ISSN 0439-755X
CN 11-1911/B


    25 June 2013, Volume 45 Issue 6 Previous Issue    Next Issue

    For Selected: Toggle Thumbnails
    Grammatical Planning Scope in Sentence Production: Evidence from Chinese Sentences
    ZHAO Liming;YANG Yufang
    2013, 45 (6):  599-613.  doi: 10.3724/SP.J.1041.2013.00599
    Abstract ( 859 )  
    The grammatical planning scope in sentence production is an important issue and investigated by many psycholinguists. Regarding the grammatical level proper, the planning scope is often considered to be constrained by certain grammatical units, such as a clause (a sentence fragment consisting at least of a subject and a predicate), a subject noun phrase (the phrase including all nouns associated with the grammatical role of subject), or just a content word. Recently this planning scope was specified as a functional phrase. This issue, however, remains controversial. The goal of the current study is to thus clarify the scope of planning at the grammatical processing level of sentence production, especially to test the functional phrase hypothesis. In the current study the functional phrase hypothesis was tested with the picture-word interference paradigm. Native Mandarin Chinese speakers were asked to produce sentences with a prepositional phrase modified subject (e.g., ‘‘the dog under the flower is red’’) or sentences with a conjoined noun phrase as the subject (e.g., ‘‘the dog and the flower are both red’’) in response to vertical pairs of objects, and each display was accompanied by a visual distractor. The distractor could be categorically related or unrelated to the target picture on which it was presented, and was always unrelated to the other picture. Forty-three undergraduate and graduate students in total participated in this study and were asked to name pictures with the required syntactic structures as accurately and quickly as possible while ignoring the distractor. Two separate analyses were carried out with participants and items as random factors for onset latencies and error rates. There was a semantic interference effect for the first noun in both utterance formats, showing longer latencies in categorically related conditions compared to unrelated conditions. For the second noun of both utterance formats, the semantic interference effect was found in error rates rather than in onset latencies, no matter which SOA (Stimulus Onset Asynchrony) the condition was (0, 150, or 300ms). Meanwhile in the condition of SOA=0, a semantic facilitation effect in onset latencies was observed on the second noun of sentences with a conjoined noun phrase as the subject. These results suggest that the grammatical planning scope should be specified as that the lexical selection is radically incremental while the conceptual planning scope may encompass the functional phrase.
    Related Articles | Metrics
    Preview Benefits and Word Segmentations When Reading Chinese
    LIU Zhifang;YAN Guoli;ZHANG Zhijun;PAN Yun;YANG Guifang
    2013, 45 (6):  614-625.  doi: 10.3724/SP.J.1041.2013.00614
    Abstract ( 788 )  
    Recently, a number of studies have confirmed that Chinese readers could obtain linguistic information from word n+1 when it was in parafoveal vision. However, there was no confirmation that Chinese readers could obtain preview benefits from word n+2 (Yan, et. al., 2009; Yang, et. al., 2009; Yen, et. al., 2009; Rayner, et. al., 2003). The preview benefit effects received in a boundary paradigm could not exclude the effects of word segmentation. Therefore, there is no evidence that Chinese readers process the meaning of word n+1 by the word unit. Two experiments were conducted using a new eye-movement contingent display technology to clarify the confusion in the present study. The Chinese sentences used in both experiments consisted of 7 to 10 two-character words. There were four treatments of Chinese sentences in Experiment 1. Namely, as the nth word was fixated, (1) the words located to the right were all masked by a series of “※””, (2) the words located to the right of the n+1th word were all masked by a series of “※””, (3) only the n+1th word was masked by two masks of “※””, or (4) no words were masked as a baseline comparison (the control). Among the treatments, the first and third provided the cues of word n segmentation, and the second treatment provided information of word n+1 segmentation. Moreover, the first treatment was deprived of the preview benefit effects from word n+1 and n+2, the second treatment was deprived of the preview benefit effects from word n+2, and the third treatment was deprived of the preview benefit effects from n+1. An eye-movement apparatus was used, and eye-motion indices were measured. The effects on the total reading time, mean fixation