Article | Open

On the definition and identifiability of the alleged “hiatus” in global warming

Published online:


Recent public debate and the scientific literature have frequently cited a “pause” or “hiatus” in global warming. Yet, multiple sources of evidence show that climate change continues unabated, raising questions about the status of the “hiatus”. To examine whether the notion of a “hiatus” is justified by the available data, we first document that there are multiple definitions of the “hiatus” in the literature, with its presumed onset spanning a decade. For each of these definitions we compare the associated temperature trend against trends of equivalent length in the entire record of modern global warming. The analysis shows that the “hiatus” trends are encompassed within the overall distribution of observed trends. We next assess the magnitude and significance of all possible trends up to 25 years duration looking backwards from each year over the past 30 years. At every year during the past 30 years, the immediately preceding warming trend was always significant when 17 years (or more) were included in the calculation, alleged “hiatus” periods notwithstanding. If current definitions of the “pause” used in the literature are applied to the historical record, then the climate system “paused” for more than 1/3 of the period during which temperatures rose 0.6 K.


“There was no such thing as the Scientific Revolution, and this is a book about it.”

—Steven Shapin, 1996, The scientific revolution. University of Chicago Press.

In the public sphere, the claim that global warming has “stopped” has long been a contrarian talking point1,2. After being confined to the media and internet blogs for some time, this contrarian framing eventually found entry into the scientific literature3,4, which is now replete with articles that address a presumed recent “pause” or “hiatus” in global warming5. The “hiatus” also featured as an accepted fact in the latest assessment report of the IPCC6. Despite its widespread acceptance in the scientific community, there are reasons to be skeptical of the existence of the “hiatus”5.

Recently, possible artifacts in the global surface temperature record have been noted which, when corrected, suggest that there is little evidence for a “hiatus” relative to the long-term trend used by the IPCC7. In addition, multiple other indicators such as ocean heat content point to continued warming8,9,10.

In this article, we show that even putting aside possible artifacts in the temperature record, there is no substantive evidence for a “pause” or “hiatus” in warming. We suggest that the use of those terms is therefore inaccurate. Because this conclusion appears to contradict the IPCC’s explicit endorsement of the “hiatus”, it is important to differentiate between the different ways in which the term “pause” or “hiatus” has been motivated and used in the recent climatological literature.

Research on the “hiatus” has been couched within at least 4 distinct research questions: (1) Is there a “pause” or “hiatus” in warming? (2) Has warming slowed significantly compared to the long-term trend? (3) Has warming lagged behind model-derived expectations? (4) What are the physical mechanisms responsible for the “hiatus”? Here, we are exclusively concerned with the first question: Is there, or has there recently been, a “pause” or “hiatus” in warming? We focus on this question because it is ineluctably tied to the contrarian claim that global warming has “stopped”, which has demonstrably affected the political and media landscape3 as well as, arguably, the scientific community4. The question whether there is a “pause” in global warming can be readily tested: Standard dictionary definitions of the words “pause” or “hiatus” imply that a process has been suspended or interrupted. It follows that for the notion of a “hiatus” in global warming to be scientifically well-founded, there must either be a demonstrable and statistically-relevant absence of any trend in global mean surface temperature (GMST) during the time period that is considered relevant or, minimally, the observed trend must differ in a statistically identifiable way from the historical record.

Our focus on the question whether there is a “hiatus” or “pause” implies that we do not address two related issues: First, we are not concerned with the differences, if any, between climate model projections and observed GMST trends. We have addressed the issue whether or not warming has lagged behind model-derived expectations elsewhere11, and this issue has no bearing on the existence of a “hiatus”. Second, we are not concerned with the underlying physical processes that may explain fluctuations, whether positive or negative, in GMST. This is again a different question, which is interesting in its own right but has no bearing on the existence of a “hiatus”.

We examine the status of the “hiatus” in three steps: First, we compile an inventory of operationalizations of the “hiatus” in the existing scientific literature and ask whether they converge on a consistent definition. Second, we ask whether the rate of temperature change during the “hiatus”, as it is operationalized in the literature, differs meaningfully from the set of rates for equivalent trend lengths observed during the era of modern climate change. This comparison is essential because any trend will exhibit periods of statistical insignificance when the sample size (i.e., number of years considered) is small: The existence of the presumed “hiatus” thus cannot be ascertained without a historical comparison to other comparable trend durations at earlier times during which warming was consensually thought to be present. Finally, for the same reason, we ask whether the duration of periods in which there is no significant warming has changed during the presumed “hiatus” relative to the rest of the modern period.


There is no agreed “hiatus” period in the scientific literature

We catalogued a corpus of peer-reviewed articles published between 2009 and 2014 that specifically addressed the presumed “hiatus” in global warming. Table 1 shows that the term “hiatus” was used more than 550 times in this corpus, and the word “pause” in excess of 70 times.

