Beginning scientific writers often approach the Discussion section by attempting to list any and all possible reasons for the results of or the errors in your experiments. Although this might be fine for a lab report in an introductory class, it is not usually a good practice when writing a scientific paper!

A reasonable amount of human error is assumed in any experiment, so it is rare that “human error” should  be mentioned at all when discussing measured differences from expected values. The exception to this is, in some cases, if you know of some specific human error during your experiment that you believe affected your results in a substantial way. (However, in a published paper, it would be much better practice to repeat the experiment without the error, instead.)

For example, here is an excerpt from a novice student’s Discussion section concerning a lab exploring pipette use:

Lack of precision can be explained primarily by operator error. An inexperienced operator, such as the one operating the pipette in this study, is less likely to reproduce the precision of an experienced operator.

Notice that this discussion point, in addition to being wordy and repetitive, fails to provide us with substantive science to explain the pipette error. We should already assume, as scientific readers, that humans are imperfect to a certain degree–there is no need for the students to mention it in her paper. She should instead save space in her paper to discuss information more relevant to understanding the data on a scientific level.


Audience and purpose


Every scientific writer has two main purposes to address in their Discussion: to interpret the findings and to explore broader implications of those findings. Considering that the Discussion is the bottom of the organizational hourglass, the more narrow of these purposes (to interpret) precedes the broader.


Reflective questions to help you address your purpose in the Discussion
When thinking about how to interpret each of your individual findings:

  • What likely caused this phenomenon?
  • What factors contributed to this result?
  • What do any trends in my data show?
  • How can I explain what I see using established science?
  • How do others’ work add to or detract from this finding?
When thinking about how to explore the broader implications of your work:

  • What are the implications of this work?
  • What are useful applications of this work?
  • What did I find that was fundamentally new?
  • How do my findings progress my field?
  • How does this work contribute to society?


You will learn more about how to address these purposes move by move in the Organization section of this lesson at the bottom of the page.

Because of your broadening purpose throughout the Discussion, your audience will also morph somewhat. Discussion of the meaning of a single data point, for example, at the beginning of the section will mostly be directed at people interested in replicating your experiment or using it for another research endeavor, meaning that your audience at this point will be other experts in your specific field. As you broaden your discussion to musings about the significance of your project as a whole, however, you will need to engage all scientists whose interests may connect with this project in some way–that is, potentially all scientists.


Style and conventions


Avoiding “hand-waving” arguments


When you interpret your data in a way that is not supported by evidence, scientists say that you are making a “hand-waving” argument. You can imagine the origin of this term being someone gesturing wildly with their hands while they make a point in order to make it seem more convincing, even though they aren’t backing their argument up with sufficient details. Below is an example of a result followed by a hand-waving interpretation:

Result: Neither strain grew by utilizing organic substrates in the absence of iron, and their growth on iron was not stimulated by the presence of acetate. (Adapted from Emerson and Moyer 1997)

Discussion (hand-waving): This result shows that organic substrates are complexed with high concentrations of iron in all life forms.

As we know, a result from two bacterial strains does not suggest a conclusion about “all life forms,” nor does how an absence of iron impacts bacterial growth indicate anything about how organic substrates interact with iron. This discussion point is an extreme example of a hand-waving argument because it draws large, far-reaching, and irrelevant conclusions from a single result.

Although you are unlikely to hand-wave in your papers to the same extent as in the example above, it can be very tempting to stretch the meaning of your data at least a little. After all, the results of the experiments you do in class are usually not incredibly exciting, and it can be nice to imagine that they might mean something significant. But be careful! If you make an assertion that reaches beyond the scope of your data, you will lessen your credibility as a scientist and an author. It is always better to exercise restraint when discussing your data so that the conclusions you do draw are more trustworthy and powerful.



Using hedging language


Hedging is the use of qualifiers to acknowledge that a statement is not absolute fact. The results you obtain in an experiment may suggest a certain conclusion, but they are not absolute truths. It may seem, at first, like hedging makes your argument sound weaker, but it actually makes you more trustworthy. Take, for example, the following discussion point:

Thus, it remains possible that particle size may be one of the largest contributing factors to the decrease in absorption observed.”

The words “possible” and “may be” are used intentionally to show that we cannot prove that particle size is one of the largest contributing factors, but that it is what our current scientific knowledge has led us to believe. Alternatively, trying to convince your audience of an idea makes you appear biased and therefore not an objective scientist. (Read more about objectivity here.)

