W9B. Writing a Literature Review for Research Proposals

1. Summary

1.1 The Role of the Literature Review in a Research Proposal

A literature review (LR) is a critical synthesis of existing research on a topic. Within a research proposal, the Literature Review serves a precise purpose: it establishes the intellectual context for your study and demonstrates that a research gap exists—a question that current scholarship has not yet fully answered, which your proposed research will address.

Understanding what kind of literature review you are writing matters. According to Susanne Hempel’s taxonomy (Conducting Your Literature Review), literature reviews fall into three broad categories:

1. Background section — a brief contextual overview embedded within another document.
2. Part of a research project write-up — a focused, strategic synthesis that is one section of a larger proposal or paper. This is the type you are writing.
3. Stand-alone product — a comprehensive, independent review published as a research output in its own right (e.g., a systematic review).

Because your literature review is part of a research proposal, it does not need to be exhaustive. It must be strategic: covering the most important theories, debates, and findings that directly justify and frame your proposed study.

1.2 What a Literature Review Must Do

A literature review in a research proposal has five core functions:

• Describe the most important theories and concepts relevant to your topic. A reader unfamiliar with your field should finish the LR understanding the key intellectual frameworks at play.
• Describe key debates and controversies in the field. Where do researchers disagree? What remains contested? This shows that the field is live and that there is something for your research to contribute to.
• Describe the research gap — the specific problem, question, or gap in knowledge that current literature has not addressed, which your study will fill.
• Ensure that your sources are relevant and credible — use peer-reviewed journal articles, conference papers, and recognized academic sources. Avoid non-academic websites, Wikipedia, or sources of unclear provenance.
• Synthesize and interpret your sources — do not simply summarize each source in turn. Identify patterns, agreements, contradictions, and implications across sources.

1.3 Tips for Writing a Strategic Literature Review

A common mistake is to write a literature review as a series of summaries: “Smith (2019) found X. Jones (2021) found Y. Brown (2022) found Z.” This approach—sometimes called annotation-style writing—fails to synthesize the literature and does not demonstrate critical thinking.

Instead, follow these principles:

• You do not have to provide a systematic literature review. A systematic review requires exhaustive coverage of all published work and formal methodology for source selection. A proposal LR only needs to cover the most important publications.
• Focus on the most important publications only. Prioritize foundational works, highly cited papers, and recent studies directly relevant to your RQ.
• Group the literature into major themes or conceptual strategies. Organize by idea, not by author or publication date. For example: “Studies on X can be grouped into two camps: those that argue A [1, 4, 7] and those that argue B [2, 5].”
• Avoid chronologically or methodically describing sets of studies one by one. The reader wants to understand the state of knowledge, not follow a timeline of publications.

1.4 The Five C’s of the Literature Review

The Five C’s framework provides a systematic approach to analyzing and writing about your sources. Each C represents a distinct intellectual operation you must perform:

1.4.1 Cite

Cite your sources properly using IEEE format. In IEEE citation style, references are numbered in order of appearance in the text (e.g., [1], [2], [3]). Every claim that draws on a source must be cited; uncited claims are treated as the author’s unsupported opinion.

Example of a properly cited claim: > Machine learning models have been shown to exhibit systematic bias against minority groups in facial recognition tasks [3], [7].

1.4.2 Compare

Compare the methods, outcomes, models, and arguments across your sources. Identify where authors agree. What findings are consistent across multiple independent studies? What methodological choices do researchers converge on?

Comparison shows the reader what is established in the field—knowledge that has been confirmed by multiple independent investigations.

Example: > Both Foelicker [3] and Kim et al. [9] find that transformer-based architectures outperform recurrent models on long-range dependency tasks, suggesting this advantage is robust across different dataset types.

1.4.3 Contrast

Contrast the methods, outcomes, models, and arguments across your sources. Identify where authors disagree. Where do results conflict? What methodological differences might explain divergent findings?

