Examining the Effectiveness of Mnemonics Serious Games in Enhancing Memory and Learning: A Scoping Review

Abstract

Mnemonics hold potential for promoting long-term memory. Hence, they are being leveraged in serious games aimed to support long-term retention and retrieval of information. However, there is limited work focused on synthesizing the published research and findings on mnemonics serious games with a view to uncovering the extent of their application and effectiveness. This scoping review aims to bridge this gap. Articles were retrieved from four databases (ACM Library, IEEE Xplore, Scopus, and Web of Science). The criteria for inclusion were that the papers must be user studies that focused on mnemonics and serious games at the same time, were written in English, and were published in peer-reviewed journals or conferences. Two researchers, with the guidance of a senior researcher, independently and collaboratively assessed the eligibility of the retrieved papers using the PRISMA flowchart, elicited the relevant data, and tabulated the results in tables and charts using the GPS (game play, purpose, and scope) model. There were 12 papers that were accepted in this scoping review. Overall, most of the mnemonics serious games had a positive effect on memory, suggesting that they hold potential for promoting long-term memory, especially in memorization-intensive instructions, where a good number of students still struggle to retain taught material due to pedagogical, personal, and social challenges. However, more research still needs to be conducted, especially in the area of player-created mnemonics and teaching users how mnemonics can be effectively created using visualization and elaboration techniques.

1. Introduction

It is increasingly becoming challenging for students at all levels of education to learn and remember material taught in classrooms, especially in memory-intensive subjects and courses, as a result of pedagogical, personal, social, economic, technological, and environmental factors [1,2,3,4,5]. Apart from the way memory-intensive subjects such as biology [6], psychology [7], history [8], and English/foreign languages [9] are taught, what students bring to the learning task, and the very nature of these subjects, which make them difficult to learn [6], several other factors can be identified. One of these factors is the fact that some college students have to work and study at the same time to enable them to pay for their education, leaving them with little or no time to study [10]. A second factor is that the volume of teaching materials is increasing over time due to the ever-growing body of knowledge. A third factor is that students have limited opportunities to study the taught material in a collaborative context, partly due to the prevailing culture of individualism that promotes independence and individual learning over cooperative learning [4]. A fourth factor is that the modern generation’s memory is declining under the distracting influence of addictive technologies such as social media and video-streaming services, which are available to them 24/7 [11,12]. For example, in Li and Nietfeld’s [3] study, a participant remarked, “I understand the material but am not sure if I can remember it later”. The challenges faced by students in learning and remembering taught material in the modern world have resulted in the search for and research of effective learning strategies that can help students learn better and remember learned content easily. Remembering, despite being a low-order cognitive process in Bloom’s taxonomy, is the foundation upon which higher-order cognitive processes such as analysis and evaluation build [13]. For example, if you cannot remember what you have learned, despite understanding it, it is difficult to describe or discuss that which you have learned. Therefore, remembering remains an indispensable activity in the learning process. As Jensen put it, “learning and memory are two sides of a coin. You can’t talk about one without the other” [14]. Prime among the many available techniques aimed at enhancing long-term memory and life-long learning is the use of mnemonics [1,15].

1.1. Mnemonics

Mnemonics are memory-aiding techniques that help to improve long-term memory and recall with the the aid of cues, concreteness, associations, and/or other types of connections that can be made with the target information to be remembered [5]. They are increasingly being used in memorization-intensive subjects and subjects such as biology [16,17] and psychology [18,19] to help students in high schools and tertiary institutions to learn taught material and remember it easily. Mnemonics have four main properties that make them effective in promoting long-term learning: meaningfulness, elaboration, association, and visualization [11]. These properties are utilized by instructors and students aware of their utility to develop useful, meaningful, and effective mnemonics that help students learn more effectively and remember the taught material in the long run. Research shows that mnemonics are effective in enhancing long-term memory. For example, in 1967, Miller found that students who regularly used mnemonic devices to support their study increased their test scores up to 77% [20]. In more recent times, mnemonics have been shown to enhance memory in a wide range of subjects, including but not limited to engineering [21], chemistry [22], psychology [18], mathematics/statistics [23], and economics [11].

1.2. Serious Games

Moreover, technological tools such as serious games designed for the purpose of education are being used to support students to help them learn and remember taught materials effectively. Serious games are applications designed to educate rather than solely entertain using game elements [24]. While entertainment is certainly not the primary goal, serious games leverage entertainment elements to motivate players to take action in and outside of the game that can enhance their learning. Serious games have been shown to be helpful in various fields, leading to their increased use. Sharifzadeh et al. [25] found a rise in health education games over the second half of the last decade and they have become more generalized as opposed to focusing only on one medical condition. Graafland et al. [26] found that serious games about medical care have improved teamwork, multitasking, and visual and cognitive skills. Traver et al. [27] evaluated a chemistry game named E-CHEMMEND. They found that the games helped players to memorize aspects of the periodic table. Moreover, they reported that the game received players’ interest and appreciation to the extent that they continued playing it in their free time and competed with one another. Micallef and Arachchilage [28] also assessed the efficacy of a modified “4 Pics 1 Word” game in enhancing memory. They found that the game improved short-term memory of answers to security questions and showed potential for accomplishing the same with long-term memory. Similarly, Johnson et al. [29] found that games about domestic energy consumption have generally led players to improve their real-life behavior regarding energy conservation.