In scientific writing, hedging is a critical tool to ensure that you don’t overstate your knowledge and to what extent it applies.


Hedging words
  • Can
  • Could
  • May
  • Might
  • Should
  • Would
  • Possible/possibly
  • Probable/probably
  • Potential/potentially
  • Typically
  • Generally
  • Broadly
  • Likely
  • Suggest(s)
  • Indicate(s)
  • Seem(s)
  • Support(s)


This is list is not exhaustive but it should help you identify and construct statements that don’t inaccurately overstate what science can tell us. Never use the word “prove” in your discussion–experiments can “demonstrate” or “suggest” an explanation, but they can never prove that something is true.

Below we provide examples of hedging from published discussion sections. Note which words they use to hedge their arguments and how frequently they are used.


“MAT sequences may be useful for improving phylogenetic resolution of species in the Gibberella/Fusarium complex. MAT appears to be more variable than most other gene sequences in a given genome (51) and therefore should enhance phylogenetic resolving power, a clear advantage in unraveling species relationships in the taxon-rich Gibberella/Fusarium complex. In contrast to high MAT gene variability between species, MAT appears to be nearly homogeneous within species (52). Thus, MAT genes have the potential to mark species boundaries.” (Adapted from Yun et al. 2000)
“As mentioned above, dialkyl malonates 1 were monosubstituted successfully by the poorly reactive polyfluoroalkyl halides, indicating that anions derived from malonic esters react very well in this nucleophilic displacement. In addition, among the base-solvent combinations used for performing such perfluoroalkylations, it appears that weak bases such as potassium carbonate and aprotic solvent such as THF or strong bases such as sodium hydride in THF should be used.” (Adapted from Trabelsi, Szönyi, and Geribaldi 2001)
“The mechanical weakness of the Eastern Cordillera basal detachment, as inferred from the presence of melts and fluids, probably separates crustal shortening in an upper-crust imbricate belt [7-9] from a deeper crust with an unknown mode of internal deformation. The continuity of the deep crustal structure across the entire plateau suggests a single crustal thickening process rather than important additional contributions from other processes that involve the mantle [23].” (Adapted from Yuan et al. 2000)


You should be sure to hedge your statements whenever your point is not an absolute fact. Hedging is used most often in the discussion section of a paper or poster because the authors are speculating about, but cannot be certain of, the meaning of their results. Hedging is also commonly used in introductions (to indicate the uncertainty associated with talking about others’ results) and in results sections (to describe findings without making them sound like universal truths).


Pro-7: Practice exercise

Choose any primary research article and read its discussion section. Underline all hedging words and phrases and note the contexts in which they are used.



Using “we”


When you think about the purpose of your paper or poster, is it usually to talk about you (doing the science) or about the science (that you did)? In most cases once you get to the undergraduate level and beyond, the science is the most important part of the project. For that reason, scientific writing should usually focus on what was done rather than who did it.

“We” is appropriate in some instances, but not in others. Taking the scientists out of the science helps a statement become more objective, but taking them out of a personal insight or decision can be downright confusing. For example, saying “It was believed that this project was important” isn’t any more objective than saying “We believed that this project was important,” but the latter statement is less awkward and more straightforward.

“We” is used more frequently in the Discussion than in other sections of a journal article. This should make sense, considering that the discussion section is about interpretation, which unavoidably involves scientists and their personal contributions to the project. However, you should use “we” minimally in the other sections, especially the Methods and Results, in which objectivity is most important.



Supporting claims with the literature


Literature references appear most often in two sections of a journal article: the Introduction and the Discussion. While in the former we use the literature to help set the context of our study, in the Discussion we use other published work to provide scientific explanations for our own data. In fact, it would be nearly impossible for you to decide how your data should be interpreted without referring to the literature yourself! Any claim you make in your discussion must be grounded in science.


Adapted from Armstrong et al. (2012):

Experiments with unsymmetrical π-allyl substrates and PHOX ligands have shown that the preference for nucleophilic addition trans to phosphorus is very high (>104).6b This strong preference suggests that the difference that allows high enantioselectivity with PHOX ligands is primarily the lower reactivity of the endo diastereomer.


Another key use for the literature in the Discussion section is to support or expand upon the conclusions you draw. Using multiple reputable sources will bolster your assertions with those of many other reputable scientists.


Adapted from Kaspi et al. (2005):

The slope of the BLR size–Hβ luminosity relation found here is almost identical to the slope found using λLλ (5100 Å). This is in accord with studies that show that the Balmer line luminosity scales as the optical luminosity between different AGNs (e.g., Yee 1980; Shuder 1981).