Table 1: Summary of literature on the “hiatus”.

Many articles assumed that the “hiatus” commenced around 1998, at which time temperature anomalies were considerably above the long-term trend. There is, however, considerable heterogeneity in published onset times, with the range spanning a decade (1993–2003). Similarly, there is considerable heterogeneity in the presumed duration of the “hiatus” across the same corpus of articles, with a range 10–20 (median 13 years, , ). For each article, we took the duration to be the number of years since the assumed onset of the “hiatus” to the end of the period being analyzed. This constitutes a lower bound on the presumed duration of the “hiatus” as some authors may have presumed that the “hiatus” was ongoing at the time they published an article. Figure 1 shows the modern global temperature data together with a histogram of the distribution of presumed onset times of the “hiatus” derived from the corpus.

Figure 1: Global mean surface temperature (GMST) anomalies estimated by NASA’s Goddard Institute for Space Studies (GISS) data set (40, all analyses based on dataset downloaded on 17 January 2015).
Figure 1

The histogram at the bottom represents the distribution of presumed start years for the presumed “hiatus” in the corpus of articles (; see Table 1) considered for this analysis. The vertical lines represent the 5th (1993) and 95th (2001) percentile, respectively, of presumed starting years for the “hiatus”. The small inset shows the overall historical temperature anomalies recorded since 1880.

The heterogeneity in onset and duration raises the possibility that the use of the term “hiatus” departs from normal scientific practice, which strives to define phenomena on the basis of clear and generally accepted criteria. The heterogeneity may be explained by the supposition that authors defined the “hiatus” retrospectively, via an ad hoc analysis of the recent trend leading up to the time of writing, rather than on the basis of a priori criteria. This apparent lack of clear and a priori criteria must be of concern in the statistical environment in which the “hiatus” has unfolded, which is known to be sensitive to the particular choice of start and end points that define short-term trends and the comparison baseline12.

The “hiatus” is an unexceptional fluctuation

If the definitions of the presumed hiatus are highly variable, with many different time periods proposed in the literature, how can we determine whether or not there is one? In order to answer this question, we compared the distribution of decadal warming trends during the “hiatus”—as defined by the articles in the corpus—against the distribution of all possible trends that have been observed during the period of modern global warming. The results are shown in Fig. 2, using three different onset dates for global warming.

Figure 2
Figure 2

(A) distribution of observed decadal temperature trends (GISS) within the “hiatus” windows defined by the corpus of articles considered for this analysis (blue), compared to the distribution of all possible temperature trends from 1950 till 2012, the reference period used by the IPCC to establish the long-term warming trend (pink). (B) same distribution of temperature trends within the “hiatus” windows (blue) compared to the distribution of all possible temperature trends from 1964 till 2012 (pink). The year 1964 is the lower bound for the 95% confidence interval of a recent change-point analysis that sought to identify the onset of modern global warming. (C) same distribution of temperature trends within the “hiatus” windows (blue) compared to the distribution of all possible temperature trends from 1976 till 2012 (pink). The year 1976 is the upper bound for the 95% confidence interval of a recent change-point analysis that sought to identify the onset of modern global warming. In all panels, the distribution of all possible trends is obtained by computing all trends of a given duration from all possible years within the time period considered. The duration of trends is weighted by the propensity of presumed “hiatus” durations in the corpus. Thus, each 10-year trend is replicated 8 times (as 8 articles in the corpus presumed the “hiatus” to extend over 10 years), each 11-year trend 5 times, and so on. See Table 1 for details of the distribution of presumed “hiatus” durations in the corpus. The vertical red lines in each panel represents the long-term trend (1951–2012) that was used by the IPCC in their Fifth Assessment Report as a benchmark for comparison with the “hiatus.” The solid line is for the GISS dataset40 analyzed here, and the dashed line is the same long-term trend using the UK Met Office’s HadCRUT4 data set39 used by the IPCC.

The question of when, precisely, greenhouse-driven warming began to be observable against background natural variability is itself contested. An early review13 that examined the literature back to 1824 finds that scientific concern about global warming arose as early as 1938. Every decade since then has seen increased scientific attention and concern13, although no consensual onset date for global warming has been identified. Figure 2 therefore uses three different onset dates for the computation of all possible trends. Panel A uses the period 1951–2012, which was used by the IPCC in AR5 as the long term trend against which to define the “hiatus”6. Panel B uses 1964 as the onset of modern global warming, whereas Panel C uses 1976. Those two years are two standard deviations () below and above, respectively, of the best estimate (1970) of the onset of modern global warming in the GISS data set reported in a recent change-point analysis14. Panels B and C therefore approximate the lower and upper bound, respectively, of the 95% confidence interval for the onset of modern global warming by the change-point measure. All panels include data through 2012 because many of the articles in the corpus were written when the latest available data were for 2012 (or even earlier). (See the Online Supplementary Material for an extension of our analysis to the entire instrumental record.)