duration, mean gaze duration, probability of refixation and word skipping were significant (p<0.05). The differences among the second and third treatments and the control for these indices showed significant preview effects on word n+1 and n+2. The mean gaze duration in the second treatment was observed to be less than the control, although the total reading time increased. However, we cannot exclude the effect of the mask paradigm itself, which influenced the patterns of eye movements and affected the results of Experiment 1. Therefore, a second experiment was conducted. Experiment 2 adopted similar treatments as Experiment 1, but the two adjacent characters that did not belong to a word were masked together. The manipulations did not provide cues to facilitate word segmentation. Consequently, all the masked treatments in Experiment 2 influenced the total reading and fixation times more significantly than those in Experiment 1. The mean gaze duration in the second treatment was greater than the control, which was an opposite result to that of Experiment 1. The patterns of eye movements in Experiment 2 differed from those in Experiment 1. Thus, the results obtained in Experiment 1 were not because of the influence of the mask paradigm. Because the second treatment in Experiment 1 emphasized the boundary of word n+1, the differences in the eye movement patterns between this condition and the control revealed that the beneficial effects from the text in word n+1 were based on the word unit. Thus, the data from the current experiment better fit with the theoretical assumptions of parallel processing models in reading.
    Related Articles | Metrics
    Retroactive Effects of Irrelevant Sound on Serial Recall from Short-Term Memory of Chinese Characters
    MU Defang;SONG Yaowu;GAO Jiangfeng
    2013, 45 (6):  626-635 .  doi: 10.3724/SP.J.1041.2013.00626
    Abstract ( 1623 )  
    Even when stimuli are presented visually, the presence of irrelevant sound (IS) significantly impairs performance on immediate serial recall. Theoretical models provide conflicting accounts of the origins of irrelevant sound effects. The detailed pattern of interaction between IS and other factors influencing memory performance has been seen as placing important constraints on the development of models of short-term memory such as the working memory model and object-oriented episodic record model, but the fundamental question of the locus and mechanism of the IS effect has still not been satisfactorily resolved. To explore whether irrelevant sound has a significant influence on serial recall from short-term memory of Chinese characters, and to adjudicate between conflicting accounts, one experiment with varying levels of IS was conducted by 40 participants with Chinese characters. In the experiment, participants performed an arithmetic task follow the visually presence of Chinese characters list. This prevented them from memorizing the characters during the retention interval. The arithmetic task was three single digits drawn from 1 to 9. Participants were instructed to be as accurate as possible in calculating the sum of the digits. There were four IS conditions: in one condition participants heard no speech during the trial (no IS); in a second condition they heard three seconds of speech that started 50 ms before the visual presentation of the first item and ended 50ms before the onset of the first digit (IS during list); in a third condition participants heard three seconds of speech that started 50ms before the onset of the first digit and continued until 50 ms before the word "Recall'' (IS during sum); in the fourth condition, participants heard six seconds of speech that started 50 ms before the first item and continued until 50 ms before the recall cue (IS throughout). The four conditions were distributed equally and randomly around the 120 lists resulting in 30 lists per IS condition. The results were as follows: (1) Irrelevant sound had a retroactive effect on material in memory. Given the constraints placed on the participants by the requirement to perform the arithmetic task (correctly) during the retention interval, IS presented during a filled retention interval and IS throughout had a reliable effect on list recall. (2) Performance with no IS was not reliably better than that on the IS during list. (3) Serial position effects were observed in serial recall. Accuracy steadily declined towards the latter part of the list, but the performance of the last one was better than that of the former one. The results confirmed that a retroactive effect of IS persists even when we use an overt speeded digit-reading task to prevent rehearsal. The Object-oriented episodic record model for the irrelevant-speech effect needs to be reconsidered.
