CHAPTER 2 LITERATURE REVIEW
2.4. Literature Summary and Research Hypothesis
國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
25
(tonal syllable) is a possible unit in word production and orthography and the word’s morpheme do not seem to play a role in language planning; (3) tone could be a metrical frame for the segmental syllable to attach to in Chinese; (4) syllable-onset cannot induce priming effect either, so it couldn’t be regarded as a planning unit in underlying.
2.4 Literature Summary and Research Hypothesis
Researchers have investigated sounds activation in production mechanisms, as proposed by the former serial and interactive model, by means of error-inducing experiments (Dell 1986 & 1990 & 1992, Schwartz 1994, Chen et al. 2002, Martin et al. 2004, Schwartz 2006) and speech errors in a naturalistic setting (Fromkin 1971, Buckingham 1980, Stemberger 1984, Bloch 1986, Schwartz 1994). Both production models should be compared to identify the role of the sound, and how the elements are activated in these models.
Theoretical frameworks of serial-ordering and connectionist have debated on whether the error sounds are produced from a sequential level of processing or the consequence of levels interacting altogether. As to serial approach, the sound error was dealt with the interference within the chain of linguistic components, including semantic, syntactic, morphological or phonological generators (Fromkin 1971). In order to explain how the retrieved lexicon is proceeded to the next level to retrieve the sound, which influenced Levelt (1989 & 1999) to propose lemma and lexeme levels in explaining the phonological units are encoded falsely to the retrieve a similar lexicon (at lemma level), thus the errors are produced. Either target sound or error sound is uttered; serial-ordering model prohibits the possibility of backward activation in lexical processing.
On the other hand, interactive approach proposes bi-directional activation: not
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
26
only spreading but also feedback during lexical activation. Speech error seems to be the noise between lexeme and lemma nodes, or we may say it is also a false connection (Stemberger 1985), on the way to the target lexicon which receives enough activation falsely. The phonetic output in the interactive mechanism is the product of mutual interaction among syntactical, lexical and phonological levels, which in serial-ordering model is merely the product of single source, route, or Fromkin said the false encoding of morphophonemic rule results in error, or the Levelt said it is the false retrieving from lemma level to lexeme level. In Dell’s connectionist model (1986), the output comes from the lexical-network organized in mind, also comprises semantic, lexical, and phonological layers, but is the co-effect of these levels. Sometimes speech errors result from the latter level spreading weight back to the former level. The feedback strength goes back from phonological layer to word layer and retrieves a lexicon framework with similar morphophonemic organization, spreads again to phonological layer, and then induces speech errors. The interactive simulation might imply the information of phonological units (feature, segment, or phonological constituent), pure semantic substitution, or the mixed error (interacting between phonological and semantic layers) about speech errors.
In current Stroop-technique study, there are some research hypothesis about the relation between sound activation and visual representation, which need to be dealt with in this study:
(1) Through the control of phonological similarity, we would like to know the trials of high phonological similarity would induce more speech errors than trials of low similarity, and see whether the controlled factor would cause difference of error amount among visual tasks. In addition, we should observe and compare the response time between the two groups. We would like to know whether phonological similarity serves as interfering or facilitation
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
27
effect in these tasks.
(2) Among speech errors, we plan to grade phonological similarity between target and error. We would like to examine which linguistic effects dominate in lexical encoding: initial, vowel, rhyme, tone, syllable structure, phonotactic regularity, and Stroop effects in this study.
(3) In order to look further into the issue of advanced planning unit during language production, carriers with diverse structures of shared units were recruited, such as the units of onset, vowel, rhyme, tone, syllable, tonal syllable. Response time and amount of speech error will be compared and discussed on their possibilities of serving as planning units.
(4) The status of tone will be concerned in shared unit test. If it is a phonological tone, it will induce facilitation in lexical process. If it is a lexical tone, the unit of tone should not induce any facilitation.
(5) According to the result of error amount and reaction time in the five tasks, where processing model could well account for the generated patterns. Serial and interactive accounts will be testified respectively.
This modified Stroop-technique experiment could provide supportive and empirical evidence to examine whether serial-ordering model or connectionist model could tell us more about sound activation during human lexical production. If we could take speech error, reaction time, and basic planning unit into consideration by means of controlling phonological similarity between linguistic forms (color term / carrier)and visual concepts (visual color), we could obtain more insights on how concept and sound are processed when human encode lexicon, especially there are dual inputs coming into visual channel.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
28
Chapter 3 Methods
In this experiment, we modified the test items in Stoop’s experiment technique (1935) and conducted experiments which we controlled the variety of phonological similarity and added homophonous materials into experiment. In the original work, Stroop designed the color term as visual stimulus with ink color other than the term indicated, such as the test item of “red” in color green, as reviewed in former chapter.
