近年來，數位遊戲式語言學習已逐漸由探討遊戲是否具學習成效，轉向探討不同遊戲設計機制如何影響學習歷程。其中，不確定性（Uncertainty）被視為維持玩家投入與驅動探索行為之核心遊戲機制，然其對於日文語言學習，尤其是高度情境依賴、語意細膩之日文擬聲擬態詞學習中的應用與實證研究仍相對有限。基於此，本研究開發一款以2D角色扮演遊戲（RPG）為基礎之數位遊戲式日文擬聲擬態詞學習系統「歸來」，並結合《得與失》、《隱藏區域與道具》及《劇情選擇》三種不確定性機制，設計有不確定性機制與無不確定性機制兩個版本，藉以探討有無不確定性機制對學習者之日文擬聲擬態詞學習成效、遊戲成效、學習動機、遊戲心流及好奇心之影響，以及上述變項間之相關性。 本研究為準實驗研究設計，總共有46名參與者，並隨機平均分派至有不確定性機制組與無不確定性機制組（各23人），於系統體驗前後分別進行日文擬聲擬態詞能力前後測，並於系統體驗後填寫學習動機、遊戲心流及好奇心量表，並蒐集學習者於系統內之各項歷程行為資料。資料分析採無母數統計方法，組內差異以魏克生符號檢定（Wilcoxon Signed-Rank Test）進行檢驗，組間差異以曼－惠特尼U檢定（Mann-Whitney U Test）進行比較，並以斯皮爾曼等級相關分析（Spearman’s Rank Correlation）探討各變項間之關聯性。 研究結果顯示，兩組學習者於日文擬聲擬態詞學習成效上皆有顯著進步，惟組間並未呈現顯著差異，顯示敘事驅動RPG基礎架構本身即具備促進詞彙習得之效力。在遊戲成效方面，有不確定性機制組之戰鬥測驗發起頻率明顯高於無不確定性機制組，惟兩組之答題正確率與單字習得總量則無明顯差異，顯示不確定性機制主要提升行為投入頻率而非答題精準度。在學習動機與遊戲心流方面，有不確定性機制組之外在動機、整體動機，以及心流之控制、參與與整體心流表現，均明顯優於無不確定性機制組，而內在動機與樂趣感則無明顯差異。在好奇心方面，兩組之興趣型、匱乏型及整體好奇心表現則相近，未見明顯差異，顯示敘事探索情境本身已足以激發學習者之好奇心。此外，相關分析結果顯示，兩組學習者之學習動機、遊戲心流與好奇心彼此皆呈正向關聯，且無不確定性機制組之心理構面間關聯程度普遍較為緊密；而系統內各類行為指標與心理構面之關聯，則因有無不確定性機制而呈現不同之模式。 整體而言，本研究證實不確定性機制能在不損害學習成效之前提下，有效提升學習者之行為投入、外在動機及遊戲心流體驗，為數位遊戲式語言學習中不確定性機制之設計提供實證依據，亦可作為語言教師、教材設計者及教育遊戲開發者規劃日文擬聲擬態詞及其他高情境依賴詞彙之數位學習系統時之重要參考。;In recent years， research on Digital Game-Based Language Learning (DGBLL) has gradually shifted from verifying whether games are effective for learning toward investigating how different game design mechanisms influence the learning process. Among these mechanisms， uncertainty has been regarded as a core design element that sustains players’ engagement and drives exploratory behavior; however， empirical research on its application in Japanese language learning—particularly for Japanese onomatopoeia and mimetic words， which are highly context-dependent and semantically nuanced—remains relatively limited. To address this gap， the present study developed a 2D role-playing game (RPG)-based digital game-based learning system， “Return” for Japanese onomatopoeia and mimetic words， incorporating three uncertainty mechanisms: Gain and Loss， Hidden Areas and Items， and Story Choices. Two versions of the system—with and without uncertainty mechanisms—were designed to examine their effects on learners’ Japanese onomatopoeia and mimetic word learning performance， in-game performance， learning motivation， game flow， and curiosity， as well as the correlations among these variables. A quasi-experimental design was adopted， there were 46 participants randomly and evenly assigned to the uncertainty-mechanism group and the non-uncertainty-mechanism group (23 each). Pre- and post-tests of Japanese onomatopoeia and mimetic word ability were administered before and after system use， and questionnaires on learning motivation， game flow， and curiosity were completed afterward. Learners’ in-system process data were also collected. Non-parametric statistical methods were employed: within-group differences were examined using the Wilcoxon Signed-Rank Test， between-group differences were compared using the Mann-Whitney U Test， and Spearman’s Rank Correlation was used to examine the relationships among variables. The results showed that both groups demonstrated significant improvement in Japanese onomatopoeia and mimetic word learning performance from pretest to posttest， but no significant between-group difference was found， suggesting that the narrative-driven RPG framework itself was sufficient to promote vocabulary acquisition. Regarding in-game performance， the uncertainty-mechanism group initiated combat quizzes significantly more frequently than the non-uncertainty-mechanism group， although no significant differences were found in answer accuracy or total vocabulary acquired， indicating that the uncertainty mechanism primarily enhanced behavioral engagement frequency rather than answering precision. In terms of learning motivation and game flow， the uncertainty-mechanism group showed significantly higher extrinsic motivation， overall motivation， and the flow dimensions of control， engagement， and overall flow than the non-uncertainty-mechanism group， while intrinsic motivation and enjoyment showed no significant difference. Regarding curiosity， both groups exhibited similar levels of interest-type， deprivation-type， and overall curiosity， with no significant between-group difference， suggesting that the narrative exploration context itself was sufficient to stimulate learners’ curiosity. Furthermore， correlation analyses revealed positive associations among learning motivation， game flow， and curiosity in both groups， with the non-uncertainty-mechanism group generally exhibiting stronger associations among these psychological constructs; the relationships between in-system behavioral indicators and psychological constructs also differed depending on the presence of uncertainty mechanisms. Overall， this study confirms that uncertainty mechanisms can effectively enhance learners’ behavioral engagement， extrinsic motivation， and game flow experience without compromising learning performance， providing empirical evidence for the design of uncertainty mechanisms in digital game-based language learning. The findings also offer practical implications for language teachers， instructional material designers， and educational game developers in designing digital learning systems for Japanese onomatopoeia and mimetic words， as well as other highly context-dependent vocabulary.