1.3. Mnemonics Serious Games and Lack of Reviews

More recently, serious games that use mnemonics to teach memorization-intensive subjects have started gaining traction in the promotion of long-term learning and recall of taught material. However, mnemonics are still being perceived by some as having a low utility, with reasons such as they are difficult to implement in certain contexts and only benefit students for a limited number of materials and for a limited amount of time regarding retention being put forward [1]. Hence, it becomes pertinent to know the extent to which they have been used and are effective in enhancing students’ learning, particularly in the context of serious games that are becoming more and more popular and entrenched in students’ education and enhancing episodic and long-term memory [30,31,32]. Unfortunately, there is a lack of systematic or scoping reviews of mnemonics serious games synthesizing and summarizing the various research outcomes and results, which are currently fragmented, to help researchers, educators, and developers know the extent to which these types of games help enhance students’ learning and long-term memory, especially in memory-intensive subjects. To bridge this gap, we set out to map out the landscape of mnemonics serious games aimed to promote education and long-term memory in general. It is our hope that the synthesized and summarized findings will provide stakeholders with a bird’s-eye view of the research landscape by uncovering the research gaps and open questions yet unanswered and bringing to the fore research opportunities for future work.

1.4. Research Questions

Based on the foregoing, the objective of this scoping review is to summarize and synthesize the characteristics and research findings of the existing body of work on mnemonics serious games. Specifically, we sought to answer the following research questions:

• How have mnemonics been utilized in serious games to promote learning and retention in memorization-intensive subjects?
• How are the mnemonics in serious games generated and by whom?
• How effective is the use of mnemonics in serious games aimed to promote learning and content retention in memorization-intensive subjects?
• To what extent have serious games been utilized to teach mnemonic techniques?

2. Methods

In this section, we discuss the methods we used in our scoping review. First, we explain our eligibility criteria for papers. Second, we state which databases we searched, what search strings we used, and what data we collected and how. Third, we discuss our PRISMA-based screening process performed on found papers. Fourth, we explain our adaptation of Djaouti et al.’s GPS model. Fifth, we describe our synthesis of results and our risk-of-bias analysis.

2.1. Eligibility Criteria for Papers

We used the eligibility criteria shown in

Table 1. Eligibility criteria.

Inclusion Criteria	Exclusion Criteria
The article must be a journal or conference paper.	The article is a book chapter, literature review, etc.
The article must be written in English.	The article is not written in English.
The article must be about a mnemonics serious game aimed to improve memory or learning easily forgotten information.	The article is not about a mnemonics game aimed to improve memory or learning easily forgotten information.
The article must be based on primary and real user data.	The article is based on secondary or simulated data.
The article can relate to any subject matter, provided the game is aimed at promoting memory and retention.	The article is not not related to education or learning.

to determine the eligible papers during the screening process. For example, the scoping review only focused on papers written in English and serious games whose goal is to use mnemonics to promote learning and retention in any subject matter.

2.2. Databases, Search Strings, and Data Collection Process

To find relevant papers, 4 databases (ACM Digital Library, IEEE Xplore, Scopus, and Web of Science) were searched on 3 May 2024 and 4 May 2024. We used the search strings provided in

Table A1. Search strings for each database.

Database	Search Strings
ACM Digital Library	[Title: “serious game*”] AND [[Title: mnemonic*] OR [Title: memor*]]
	[Abstract: “serious game*”] AND [[Abstract: mnemonic*] OR [Abstract: memor*]]
	[Keywords: “serious game*”] AND [[Keywords: mnemonic*] OR [Keywords: memor*]]
IEEE Xplore	(“Document Title”: “serious game*”) AND (“Document Title”: mnemonic* OR “Document Title”:memor*)
	(“Abstract”: “serious game*”) AND (“Abstract”: mnemonic* OR “Abstract”:memor*)
	(“Author Keywords”: “serious game*”) AND (“Author Keywords”: mnemonic* OR “Author Keywords”: memor*)
Scopus	TITLE-ABS-KEY(“serious game*” AND (mnemonic* OR memor*))
Web of Science	(TI = (“serious game*” AND (mnemonic* OR memor*))) OR (AB = (“serious game*” AND (mnemonic* OR memor*))) OR (AK = (“serious game*” AND (mnemonic* OR memor*)))

. We added “memor*” in case there were papers about mnemonics that did not say so explicitly. However, afterwards we also became concerned with synonyms for “mnemonic”, so for all our original search strings, we replaced “mnemonic*” with “acronym*” and later “initialism*”. Yet, the new search strings returned almost no new papers that were found with the original search strings, and the few that did appear were clearly irrelevant to our topic even at a quick glance. As such, we maintained our original search strings.

The retrieved papers were based on titles, abstracts, and keywords. Using each database’s export feature, we exported the papers to 4 .csv files (1 for each database). We then combined the data into one larger .csv file (which included duplicates) before transferring it into a Google Sheet. We had the columns “Document Title”, “Publication Year”, “Abstract”, “Authors”, and “Paper Type”. Next, for sets of rows that had the same document title, we marked the second and onward rows as duplicates. All rows not marked as duplicates had their featured papers screened.

Due to the lack of papers that focused on mnemonics serious games, we sought additional papers by typing in “mnemonics serious games” on Google Scholar and using the snowballing method. Finally, we proceeded to conduct a full-paper review on any papers that seemed relevant based on their title alone.

2.3. PRISMA-Based Screening of Papers

We used the PRISMA-ScR (Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews) flowchart [33] to screen the papers that we found from the 4 databases. Two (2) researchers (KF and OI), guided and supervised by the senior researcher (KO), screened all unique papers independently. The papers were first screened based on their title and/or abstract, and those that passed the screening were subjected to a full-text review. During both phases of screening, papers that KF and OI disagreed on in terms of passing the inclusion criteria were discussed and resolved by them. KO oversaw both phases of the screening process. Moreover, the Google Scholar papers were found by KO and passed to KF, who performed a full-paper review on them. Upon completion of the full-paper review, KF notified KO about which papers were rejected as well as why they were rejected. Both researchers then discussed the rationale for acceptance and rejection of each paper and came to a consensus.