To permit a commensurable comparison, in all panels the distribution of all possible trends has the same propensity of trend durations as the “hiatus” in the corpus. Thus, each possible 10-year trend is replicated 8 times (as 8 articles in the corpus presumed the “hiatus” to extend over 10 years), each 11-year trend 5 times, and so on as determined by the propensity of trend durations in the corpus. The distribution of trend durations is therefore identical between the two histograms in each panel.

Figure 2 demonstrates that, although the distribution of trends during the “hiatus” is shifted downward compared to the overall distribution of trends of the same durations, the “hiatus” distribution falls within the overall envelope of historically observed trends. For the IPCC base period (1951–2012; Panel A) there is little discernable difference between the two distributions. For the two years that bracket the most likely change-point onset of the modern warming period (Panels B and C), the “hiatus” distribution is more clearly offset towards the lower end but it is by no means unusual or extreme.

Moreover, for nearly 15% of imputed “hiatus” trends (5 out of 40 articles in the corpus), the warming exceeded the long-term trend used by the IPCC (1951–2012; vertical red lines in Fig. 2). Similarly, nearly 20% of operationalizations (7/40) referred to a period during which temperatures increased significantly (i.e., in OLS regression), which is not consistent with a “hiatus.”

The results in Fig. 2 show that all operationalizations of the “hiatus” in the literature are unexceptional in the context of equivalent-length trends in the record of modern global warming. At most, the operationalizations in the literature support the conclusion that the rates of warming over some recent intervals have been toward the lower end of the historically-observed surface temperature record. However, they do not support the conclusion that there is a “pause” or “hiatus” in the warming.

The “hiatus” has always been there when sample size is small

We next analyzed the GMST data from all possible different vantage points (end years looking back in time) to examine whether a scientist in, say, 2014 or 2010 would have been justified in accepting the existence of a “hiatus” in warming relative to what would have been detectable at any other prior point in time.

Figure 3(A) shows the warming trends that were observable, given the available data at the time, for any vantage point between 1984 and 2014 (horizontal axis). For each vantage point, between 3 and 25 years were included in the trend calculation (vertical axis). The Online Supplementary Material extends this analysis to even longer time scales. Timescales of at least 17 years are known to be necessary for noise reduction and detection of a signal15.

Figure 3
Figure 3

(A) Observed magnitude of temperature trends (GISS, K/decade) as a function of vantage year and the number of years included in the computation of the trend. Trends are capped at ±1 K for plotting. For each vantage year, trends are computed for all possible windows between 3 and 25 years duration, all of which end with the particular vantage year. The dots indicate which trends are significant () in an ordinary least squares analysis of annual means, and the horizontal dashed line indicates the number of years that must be included () for the trend to be significant from all vantage points. The open circles identify combinations of onset and duration that have been used to identify the “hiatus” by articles in the corpus. Multiple articles may contribute to a given circle. The Online Supplementary Material shows that the basic conclusions are unaffected by consideration of autocorrelations, although an additional 2 years are required to reach significance for all vantage points across the entire 30-year period. (B) Level of statistical significance for trends (GISS, K/decade) as a function of vantage year and the number of years included in the computation of the trend. Trends that are clearly non-significant () are shown in beige, those that approach significance () are shown in shades of gray, and significant trends () are shown in shades of terracotta. The diagonal lines identify calendar years that contribute to the analysis. Any observation in the grid that lies to the Southeast of a given line includes only observations from the stated year onward, and any observation to the Northwest also includes earlier years. The observation in the top-left corner is 1960 (i.e., looking backward 25 years from 1984).

Figure 3(A) shows that at every year (vantage point) during the past 30 years, the immediately preceding warming trend was always significant when 17 years (or more) were included in the calculation (dots denote ). Figure 3(B) presents the same data using a ternary classification of p-values for the linear trend into non-informative (beige), partially informative but not conventionally significant (gray), and significant (terracotta). This panel also includes three diagonal lines that identify the earliest calendar year included in the analysis. Thus, any observation to the Southeast of the line labeled “1975” only includes observations later than that, and so on for the other two lines. The observations to the Northwest of “1965” go back to 1960 (top-left corner; looking back 25 years from 1984 inclusive).

The large beige area in Panel B highlights the well-known fact that when sample size is small, statistical power is often insufficient to differentiate signal from noise. Conversely, the large terracotta area highlights the fact that when power is sufficient, the warming signal has been detectable at any point during the last 30 years, irrespective of vantage point. When one extends the period looking backwards in time, the warming trend is always significant, and the most recent vantage point(s) do not differ systematically from earlier vantage points. It follows that the data do not permit identification of a “pause” or “hiatus” during the last 10–20 years. Significantly, this conclusion is unaffected by the choice of year taken to represent the onset of modern warming (i.e., areas to the Southeast of all 3 diagonal lines in Figure 3(B) permit the same conclusion). The conclusion is also unaffected by the choice of the year during which the “pause” was examined (i.e., the vantage point).