    Related Articles | Metrics
    Effect of Cognitive Flexibility on Language Switching in Non-proficient Bilinguals: An ERPs Study
    LIU Huanhuan;FAN Ning;SHEN Xiangying;JI Jiangye
    2013, 45 (6):  636-648.  doi: 10.3724/SP.J.1041.2013.00636
    Abstract ( 1052 )  
    Recently, the effect of individual difference on language switching and switching cost has become a focus of many studies. Among the factors involved in the effect, L2 proficiency and age of acquisition are shown to be two important factors. Whilst the switching cost in proficient bilinguals is symmetrical, it is not in non-proficient bilinguals. Similar findings are obtained from bilinguals who acquire L2 early and those who acquire late. Moreover, some studies indicate that the switching cost may result from a system that is outside the word-recognition system in the mental lexicon. That is, inhibitive control also plays an important role in language switching as in general task switching. Previous EEG evidence shows that N2 (about 320 ms) that is usually found in the central frontal areas can be used as an indicator reflecting the top-to-down inhibitive control. In this study, the effect of cognitive flexibility (CF) on language switching was examined in non-proficient bilinguals using EEG method. Thirty non-proficient Chinese-English bilinguals were selected from 119 non-English major undergraduates by the Wisconsin Card Sorting Task (WCST). Among them, 15 participants were scored as high cognitive flexibility (CF), and another 15 were scored as low CF. A 2 (high-CF and low-CF) × 2 (L1 and L2) × 2 (repetition and switch) mixed design was employed. A picture naming task and a semantic-decision task were performed by the participants in order to examine the mechanism of language switching during language production and comprehension, respectively. In experiment 1, an immediate naming paradigm was used, and behavioral data were recorded. In experiment 2, a delayed naming paradigm was used, and only ERPs data were collected. In experiment 3, the participants were required to decide whether the presented word indicated a living meaning by pressing buttons. Both behavioral and ERPs data were obtained. The results showed that: (1) The switching cost in high-CF participants was symmetrical, whereas that in low-CF was not. Meanwhile, the mean amplitude and peak value of N2 was larger significantly in high-CF than in low-CF. (2) For high-CF participants, the peak value of N2 in repetitive language production task was larger in L2 repetitive condition than in L1 repetitive condition. (3) In contrast, the peak value of N2 in repetitive language comprehension task was larger in L1 repetitive condition than in L2 repetitive condition. These results suggest that CF affects language switching and its cost. The capability of inhibitive control have a key role in bilingual’s language switching of language production and comprehension. Language switching and its cost in non-proficient bilinguals result from a system that is outside the word-recognition system in the mental lexicon.
    Related Articles | Metrics
    The Role of Surface Color Feature Cue in Object Persistence under Two Different Spatiotemporal Conditions
    XU Baihua;LI Yuming;CUI Xiangyu
    2013, 45 (6):  649-657.  doi: 10.3724/SP.J.1041.2013.00649
    Abstract ( 627 )  
    Whether surface feature information could become a cue in object persistence has been a controversial issue. According to the object file theory by Kahneman, Treisman and Gibbs (1992), spatiotemporal information, rather than surface feature information, is the key cue in object persistence. This viewpoint was supported by Mitroff and Alvarez’ (2007) experiments using an object-reviewing paradigm. They found that participants could not establish object correspondence based on surface feature cues. However, Moore, Stephens and Hein (2010) exchanged the features of two objects abruptly in their experiments and found that the object correspondence operation could be established on the basis of surface feature cues. In the present study, the role of surface feature cues in object persistence was examined under two different conditions. In one condition, the spatiotemporal cue was ambiguous; in another condition, the spatiotemporal cue and the surface color feature cue were either congruent or conflicting. The present study used an object-reviewing paradigm and added a “tunnel” stimulus in order to control the role of spatiotemporal cues. Each “tunnel” stimulus was composed of a large circle with rectangles located at four points around the circle, to the left and right of the circle and above and below it. The left and right rectangles were entrances and the upper and lower rectangles were exits. On each trial, two objects entered the tunnel at same time from the left and right entrances, and then exited from the upper and lower exits respectively. The objects were not visible when they were in the tunnel. There were two kinds of tunnels. One was “dual-route tunnel”; another was “single-route tunnel”. When the dual-route tunnel was used, the two objects moving into the tunnel would leave the tunnel from the upper and lower exits at random, therefore the spatiotemporal cue was ambiguous. When the single-route tunnel was used, each object would leave the tunnel from one specific