2.4. Adaptation and Usage of GPS Model

For the papers from the 4 databases that passed both phases of screening, as well as the Google Scholar articles that passed the full-paper review, we used an augmented/adapted version of Djaouti et al.’s [34] GPS model as an analytical framework in extracting and tabulating the data shown in

Table 2. Characteristics of included papers and their evaluated games.

Characteristic	Description
Authors	Who wrote the paper
Year	What year the paper was published
Title	Title of the paper
Region of study	Which region the study featured in the paper took place in
Target population	Number and characteristics of participants in the study
Name of game	Name of the game featured in the paper
Game type	The type of game
Mnemonic type	What mnemonic type was used by the game
Target subject	What field the game and/or study was about
Design approach	Whether the design of the game was based on literature or user studies
User-evaluated	Whether users provided feedback regarding the game
Game mode	Whether the game involved interaction with other users or not
Age group	What age group the participants belonged to
Level of education	Whether the games were meant for people in kindergarten, elementary school, middle school, high school, or university
Game outcome	How effective the game was in improving users’ memory
System	Whether the game is for research purpose or available to the public to use
Aim/objective of paper	What was done in the paper’s study
Takeaway of findings	A brief summary of the game’s effect on its users
Purpose	Whether the game was meant to impart knowledge on a particular subject (educative), provide information such as news (informative), persuade people to behave a certain way (persuasive), or does not seem to focus on any of these (subjective)
Type of study	Whether the study’s data was qualitative, quantitative or mixed
Comparison	Whether an experimental group was compared against a control group, another experimental group, both, or there was no splitting of participants into groups
Mnemonic generation	Whether the mnemonics were created by people or computers
Mnemonic source	Whether the mnemonics were created by players or games (or their creators)

in a Google spreadsheet, e.g., purpose of game, target population, etc. The GPS model provides a consise taxonomy of games based on three criteria: Game Play (G), Purpose (S), and Scope (S). Each of these criteria is further divided into types. For example, purpose is composed of two types: message and training, which are further subdivided into subtypes. Message represents the purpose of the message communicated by a game, which can be educative, informative, persuasive or subjective. On the other hand, training can either mental or physical [34].

2.5. Synthesis of Results and Risk-of-Bias Analysis

For the target subject and all other characteristics shown in Table 2, we kept track of their respective values, which papers are associated with each value, and the number and percentage of all papers that are associated with each value. We also counted the number of papers published on each year from 1996 to 2023. In addition, we counted the number and percentage of papers’ studies for every region that had at least one. Moreover, we conducted a risk-of-bias analysis on the studies. We kept track of each study’s risks of selection bias (bias arising from the randomization process), performance bias (bias due to deviations from intended intervention), attrition bias (bias due to missing outcome data), detection bias (bias in measurement of the outcome), and reporting bias (bias in selection of the reported result) [35]. We also kept track of the study risk distributions for each bias.

3. Results

In this section, we present the results of the scoping review, including the number of included papers and summary tables showing the main characteristics of the papers, which relate to the four main research questions. More specifically, we accomplish the following four things. First, we present the statistics of our PRISMA-based screening process. Second, we discuss each study’s authors, population, game details, goals, and findings. Third, we show the value distributions of the remaining characteristics (as well as year and region). Fourth, we discuss the results of our risk-of-bias analysis.

3.1. Statistics of PRISMA-Based Screening

The review followed the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) flowchart for the screening and inclusion of articles in the final analysis (Figure 1). A total of 729 papers were retrieved from the four databases that we searched. Removing around 35.7% (260/729) of them for being duplicates, we had 469 left. Next, we commenced the first proper phase of screening, which was based on title and/or abstract. As shown in

Table 3. Agreement between researchers during screening.

	Title/Abstract Screening	Full-Paper Screening
Both Accepted	16	7
Both Rejected	415	15
Only KF Accepted	25	3
Only OI Accepted	13	0
% Agreement	91.9	88
Cohen’s Kappa	0.415	0.737

, there were 38 disagreements between the two researchers (KF and OI), in which 25 were only accepted by the former and 13 by the latter. There were 431 agreements between them, in which 16 and 415 were accepted and rejected, respectively, by both researchers. This resulted in an agreement rate of around 91.9% (431/469) and a Cohen’s Kappa of around 0.415. We screened out around 94.7% (444/469) of them by title and abstract, which meant we had 25 remaining.

Finally, we performed the second phase of screening, which was based on full-paper review. As shown in Table 3, there were three disagreements between the two researchers (KF and OI), with all three associated papers accepted by the former only. There were 22 agreements between them, with 7 of the associated papers accepted by both and 15 rejected by both. This resulted in an agreement rate of 88% (22/25) and a Cohen’s Kappa of around 0.737. We screened out 64% (16/25) of those based on a full-paper review, which would have resulted in us being left with 9 papers to report. However, the senior researcher (KO) realized that around 88.9% (8/9) of these papers did not feature mnemonics as they were only retrieved due to the generic “memor*” term in our search strings aimed to ensure we retrieved as many papers as possible that focused on improvement of memory, but did not explicitly use the term “mnemonics” in the title, abstract or keywords. This meant we only had one paper left from the formal screening process. Moreover, 11 papers were retrieved through the Google (Scholar) search and the snowballing method by the senior researcher (KO), resulting in 12 papers in total for the final analysis.

3.2. Authors, Population, Game Details, Goal, and Findings of Each Study

The major characteristics for the papers, as described in Table 2, will be shown in charts and text. Each row in

Table 4. Summary of included articles and findings.