Stroop technique was utilized into current study. In current study, all the stimuli were replaced with Chinese characters. In addition to the language variety, phonological similarity was an independent variable controlled in this study. The amount of speech error, types of speech error, and reaction time in each trial were dependent variables in the experiments. The amount of speech error and response time were served as judging facilitation or inhibition effect during color processing. Linguistic effects between target and error could help us observe whether there was any structural effect in lexical encoding. Phonological similarity was utilized to observe whether it would increase difficulty or error amount in lexical processing. Therefore, the criteria for phonological similarity, stimuli design, task procedure, and error analysis will be mainly concerned in this chapter.
There was a series of tasks which were tested for different purposes, which include: (a) Task 1: Color Naming Test, (b) Task 2: Color Reading Test, (c) Task 3:
Stroop Naming Test, (d) Task 4: Homophonous Naming Test, and (e) Task 5: Shared Unit Test. The five tests were designed with different visual tasks. Subjects were asked to identify the colors they perceived or to read the words they saw, according to the conditions those tasks required. During task 1, subjects were presented a strip of color squares, and they needed to call the colors in order for each strip. In task 2, subjects had to read the color terms for each trial, and they didn’t need to name the
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
29
color of each term. Task 3 modified the Stroop’s work (1935) into a Chinese edition.
Subjects were required to name the ink color for each term sequentially. This was aimed to testify whether interfering effect of color stimuli upon reading color terms exists in Chinese, as that mentioned in Stroop’s (1935). In present study, we created a list of homophonous stimuli corresponding to the color terms. Subjects had to name the colors for stimuli in task 4, as they did in task 3. Homophonous Naming Test used these colors’ homophones to substitute for corresponding colors. As to the reason which we recruited in task 4, if homophone induces similar response time and error types, Stroop’s effect might not be specialized to color terms, even though phonological similarity might have effect on both processes. If it induces various patterns of time span and error types, it appears that color terms might be enclosed as a semantic group at the lexicon process for further phonological encoding. If not, it implies that homophone serves the same role as color term in lexical process.
In task 5, each trial was replaced with other kind of characters, which shared one certain structure of phonological units with color terms. Subjects were trained to read and name the stimuli’s color for each trial according to the tasks. Diverse shared units were supposed to induce interfering or facilitation during lexical process. Shared Unit Test was aimed to focus on the issue of advanced planning units among these color terms. We created a list of characters which shared different types of structure with color terms, such as initial, vowel, rhyme, bare syllable, bare tone, or whole syllable.
The subjects’ response time and induced speech errors among the four visual tasks were record for later analysis.
3.1 General Methods of Task 1 ~ 4 3.1.1 Subjects
We’ve recruited 22 college students (7 males and 15 females) whose native
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
30
tongue was Taiwan Mandarin, aged 20-30 years old. All these recruited subjects took the four tests: naming test, reading test, Stroop’s Task, and Homophonous Naming Task. In order to balance the practice effect and to avoid it inducing practice bias, half of the subjects were provided the order of color square naming, color term reading, Stroop’s reading, and homophonous naming tests in order. The others were also provided the four tasks, but were presented as the order of homophonous naming task before Stroop’s naming task. The ones with psychology, linguistics and art background were excluded from this study, especially the art major students who might be professional and sensitive to the subtle difference. When they identify the colors, they might probably generate many low-frequent color terms as their response, which might cause difficulty in data analysis. This study was aimed to recruit ordinary as well as normal speakers in color identification and to analyze the way they process concepts in color naming, instead of the way how they sort these colors sophistically and precisely.
Besides, in order to ensure that subjects had normal ability to identify and categorize the colors they perceived, they were asked to accept a pre-test to examine their knowledge about sorting colors to be at or above average level. Those subjects who did not participate in this study were color blindness or weak identification toward colors.
At training stage, we provided a conception formation test to help subjects set correct categorization for the colors recruited in the study. The target colors included:
red, orange, yellow, brown, green, blue, purple, black, gray, and white. Subjects were presented a color square and a description on computer screen, and they needed to press “yes” or “no” button to confirm whether the description is correct. If they provided correct answer, screen would show a positive response, and a negative response was given when they provided incorrect key. The experiment operator
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
31
evaluated the subject’s performance. If subject’s precision rate achieved to 95%, they could attend the main experiments. If they failed to get to the threshold, they had to take the concept exercise again, until they passed the training requirement.
