To our knowledge this is the first study to examine the combined contribution of changes in PS, ED, and the frequency of EOs to changes in total daily energy intake in any free-living population. Annualized daily energy intake for US adults increased by 28 kcal/d/y between 1977–78 and 2003–06. Among adults in the US, we show that for the full period between 1977 and 2006 the largest contributor to change in annualized total daily energy intake was change in the number of EOs, accounting for roughly 22 kcal/d/y. Over the full time period, changes in PS accounted for the next largest proportion, at 10 kcal/d/y, and our results suggest that the ED of the average EO has actually decreased, offsetting increases in PS and the number of EOs and accounting for roughly −4 kcal/d/y of the annualized change in TE. It is important to note that this is a macro level analysis; these results do not negate the issue of how individual diet and weight change are affected by ED, PS, or eating frequency. However, to the extent that all energy intake is equal, and has an equal impact on energy imbalance, this approach to studying changes in energy intake helps guide us to interventions to reduce intake.
Our findings are in line with more detailed EO-specific research from our group, which documents increases in the frequency of reported EOs among US adults 
. At the 50th
percentile, the average number of EOs increased from 3.5/d to 5.0/d, a change that was accompanied by an increase of 400 kcal/d from meals and snacks combined 
The supersizing of portions of various food and beverage items has been the subject of much scientific research and both scientific and popular books and films 
. Our group has shown actual consumption of key food and beverage portions increased across all EOs 
. Other research shows similar increases for selected foods in the last decade, although much of this research has focused on specific food items 
(i.e., soft drinks, hamburgers, and pizza) as opposed to changes in total meal size. Rolls et al. 
have also studied combinations of water and food at meals. They report that decreasing the ED (and increasing the volume) of a meal preload by adding water results in a reduction of energy intake at lunch; giving the equivalent amount of water as a beverage separate from a meal did not affect satiety and was associated with greater energy intake compared to the condition in which water was incorporated into the meal 
The present study employs just one possible method of decomposition; utilizing the fact that daily energy intake can be defined as the number of EOs (number) multiplied by PS (grams) and ED (kcal/g) to calculate the partial and full derivatives of each of these components as contributors to overall change. Decomposing change has been previously employed, especially in the sociological literature (i.e., examining changes in fertility rates), and other methods exist 
and should be examined. In addition, the use of nationally representative cross-sectional dietary data has its limitations, particularly for trend analyses. Perhaps the most notable limitation is the introduction of the five-step multiple-pass method of 24-h recall collection, which was implemented in the NHANES 2003–04 (and subsequent 2005–06) survey. This differs from previous USDA methodologies, and could result in more accurate intake reported by individuals in later years since there is additional prompting by the interviewer. Since there are no bridging studies to determine the extent to which these methodological shifts may have resulted in systematic changes in individuals' reporting, it is not possible to know whether such confounding by time exists. Bridging studies conducted as a result of methodological shifts between the 1970s and 1980s, however, found that shifts in TE and food composition did not significantly impact results 
Additionally, in NHANES 2003–04 and 2005–06, respondents were specifically asked if they consumed water (e.g., it was added as a food item). This probing resulted in the reporting of 26,000 “water only” EOs in 2003–06 (approximately 24% of reported beverage consumption; data not shown), an increase compared to previous years, which had 200–300 observations each (range 0.06%–0.02% of reported beverage consumption; data not shown). This survey change had significant implications for the calculation of ED and the number of EOs specifically, and inclusion dramatically alters our decomposition results between 1994–98 and 2003–06, inflating the number of reported EOs and decreasing the ED of the average EO. Since this additional probing was not consistent between exams or across exam years, we excluded these additional water-only EOs to maintain consistency across time.
Another important limitation with using cross-sectional data is the possibility that results are confounded by population-level changes in factors that might influence our relationship of interest. Although we can at least partially control for some of these changes (through age, gender, and race/ethnic standardization to the earliest time point), we are unable to adjust for others (such as physical activity or the prevalence of chronic disease) because of a lack of detailed and comparable measures over time. Finally, we were limited by the possibility of increasing underestimates of actual food intake over time. Scholars have shown that adults tend to underestimate TE intake, particularly from “junk foods” and other foods that are considered to have negative health connotations 
. This is particularly true for overweight individuals 
Using cross-sectional nationally representative samples of US adults, this study documents marked increases in the number and PSs of EOs and steady overall ED per EO over the past 30 y. During the most recent period, from 1994–98 to 2003–06, there were large increases in the number of EOs, but, equally important, there were no changes in the average ED of each EO. The results suggest that as contributors to increased caloric intake over both this most recent decade and over the full 30-y period, increased EO contributed significantly more to the shift in TE intake than the other two components, although PS positively and ED negatively contributed to some extent. To the extent that energy imbalance as a result of increased energy intake contributes to obesity and its associated co-morbidities, prevention efforts should focus more on reducing EOs as a way to reduce energy imbalance.