Authors (Year of Publication)	Population (Sample Size and Demographics)	Game Name, Game Type, Mnemonic Type	Aim and Objective of Study	Summary of Findings
Costuchen et al. (2022) [36]	Spain: 48; 21 experimental, 27 control	Roman Palace; puzzle and simulation; loci method	Compare the effectiveness of Roman Palace based on the loci method in remembering words with simply receiving a list of said words.	The loci method was more entertaining and efficient in terms of accuracy and time than merely memorizing a list of words.
Oyen and Bebko (1996) [37]	Canada: 120; 20 F, 20 M in JK, SK, Grade 1	GrowWorm and ShipWrecks; cognitive; not stated	Find out whether the games “GrowWorm” and “ShipWrecks” are more effective in recalling pictures than traditional lessons.	Rehearsers recalled more items than non-rehearsers, with recall being less in the games than lessons. Though users found the games interesting, their difficulty, likely a result of multiple goals and components, may have distracted the participants.
Rawendy et al. (2017) [38]	Indonesia: 30; aged 6–12	Mandamonik; card, puzzle; loci, keyword, peg-word, picture	Evaluate a game that can help kids in Indonesia learn the Chinese language with various mnemonic methods.	The games resulted in significant differences between pre-test and post-test scores.
Keehl and Melcer (2019) [39]	United States: 24; 11 F, 12 M, 1 NB; aged 19–51; 12 experimental, 12 control	Radical Tunes; cognitive; musical	Evaluate: (1) the musical version of Radical Tunes that helps user remember how to draw kanji; (2) the version that uses non-musical sound effects; and (3) which versions have significant effects on remembering kanji and their significant difference.	Both versions of Radical Tunes resulted in significant differences between pre- and post-test scores, showing it effectively taught kanji. It was also found to be immersive. However, there was no significant difference between the two versions, and any long-term effects have yet to be investigated.
Economou et al. (2020) [40]	United Kingdom: 24; aged 19–53	Signum Battle; role playing; not stated	Develop a game to facilitate lerarning British Sign Language (BSL) to bridge the communication gap between adults with and without hearing impairment.	Signum Battle is more effective at teaching users BSL than video-based material is, based on accuracy. Its mythological environment was also well liked.
Schoen (1996) [41]	United States: 24; 3 groups of 8	Mnemopoly; board; acrostics, rhymes	Compare the effectiveness of the Mnemopoly board game, based on acrostics and rhymes, with other game types based on loci and phonetic peg.	Mnemopoly had significantly higher scores for both immediate recall and 1-week delay recall than both loci and phonetic peg, but the delayed recall score was significantly lower than the immediate recall score, like phonetic peg and unlike loci.
Balbuena and Buayan (2015) [42]	Phillipines: 46; Grade 7; 23 control, 23 experimental	Integer Operation (INTOP) Card Game; card; LAUS, LPUN	Create a card game that uses mnemonics to help facilitate understanding concepts and principles of integers.	In the INTOP and traditional teaching conditions, there were significant differences between pre- and post-test scores. Moreover, INTOP’s post-test scores were significantly higher than those of traditional learning.
Heidari et al. (2023) [43]	Iran: 75; all F, aged 13–17, 3 groups of 25 (1 control, 2 experimental groups)	Vocabulary Builder; cognitive; not stated	Test whether Vocabulary Builder is more effective at teaching English to Iranian EFL students than traditional methods.	Vocabulary Builder was effective in teaching vocabulary to Iranian EFL students, but it was not significantly better than traditional methods.
Weng et al. (2015) [44]	Taiwan: 135; 64 M, 71 F, Grade 8	Digital Games-Based Learning (DGBL); cognitive; not stated	Create a game that can help junior high school students remember several properties of chemistry elements.	DGBL is effective at reducing learning difficulties and enhancing student learning outcomes. Girls in particular benefit from it more significantly.
Ramlow and Little (2020) [45]	United States: Study 1: 19; 12 F, 7 M; aged 4–5.6. Study 2: 31; 17 female, 14 male; aged 4–5.1.	Rainbow Game; card, rainbow mnemonic (a new peg-type technique)	Introduce a new mnemonic type that uses various colors, then create a game that uses this method and compare it with a word game and later a mental mnemonic version that does not show the colors but discusses them.	Participants who used the rainbow mnemonic while shown colors recalled more words than those who played the word game, and so did those who used the mental version of the rainbow mnemonic. There was not much discussion comparing the two versions of the rainbow mnemonic.
Shen et al. (2011) [46]	Taiwan: 53; aged 20–24; 29 experimental, 24 control	3 unnamed memory games; puzzle; loci, number/rhyme, grapheme-based, link	Test three Flash-based puzzle games that use mnemonics and see if they are more effective at teaching various concepts than traditional methods are.	The experimental group, which played all three games, had significantly better Working Memory Capacity Test (WMCT) post-test scores than the control group, which learned from traditional methods.
Nordin and Jaafar (2019) [47]	Malaysia: 50	ATOMICfrenzy AR; card; not stated	Evaluate a game that can help improve students’ memory on properties of chemical elements.	The game resulted in an average of 2 more questions answered correctly after a 15-min play, with 84% of students agreeing that it helped them memorize important information in the periodic table and 98% agreeing that combining it with Kahoot was more fun.

corresponds to a different study. The first column shows the authors of the study and what year the paper was published. The second column shows the region the study was conducted in, how many participants there were in total, and any participant categories and how many participants were in each one. The third shows the game’s name, what type of game it is, and what mnemonic types it used. The fourth shows a brief description of what the study is trying to achieve. Lastly, the fifth column shows a summary of important findings from the study.