3.1.2 Colors and Phonological Similarity
Phonological similarity was controlled as an independent variable in this study.
The number of speech errors and the response time were dependent variables in this study. It was controlled to maintain colors of each trial in a state of phonological homogeneity. In the following experiments, we needed to sort the phonological relation for each color pair. Jaeger (2005) had evaluated the phonetic similarity between targets and error words for English children, and used phonological and phonetic criteria to judge whether the lexical errors could be considered phonologically related. It appeared to be an ideal basis for us to adapt and extend to evaluate the phonological similarity in Chinese system. Jaeger provided a list of criteria for grading the degree of similarity in English speech errors, which is provided in Table3-1. Jaeger supposed that a lexical error is considered
“phonologically related” if its score achieved to at least 3 points. This evaluation system is also the criterion of phonological similarity in present study.
‧
Table 3-1. Jaeger’s Phonological Similarity Grading
Criteria for grading phonological similarity in lexical errors.1. Same number of syllables = 1 point 2. Same stress pattern:
a) if both words are stressed, but each only one syllable = 0.5 points.
b) if both words have 2 or more syllables and the same number of syllables, and same primary stress location = 1 point.
c) if one word is one syllable and stressed, and the other is two-syllable and has first-syllable stress = 0.5 points.
d) if both words have 2+ syllables, but a different number of syllables, but the same general word-stress pattern = 1 point (e.g. ‘dictionary’ and ‘library’).
3. Same initial phoneme = 1.5 points.
4. Same final phoneme = 1 point.
5. Same primary-stressed vowel = 1.5 points (if only one of them occurs before /r/
or a nasal, only 1 point.)
6. Other same phonemes in same position in primary-stressed syllables = 1 point each.
7. Other same phonemes in same position in non-primary-stressed syllables = 0.5 points each.
8. Same phonemes in same position in primary-stressed syllable in one word, but non-primary stressed syllable in the other word, if syllables are in same position in word = 0.5 points each.
9. If a vowel is both word-initial and primary stressed, or both word-final and primary stressed = 2 points total.
According to Table 3-1, we noticed that most of these criteria reflected the structure in English phonology, such as multi-syllable pattern in lexicon and stress pattern in syllable, where as Chinese might not have them. Criterion 2 and 5 to 9 are stress-related grading for lexical errors, since stress assignment plays an important part in English syllable. However, in Chinese, we have pitch pattern within a syllable, that refers to the level tone and contour tones. English has multi-syllable lexicons where as Chinese is viewed as a mono-syllabic way. Therefore, criterion 2 and 5 to 9 might not be necessary for Chinese. Furthermore, tone is an important organizing
‧
structure in Chinese and thus there should be a system for it. The following is a revised list of evaluating criteria as shown in Table 3-2, the evaluation of which is hoping to reflect the sound patterns in Chinese.
Table 3-2.The Revised Criterions on Sound Similarity in Chinese
Strucrure Shared Units Score
A (syllable)
1. Whole Syllable (with tone)
=Word 4.5-5.5
2. Whole Syllable (without tone) 4-5
B (sub-syllable)
3. Syllable Structure 1
4. Syllable Number 1
10. Pre-nuclear Glide 0.5
In order to make the criteria to well explain the similarity between error and target in Chinese, Table 3-2 considered the feature of syllable structure and the numbers in Chinese, and it modified Jaeger’s methodology when grading phonological similarity. For there exists typology diverse, the first difference is to simplify the way of syllable counting. We don’t need to count syllable amount for Chinese. All the character in Chinese is phonologically monosyllabic, so we count the lexical pair sharing the same number in one point, especially for the monosyllabic case of color terms in Chinese. Second, tone is another issue that its phonemic significance should have independent criterion for grading. According to Jaeger’s criteria, the monosyllabic words which share the same stress pattern, regardless of the vowel or syllable number factor, could get 0.5 points. Therefore, lexical pair which
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
34
shares the same tone could get 0.5 points. Even though we know that syllable number and pitch pattern are the main differences between English and Chinese, Table 3-2 seems to help deal with the typological differences and grade for Chinese pairs.