exit. Several unfamiliar figures were used as the test stimuli to minimize the possibility of verbal encoding of the stimuli by participants. Participants were instructed to judge whether the two figures on the objects after they left the tunnel were the same as the two figures on the objects before they entered the tunnel. Object specific previewing benefit was calculated to estimate the role of the surface feature in object persistence. Two experiments were conducted. In experiment 1, the dual-route tunnel was used and evidence for OSPB was found when the spatiotemporal cue was ambiguous. In experiment 2, the single-route tunnel was used and evidence for OSPB was also found, but the size of the effect was smaller in the conflicting condition than in the congruent condition. These results suggest that object persistence could be based on the surface feature cue when the spatiotemporal cue is ambiguous, and that the surface feature cue could also play a role in object persistence when the spatiotemporal cue was present.
    Related Articles | Metrics
    The Development of Future Planning and Its Relation with Parenting Behaviors During Early and Middle Adolescence: The Mediating Effects of Behavioral Autonomy
    YU Fengjie;ZHAO Jingxin;ZHANG Wenxin
    2013, 45 (6):  658-671.  doi: 10.3724/SP.J.1041.2013.00658
    Abstract ( 1570 )  
    An important feature of individuals’ thinking and behavior is that they are future-oriented. Future planning, a process that involves constructing plans and realizing these plans, has been extensively studied by researchers according to two dimensions, i.e., exploration and commitment. Literature indicates that future planning is particularly important for many aspects of adolescents’ current and later adjustment. In recent years, with the inspiration of the notions of Developmental Contextualism, researchers have become increasingly interested in adolescents’ future planning in specific cultural and/or social contexts, especially in those contemporary societies with rapid social and economic changes. However, the overwhelming majority of the research in this area was conducted in Western cultures and research on how contemporary Chinese adolescents plan for their future has been rare. Methodologically, existing research findings were mainly based on studies of cross-sectional design, resulting in very limited information available about the developmental changes of future planning with age. Besides, family is the more proximal context that is salient for adolescents’ development, in what ways that parenting behaviors influence adolescents’ future planning and whether the associations between parenting and future planning are mediated by individual characteristics such as autonomy are important questions of interest. The present study was designed to examine the development of future planning, the concurrent and longitudinal relationships between future planning and parenting behaviors, and the possible mediating role of behavioral autonomy in the associations with a sample of Chinese adolescents. Using a short-term longitudinal design, a total of 948 adolescents (499 males, 449 females) of grades 8 and 10 attending junior and senior high schools respectively in Shandong province of China were investigated during a period of one and a half years. At Time 1, those adolescents completed a series of questionnaires assessing future planning (i.e., exploration and commitment) concerning education and occupation, parenting behaviors (i.e., acceptance/involvement and strictness/supervision), and behavioral autonomy. A year and a half later, at time 2, they completed the future planning questionnaire again. The main findings were as follows: (1) the developmental patterns of future planning for education and occupation varied between early and middle adolescents. While middle adolescents reported increased exploration for education with age, no significant changes were observed for early adolescents. Both early and middle adolescents reported no significant changes in their commitment to future education, and their exploration for and commitment to future occupation between Time 1 and Time 2. Early adolescents reported more exploration for and commitment to occupation than middle adolescents at both Time 1 and Time 2. (2) At Time 1, for both early and middle adolescents, parental acceptance/involvement was strongly associated with future planning (exploration and commitment) concerning education and occupation in a positive way, whereas parental strictness/supervision only positively predicted adolescent exploration for and commitment to future education. Even after controlling for future planning at Time 1, parental acceptance/involvement still positively predicted future planning concerning education and occupation at Time 2; parental strictness/supervision did not predict adolescent future planning longitudinally. (3) Adolescents’ behavioral autonomy served to partially mediate the concurrent and longitudinal associations between parental acceptance/involvement and future planning concerning education and occupation. The results demonstrated that the development of Chinese adolescents’ future planning was shaped by the specific culture and age-graded tasks, which further supports the notion that future planning is the process of socialization.