Adapted from Boyer et al. (2002):

No other optical method, not even near-field optics (19), is able to detect nonfluorescing objects as small as 2.5 nm. This represents a gain of more than three orders of magnitude in volume over the current optical detection by plasmon Rayleigh scattering.


Finally,  your work should be expanding on the work of others, as you should have probably delineated in your Introduction section. Although you should certainly circle back to these ideas when you draw the broader conclusions of your work (see Organization, below), you may find that you want to elaborate on this more specifically in discussing a result by referencing a specific work or works that your findings corroborate, refute, or expand upon. In other words, you are showing how you are building upon others’ work, just as you promised you would in your Introduction. This strategy therefore also allows you to exemplify how or why your findings are important.


Adapted from Najafi, Zwiers, and Gillett (2015):

Although a role for aerosols in driving mid-century Arctic cooling has previously been proposed on the basis of analysis of model simulations,11 these results demonstrate for the first time that an aerosol contribution to temperature variations is detectable in observations as well.

Adapted from Costanza-Robinson and Brusseau (2002):

Kim et al. [1999] noted that the gaseous and aqueous tracer methods may probe different physical domains within a porous medium, due to limitations in hydrodynamic access of the aqueous-based surfactant tracers. It is further hypothesized here that this difference in domain access may account for the much larger interfacial areas measured by the gaseous tracers.



By assessing your findings in the context of other published work you connect it to a larger body of evidence. This ultimately moves the field in the direction of scientific coherence, the goal of all research.


Test yourself

Read each of the following excerpts from published journal articles. Explain how the literature reference is used to evaluate the finding. Does it support the finding? Does it show that the finding is important? How?

  1. Using this technique, the extracted Eg ~0.494 eV, as shown in Fig. 7 for the experimental CNFET of Javey et al.,5 is in close agreement with the reported 0.5 eV. (Adapted from Akinwande et al. 2008)
  2. Our findings make available the MAT sequences of Gibberella/Fusarium, which allows investigation of several key issues concerning the biology of this group. First, the relative importance of sexual spores in the G. zeae disease cycle (Bowden and Leslie, 1999) can be examined. (Adapted from Yun et al. 2000)
  3. The low degree of asymmetry between the upper and lower branches of the Dirac cone implies a nearly constant band velocity vB = 0.38(3) × 106 m/s, consistent with previous estimates by Jianget et al. [25] (0.46 × 106 m/s) and Riemann et al. [29] (0.25 × 106 m/s). (Adapted from Johannsen et al. 2015)
  4. Strong correlations between diurnal peaks in moulin hydraulic heads and ice velocity suggest that pressure variability in the channelized drain-age system reduces basal friction in an adjacent active but unchannelized component of the hydraulic system and drives diurnal ice acceleration, as observed in alpine glaciers (14). (Adapted from Andrews et al. 2014)
  5. Since the G. zeae MAT locus does not carry a MAT gene fusion, a molecular mechanism for evolution from one reproductive mode to the other for the genus Gibberella cannot be proposed at this time. A plausible strategy to examine lifestyle evolution in Gibberella might be to compare G. zeae MAT genes with those of the closely related species F. cereali (Aoki and O’Donnell, 1999).


  1. The agreement of the finding and published experimental data supports the effectiveness of the technique.
  2. The published work gives an example of the usefulness of the finding.
  3. The published data support the accuracy of the finding.
  4. The published work lends credibility to the author’s suggestion about the meaning of the finding.
  5. The published work provides a foundation to help find the answer to a question that couldn’t be answered using the current findings.


Remember: to write concisely, you should reference as many sources as necessary to support your claims—but no more! If you need more help with paraphrasing and other aspects of conciseness, visit this page.





Remember that your discussion should move from the specific to the general in order to situate your findings in context. You can accomplish this by first discussing smaller, specific aspects of your results and gradually moving to larger portions of your research, such as what your overall conclusions mean to your field or to the world.

Some authors choose to help accomplish this organizational structure though ending with a separate section called a  Conclusions section, in which they discuss the overall implications of, impacts of, and questions that arise from the project at hand. Another option is to designate a “Conclusions” sub-heading within your Discussion. But this broadness does not have to be so explicit–as long as you begin your discussion with remarks about experiments within your project and end with discussion about the project as a whole, you have achieved at least a basic “base of the hourglass” shape.


