A (Personal) Note on Research Writing
By Dipanjan
By Dipanjan
This note outlines a few suggestions on the scientific writing process. Please treat them as such—suggestions, not rules.
Research writing is, at its core, an individual endeavor. Your unique perspective and voice are what make your work interesting. In that sense, a "how-to" guide like this can seem counterproductive.
However, when you're starting, having a few pointers can be a helpful scaffold. It is worth remembering that finding your own creative and individual style—a path that is inherently more unpredictable than following a rigid template—paradoxically requires more discipline, not less. It's like pursuing a passion such as hiking, which demands rigorous self-discipline and adherence to strict safety rules, whereas a conventional, 'mundane' life path does not require the same level of personal vigilance.
This is especially true during the learning years. That structured practice is the foundation that gives you the freedom to explore and find your own rhythm. Once you have assimilated that discipline and gained expertise, you can and should ignore any advice that doesn't work for you—including the suggestions made right here. If something doesn't fit your style, it's not for you.
Sometimes, the hardest part is starting. Many find it helpful to begin writing the Methods and Results sections first, even before the Introduction.
Once you have all your results, figures, and tables organized, you can see the whole picture more clearly. This "bird's-eye view" makes it much easier to construct the overarching "story" you'll need to tell in the Introduction (what problem you're solving) and the Discussion (what your results mean).
For me, waiting to find the perfect language or the most elegant sentence can kill my motivation. I find it's better to first populate the space with the main ideas, even if the language is clumsy or incomplete. Get a "draft zero" down on paper.
Once the entire structure is there, I can then improve the language, fix the flow, and polish the text over multiple iterations. The goal is to separate the creative/generative process (getting ideas down) from the critical/analytical process (editing). This might or might not be the case for you, but it's a useful strategy if you find yourself stuck.
A critical challenge for many students is keeping interpretation out of the Methods and Results sections. These sections must be a direct, objective report of what you did (Methods) and what you found (Results).
The "story" and the "why" should always be in the back of your mind, guiding which results you present, but the writing itself should remain descriptive. You will interpret these results later, in the Discussion.
Example:
❌ Avoid (Interpretive):
"To see how the brain's emotional centers were controlled, we ran a connectivity analysis. The result (Fig. 2) showed a significant increase in coupling between the prefrontal cortex (PFC) and the amygdala, which proves that the PFC was actively suppressing the emotional response driven by the amygdala."
✅ Write (Descriptive):
"We performed a Dynamic Causal Modelling (DCM) analysis to assess task-based changes in effective connectivity. The analysis revealed a significant increase in the connection strength (p < .05, FWE-corrected) from the left dorsolateral PFC to the right amygdala during the 'Reappraisal' condition compared to the 'View' condition (see Figure 2 and Table 1 for parameters)."
Save the part about "suppressing the emotional response" for the Discussion.
Beyond objectivity, the Methods section must serve as a complete "recipe" for your experiment and analysis. The ultimate goal is to provide enough detail for another researcher to understand and, in principle, replicate your work precisely. This is no longer a "nice-to-have"; it is a core requirement of transparent and rigorous science.
In practice, this means:
Specify All Tools: List all software, toolboxes, and packages used, including their version numbers (e.g., 'SPM12 (v7771)', 'Python 3.9 with scikit-learn v1.2.1'). This prevents ambiguity, as software updates can and do change results.
Detail All Parameters: Do not assume any setting is "default" or "standard." Report all key parameters, thresholds, and choices.
fMRI Example: "Data were smoothed with an 8mm FWHM Gaussian kernel. Statistical maps were thresholded using a cluster-forming threshold of p < .001 (uncorrected) with a cluster-level FWE-corrected threshold of p < .05."
Modeling Example: "The model was trained for 100 epochs with a batch size of 32 and a learning rate of 0.001 using the Adam optimizer."
Share Code and Data: The gold standard for reproducibility is to share your analysis code and, where possible, your data. In the Methods, provide a statement on data and code availability.
Example: "All analysis code used in this study is available at [Link to GitHub/OSF Repository]. The anonymized data and task materials are available at [Link to Data Repository]."
Being this explicit not only builds trust with reviewers and readers but also serves as an invaluable record for your future self when you return to the project months or years later.
In neuroimaging and computational neuroscience, clarity must always take precedence over stylized language. I enjoy stylized prose, but only if it doesn't hamper understanding.