    Related Articles | Metrics
    The Effects of Moral Self-Regulation on Prosocial Behavior and Rule Infraction
    LI Gu;ZHOU Hui;DING Ruyi
    2013, 45 (6):  672-679.  doi: 10.3724/SP.J.1041.2013.00672
    Abstract ( 2247 )  
    Moral self-regulation, the dynamic process through which an individual’s moral behavior history influences subsequent moral action, includes a negative-feedback mechanism. When people recall their own moral (immoral) behavior, moral self-perception will deviate from the ideal moral self-image, motivating them to engage in decreased (increased) moral action or increased (decreased) immoral action to compensate their bolstered (threatened) moral self-perception, and thereby activating a moral licensing (cleansing) effect. Whereas moral licensing and cleansing have been frequently reported in studies on prosocial behavior, less is so in rule infraction (e.g., cheating). In the current study, the authors designed two experiments to study the effects of moral self-regulation on prosocial behavior and rule infraction. We predicted that after moral (immoral) priming, participants would demonstrate less (more) charitable giving and more (less) cheating. Moreover, because people are more motivated to avoid harms than to approach benefits, the moral self-regulation process may be asymmetric in terms of prescriptive morality (what we should do) versus proscriptive morality (what we should not do). Specifically, we hypothesized that moral licensing would be more readily observed in charitable giving than in cheating; in contrast, we expected the reverse in moral cleansing. In Study 1, 155 undergraduate students were randomly assigned into positive-traits, negative-traits, or neutral-words condition. In each condition, participants were asked to copy a list of positive or negative moral traits, or inanimate objects; they were also instructed to recall past self-experience relevant to the keywords that they had copied. After the priming task, participants were invited to make a hypothetical donation. A one-way Analysis of Variance (ANOVA) was conducted to examine the influence of moral priming on the amount of money that participants intended to donate. In Study 2, 96 undergraduate students’ moral self-perception was manipulated using the same method as that in Study 1. Then, participants’ cheating behavior was observed. A 3×2 contingency table and a one-way ANCOVA were conducted to explore the effects of moral priming on the occurrence rates and severity of cheating among the three conditions, respectively. Results indicated that participants in the positive-traits condition were more willing to donate higher amount of money than those in the negative-traits and neutral-words conditions (Study 1). Moreover, participants in both experimental conditions were less likely to engage in cheating than those in the control condition, and their cheating behavior was less severe (Study 2). Taken together, the current study partially confirmed the moral cleansing effect in the proscriptive, but not the prescriptive morality system. However, the opposite of moral licensing effect was observed in both conditions. In conclusion, the motivational mechanism of moral action could be far more complex than previously suggested. For one thing, the self-regulation process could be different in prescriptive versus proscriptive morality. For another, people’s moral action may be influenced by both moral self-regulation (a compensatory process) as well as moral identity (a consistent process). To better understand and predict moral actions, future research should investigate both processes in the contexts of proscriptive and prescriptive moralities.