Conversely, if one uses shorter time periods of analysis, one can find many “pauses.” Using the operationalizations found in the corpus (mean duration 13.5 years), and a null hypothesis of no warming, we find that the climate “paused” strikingly often during the last 30 years. During that period, the 14-year trend escaped significance 10 times and the 13-year trend 13 times, suggesting that a “pause” occupied between 30% and 43% of a time period during which the climate warmed 0.6 K overall (Fig. 1). If the duration of the defined “hiatus” drops to below 12 years—which applies to 13 out of 40 articles (i.e., 32.5%) in the corpus—then almost everything is a “hiatus”, as signified by the preponderance of beige for trends of this duration in Fig. 3(B). Anyone making a “hiatus” claim of this duration will almost always find one, not because something new and different is happening, but because of the fundamental fact that small sample sizes provide insufficient statistical power for the detection of trends. Thus, a third of the articles in the corpus either presumed that the climate has nearly always “paused” during the last 30 years (rendering the term meaningless), or they inconsistently highlighted only one of many events that would qualify with their definition.

These results have been replicated using a variety of additional methods that incorporate autocorrelations in the time series (see the Online Supplementary Material). The results are not sensitive to the trend detection methods employed, and they are also not sensitive to the choice of GMST data set (see the Online Supplementary Material).

We conclude that the evidence does not support the notion of a “pause” or “hiatus” as an identifiable phenomenon that is implied by standard dictionary definitions and common understandings of these terms.


We recognize that our claim that there is no “hiatus” will be controversial, particularly in light of the widespread embrace of the “hiatus” in public and scientific discourse. Therefore, it is important to clarify what we are not claiming. First, and perhaps most important, we do not argue against the merit of research on decadal-scale variation in the climate. On the contrary, the numerous articles on the “hiatus” have contributed to our understanding of what drives decadal fluctuations in climate, including for example its seasonal aspects16. Notably, none of the articles in our corpus indicate that they expect the “hiatus” to continue indefinitely, implying that they do not support some public interpretations that recent fluctuations in the GMST undermine the scientific basis for understanding anthropogenic climate change17.

Second, our exclusive focus on GMST relative to the null hypothesis of no trend was mandated by our goal to examine the notion of a “pause” or “hiatus” with respect to the observations alone. It does not follow that global trends constitute the only—or even preferable—level of analysis for the climate system.

Third, we do not explicitly address the question whether warming has slowed significantly during the presumed “hiatus” period, although we have suggested elsewhere that it has not5. In confirmation, a recent change-point analysis of GMST has shown that there is no statistically-identifiable change in warming trend after the 1970s14.

Fourth, our analysis does not speak to the apparent or presumed discrepancy between model projections and GMST trends. Research on this question has identified several effects and variables that can reconcile apparent differences between modeled and observed temperatures during the recent fluctuation, such as model-versus-observed differences in the phasing of internal variability11,18,19, systematic errors in some of the external forcings used in CMIP5 simulations20,21, and incomplete coverage and quality of observations7.

Finally, our demonstration that the “hiatus” is statistically indistinguishable from previous fluctuations has no bearing on the question of the physical causes of fluctuations in surface temperature trends. Such fluctuations can be due to internal variability alone12,22,23, or they may involve variations in external forcings on the climate system such as solar cycles or volcanic eruptions, or both24,25,26. We have no commitment to a particular causal model of those fluctuations.


We have shown that there is a wide range of different operationalizations of the “hiatus” in the literature. For none of these operationalizations is the rate of temperature change meaningfully different from the set of rates of equivalent trend lengths over the modern period. That is, the “hiatus”, however defined, is not unusual or unprecedented27. Further, the duration of periods over which trends must be extended to generate significant warming trends has not changed noticeably in the “hiatus” periods relative to the rest of the modern warming period. We conclude that there is no “hiatus”, and neither has the climate system “paused.”

Our conclusion raises at least two questions. First, why has so much research been directed at the “hiatus” when it does not exist? We have addressed the likely reasons for this in detail elsewhere4. The notion of a “pause” or “hiatus” demonstrably originated outside the scientific community3, and it likely found entry into the scientific discourse because of the constant challenge by contrarian voices that are known to affect scientific communication and conduct4,28,29.

The second question pertains to the broader implications of this apparent discord between data and the discussion in the literature. We suggest that discussing climate change using the terms “pause” or “hiatus” creates notable hazards for the scientific community.