3.3. Remaining Characteristics’ Value Distributions

In this section, we present other important descriptive characteristics than those presented in Table 4, as well as year and region, despite them already being in Table 4. The papers were all published between 1996 and 2023 inclusive. The years 1996, 2015, 2019, and 2020 recorded the highest number of papers (2/12, 16.7%), while the years 2011, 2017, 2022, and 2023 recorded the lowest non-zero number of papers (1/12, 8.3%). As for regions and their number of papers, the United States has the highest number of relevant papers, followed by Taiwan, followed by several regions tied for third.

Table 5. Synthesized characteristics.

Characteristic	Value	Papers	Count (%)
Target Subject	Language learning	[36,38,39,40,43]	5 (41.7%)
	Mnemonic type comparison	[41,46]	2 (16.7%)
	Chemistry	[44,47]	2 (16.7%)
	Rehearsal and recall	[37]	1 (8.33%)
	Mathematics	[42]	1 (8.3%)
	Testing a new mnemonic	[45]	1 (8.3%)
Design Approach	User-study driven	[37,38,42,43]	4 (33.3%)
	Literature driven	[36,37,38,39,40,41,42,43,44,45,46,47]	12 (100%)
User-Evaluated	Yes	[36,39,40,44,46,47]	6 (50%)
	No	[37,38,41,42,43,45]	6 (50%)
Game Mode	Personal	[36,37,38,39,40,43,44,45,46]	9 (75%)
	Collaborative	[41,42]	2 (16.7%)
	Hybrid	[47]	1 (8.3%)
Age Group	Adults	[36,39,40,41,46,47]	6 (50%)
	Kids	[37,38,45]	3 (25%)
	Teenagers	[42,43,44]	3 (25%)
Level of Education	Kindergarten	[37,45]	2 (16.7%)
	Elementary School	[37,38]	2 (16.7%)
	Middle School	[42,43,44]	3 (25%)
	High School	[43]	1 (8.3%)
	University	[36,39,40,41,46,47]	6 (50%)
Game Outcome	Positive	[36,38,39,40,42,44,45,46,47]	9 (75%)
	Mixed	[41,43]	2 (16.7%)
	Negative	[37]	1 (8.3%)
System	Prototype	[36,37,38,39,40,41,42,43,44,45,46,47]	12 (100%)
	Product	None	0 (0%)
Purpose	Educative	[36,38,39,40,42,43,44,47]	8 (66.7%)
	Subjective	[37,41,45,46]	4 (33.3%)
Type of Study	Quantitative	[38,41,42,43,45]	5 (41.7%)
	Mixed	[36,37,39,40,44,46,47]	7 (58.3%)
	Qualitative	None	0 (0%)
Comparison	Vs. control	[36,42,44,46]	4 (33.3%)
	Vs. other experimental	[41]	1 (8.3%)
	Both	[37,39,43,45]	4 (33.3%)
	No comparison	[38,40,47]	3 (25%)
Mnemonic Generation	Manual	[36,37,38,39,40,41,42,43,44,45,46,47]	12 (100%)
	Automatic	None	0 (0%)
Mnemonic Source	Player	[36,37,41,44]	4 (33.3%)
	Game/designer	[38,39,40,42,43,45,46,47]	8 (66.7%)

shows key data extracted from the 12 included papers. They relate to target subject, design approach, purpose, effectiveness, to mention but a few. For example, regarding target subject, the majority of the studies (5/12, 41.7%) were about language learning. Regarding design approach, all of the serious game designs were informed by the literature, with one-third of them being informed by user studies as well. Half of the games were evaluated by users; the other half were not. Three in four of the games were personal, while one in six was collaborative. Similarly, regarding effectiveness, three in four of the games were positive, while one in six was mixed.

3.4. Risk-of-Bias Analysis

We also present a traffic light chart (Figure 2) showing each study’s risk of selection bias, performance bias, attrition bias, detection bias, and reporting bias. In addition, we present a bar chart (Figure 3) showing the distributions of studies’ risks of each bias. Both charts were created with the use of robvis [48]. Most notable is that the studies’ overall risks of bias are almost evenly split between low (6/12, 50%) and some concerns (5/12, 41.7%), with only Costuchen et al. [36] having a high overall risk of bias.

4. Discussion

Based on the presented results, we discuss the key findings emanating from our analysis of the 12 included papers considering the four main research questions that underpinned the scoping review. We also discuss the studies’ risks of bias, future trends, and the limitations of our own study.

4.1. Utilization of Mnemonics Serious Games in Education

Here, we answer the research question, “How have mnemonics been utilized in serious games to promote learning and retention in memorization-intensive subjects?”

4.1.1. Year Gap with Game Outcome as Rationale

A total of 12 eligible papers between 1996 and 2023 inclusive were analyzed in the scoping review. An interesting observation worthy of note is that while two papers regarding mnemonics and serious games were published in 1996, no additional papers were published until 2011, which is 15 years later. It turns out for both of these two papers, their games’ effectiveness was unfortunately not positive. Schoen [41] had mixed results regarding an attempt at a mnemonic board game with Mnemopoly, while Oyen and Bebko [37] came to an outright negative conclusion regarding their games GrowWorm and ShipWrecks. Perhaps this may have discouraged others from attempting to create mnemonics serious games for a considerable amount of time. However, Shen et al. [46] eventually made their own attempt and were fortunate to discover positive results regarding their three unnamed games, which may have encouraged more researchers to follow in their footsteps. It is also worth noting that Schoen [41] and Oyen and Bebko [37] both conducted studies in North America, while the studies from 2011 onward vary in terms of continent and include Europe and Asia. This could mean that word of the potential of mnemonics had spread across the world. However, the fact that Shen et al. [46], who put an end to the year gap, conducted their study in Taiwan, an Asian region and not a North American one, is surprising. Yet it is not as if mnemonics are a concept only known to certain cultures, so this fact is not far-fetched. The fact that the resurgence started in Taiwan could also be why Taiwan has the second-highest number of relevant papers.