Move structure




1. Discuss specific results. For the sake of clarity and consistency, you should discuss results in the order in which you presented them in your Results section. You should spend only a sentence or so accomplishing sub-move 1, reminding your reader of a set of results before discussing it. (In a combined Results and Discussion section, you can skip sub-move (i) altogether; instead, after presenting a particular set of results, make Discussion Move 1(ii) before presenting the next set of results. And repeat!)


An example of Move 1(i)

1(i) is in bold and is followed by 1(ii). Adapted from Laws et al. (2000):

All of the test chemicals evaluated were capable of inducing an uterotrophic response within 3 days in the prepubertal Long Evans rats. Based upon the dose-response data for 4-ter-octylphenol and 4-nonylphenol, the 3-day uterotrophic assay in the prepubertal rats was the best indicator of estrogenic activity when compared with the age at vaginal opening in prepubertal rats, estrous cyclicity in intact adult rats or changes in vaginal cytology.


Sub-move (ii) is arguably the most important part of your published paper and is certainly the meat of your discussion section. Because in broad terms the goal of your project should be to contribute to a greater understanding of a broader scientific concept, your discussion should connect your project to those of others by considering how your work supports, refutes, expands, or otherwise connects to previously published work. It is critical that you avoid “hand-waving” arguments in which you make claims about your results that have little scientific evidence to support them.

Repeat Move 1 for each result you think is important to discuss. Usually, each subsequent Move 1 iteration is begun in a new paragraph. Some results can be grouped together by discussing trends or patterns in the data; this strategy should be obvious if you already have identified data in such groups in your Results section. Remember to gradually broaden the focus of each paragraph as you move throughout your Discussion, if possible.


2. Draw broader conclusions. This move allows you to broaden the scope of your paper again to the general scientific work. If it comprises multiple paragraphs, sometimes all or part of Move 2 is sequestered to a separate Conclusions section. In this section you should attempt to answer such questions as: What should other scientists consider when thinking about the impact of your paper, including its limitations and applications? What do you think should be done next to further the accomplishments of your project? Although not every one of these sub-moves is necessary depending on your discipline and project, this move is critical for showing the importance and meaningfulness of your work. As usual, the more concise and scientifically grounded you are throughout this move, the more convincing your writing will be. It will greatly help your reader here to tie your conclusion back to your introduction, showing your reader how you finally addressed your gap.


Org-8: Test yourself

Each of the following excerpts are taken out of context from journal article Discussion sections. Assuming each excerpt as a whole accomplishes Move 2, identify which of the four sub-moves is accomplished by each sentence. Note that not every excerpt will contain all of the sub-moves. (Sentences are numbered with italicised superscript numerasl for reference in the Solutions.)

1. Adapted from Steelman, Kennedy, and Parker (2015):

1 The riverbed along the Eramosa River is an intermittently-exposed bedrock surface, where groundwater and surface water are variably connected through networks of open and sediment-filled discrete fractures and conduits. 2 The identification of epikarst partially covered by fluvial sediment and its relationship to the formation open and exposed fracture conduit networks and suspected sinkholes, advanced our understanding of hydraulic and temperature responses in the shallow bedrock aquifer. 3 Our ability to estimate physical properties of the bulk rock formation to 15–20 m depth using GPR resulted in a better understanding of hydraulic conditions and aquifer properties across the study area, improving our understanding of spatial variability of advective pathways along this river system. 4 The approach used in this study could be used to construct larger-scale 2D layer models of shallow groundwater aquifers enhancing the development of local flow models capable of assessing groundwater–surface water dynamics.

5 While our geophysical approach was effective in detecting the position and frequency of vertical and subvertical fracture zones and exploring the general characteristics of a bedrock river system, it did not provide direct evidence of the mechanisms governing spatial variations in bedrock riverbed features (e.g., buried channels, fractures and conduits, sinkholes) or their hydrogeologic significant to the local flow system. 6 Based on the available geophysical, hydrogeological and geological information, we suspect that lateral variations in erosion-resistant layering (e.g., fine-grained, well-cemented, thinly-laminated rock) are likely controlling channel morphology, as well as the occurrence and development of discrete fracture networks and conduits which transiently control surface water–groundwater connectivity from low to high stage periods. 7Nevertheless, this initial work represents a basis for future studies aimed at characterizing groundwater–surface water exchange patterns in a bedrock river environment.