Contrast reveals the contested areas in the field—exactly where your research can make a contribution.

Example: > Whereas Eshol [45] argues that gradient clipping is unnecessary when batch normalization is applied, Sigurdarsson [34] found that models trained without gradient clipping diverged on large learning rates regardless of normalization strategy.

1.4.4 Critique

Critique the literature. This does not mean being hostile or dismissive. It means evaluating which ideas are most valid (logically sound), suitable (appropriate for the problem at hand), and reliable (consistent and reproducible). Identify the limitations of existing work and explain which approaches are most convincing and why.

Example: > While Kickfeis [23] provides a large-scale dataset, the evaluation is limited to a single domain. The generalizability of the findings therefore remains unclear, making it difficult to apply these results directly to cross-domain tasks.

1.4.5 Connect

Connect the literature to your own particular study. After comparing, contrasting, and critiquing the existing work, explain how your proposed research relates to it. Does your study:

• Fill a gap that existing work leaves open?
• Apply a known method to a new domain or context?
• Directly test or challenge a contested claim?
• Synthesize two previously separate lines of inquiry?

This is the most important of the Five C’s for a research proposal, because it is where you explicitly justify your research question.

Example: > Given the disagreement about cross-domain generalizability noted above, this study proposes to evaluate the model described by Kickfeis [23] on three additional domains, providing a more comprehensive test of its performance boundaries.

1.5 Tenses in the Literature Review

Tense choice in the literature review is not arbitrary—it signals the status of the knowledge being reported. The three tenses used in academic LR writing are Present Simple, Past Simple, and Present Perfect Simple.

1.5.1 Present Simple

Use the Present Simple when:

• You refer to recent research whose findings are current and still widely accepted: > Example: Foelicker [3] shows that transformer models require significantly less fine-tuning data than LSTM-based architectures.
• You wish to show that the author still actively supports the cited ideas: > Example: Eshol [45] persistently argues that symbolic reasoning must be integrated into deep learning systems for true generalization.

1.5.2 Past Simple

Use the Past Simple when:

• You refer to a specific past study and its discrete results: > Example: Kickfeis [23] found that accuracy dropped by 12% when training data volume was reduced by half.
• The cited author is deceased, and the claim belongs to a historical moment: > Example: Sigurdarsson [34] infamously claimed that neural networks could never approximate continuous functions with arbitrary precision.

1.5.3 Present Perfect Simple

Use the Present Perfect Simple when:

• The research occurred in the past but remains directly relevant to the current state of the field: > Example: Tricksters [1] has demonstrated that the X effect can manifest itself under the right conditions.
• You use time expressions such as ever, never, just, already, yet, so far, up to now, recently, since, or for (to indicate duration or relevance up to the present): > Example a): So far, no evidence has been found that adversarial training fully eliminates distributional shift. > Example b): Since 2020, the performance gap between these two approaches has become increasingly clear.

1.6 Useful Phrases for the Literature Review

Academic writing relies on a shared repertoire of citation phrases—formulaic expressions that introduce, attribute, and evaluate sources. Using these phrases correctly signals academic fluency.

The Academic Phrasebank (University of Manchester) provides an extensive, organized list of phrases for all academic writing functions. For the literature review, consult the “Referring to sources” section at: https://www.phrasebank.manchester.ac.uk/referring-to-sources/

The following categories of phrases are particularly useful:

Introducing a source’s claim:

• X argues that…
• According to X,…
• X contends that…
• X maintains that…

Showing agreement between sources:

• This view is supported by X, who found that…
• Similarly, X demonstrates that…
• In line with X’s findings, Y also reports that…

Showing disagreement between sources:

• However, X challenges this view by arguing that…
• In contrast, X found that…
• This finding contradicts earlier work by X, who reported that…

Introducing a gap or limitation:

• However, few studies have examined…
• The existing literature has not yet addressed…
• A limitation of this research is that…
• To date, no study has investigated…