4.1.2. Lack of Relevant Papers

While it is good that the year gap has been closed, there are still not as many studies about mnemonics serious games as we had hoped for. As stated before, even amongst the 469 non-duplicate papers found with our search of four databases, only 1 was about mnemonics serious games. The high number of papers that were rejected by the two researchers (KF and OI) turned out to be about reducing cognitive decline in older adults, likewise for catering to people with certain disabilities. It seems the research community has given a greater priority to addressing these issues. As such, improving memory in people without such impairments tended to have taken the backseat. Many other papers did try to improve memory in people without impairments, but not by using mnemonics, which might suggest mnemonics are not exactly seen as traditional. Even some of the mnemonics papers, such as those by Balbuena and Buayan [42], as well as Heidari et al. [43], explicitly compared mnemonics serious games to methods regarded as traditional.

4.1.3. Focus on Personal Use

The majority of mnemonics serious games seem to focus on personal use, as we have shown in Table 5. Some studies encourage a participant to compete against their own past self with pre- and post-test scores, such as the one by Rawendy et al. [38]. Some merely compare groups who used or did not use the game, but without explicit interaction between the groups; the one by Costuchen et al. [36] is an example. Introducing competition may be beneficial, even if a particular non-competitive game already has positive effectiveness. Hu et al. [49] tested five versions of a cognitive-based Whack-A-Mole game (not involving mnemonics) that differ in terms of reward conditions; the leaderboard condition correlated with the highest enjoyment, and it also improved cognitive performance in easy but not difficult tasks. Traver et al. [27] noted that their (also non-mnemonic) chemistry game E-CHEMMEND is single-player, but users ended up competing against each other and playing more than expected, leading the authors to consider adding a leaderboard. As mentioned in Table 5, ATOMICfrenzy AR [47], which is about mnemonics, was found to be more fun when combined with the well-known competitive game Kahoot, according to 98% of users.

4.1.4. Games Remaining as Prototypes

As seen in Table 5, all 12/12 (100%) studies’ games were prototypes. While there was a possibility that these games may have been released to the public after the papers were published, a Google search for each game by typing in that game’s name failed to find it. In addition, none of the papers appeared to have follow-ups that mentioned their games becoming available to the public, let alone having follow-ups at all. Regardless, the games were certainly prototypes at the time of their respective studies. Among the papers that were accepted by the two researchers (KF and OI) but rejected by the senior researcher (KO), only one, by McGregor et al. [50], involved a product, namely Vocabulary.com; even so, they were not the ones who created it. It appears that researchers who design serious games primarily do so just to give them to a select group of people and gather specific data from said people based on said games; otherwise, the games remain shelved. Releasing these games to the public afterwards may be beneficial, especially for long-term memory enhancement, since 9/12 (75%) studies’ games had a positive outcome, and 2/12 (16.7%) had a mixed outcome as opposed to an outright negative outcome (Table 5). Doing so would spread the good news about mnemonics, allow for more data collection due to the increased number and diversity of users, and can likely generate profits for further research.

4.2. Mnemonic Generation Approaches and Creators

Here, we answer the research question, “How are the mnemonics in serious games generated and by whom?”

4.2.1. Mnemonics Generated by People

As shown in Table 5, all of the studies’ games involved people creating mnemonics by themselves rather than using any computers to do so automatically. The closest study regarding automatic mnemonic generation in the context of serious games is Costuchen et al. [36], who had an external tool generate pseudowords for participants to memorize. Perhaps researchers were skeptical of machines’ ability to automatically generate coherent mnemonics, even in spite of the ever-growing power of AI. Alternatively, maybe they saw no need to use such technology to develop arguably simple mnemonics. A third possibility is that the people developing the mnemonics are more likely to remember them better, which is supported by Dave and Awasthi [21].

4.2.2. Mnemonics Mostly Generated by Games or Creators

As seen in Table 5, the majority (8/12, 66.7%) of studies’ mnemonics were provided by games or their creators, as opposed to having users create the mnemonics themselves, which research has shown is more likely to be effective [21,22,23]. Moreover, manual mnemonic generation is favored over automatic as all of the mnemonics in the reviewed serious games were hand-created as shown in Table 5. There are a number of plausible explanations for the choice of manual generation by most serious game designers and/or researchers. Apart from the lack of technical expertise, for example, using machine learning, perhaps the researchers sought to reduce the number of possible extraneous variables by having fixed mnemonics for all users. Maybe they wanted to provide the mnemonics as examples to make it easier for the users to learn about them and their usage and effectiveness. This is most likely the case for Ramlow and Little [45], as they were introducing a new mnemonic type and most likely saw the necessity to demonstrate how to use it using a serious game. There is also the possibility that at the time, researchers recognized the less-than-positive outcomes of the 1996 studies [37,41], which both had users create the mnemonics themselves. Moreover, the target users, some of which are kids [37,38,45] and teenagers [42,43,44], may not be well equipped to create effective and meaningful mnemonics, which require critical thinking, creativity, relatively broad vocabulary, and foundational knowledge [51,52]. As a result, most of the researchers might have attempted to avoid the ineffectiveness associated with user-created mnemonics in prior research [37,41] by creating the mnemonics themselves or by the instructor on behalf of the target users.

4.2.3. The Importance of Self-Generated Mnemonics

Nevertheless, given that research has shown that self-created mnemonics are likely to be effective in promoting long-term memory [21,22,23], we recommend that instructors teach their students how to effectively create their own mnemonics, for example, by creating mnemonics in front of them and providing adequate examples [53]. We also recommend that tools be developed to both support the training of students on how to create effective mnemonics and help them create their own mnemonics. Creating their own mnemonics help students to connect them strongly to what they are already familiar with and their personal experiences, thereby strengthening the encoded information and enhancing memory recall in the long run [21,54]. In other words, the way students encode information when they first study enhances memory recall [53]. Hence, there is a need for tools that can support them in creating their own mnemonics easily and effectively.