    Related Articles | Metrics
    The Impact of Employee Career Adaptability: Multilevel Analysis
    YU Haibo;ZHENG Xiaoming
    2013, 45 (6):  680-693.  doi: 10.3724/SP.J.1041.2013.00680
    Abstract ( 1801 )  
    With high speed of mobility and end of life-long employment of Chinese employees, it is critical to investigate the impact of employee’s career adaptability on organizational success. However, the relationship between career adaptability and organizational success is mixed due to the so-called management dilemma of career adaptability (when the employee’s career adaptability is high, his job performance and turnover intention are also high.). Moreover, the role of career adaptability as the core of self-career management capacity among the relationships between organizational career management and individual turnover intention/job performance have not been known. So, in this study, we investigated the relationships between employee’s career adaptability and their turnover intention, and job performance, and the moderating role of tenure over these relationships. And based on the multilevel theory, we propose that career adaptability is a mediator between organizational career management and individual turnover intention/job performance. We collected data from 485 employees in 54 Chinese enterprises. Employees completed the career adaptability questionnaire at the first time wave, and completed the turnover intention questionnaire two weeks later, while one direct supervisor of each employee was also invited to complete the job performance questionnaire and organizational career management questionnaire. For data analyses, we used SEM (structural equation model), HRA (hierarchical regression analysis) and HLM (hierarchical linear model). The results of SEM showed that career adaptability was negatively correlated to turnover intention, and positively associated with supervisor-rated performance. The results of HRA showed that tenure moderated the relationships between career adaptability and turnover intention/supervisor-rated job performance. This means that the correlation between career adaptability and turnover intention was significantly negative for employees with short-tenure, but not for employees with long-tenure. The correlation between career adaptability and supervisor-rated job performance was significantly positive for employees with short-tenure, but not for employees with long-tenure. So the findings were different from the current literature. The HLM results indicated that career adaptability fully mediated the relation between organizational career management and supervisor-rated job performance, but had no mediating effect on the relation between organizational career management and turnover intention. Therefore, individual career adaptability is the key mediating variable between organizational career management and self-career management. According to the career construction theory, the complex relationships between employee’s career adaptability and organizational success (for example turnover, job performance) and individual career success (for example subjective and objective success) should be tested in the future. More moderators of the relationships between career adaptability and organizational success / individual career success should be explored. Meanwhile the mediating role of career adaptability among the relationships between organizational career management/self-career management and individual career success /organizational performance could be further explored, so that organizations could establish the career adaptability-based human resource development system by integrating organizational career management and self-career management.
    Related Articles | Metrics
    A Comparison of Item Selection Methods for Controlling Exposure Rate in Cognitive Diagnostic Computerized Adaptive Testing
    MAO Xiuzhen;XIN Tao
    2013, 45 (6):  694-703.  doi: 10.3724/SP.J.1041.2013.00694
    Abstract ( 715 )  
    Item exposure rate is the utilization frequency of an item. When the exposure rate is high, examinees will likely share item content. If there are too many over-exposed items, test security and hence the validity of the assessment will certainly be compromised. Furthermore, with a lot of under-exposed items having low or zero item-exposure rates, the manpower and financial resources spent on item construction will be wasted and the item pool construction will become more challenging. Item exposure control is, therefore, an important issue in computerized adaptive testing (CAT). Cognitive diagnostic CAT (CD-CAT) combines and makes use of the strengths of cognitive diagnosis theory and CAT. The system will be able to provide information on the knowledge competence of the examinees by administering fewer items than traditional assessment. Based on the a-stratified method and the item eligibility method in regular CAT, the present study proposed and compared the performance of six techniques, namely, (a) the item eligibility (IE) method, (b) the stratified multistage (SM) approach, (c) the stratified multistage-item eligibility (SMIE) method, (d) the restrictive threshold (RT) method, (e) the maximum modified priority index (MMPI) method, and (f) the restrictive progress (RPG) method. With noting it that the SM approach is similar to the a-stratified method in item selection steps. The SM approach, however, different