以電腦化複雜廣度作業測量兒童工作記憶

Computerized Complex Span Measurement of Children’s Working Memory

本研究從記憶理論和認知發展的角度，探討如何以電腦化廣度作業有效評估兒童的工作記憶發展。據此，本研究檢驗改編自Camos與Barrouillet（2011）的動物顏色廣度作業的信度與效度，同時，藉由以此作業了解4至7歲兒童工作記憶的發展情形。臺灣北部典型發展兒童共333名（男生167名）參與本研究，依年齡分為四歲、五歲、六歲和七歲四組。在此作業中，兒童需在電腦控制的時間限制下，交替完成視覺圖像呈現的儲存作業（動物記憶）與處理作業（顏色叫名），最後依序回憶記憶項目。信度分析顯示此作業有理想的內在信度與折半信度。在效度方面，幅合效度分析顯示此作業與數字順背廣度作業、數字逆背廣度作業、魏氏幼兒智力測驗工作記憶分測驗和托尼非語文智力測驗有顯著正相關；試題分析以及Rasch模式分析指出，此作業的難度隨廣度增加而增加，試題鑑別度適中且與Rasch模式有良好的適配度。此外，單因子變異數分析發現工作記憶廣度隨年齡增長而有顯著增加，顯示此作業能夠反映成熟造成的發展差異。綜合上述結果本研究建議，動物顏色廣度作業可有效評估幼小兒童的工作記憶發展，而電腦化複雜廣度作業準確控制時間與測驗程序，可增進兒童工作記憶測量的信度與效度。

Psychology researchers and educators have determined that many behavioral and cognitive abilities that we want to understand, measure, and manipulate are related to the function of working memory. Working memory, a cognitive system with limited capacity, temporarily stores and processes information relevant to ongoing activity (Baddeley, 2000). Unlike shortterm memory, working memory has a longer developmental period and is affected by the maturation of the brain’s frontal lobes (Cowan, 1997). This cognitive ability is essential not only for children’s interactions with the external world and acquisition of new knowledge but also for their daily functioning.
Although different theories provide various explanations for the development of working memory, most theories hold that working memory involves two cognitive operations: storage and processing (Baddeley, 2000; Engle & Kane, 2004; Gavens & Barrouillet, 2004). Therefore, tasks that measure a person’s working memory span typically require the person to temporarily remember a memory item while performing another processing task or to engage in some form of processing related to the memory item. They are specifically called dual-task or complex span tasks (Redick et al., 2012). Such tasks differ from simple span tasks, which require the testee to retain memory items for a period of time and then recall them sequentially.
Assessing children’s working memory by using complex span tasks requires clear, simple, and easy-to-understand instructions. The tasks must be designed with stimuli, content, presentation formats, and response methods that are consistent with children’s cognitive abilities. In addition, test procedures should account for children with varying levels of working memory capacity. For those with low working memory capacity, the tasks should avoid inducing a sense of frustration or helplessness. By contrast, for children with high working memory capacity, the test should provide opportunities for them to fully demonstrate their abilities. A common approach involves gradually increasing the difficulty of the tasks, requiring the testee to memorize more items, and terminating the task when a specific level of difficulty cannot be completed (Gonthier et al., 2018; Pickering, 2006).
Memory span tests use the all-or-none scoring method (Conway et al., 2005). In this approach, if a child fails to recall a memory item or recalls it in the incorrect order, that item is marked as a failure. Such failures can lead to task termination, potentially affecting the accuracy of assessments of children’s working memory abilities. Alternatively, one procedure involves presenting all test questions of various difficulty levels in random order to each child (Conway et al., 2005; Unsworth et al., 2009). However, this method may frustrate younger children or those with lower working memory capacity when they encounter difficult questions, reducing their motivation to complete the task (Gonthier et al., 2018).
This study introduced a computerized complex span task designed to measure the working memory abilities of children aged 4 to 7 years and analyzed its reliability and validity by integrating data from multiple research projects and a master’s thesis. In addition, this study explored the development of working memory in Taiwanese children aged 4 to 7 years. A total of 323 typically developing children (164 boys) from North Taiwan participated in the study, and they were divided into age groups of 4, 5, 6, and 7 years. The study employed the animal-color span task, adapted from Camos and Barrouillet (2011), in which the children alternated between a storage task (animal memory) and a processing task (color naming) within a limited time. At the end of the task, the children were required to recall the memory items in the correct sequence. Scoring for the task was strict. Points were awarded only if both the color naming and animal recall sequences were correct. In addition to the animalcolor span task, all participants completed four additional memory and cognitive assessments, namely the Digit Forward Span Task (DF), the Digit Backward Span Task (DB), the Working Memory Subtest of the WPPSI-IV, and the TONI-4 Nonverbal Intelligence Test.
The distribution of the children’s performance on this task followed a normal curve, suggesting that the sampling was effective. Reliability analysis demonstrated strong internal consistency for the task, with a Cronbach’sαof 0.76 and good splithalf reliability (r = .77). Age-based analyses revealed consistent reliability across the age groups. Convergent validity analysis indicated significant positive correlations of this task with the DF task, the DB task, the working memory subtest of the WPPSIIV, and the TONI-4 Nonverbal Intelligence Test. These findings indicate that the task effectively measures individual differences in short-term memory, working memory, and nonverbal intelligence.
Age discrimination and item response theory (IRT) were employed to evaluate the construct validity of the animal-color span task. On the basis of IRT, the task’s difficulty increased progressively with the span levels, aligning with the Rosch model and demonstrating strong construct validity. Moreover, the conversion of item pass probabilities into item difficulty and a participant ability distribution yielded results that approximated a normal distribution. According to classical test theory, discrimination indices for spans 2 and 3 were high, indicating strong item discrimination at these levels. Most children successfully completed span 1 but faced limitations at spans 2 and 3. They considered span 4 to be challenging. Age-based analysis further revealed that children aged 4, 5, 6, and 7 years achieved the approximate maximum spans 1, 2, 2, and 3, respectively, which aligns with expected developmental trends. One-way ANOVA showed a significant increase in working memory capacity from ages 4 to 7 years, indicating that the task effectively captures the maturation of working memory.
These results demonstrate that the animal-color span task developed in this study, a computerized complex span task, is suitable for assessing working memory in children aged 4 to 7 years. The task demonstrates good reliability and validity and reflects developmental differences in working memory. In addition, the computerized complex span task offers several other advantages. (1) Theoretically, it aligns with the working memory model by incorporating both storage and processing tasks, controlling the cognitive load of processing tasks (Camos & Barrouillet, 2011), and measuring children’s ability to resist interference and maintain attentional control during the memory retention stage (Jarrold et al., 2011). (2) Methodologically, the stimulus size, presentation time, interval time, and task termination for the storage and processing tasks are all computercontrolled, which ensures standardized test procedures and reduces errors caused by human subjectivity. (3) In terms of practical application, the standardized format enables researchers in different locations to implement consistent testing procedures (Bailey, 2012), facilitating cross-cultural and cross-study comparisons. (4) In future applications, the task’s standardized features can be further developed into online or group testing formats that do not require experimenter assistance.
This study adapted only one version of the animal-color span task designed by Barrouillet et al. (2009) and Camos and Barrouillet (2011). The other two versions of the task, which involve different processing conditions, were not included. In addition, the complex span tasks developed in the aforementioned studies were originally intended to investigate the mechanisms of working memory development in children aged 5 to 7 years. This study extended the applicability of one version of the task downward to include 4-year-old children. However, additional research is needed to determine whether this task is suitable for children younger than 4 years or those older than 7 years.