4.3. Effectiveness of Mnemonics Serious Games

Here, we answer the research question, “How effective is the use of mnemonics in serious games aimed to promote learning and content retention in memorization-intensive subjects?”

4.3.1. Overall Positive Outcome

Overall, using mnemonics in serious games to help with learning and remembering information in memorization-intensive subjects has fortunately been successful. The review showed that 9/12 (75%) studies’ games had a positive outcome, and 2/12 (16.7%) had a mixed outcome (Table 5). For the only one that had a negative outcome (GrowWorm and ShipWrecks), Oyen and Bebko [37] attributed it to difficulty arising from multiple goals and components. Some studies showed that mnemonics serious games can help individuals improve themselves, as proven by comparing pre- and post-test scores [38,39,42,47]. Others showed that these games can allow users to fare better when compared with people who used traditional methods [36,40,42,46].

4.3.2. Game Enjoyability

The overall positive outcome of the games is more focused on performance results as opposed to user experience. It would be worthwhile to focus on specific aspects based on feedback, such as enjoyment. Meanwhile, we consider the 7/12 (58.3%) studies that involved qualitative data analysis. In some of the studies, users noted how the games were enjoyable to them [36,37], even if the games were difficult and had negative outcomes [37]. In other studies, more specific statistical data were received, suggesting that users did enjoy the games [39,40]. Both may occur in the same study [47]. A total of 5/7 (71.4%) studies that involved collecting qualitative data had users expressing their enjoyment. The other 2/7 (28.6%) did not have users express the opposite; there were no sections gathering data about enjoyment, nor were there any open-ended sections.

4.3.3. Lack of Memorability

Similarly, it would be nice to see how memorable these games are to their users. Once again, we will focus on the 7/12 (58.3%) studies that involved collecting qualitative data. In 2/7 (28.6%) studies [36,40], the researchers included making their games more memorable as a future concept, implying their current games were not particularly memorable despite being enjoyable. The other 5/7 (71.4%) studies made no mention of whether or not their games were found to be memorable. With the data provided, it can be inferred that considerable effort will need to be made to make mnemonics serious games more memorable overall.

4.4. Using Serious Games to Teach Mnemonic Techniques

Here, we answer the research question, “To what extent have serious games been utilized to teach mnemonic techniques?”

4.4.1. Lack of Guidance in Games That Only Provide Mnemonics

Unfortunately, there does not seem to be emphasis on teaching players how to develop good mnemonics when it comes to games that provide their own mnemonics. In the 8/12 (66.7%) studies shown in Table 5 that had mnemonics provided by the games or their creators, there were no later sections where players were guided or even asked to create their own mnemonics, even though the given examples could have helped. Rather, the players were often required to practice or at least make use of the given mnemonics. A clear example of this is Balbuena and Buayan’s [42] pair of mnemonics, LAUS (“if two integers have like (L) signs, add (A) their absolute values” and “if two integers have unlike (U) signs, subtract (S) their absolute values.”) and LPUN (“when two integers have like (L) signs, their product or quotient is positive (P); unlike (U) signs, negative (N).”), which players must memorize to better answer integer questions that they create by choosing cards with numbers or operators. Players are not asked to create other mnemonics that can help them. We might say that at best, researchers are implicitly nudging users to create more mnemonics by providing some examples, but the lack of explicit teaching is rather unfortunate.

4.4.2. Varying Degrees of Guidance in Games That Involve Self-Generated Mnemonics

Games where players create mnemonics apparently have an increased chance of guiding users through creating mnemonics. Costuchen et al. [36] clearly taught the loci method by outright telling users to imagine what every new word reminds them of (even though the words are actually pseudowords) and remember where in their game Roman Palace it appears. Schoen [41] instructed all three participant groups on their respective mnemonic techniques (Mnemopoly, loci, phonetic peg) for 20 min, then gave them 40 words and 10 min to encode them with tools (game boards, campus maps, peg word lists) for reference. However, not all such games clearly teach how to create mnemonics. For instance, Oyen and Bebko [37] stated that “the experimenter did not help the children with mnemonic clues”, though in their defense, they wanted to investigate children’s strategies and classify them as rehearsers or non-rehearsers.

4.4.3. Lack of Simultaneous Use of Text and Images

An additional thing to note is that the studies generally did not make much use of both images and text simultaneously to improve learning and long-term memory. To the best of our knowledge, the closest a study has come is the use of the loci method [36,38,46], which mainly involves thinking of a place and associating it with given information. Research has shown that contrary to fears of images hindering the reading of text, images actually compliment the text and may even provide additional details as the dual coding theory holds [5,55,56]. This is why picture books are sometimes called “twice-told tales” [56]. With regards to mnemonics, associating both images and text with given information will likely establish a greater connection rather than overwhelm the user. As such, mnemonics serious games should leverage the simultaneous use of both images and text if possible.

4.5. Studies’ Risks of Bias

Here, we discuss the risks of bias of the analyzed studies, namely, selection bias, performance bias, attrition bias, detection bias, and performance bias.

4.5.1. Considerable Issues with Selection Bias and Performance Bias

Out of the five types of bias we focused on, the analyzed studies overall seem to raise the most concerns or confusion about selection bias. Several of the studies [37,40,41,44,47] did not describe their recruiting process at all, apart from possibly stating the target demographic. Several others [42,43,46] clearly displayed a lack of complete randomness with regards to either the recruiting process or the splitting of participants into groups but still had some random elements. The study by Costuchen et al. [36] was the most concerning since the control group had six more people than the experimental group but the latter had more men, and no explanation was provided. The studies were also overall somewhat susceptible to performance bias. Some of them [41,42,43,46] had multiple sessions with considerable breaks in between, meaning participants might have tried to prepare themselves more during these breaks. However, they most likely did not know what exactly to prepare for.

