Chapter 1 Introduction
1.2 Research Questions
國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
sounds could be mapped to specific phonetic radicals (e.g., /nɑw3/ sounds can be orthographically represented by “腦”, “瑙” and “惱” with the same phonetic radical).
The investigation concerning the mappings from phonologic codes to orthographic codes in non-alphabetic writing system such as Chinese is rather limited and lacking of a wide range of supplementary researches from other various perspectives and empirical evidences.
1.2 Research Questions
The primary research goal of the present study is to conduct two eye movement experiments to investigate whether the mappings between phonology and orthography influences the visual word recognition as well as spoken word recognition. In addition, through the utilization of different paradigms, we are capable of getting further insight
into how the patterns of mappings from phonology to orthography affect the processing from the aspect of different input modalities in human’s cognitive
mechanism.
From inspecting the P-O consistency effect and homophone density effect in processing Chinese phonogram characters, two eye movement experiments were conducted to investigate the phonological-to-orthographic mappings in Chinese from
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
both visual and auditory modality. In addition, time course of P-O consistency effect and homophone density effect were further prudently examined through the employment of visual world paradigm. Specific research questions of two experiments to be addressed are as follows:
(1) In Chinese writing system, whether P-O mappings take place at the stage of lexical processing in visual and auditory modality?
(2) If P-O mappings certainly make active in the Chinese word recognition mechanism, whether this orthographic activation from phonology influences word recognition differentially for reading and listening process?
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
Chapter 2
Literature Review
2.1 Two-way mapping between orthographic and phonological representations
2.1.1 The bidirectional nature of consistency
In traditional information processing models, research often placed focus exclusively on the one-way consistency effect. In this traditional genre of framework, information simply flows in the manner of “downstream”, as from orthography to
phonology. This kind of orthography-to-phonology (O-P) mappings is generally titled as “feedforward consistency”. The theoretical issue of feedforward consistency
concerns whether there exists more than one way to pronounce a spelling body (e.g., INT as in HINT and PINT versus EAP as in HEAP and LEAP). Inconsistency in the mapping between spelling and sound arises when a spelling has more than one possible pronunciation (e.g., _INT can be pronounced as in PINT or MINT). In English, a number of inconsistent spelling bodies may approximately have up to five possible pronunciations (e.g., _OUGH in COUGH, DOUGH, THROUGH, BOUGH, TOUGH) (Ziegler, Van Orden, & Jacobs, 1997). In many behavioral studies, the property of O-P consistency effect has been evidenced in several empirical
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
investigations. The inconsistency words in general had longer reaction time and are more error-prone when making lexical decision and naming tasks. On the other hand, the consistency words usually consume shorter reaction time, which reflect the relatively rapid connection between spellings and sounds in consistent condition (Andrews, 1982; Glushko, 1979; Lee et al., 2005; Van Orden, 1990).
In addition to the consistency from spelling to sound, accumulating researches had pointed out the bidirectional portrait of the intrinsic nature of consistency between phonological and orthographic representations. For both visual and auditory modalities, the fundamental pattern of phonology-to-orthography (P-O) consistency is referring to the consistency relations of mapping from phonology to orthography, which is occasionally named as “feedback consistency”. In the construction of P-O
mapping, spelling enemies are words with the same pronunciation body but different spelling bodies. These false rhymes are the “enemies” from the P-O consistency literature. On the other hand, “friends” refer to words which share the same
pronunciation and the same spelling pattern (e.g., /_Ob/ as in PROBE and GLOBE).
The cross-code consistency flow of activation is gradually turning out to be a major determinant in the word recognition system. Accordingly, the interactivity between consistency leads to the prediction that word recognition should not simply depend on the characteristics of the mapping and consistency from the direction of
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
orthography-to-phonology, but also hinge on the correspondent associations from phonology-to-orthography.
2.1.2 Construction of P-O consistency effect
2.1.2.1 P-O consistency in visual modality
Supporting the interactive view of word recognition, Stone et al. (1997) claimed that phonological activation reverberates to orthographic processing units and consequently constrains orthographic encoding. They conducted two experiments and demonstrated a robust phonological effect in visual lexical decision. Experiment 1 tested for consistency effect of spelling to pronunciation correspondence (e.g., HEAP vs. PROBE) on correct ‘‘yes’’ responses to words, in which all experiment stimuli were low-frequency, O-P consistent words. The usage of only O-P consistent words for experimental material allowed a constricted focus on the P-O consistency effect.
