the syllable (e.g. โรค written as /o:rk/ but spoken as /ro:k/ disease), as they possibly form salient syllabic segmentation cues to the reader. In addition, tone markers that occur above the syllable or lexeme (e.g. หน้าต่าง /nâ:tà:ŋ/ window) may form effective segmentation cues to the skilled reader. Future research can investigate whether children's reading benefits from the use of spaces or other segmentation cues such as alternating colour words (Perea, Tejero, & Winskel, 2015) when learning to read unspaced scripts such as Thai or Chinese.
(2) Letter position coding and initial letter advantage In order to understand the fundamentals of reading, it is necessary to examine the early stages of visual-word recognition (Finkbeiner & Coltheart, 2009). Registration of both the identity and position of a letter in a word forms the preliminary but highly significant stage of visual-word recognition in alphabetic orthographies (Chanceaux & Grainger, 2012; Frost, 2012; Grainger, 2008; Tydgat & Grainger, 2009). In order to recognise or distinguish between words (e.g., cat and act), it is essential to identify which letters are where in a given word (i.e., the order of letters in a word). Research on Roman script indicates that there can be quite a degree of flexibility in the coding of letter position when transposition letter effects are examined (e.g., jugde is readily read as judge) (O’Connor & Forster, 1981; Schoonbaert & Grainger, 2004; Perea & Carreiras, 2006; Perea & Lupker, 2004, in English, French, Basque, and Spanish, respectively). Typically, when investigating letter transposition effects, the masked priming paradigm (Forster & Davis, 1984) is used with a lexical decision task (where participants decide as quickly as possible if a stimulus is a word or non-word). It has been found that a
10
given target word (e.g., JUDGE) is responded to more rapidly when it is briefly preceded by a forwardly masked transposed-letter prime (e.g., jugde) than when it is preceded by an orthographic control (e.g., the replacement-letter prime jupte). In Roman script, letter position coding is particularly noisy in middle positions, but not in the initial letter position, as an illustration of this, jugde closely resembles judge while ujdge does not. Research on Roman script has consistently failed to find a masked transposed-letter priming effect when the initial letter is involved (e.g., Perea & Lupker, 2007). The commonly held view that has emerged from this research and related research on Roman script is that initial letter position has a privileged role for word recognition in comparison to internal letters (e.g., Chambers, 1979; Estes, Allmeyer, & Reder, 1976; Gómez, Ratcliff, & Perea, 2008; Jordan, Patching, & Thomas, 2003; Perea, 1998; Rayner & Kaiser, 1975; White, Johnson, Liversedge, & Rayner, 2008). An initial letter advantage is also found when participants identify briefly presented strings of five letters (e.g., T G H K N) using a two-alternative forced choice (2AFC) procedure. In the 2AFC procedure, a string of five characters are briefly presented followed by a choice of two characters, one occurring above and one below one of the characters in the array. Participants are then required to decide which character, either above or below, was present in that corresponding position of the preceding array. A W-shaped serial position function and initial letter advantage (and to a lesser extent final letter advantage) has been found for Roman script (Tydgat & Grainger, 2009; Winskel, Perea, & Peart, 2014; Ziegler, Pech-Georgel, Dufau, & Grainger, 2010). However, a different Λ-shaped serial position function is typically found when identifying symbols or shapes (e.g., & @ $ %