Adoption of the terms “hiatus” or “pause” is not inconsequential because the way in which environmental issues are linguistically and semantically framed contributes crucially to public (mis-)understanding30. Scientists may argue that when they use the terms “pause” or “hiatus” they know—and their colleagues understand—that they do not mean to imply that global warming has stopped. Indeed, the use of scare quotes in some articles (Table 1) is clearly intended to imply this. The problem, however, is that these terms have vernacular meanings, and when scientists use a term from the public vernacular to describe a feature of science, confusion results when the vernacular term is not an appropriate description of that feature. This misunderstanding may be particularly acute in this instance because the terms “pause” and “hiatus” originated as contrarian talking points3,4. Hence, we argue that scientists should use the term that most appropriately describes what they are studying. In the present case, that implies the use of “fluctuation”, not “hiatus,” because when scientists use the term “hiatus”, this sends a confusing and potentially misleading message to the public. Scientists might tacitly understand that global warming continues notwithstanding the alleged “hiatus”, or they may intend the “pause” to refer to differences between observed temperatures and expectations from theory or models, but the public is not privy to that tacit understanding.

Moreover, acceptance and use of scientific propositions carries ethical implications and responsibility31,32. Some philosophers argue that holding a belief without sufficient “warrant”—i.e., without support by strong evidence—engenders a moral hazard33. An important element of this argument is that any belief, no matter how innocuous or inconsequential, creates the enabling conditions for similar and related beliefs. Any belief or opinion thus contributes to shaping an epistemological landscape, which in turn implies a responsibility—or when the belief is unwarranted, a moral hazard—for “downstream” intellectual consequences. Specifically, if unwarranted acceptance of a “hiatus” in global warming contributed to the delay of political action to mitigate climate change, with potentially adverse consequences on innocent parties, then the scientific status of the “hiatus” could become a matter not just of science and philosophy, but also ethics and even law. Lest one consider such a potential hazard remote, the legal aftermath of the earthquake in L’Aquila, Italy, which embroiled scientists in charges of manslaughter for their alleged failure to warn the community34,35, vividly illustrates the legal and moral hazards that are incurred when the public is not adequately informed of the full envelope of identifiable risks arising from scientific findings. In this context, it is notable that in a blind expert test, the notion that global warming has “stopped” was found to be misleading in light of the data5.

Those hazards can be largely avoided in this case by clear communication, which includes (although to be sure is not limited to) avoiding the unsubstantiated use of “pause” or “hiatus” when referring to fluctuations of GMST about the longer-term warming rate.


Corpus of articles

Table 1 summarizes the corpus of 44 articles that explicitly addressed the “hiatus”, either by seeking an explanation or by reconciling it with model output. Only articles addressing global (as opposed to regional) temperatures were included. Articles were sourced by the authors with the help of a number of other researchers and climate experts who are conversant with the current literature.

For each article, the table records the number of times that keywords such as “slowing”, “pause”, or “hiatus” occurred in the text. Occurrences in the reference section, in running heads, or in metadata were not counted. All forms of the stem were accepted; e.g., “slow”, “slowed”, “slowing”, and so on. Note that Crowley et al.36 used another term, namely “plateau”, 13 times. In addition, the word “stop” appeared 4 times in two articles37,38. Wherever a number is put into quotation marks (e.g., “1”) this refers to the number of times the term was put into “scare quotes,” implying that the term was not necessarily accepted by the author. When scare quotes were used together with unquoted occurrences, those other occurrences are provided after the “+” symbol.

Where applicable, the table also presents a quotation (usually from the abstract or first paragraph) that was judged to be indicative of the “framing” of the article. Citations or acronyms (or clauses not relevant to the meaning) in the quotation are omitted and replaced by…. When the quotation is absent for an article, a clear identification of framing was not possible. The Focus column indicates whether the “hiatus” was discussed primarily with respect to the observations (O) or with respect to the match between models and observations (M), or both (OM). The Data column indicates which data set was used by the study, where H = HadCRUT439; G = GISS40; N = NCDC41; CW = Cowtan & Way42; C3 and C5 refer to CMIP343 and CMIP544 model ensembles, respectively; and “o” refers to other data sets.

The table also records the presumed onset of the “hiatus” as stipulated in each article (column labeled From) and the end of the “hiatus” (To). Concerning onset, articles sometimes use fuzzy terminology such as “first decade of 21st century” (interpreted to mean 2000–2009) or “2000s” (also taken to mean 2000–2009), or they contain several explicit and mutually incompatible onset times (in which case the first or more explicit one was taken as the article’s declaration of onset). Similarly, the presumed end of the “hiatus” sometimes remained unclear as it was often (but not always) the “present” or time of writing of the article. It was not always possible to unambiguously identify the last observation in the data set. Because of those potential ambiguities, a second independent reader who was blind to the purpose of the study audited, and confirmed, the values derived by the first author. Unambiguous identification of onset and duration proved impossible for 4 articles, and the main analyses are therefore based on . The corpus reported in Table 1 does not claim to be exhaustive; note, however, that the inclusion of further articles cannot reduce the range of onset times—it could only extend it.