with the a-stratified method firstly in that it stratifies the remaining item pool based on the values of item information at the estimated attributed mastery pattern while the a-stratified method is based on the values of item discrimination parameter a. Secondly, in the SM method, the remainder item bank are stratified into a number of levels before the selection of each item, whereas in the a-stratified method, the item pool is stratified only once before the test and all the examinees have the same item strata. The SMIE method combines the SM and the IE method. MATLAB (R2010a) was used in the simulation experiments to write the CD-CAT code and the deterministic inputs, noisy “and” gate (DINA) model was applied in this study. Results showed that: (a) the SM method used in CD-CAT produced widely distributed item exposure by increasing the exposure rates of most items and fully utilizing the item pool but without greatly diminishing the maximum exposure rate and measurement accuracy; (b) other than a few items, the exposure rates of the IE method were lower than the setting maximum exposure rate, but most items still had extremely low exposure rates and hence resulting in a narrow distribution of item exposure and the highest measurement precision; (c) SMIE and RT methods behaved similarly in that not only could they increase the utilization frequency of the under-exposed items but they could also decrease the maximum exposure rate to a certain extent; (d) the MMPI and the RPG methods performed similarly with almost evenly distributed item exposure but at the great sacrifice of the measurement precision. As a whole, the performances of different methods in the order of their measurement accuracy are IE, SM, SMIE, RT, RPG and MMPI. The order in terms of their performances in exposure control is: MMPI, RPG, SMIE, RT, SM and IE. All in all, the SMIE and RT methods are able to balance measurement accuracy and item exposure well.
    Related Articles | Metrics
    Modeling Self-reported Instrument Data with Gene Expression Programming
    QIAN Jinxin;YU Jiayuan
    2013, 45 (6):  704-714.  doi: 10.3724/SP.J.1041.2013.00704
    Abstract ( 766 )  
    It is often difficult to represent the complex relations among psychological variables with traditional analytical models like regressions. Supposedly, neural networks and support vector regression machine can be used instead. However, the limitation is that these models are recessive. Gene expression programming (GEP) can be used to handle these models with observable variables. At present, most of the data using GEP models are obtained with objective methods. But a lot of the psychological measurement data are obtained from self-report instruments and are affected by many subjective factors. Could these kinds of data be used in GEP models? How large is the modeling error? Is there any advantage in using the GEP modeling as compared with the multivariate linear regression or the polynomial regression modeling? Is the GEP modeling more accurate than neural networks and support vector regression machine modeling? All the above issues would be explored in this paper. The responses of 400 middle school students were obtained with the Williams creativity assessment packet and the need for cognition scale. A total of 17 students were deleted because of the abnormality in responses and the data from 383 students were retained for modeling. Common method biases had not been found with the Harman’s single-factor test. Five parameters of gene expression programming were optimized with the uniform design. These parameters were head length, gene number, fitness function, chromosome number and mutation probability. There were nine levels for each parameter, each established under different testing conditions respectively. The condition with maximum fitness was obtained through experiments. The GEP program was repeated 10 times under this condition. The accuracy of the models was calculated and the model with the minimum error was found, of which the expression tree was drawn. The models of the relations between need for cognition and creativity personality traits were established using BP neural networks, support vector regression machine, multivariate linear regression and polynomial regression respectively. These models were compared with the model using gene expression programming. The results showed that: (a) the accuracy of model 10, with four independent variables, was the highest; (b) the expressions of these ten models were different but their predictive errors were very close, thus supporting the robustness of the GEP modeling method; and (c) the predictive errors of different models were: GEP, 1.28; BP networks 2.76; support vector regression machine 2.31; polynomial regression 3.21; multivariate linear regression 3.86 respectively. It can be concluded that: (a) data from self-reported instruments can still be modeled with gene expression programming even though these data are affected by many subjective factors; (b) the GEP modeling is more accurate than the other intelligent computing methods (neural networks, support vector regression machine, etc.) and traditional statistical methods (multivariate linear regression, polynomial regression, etc.), and (c) the models established with GEP are robust; their predictive accuracy is similar even though their mathematical formulae are quite different.
    Related Articles | Metrics