The experiment result contradicted the unidirectional feedforward models that P-O consistency should not affect performance. It suggested that the O-P consistent pronunciation bodies of the stimuli could have been generated by another spelling body make the information flow backward from phonology to orthography. By factorially manipulating O-P and P-O consistency, Experiment 2 tested the artifact
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
hypothesis which stated that, when fully consistent (O-P and P-O consistent) words was used, the feedforward O-P consistency effect was robust and reliable.
Nevertheless, when the value of P-O consistent was ignored, the unreliable O-P consistency effect was replicated.
Table 1. Example words for Experiment 2 (Stone et al., 1997)
The result of Experiment 2 replicated the P-O consistency effect of Experiment 1 using all new spelling and pronunciation bodies. In total, 80 different spelling / pronunciation bodies were tested for a P-O consistency effect, given O-P consistent of the spelling bodies. The P-O consistency effect appeared to be robust. Stone et al.
(1997) furthermore tested three possible confounds: orthographic neighborhood, body neighborhood, and vowel ambiguity. The statistical analysis confirmed that there was a significant P-O consistency effect. The substantial differences between reaction time as well as error rate of consistent and inconsistent words cannot be attributed to
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
confounding with other lexical properties. In sum, the experimental result revealed that when the proper contrast with consistent words was used, both O-P consistency and P-O consistency effect were reliably obtained. This suggested that previous research proposed that phonology only weakly influenced performance in visual word perception tasks may have resulted from the failing to control the property of P-O consistency. Stone et al. (1997) lined out the broad issue of feedforward versus feedback perspectives on the perception regarding to the visual word recognition mechanism.
Ziegler, Montant, and Jacobs (1997) concentrated on the subject of the P-O consistency effect in French, which mapping sound to spelling is less inconsistent than English. In order to identify whether the P-O consistency effect was capable of generalizing to different language and to further examine the importance of bidirectional consistency effect, a replication and extension of Stone et al. (1997) experiment in French was conducted. The experiments used a factorial design to investigate the effects of O-P and P-O consistency effect in a lexical decision task, an immediate naming task and a delayed naming task. Four types of stimuli were manipulated: (1) words that were both O-P and P-O consistent, (2) words that were O-P consistent but P-O inconsistent, (3) words that were O-P inconsistent but P-O consistent, and (4) words that were both O-P and P-O inconsistent. The experiment
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
result replicated the classic O-P consistency effect: ambiguity in the mapping between spelling and phonology slows down naming latencies and increased error rates. The research analyses further supported the concept that P-O consistency effect is generally consistent with the idea that reading is a dynamically interactive system (Van Orden & Goldinger, 1994; J. C. Ziegler, Van Orden, et al., 1997).
However, certain research raised question regarding to the P-O consistency effect in the visual word recognition. Peereman, Ronald, Content, Alain, Bonin and Patrick (1998) challenged the P-O consistency hypothesis for potential familiarity and frequency confounds. They conducted several experiments to investigate whether sound-to-print consistency influence lexical decision in French. The same sets of consistent and inconsistent words matched on word frequency were used in three different tasks. In the result, they observed P-O consistency effects in the task of writing down the word which was auditory delivered. However, no significant effect was found in lexical decision or in naming tasks, except for a small effect on error rates, which was significant only by subject. The same stimuli were used in a following experiment under degraded visual conditions with a stricter criterion for stimulus selection in order to enhance the contribution of reverberating activation from phonology to orthography. However, the pattern of result did not reveal a stable P-O consistency effect. Peereman et al. (1998) cast doubt on the existence of
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
reciprocal constraints between phonology and orthography at pre-lexical stage of processing. In addition, they pointed out that although the cross-linguistic differences and task specific decision strategies might explain the absence of a robust P-O consistency effect, further studies are still needed to ascertain the P-O mapping consistency in visual modality.
Nonetheless, the findings of Peereman et al. (1998) have been criticized by Perry (2003) for poor item selection when inspecting the P-O consistency effect. In order to investigate whether there would be an robust effect of P-O consistency at the phoneme-grapheme level, Perry (2003) conducted the experiment with a new set of items. Three groups of word items were used for the experiment stimuli: an O-P consistent group that contained a vowel that was the most common spelling of that vowel (C-C); an O-P consistent group that contained a vowel correspondence that was not the most common spelling for that vowel (C-I); and a O-P irregular vowel group that therefore contained a vowel that was not the most common spelling (I-I). The results reported that, a significant P-O consistency effects revealed at least at the phoneme-grapheme level in English with frequency and subjective familiarity being controlled in a normal lexical decision task. Furthermore, the strong interactivity between phonology and orthography during reading process are positively suggested.