The Trend column indicates if the trend in the observations (NASA’s GISS data set;40) was significant for the time period specified (* denotes ) and whether it exceeded the IPCC’s long-term reference trend (1951–2012), denoted by >I. Entries in this column that are labeled NA are not included in the quantitative analysis because computation of the trend was prevented by ambiguity in the operationalization of the “hiatus.”

The table omits articles that did not address global mean surface temperature (GMST) but exclusively focused on other indicators such as ocean heat content or temperature9,45,46; sea level rise47; or wind48.

Additional Information

How to cite this article: Lewandowsky, S. et al. On the definition and identifiability of the alleged “hiatus” in global warming. Sci. Rep. 5, 16784; doi: 10.1038/srep16784 (2015).


  1. 1.

    There IS a problem with global warming... it stopped in 1998. URL . Accessed 18 August 2010. (2006).

  2. 2.

    2010 one of the hottest on record. URL . Accessed 2 November 2014. (2011).

  3. 3.

    Media discourse on the climate slowdown. Nature Clim. Change 4, 156–158 (2014).

  4. 4.

    , , , & Seepage: Climate change denial and its effect on the scientific community. Global Environ Change 33, 1–13 (2015).

  5. 5.

    , & The “pause” in global warming: Turning a routine fluctuation into a problem for science. B Am Meteorol Soc, 10.1175/BAMS-D-14-00106.1 (2015).

  6. 6.

    et al. Technical summary. In et al. (eds.) Climate Change 2013: The Physical Science Basis. Contribution of Working Group I to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change (Cambridge University Press, Cambridge, UK, 2013).

  7. 7.

    et al. Possible artifacts of data biases in the recent global surface warming hiatus. Science 348, 1469–1472 (2015).

  8. 8.

    et al. A review of global ocean temperature observations: Implications for ocean heat content estimates and climate change. Rev Geophys 51, 450–483 (2013).

  9. 9.

    , & Distinctive climate signals in reanalysis of global ocean heat content. Geophys Res Lett 40, 1754–1759 (2013).

  10. 10.

    , , & Quantifying underestimates of long-term upper-ocean warming. Nature Clim. Change 4, 999–1005 (2014).

  11. 11.

    et al. Well-estimated global surface warming in climate projections selected for ENSO phase. Nature Clim. Change 4, 835–840 (2014).

  12. 12.

    Has there been a hiatus? Science 349, 691 (2015).

  13. 13.

    & An annotated bibliography on the greenhouse effect and climate change. Climatic Change 21, 97–255 (1992).

  14. 14.

    , & Change points of global temperature. Environ. Res. Lett. 10, 084002 (2015).

  15. 15.

    et al. Separating signal and noise in atmospheric temperature changes: The importance of timescale. J Geophys Res-Atmos 116, D22 (2011).

  16. 16.

    , , & Seasonal aspects of the recent pause in surface warming. Nature Clim. Change 4, 911–916 (2014).

  17. 17.

    Whatever happened to global warming? URL . Accessed 2 November 2014. (2014).

  18. 18.

    et al. Recent intensification of wind-driven circulation in the pacific and the ongoing warming hiatus. Nature Clim. Change 4, 222–227 (2014).

  19. 19.

    & CMIP5 multi-model hindcasts for the mid-1970s shift and early 2000s hiatus and predictions for 2016-2035. Geophys Res Lett 41, 1711–1716 (2014).

  20. 20.

    , , , & Surface response to stratospheric aerosol changes in a coupled atmosphere-ocean model. Geophys Res Lett 40, 584–588 (2013).

  21. 21.

    , & Reconciling warming trends. Nat. Geosci. 7, 158–160 (2014).

  22. 22.

    , & NAO implicated as a predictor of Northern Hemisphere mean temperature multidecadal variability. Geophys Res Lett 40, 5497–5502 (2013).

  23. 23.

    , & A delayed oscillator model for the quasi-periodic multidecadal variability of the NAO. Clim. Dynam. 10.1007/s00382-014-2459-z (2015).

  24. 24.

    et al. Recent anthropogenic increases in SO2 from Asia have minimal impact on stratospheric aerosol. Geophys Res Lett 40, 999–1004 (2013).

  25. 25.

    et al. Volcanic contribution to decadal changes in tropospheric temperature. Nat. Geosci. 7, 185–189 (2014).

  26. 26.

    et al. The persistently variable “background” stratospheric aerosol layer and global climate change. Science 333, 866–870 (2011).

  27. 27.

    , & Determining the likelihood of pauses and surges in global warming. Geophys Res Lett 42, 5974–5982 (2015).