With regard to the counting methodology for the similarity, Table 3-2 has another rationale from the one of Jaeger’s (2005). There are three structures in Table 3-2: (A) syllable, (B) sub-syllable, and (C) phonemes. Each lexical pair should be graded according to the structural hierarchy (from A, B, to C), and each pair can only be graded within one structural criteria. Lexical pair could only be graded by the criteria in A, B, or C independently, and can’t be graded in any two of the structures at the same time (except for syllable structure and syllable number). For example, if there is a pair of homophones, xong (means red or great); they could only gain the similarity score within the criteria (shared units) of structure A (syllable). The pair couldn’t get any score from either structure B (sun-syllable) or C (phoneme). Sub-structure means that the lexical pair shares unit which is smaller than syllable and larger than segment.
Criteria of sub-syllabic structure include syllable structure, syllable number, initial onset, rhyme, and bare tone. The remaining phonemic structure, such as vowel, coda, and prenuclear glide, should be attributed to structure C.
As to the specific phonological scores in Table 3-2, we mainly refer to Jaeger’s (2005) phonemic and syllabic grading criteria. In structure C (phoneme), if the pair shares the same vowel, it could get 1 point in phonological similarity. Take the pair
huang2 [xw
ɑŋ35] (yellow) and gao1 [kɑw55] (tall) as example in Table 3-3.‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
35
Table 3-3. Lexical Pair of Huang2 and Gao1 (yellow-tall)
Criterions Factors Score
4. Syllable Number monosyllable 1 point
8. Vowel [ɑ]-[ɑ] 1 point
Total (Medium
similarity)
2 points
Their monosyllable similarity could get 1 point according to the third criterion.
In addition, there is not any similarity except for their vowel [ɑ], so it can also get 1 point. The total score of their phonological similarity is 2 points.
Regarding the score of coda, there are four phonemes which can serve at coda position in Chinese, such as [n], [ŋ], [j], and [w]. If the lexical pair shares the final coda, it can get 0.5 points in phonological similarity. The example can be seen in Table 3-4, the pair hei and kwai.
Table 3-4. Lexical Pair of Hei1 and Kwai4 (black-strange)
Criterions Factors Score
4. Syllable Number monosyllable 1 point
9. Coda [j]-[j] 0.5 points
Total (Low similarity) 1.5 points
The pair hei1 [xej55] (black) and kwai4 [kwaj51] (strange) shares the same final consonant [j]. Beside 1 point the pair could get from the criterion of syllable number, another 0.5 points is gotten because of the shared final coda. The total of the lexical pair is 1.5 points. As the former mentioned, each pair should be graded within the same structural criteria. Criteria 8 and 9 should not overlap with criterion 6; that is, if the lexical pair doesn’t get any score from the rhyme (criterion 6), then the pair is
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
36
allowed to be graded under criterion either 8 or 9.
The issue on the status of prenuclear glide has been disputed in Chinese. The related argumentation can be referred to: prenuclear glide as onset’s secondary articulation (Duanmu 1990, 2002), forming consonant cluster with initial (Bao 1990), as an element of rime (Jiang2001, Wang & Chang 2001), being in indeterminate ststus (Bao 1995, Huang 2001, Wan 1997), existing in independent structure (Shen 1992, 1993), supposing in approach of mora structure (Yip 2003, Ma 2003), and arguing under X-bar approach (van & Zhang, 2008). The status is not the main issue in present study. To avoid disputing, we follow Bao’s framework (1990, 1995) to be a tentative assumption in this study. In this study, we tentatively assume that prenuclear glide might belong to the initial structure to form a consonant cluster in some Chinese syllables. However, on the surface representation, prenuclear glide sometimes interacts with vowel, such as /j-an/ into [jɛn], /w-əŋ/ into [woŋ], and /ɥ-an/ into [ɥɛn]
etc. In the case of [j], [w], and [ɥ] interacting with vowel, we should go back to their underlying forms to give score to the lexical pair. Jaeger gave the same initial 1.5 points, which consists the part of the main first consonant and the following ones. In criterion 10, if the lexical pair shares the same prenuclear glide, the phonological score is valued 0.5 points. Therefore, the range of initial score could be 1 to 1.5 points.
Basically, the pair could get at least 1 point for sharing the part of onset. If the pair also shares the following glide, it could get 1.5 points in total. The following is the details for the score of initial position.
Initial onset and rhyme belong to the case. In Chinese, the initial part includes a
Initial onset and rhyme belong to the case. In Chinese, the initial part includes a