Lacruz (2004) investigated O-P and P-O consistency effects for both high- and
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
low-frequency words in the lexical decision and naming tasks. The experiment items were either O-P inconsistent, P-O inconsistent, or both. The three groups were controlled not only for frequency but also for subjective familiarity and a number of other lexical properties, including word length, orthographic neighborhood, bigram frequency, and summed frequency of friends. For naming latencies, there was a main effect of frequency, as subjects named high frequency words significantly faster than low frequency words. In addition, there was a main effect of consistency as O-P inconsistent words were named significantly more slowly than their consistent controls. The interaction between consistency and frequency was not significant.
The experiment outcomes provided the evidence that both O-P and P-O consistency relations were involved in the processing of monosyllabic words in the lexical decision and naming tasks. In addition, the experimental result also demonstrated a robust P-O consistency effects for both high and low frequency words.
Lacruz (2004) proposed that the interactivity from phonology to orthography was in accordance with several theories of word recognition which considered the existence of intimate relations between a word’s sound and spelling. The information do not merely flow from spelling patterns to sounds in visual word recognition, but also back from phonetic sounds to spelling representations (Colheart et al., 2001; Van Orden, 1990; Van Orden & Goldinger, 1994). Together, these results supported that P-O
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
effects are representative of a general principle in the reading system.
2.1.2.2 P-O consistency in auditory modality
Previous research has shown that literacy (i.e., the ability of reading and spelling) affects spoken language process. Word recognition is important for the reason that acquiring this skill is among the first task confronted by beginning readers during their learning stage. Ziegler and Muneaux (2007) conducted a developmental study to investigate the extent to which learning to read and write affects spoken word recognition. They reported that, prior to literacy; the spellings of spoken words will not influence the auditory lexical decisions. Once literacy is acquired, the orthographic influences on spoken word recognition and the acquisition of literacy were tightly linking with each other. Spoken word recognition is gradually affected by orthographic consistency of spoken words as literacy is developing. In addition, as early as Grade 1, the size of the P-O consistency effect was predicted by the reading level of a child. It revealed that the orthographic effects on spoken language were not artifact of the uncontrolled spoken language properties, but demonstrated the orthography influences on the mechanism of spoken word recognition. The evidence supported the view that word recognition involves a dynamic as well as interactive coupling between spoken and written words of various grain sizes (Frost, 1989; Stone
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
et al., 1997; Van Orden & Goldinger, 1994).
However, there are a number of on-going theoretical and empirical discussions regarding to the intrinsic influence of P-O consistency in relation to the associations between orthography and phonology. Ziegler and Ferrand (1998) manipulated the P-O consistency of the rime in an auditory lexical decision task in French. By analogy with previous visual word perception experiments (Stone et al., 1997; Ziegler et al., 1997), the essential prediction was that performing lexical decision to spoken words would be worse if the stimuli were P-O inconsistent (i.e., rimes could be spelled in multiple ways) than if their rimes were P-O consistent (i.e., rimes could be spelled only one way). All the P-O consistent and P-O inconsistent words were matched for a number of properties, including word frequency, familiarity, number of phonological neighbors, uniqueness point, and number of phonemes. Ziegler and Ferrand (1998) found that the auditory lexical decisions to inconsistent words took longer reaction time and yield higher error rates than did those to consistent words. The experimental result was in accordance with several research of auditory consistency effects for English (Frost, 1998), which implied that it seems unlikely that the P-O consistency effect was due to the particularities of French. The presence of P-O consistency effect confirmed a bidirectional flow of activation, which generates a coupling between phonology and orthography at various grain sizes (Frost, 1989; Stone et al., 1997; Van
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
Orden & Goldinger, 1994).
Not only in English and French, Ventura, Morais, Pattamadilok, and Kolinsky (2004) examined Portuguese in the auditory modality to determine the locus of the P-O consistency effect in speech processing. They found that words in the inconsistent condition produced longer auditory lexical decision latencies and higher error rates than did consistent words. However, the P-O consistency effect was not obtained in the shadowing task; the task difference could reflect the confinement of orthographic influences to either decisional or lexical processes. They also indicated that the consistency effect reflected the involvement of lexical but not sublexical processes in Portuguese. Furthermore, this study demonstrated that the P-O consistency effect is not a language specific phenomenon constrained by particular linguistic feature.