  28. 28.

    , , & Climate change prediction: Erring on the side of least drama? Global Environ Change 23, 327–337 (2013).

  29. 29.

    & Global warming estimates, media expectations, and the asymmetry of scientific challenge. Global Environ Change 20, 483–491 (2010).

  30. 30.

    Why it matters how we frame the environment. Environ. Comm. 4, 70–81 (2010).

  31. 31.

    Epistemic responsibility. Int. J. Philos. Stud. 16, 179–200 (2008).

  32. 32.

    A defence of epistemic responsibility: why laziness and ignorance are bad after all. Synthese 191, 3297–3309 (2014).

  33. 33.

    The ethics of inquiry, secientific belief, and public discourse. Public Aff Q 25, 197–215 (2011).

  34. 34.

    At fault? Nature 477, 264–269 (2011).

  35. 35.

    How earth science has become a social science. Hist. Soc. Res. 40, 246–270 (2015).

  36. 36.

    , & Recent global temperature “plateau” in the context of a new proxy reconstruction. Earth’s Future 2, 281–294 (2014).

  37. 37.

    , & Pause for thought. Nature Clim. Change 4, 154–156 (2014).

  38. 38.

    & An apparent hiatus in global warming? Earth’s Future 1, 19–32 (2013).

  39. 39.

    , , & Quantifying uncertainties in global and regional temperature change using an ensemble of observational estimates: The HadCRUT4 data set. J Geophys Res-Atmos 117, D08101 (2012).

  40. 40.

    , , & Global surface temperature change. Rev Geophys 48, RG4004 (2010).

  41. 41.

    , , & New surface temperature analyses for climate monitoring. Geophys Res Lett 32, L14712 (2005).

  42. 42.

    & Coverage bias in the HadCRUT4 temperature series and its impact on recent temperature trends. Q. J. R. Meteorol. Soc. 140, 1935–1944 (2014).

  43. 43.

    et al. The WCRP CMIP3 multimodel dataset: A new era in climate change research. B Am Meteorol Soc 88, 1383–1394 (2007).

  44. 44.

    , & An overview of CMIP5 and the experiment design. B Am Meteorol Soc 93, 485–498 (2012).

  45. 45.

    et al. Full-depth temperature trends in the Northeastern Atlantic through the early 21st century. Geophys Res Lett 41, 7971–7979 (2014).

  46. 46.

    & The formation of the recent cooling in the eastern tropical Pacific Ocean and the associated climate impacts: A competition of global warming, IPO, and AMO. J Geophys Res-Atmos 119, 11272–11278 (2014).

  47. 47.

    et al. The rate of sea-level rise. Nature Clim. Change 4, 358–361 (2014).

  48. 48.

    , , , & How robust is the recent strengthening of the tropical pacific trade winds? Geophys Res Lett 41, 4398–4405 (2014).

  49. 49.

    et al. Changes in global net radiative imbalance 1985-2012. Geophys Res Lett 41, 4398–4405 (2014).

  50. 50.

    , , & Top-of-atmosphere radiative contribution to unforced decadal global temperature variability in climate models. Geophys Res Lett 41, 5175–5183 (2014).

  51. 51.

    & Varying planetary heat sink led to global-warming slowdown and acceleration. Science 345, 897–903 (2014).

  52. 52.

    & The tropical Pacific ocean–Back in the driver’s seat? Science 343, 976–978 (2014).

  53. 53.

    et al. Surface warming hiatus caused by increased heat uptake across multiple ocean basins. Geophys Res Lett 41, 7868–7874 (2014).

  54. 54.

    & Is the climate warming or cooling? Geophys Res Lett 36, L08706 (2009).

  55. 55.

    , & Statistically derived contributions of diverse human influences to twentieth-century temperature changes. Nat. Geosci. 6, 1050–1055 (2013).

  56. 56.

    , & Overestimated global warming over the past 20 years. Nature Clim. Change 3, 767–769 (2013).

  57. 57.

    & Recent observed and simulated warming. Nature Clim. Change 4, 150–151 (2014).

  58. 58.

    Heat hide and seek. Nature Clim. Change 4, 158–161 (2014).

  59. 59.

    , , & Retrospective prediction of the global warming slowdown in the past decade. Nature Clim. Change 3, 649–653 (2013).

  60. 60.

    , & The impact of volcanic eruptions in the period 2000-2013 on global mean temperature trends evaluated in the HadGEM2-ES climate model. Atmos. Sci. Lett. 15, 92–96 (2014).

  61. 61.

    The cause of the pause. Nature 501, 318–319 (2013).

  62. 62.

    & Natural variability, radiative forcing and climate response in the recent hiatus reconciled. Nat. Geosci. 7, 651–656 (2014).

  63. 63.