4.5.2. Minor Issues with Other Biases

Regarding attrition bias, Oyen and Bebko [37] did not provide information on replacements for participants that failed practice trials for “GrowWorm”, and Weng et al. [44] did not state whether anyone either dropped out or was otherwise excluded from analysis, and they did not provide information that could implicitly show it, such as degrees of freedom and number of groups. Oyen and Bebko [37] also showed some risk of detection bias since participants were split into two had different lessons and games, and the participants were evaluated based on the lesson and game they specifically played, meaning not everyone’s data were collected the same way. However, the lessons and games were noted to be similar to each other. Ramlow and Little [45] showed some risk of performance bias when they failed to mention which version of their rainbow mnemonic was more successful; they only stated that both were more successful than the word game.

4.6. Future Trends: Text-to-Image Synthesis with Generative Models

Given the challenges many students face today in remembering taught information, on one hand, and the continuous advancement in machine learning (ML), artificial intelligence (AI), genetic algorithms, and text-to-image synthesis tools such as Gemini [57] and DALL-E 3 [58], on the other hand, we foresee the leveraging of these technological capabilities in the creation of memorable mnemonics grounded in the dual coding theory [5] in the near future, with serious games used for retrieval practice [59]. For example, genetic algorithms can be used in generating optimal acrostics with sufficient verbal cues that can help students recall the target information easily. Moreover, text-to-image synthesis tools such as Gemini [57] can be used to generate appropriate images that can be associated with the automatically generated mnemonics to facilitate their recall and the target information [60]. More importantly, retrieval-practice serious games have the potential of replacing repeated study in memory-intensive instructions [59], especially given that research has shown that retrieval practice is more effective than the repeated study of taught materials [61]. These AI- and ML-driven technologies have the potential of revolutionizing the teaching and learning of memory-intensive subjects such as psychology, biology, and history while fostering students’ engagement, fun, enjoyment, collaboration, and, above all, long-term learning.

4.7. Limitations

Our scoping review has five limitations. The first three have to do with the search and screening process, while the last two have to do with the subjectivity of our interpretations of the studies’ methods and findings.

4.7.1. Issues with the Search and Screening Process

The first three limitations have to do with our search and screening process. First, our search might have been over-constrained. Our inclusion criterion regarding only considering English-language papers might have led us to miss out on relevant papers in other languages. Searching only four databases might have missed papers exclusive to other databases; for example, the Google Scholar search afterwards found papers not in the four databases, but its lack of formality as opposed to the four-database search is questionable. Our search strings may have missed relevant papers in the databases that we searched due to not including synonyms for “mnemonic”. As stated before, trying “acronym” and “initialism” separately provided no new good results, but what about other synonyms? Also, terms with similar meaning to “serious game”, such as “educational game” or “learning game”, were not used, possibly leading to missing more relevant papers. Second, some of the papers that we screened out may have been relevant after all, but were only ignored because of human error or subjective opinion. Third, there may have been new relevant papers introduced between the time of the original search and the time that this paper was completed.

4.7.2. Subjectivity of Interpretations of Study Methods and Findings

The last two limitations have to do with the subjectivity of our interpretations of the studies’ methods and findings. First, some of the papers’ characteristics that we described were subjective. The “Purpose” characteristic (which coincidentally contains a value called “Subjective”) is a prominent example of our opinions, and the “Game Outcome” characteristic is an example of other researchers’ opinions. Second, since the studies we gathered were somewhat vulnerable to biases, especially selection bias and performance bias, our findings and interpretations may need to be taken with a grain of salt. Even our risk-of-bias analysis itself is based on subjective opinion; others may disagree with our ratings. Future reviews may seek to account for these limitations.

5. Conclusions

We have presented a synthesis of the research on using mnemonics serious games aimed to promote long-term learning using Djaouti et al.’s [34] GPS model as an analytical framework. The analyzed games spanned a two-decade period from 1996 to 2023, with at least one game per year in the last decade. After Shen et al. [46] in 2011 closed the hiatus owing to a lack of studies on mnemonics serious games between 1996, when Schoen [41] and Oyen and Bebko [37] first published their work on the subject matter, and 2011, the number of such studies has remained virtually constant each year. Given the consistent interest in and publications on the subject matter over the last decade, it is worth analyzing the characteristics of the published games, their strengths and weaknesses, and their effectiveness. Doing so may help create better and more effective games in future research and intervention efforts The key findings of the scoping review are as follows. First, mnemonics serious games have been slowly but steadily increasing in number, and they tend to focus on personal use. Ironically, they have not been released to the public yet. Second, mnemonics in serious games are created by people rather than computers, and the people in question are the game designers rather than the players. Third, mnemonics serious games have generally had a positive outcome with regard to improving users’ memory. They are also generally found enjoyable but not particularly memorable. Fourth, most of the mnemonics serious games that provide mnemonics do not support self-created mnemonics; at best, they only nudge players to do so outside the games. Meanwhile, games that involve players creating their own mnemonics either provide clear guidance on how to do so or possibly purposefully leave players to their own devices. In conclusion, based on the above findings, we recommend that more work be performed regarding the use of serious games to promote long-term learning, with a focus on self-created mnemonics, the simultaneous use of images and text, unexplored subject areas such as human–computer interaction, and social contexts. More importantly, there is a need for experimental research on actual games (products, not prototypes) used in promoting long-term learning in real-life settings, particularly in middle schools, high schools, and tertiary institutions.