With a graded consistency manipulation, Ziegler (2004) replicated the P-O consistency effect in the auditory lexical decision, rime detection, and naming tasks.
In order to make sure that the auditory consistency effect did not confound with phonetic differences between consistent and inconsistent items, he manipulated the degree of inconsistency of words sharing the same phonological rime. The stimuli mainly contained pairs of inconsistent monosyllabic words that possessed the same phonological rime but differed in the probability with which their phonology maps
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
onto spelling. For example, the words sign and wine have the same phonological rime:
/-ain/ in English. The “-ine” spelling is dominant, it occurs in most monosyllabic
words of this particular rime family (e.g., fine, nine, line, mine). On the other hand, the “-ign” spelling is subdominant, it occurs in few or no other monosyllabic words.
Both sign and wine are inconsistent due to the fact that their phonological rimes map onto multiple various spellings. However, wine contains the dominant spelling, whereas sign contains the subdominant spelling. Ziegler (2004) confirmed that response to words with subdominant spellings took longer and were more error-prone than words with dominant spellings in the lexical decision tasks. Several empirical conclusions were made according to their findings. First, the P-O consistency effects were not the artifacts of phonological or phonotactic properties of the stimuli. Second, the influences of orthographic codes in auditory modality were not restricted to the lexical decision tasks; the orthography effect in the mapping from phonology to orthography was strongest in lexical decision, intermediate in rime detection, and weakest in auditory naming tasks. The research finding overall indicated that acquiring the knowledge of orthography in point of fact alters the manner of processing spoken language.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
2.2 The Processing of word recognition - Bimodal interaction activation model
In recent modeling approaches, capturing the intricate spelling-to-sound interactions of consistency has gradually turning out to be a major preoccupation. The construction of word recognition model in point of fact precisely related to the inherent mapping phase regarding to phonology and orthography, in addition, it also provides a way for us to get further insight into the fundamental nature of visual and spoken word identification.
Varying principles concerning the diversity in writing systems may mean for the different mental processes involved in word recognition process. The earliest version of model which tested on the bidirectional flow of consistency effect in visual and auditory word recognition was first proposed by Frost (1989). He pointed out the important concept of orthographic depth and suggested that it is an essential element in determining the processing relationship between orthographic and phonologic systems. The orthographic system of Serbo-Croatian is an example of shallow orthography. Each letter represents only one phoneme and each phoneme is
represented by only one letter, the relation between letters and phonemes is isomorphic and exhaustive. Morphemic variations in the language due to inflection and derivation do not often produce alterations in the phonemic structure of word stems. In contrast to the Serbo-Croatian, language system of English belongs to the
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
deep orthography. Different degrees of consistency and completeness between letter
and phoneme are capable of reflecting the forms of the deep orthography in English.
Frost (1989) directly examined the cross-linguistics framework concerning the interactive processing of print and speech by the method of visual or auditory degradation in the matching task. Native speakers of English and Serbo-Croatian were presented with simultaneous printed and spoken verbal stimuli. Visual degradation was achieved by overlying a pattern of random dots on each word. Thus the whole word was embedded in a dotted background that obscured the graphemic structure.
Auditory degradation was achieved by masking each spoken stimulus with signal-correlated noise with the same amplitude envelope as the spoken stimulus.
Subjects had to decide whether the stimuli were equivalent and the decision reaction time was measured. The experiment result exhibited that both effects of degradation were much stronger in English than in Serbo-Croatian. Accordingly, Frost (1989) suggests a lexical structure which rationalizes the relationship between the orthography and phonology systems in terms of lateral connections between the constructions at all of their levels. Furthermore, different degrees of orthographic depth between spelling and phonology in the languages systems are capable of demonstrating influences on the lateral connections: in shallow orthographies there exist simple isomorphic connections between graphemes and phonemes, but more
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
complex, many-to-one connections in deep orthographies.
Grainger and Ferrand (1994) developed a more detailed and systematic architecture addressing the interconnected features of the bimodal interactive activation model (Figure 2).
Figure 2. Architecture of bimodal interactive activation model (Grainger & Ferrand, 1994)
In BIA model, orthography and phonology are connected in a mutually bi-directional manner at sublexical as well as lexical processing level. The main
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
notion of this modal is that when a printed word activates a sublexical orthographic
notion of this modal is that when a printed word activates a sublexical orthographic