    The role of natural climatic variation in perturbing the observed global mean temperature trend. Clim. Dynam. 36, 509–521 (2011).

  64. 64.

    , , & Attributing the increase in northern hemisphere hot summers since the late 20th century. Geophys Res Lett 41, 5192–5199 (2014).

  65. 65.

    , , & Reconciling anthropogenic climate change with observed temperature 1998-2008. Proc. Natl. Acad. Sci. USA 108, 11790–11793 (2011).

  66. 66.

    & Recent global-warming hiatus tied to equatorial Pacific surface cooling. Nature 501, 403–407 (2013).

  67. 67.

    , & ‘Category-6’ supertyphoon Haiyan in global warming hiatus: Contribution from subsurface ocean warming. Geophys Res Lett 41, 8547–8553 (2014).

  68. 68.

    Return periods of global climate fluctuations and the pause. Geophys Res Lett 41, 4704–4710 (2014).

  69. 69.

    , & On the possible interaction between internal climate variability and forced climate change. Geophys Res Lett 41, 2962–2970 (2014).

  70. 70.

    , & Application of the singular spectrum analysis technique to study the recent hiatus on the global surface temperature record. PLOS ONE 9, e107222 (2014).

  71. 71.

    , & Drivers of decadal hiatus periods in the 20th and 21st centuries. Geophys Res Lett 41, 5978–5986 (2014).

  72. 72.

    et al. Recent Walker circulation strengthening and Pacific cooling amplified by Atlantic warming. Nature Clim. Change 4, 888–892 (2014).

  73. 73.

    , , , & Model-based evidence of deep-ocean heat uptake during surface-temperature hiatus periods. Nature Clim. Change 1, 360–364 (2011).

  74. 74.

    , , , & Externally forced and internally generated decadal climate variability associated with the interdecadal pacific oscillation. J. Climate 26, 7298–7310 (2013).

  75. 75.

    , & Climate model simulations of the observed early-2000s hiatus of global warming. Nature Clim. Change 4, 898–902 (2014).

  76. 76.

    & Internal variability of Earth’s energy budget simulated by CMIP5 climate models. Environ. Res. Lett. 9, 034016 (2014).

  77. 77.

    et al. Total volcanic stratospheric aerosol optical depths and implications for global climate change. Geophys Res Lett 41, 7763–7769 (2014).

  78. 78.

    , , & No pause in the increase of hot temperature extremes. Nature Clim. Change 4, 161–163 (2014).

  79. 79.

    , , & Observed and simulated temperature extremes during the recent warming hiatus. Environ. Res. Lett. 9, 064023 (2014).

  80. 80.

    Has global warming stalled? Nature Clim. Change 3, 618–619 (2013).

  81. 81.

    et al. Contributions of stratospheric water vapor to decadal changes in the rate of global warming. Science 327, 1219–1223 (2010).

  82. 82.

    et al. Strengthening of ocean heat uptake efficiency associated with the recent climate hiatus. Geophys Res Lett 40, 3175–3179 (2013).

  83. 83.

    et al. Contribution of natural decadal variability to global warming acceleration and hiatus. Nature Clim. Change 4, 893–897 (2014).

Download references


We thank Didier Monselesan for helpful comments and suggestions on the analysis. S.L. receives funding from the Royal Society through a Wolfson Research Merit Fellowship. J.S.R. receives funding from the Grains Research and Development Corporation.

Author information


  1. School of Experimental Psychology and Cabot Institute University of Bristol, 12A Priory Road, Bristol, BS8 1TU, United Kingdom

    • Stephan Lewandowsky
  2. School of Psychology, University of Western Australia, 35 Stirling Highway, Crawley, WA 6009 Australia

    • Stephan Lewandowsky
  3. CSIRO Oceans and Atmosphere, Box 1538, Hobart, Tasmania 7001 Australia

    • James S. Risbey
  4. Department of the History of Science, Science Center 371 Harvard University, 1 Oxford Street, Cambridge, Massachusetts 02138, USA.

    • Naomi Oreskes


  1. Search for Stephan Lewandowsky in:

  2. Search for James S. Risbey in:

  3. Search for Naomi Oreskes in:


J.S.R. provided the initial idea for the data analysis and S.L. provided the initial idea for the quantitative literature review. S.L. and J.S.R. conceived and designed the study and co-wrote a first draft the paper. S.L. conducted the principal analysis. All authors contributed to the conceptual framing, editing, and revision of the manuscript.

Competing interests

The authors declare no competing financial interests.

Corresponding author

Correspondence to Stephan Lewandowsky.

Supplementary information


By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.

Creative CommonsThis work is licensed under a Creative Commons Attribution 4.0 International License. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in the credit line; if the material is not included under the Creative Commons license, users will need to obtain permission from the license holder to reproduce the material. To view a copy of this license, visit