Here are a few tips to achieve this:
1. Get Feedback
Ask someone in your field (but perhaps not in your specific sub-field) to read your draft. Their main feedback should be on clarity: "Is this easy to understand? Where did you get lost?" They must be critical.
2. One Idea Per Sentence
Avoid long, complex sentences with multiple clauses. Each sentence should carry one single, clear message.
❌ Avoid: "Our computational model, which included both excitatory and inhibitory populations based on known cortical microcircuitry, successfully simulated the emergence of gamma oscillations (40 Hz) when the balanced input was applied, which aligns with experimental findings showing that these rhythms are critical for information processing."
✅ Write: "Our computational model simulated a cortical microcircuit with both excitatory and inhibitory populations. We applied a balanced input to this model. This simulation successfully reproduced the emergence of gamma oscillations (40 Hz). This finding aligns with experimental work suggesting these rhythms are critical for information processing."
3. One Theme Per Paragraph
Likewise, each paragraph should contain one coherent message or idea.
❌ Avoid (Mixed Themes): "We found significant BOLD activation in the fusiform gyrus (FG) when participants viewed faces. This 'fusiform face area' (FFA) response is consistent with previous literature. However, the spatial resolution of our scanner (3T) meant we could not resolve individual sub-regions of the FG. We also observed a trend-level activation in the amygdala, which may be related to the emotional valence of the faces, but this did not survive correction. Future studies should use a 7T scanner to improve signal."
✅ Write (Single Theme): "Our primary finding was a robust BOLD activation in the fusiform gyrus (FG) when participants viewed faces compared to other objects. This activation was localized to the 'fusiform face area' (FFA), a region well-established in face processing. This result strongly replicates previous work and confirms the successful engagement of the core face-processing network by our experimental task. The specific parameters of this activation (e.g., peak voxel, cluster size) are detailed in Table 2."
(Note: The limitation about the 3T scanner and the secondary amygdala finding should be moved to separate, subsequent paragraphs in the Discussion.)
For the Introduction and Discussion, try thinking in bullet points first. Outline the logical steps of your argument, and then elaborate each point into a full paragraph.
Start Broad (The General Problem): Establish the general field and its importance.
Example: "Understanding the neural basis of memory is a fundamental goal of neuroscience..."
Review Knowns (The Specific Topic): Narrow the focus to your specific topic and summarize what the literature does know.
Example: "A large body of work has established that the hippocampus is critical for forming new memories..."
Identify the Gap (The "Known Unknown"): State clearly what is not known, what the controversy is, or what the limitations of previous work are. This is the central problem you will address.
Example: "...however, the precise mechanisms of hippocampal-cortical interaction during sleep that support consolidation remain poorly understood..."
State Your Contribution (The "Fix"): Explain what this paper does to fill that specific gap.
Example: "...Here, we used a novel combination of simultaneous EEG-fMRI and a generative model to investigate..."
State Hypothesis & Roadmap: Clearly state your specific, testable hypothesis and (optionally) preview your main findings.
Restate Main Finding(s): Start by clearly and concisely answering the primary research question you posed in the Introduction. Don't make the reader wait.
Interpret Your Findings: What do these results mean? Connect them directly back to your original hypothesis.
Contextualize with Literature: Place your primary findings in the context of existing work. Do your results support, extend, or challenge previous studies?
Discuss Secondary Findings: Address other important results or unanticipated findings.
Acknowledge Limitations: Be transparent about your study's weaknesses (sample size, design limits). Explain whythese are limitations and their potential impact. This builds credibility.
Propose Future Directions: What comes next? Suggest specific, concrete experiments that logically follow from your work.
Conclusion / Take-Home Message: End with a strong summary. What is the single most important takeaway?
Finally, always write with your most critical reviewer in mind (the proverbial "Reviewer 3"). What are the obvious holes in your argument? What alternative interpretations of your data have you ignored?
Cover these points proactively. It is far better to address a limitation yourself (e.g., "One limitation of our study is...") or explore an alternative (e.g., "An alternative interpretation of this finding could be...") than to have a reviewer point it out.
There is no substitute for a well-written paper that is logical, rigorous, conscious of its methodological limitations, and explores all alternative interpretations. This doesn't weaken your paper; it demonstrates your command of the topic and strengthens your final argument.
The clarity and rigor you bring to the writing are not separate from the science—they are the final, critical acts of the scientific process itself. Your unique voice will emerge from this structured, honest telling. Write well.