2. Adapted from Mottaleb, Woo, and Kim (2001):

1 In ELCAD-AES analysis, a stable glow discharge plasma could be obtained by applying 1-2 kV in an atmospheric air pressure and a several mm discharge gap between the Pt rod anode and electrolyte solution cathode. 2 Spectrum emitted from waters contained basic atomic lines of dissolved metals and OH bands peaks. 3 The intensity of spectral lines depends on the operating parameters such as pH of the electrolytes, discharge current and discharge gap. 4 The LOD values of Hg, Pb, Cd, Cu, Na and K were found in the range of 0.001-0.08 mg/l. 5 The LOD values of Hg, Pb, Cd and Cu were improved by more than one order of magnitude compared to closed-type ECLAD. 6 The applicability of ELCAD-AES has been assessed for fresh milk analysis. 7 The concentrations of different metals such as Cu, Pb, Fe, Mn, Ca, Zn, Mg, Na and K found in drinking water and fresh milk support the WHO and Korean Environmental Protection Agency guidelines. 8 Finally, this technique could have potential and effectively be used for analysis of other liquid samples such seawater, fruit juices, etc.

3. Adapted from Stensvold et al. (2015):

1 Future studies should monitor the prevalence, distribution, and host specificity of Babesia in ticks from different geographical regions in Denmark and neighbouring countries in order to increase awareness and to enable assessment of the potential public health risk of contracting babesiosis in this region where infections may be emerging. 2 Furthermore, it should be investigated, to which extent patients with Lyme disease are seropositive for Babesia. 3 Low diagnostic sensitivity related to traditional diagnosis of acute babesiosis can be overcome by the use of Babesia-specific PCR, suitable for direct detection of the pathogen in blood samples from patients with relevant exposure and unexplained fever.


4 This study is the first to detect of B. microti in Danish ticks. We confirmed the presence of B. microti in I. ricinus collected at multiple locations in Denmark. 5 Our data evinces endemic occurrence of potentially zoonotic Babesia in Denmark and confirms I. ricinus as a vector of multiple pathogens of public health concern. 6 Awareness of the potential impact of Babesia on public health in Scandinavia could be increased by monitoring the prevalence, distribution, and host specificity of Babesia in ticks.

4. Adapted from Pukhov and Meyer-ter-Vehn (2002):

1 In this work we have considered the wake field acceleration of electrons in the broken-wave regime. 2 This kinetic and highly non-linear regime is difficult to treat analytically, and therefore we have studied it numerically. 3 We find that the 3D wave breaking of the laser wake field can produce electron bunches with unique properties. 4 A sub-10-fs laser pulse with only 20-mJ energy is able to accelerate 109 electrons with a flat spectrum reaching 50-MeV energy. 5 A spectrum sharply peaked in energy is obtained for a 12-J, 30-fs laser pulse. 6 In this case the laser is intense enough to expel all background electrons from the first wake trough, forming a solitary cavity and a completely broken wake field further downstream. 7 Propagating through plasma, this cavity continuously traps a small portion of background electrons and generates an almost mono-energetic beam of 3 × 1010 electrons at 300 MeV.

8 These new results call for experimental verification. 9 At present we have to rely on our 3D PIC simulations. 10 However, we recall that PIC simulations have been very successful in reproducing existing experiments [12, 14, 34]. 11 The fast development of laser technology suggests that the laser pulses with corresponding parameters will be available in the near future.


1.    Sentences 1-3: Move 2(i)

Sentence 3: Move 2(iii)

Sentence 4: Move 2(iv)

Sentences 5-6: Move 2(ii) –although sentence 6 isn’t exactly identifying a limitation of the project, it is further explaining the limitation identifyed in sentence 5.

Sentences 7: Move 2(iii)

2.    Sentences 1-5: Move 2(i)

Sentences 6-8: Move 2(iii)

3.    Sentences 1-3: Move 2(iv)

Sentence 4: Move 2(i)

Sentence 5: Move 2(iii)

Sentence 6: Move 2(iv) –yes, again! Here they decided to sum up the future work in one concise, overarching statement, which helps broaden their discussion back out at the very end of their paper.

4.    Sentences 1-7: Move 2(i)

Sentence 8: Move 2(iv)

Sentence 9: Move 2(ii)

Sentence 10-11: The end to this paper is unconventional. The authors make an effort to further validify their results by explaining the accuracy of the method. This kind of discussion might be better suited to earlier in the discussion (as part of Move 1), but nonetheless somewhat implies Move 2(iv) by referring to the future of this type of project and therefore succeeds in broadening the scope of the end of the paper.


Note how none of these papers use all of the available moves or use them in the same order! The most important thing is to use the moves that make the most sense for your project and what you are trying to accomplish.