This experiment has confirmed that laboratory rats select the energy rich foods eaten by people in preference to standard chow, eat these foods to excess and become overweight. Rats exposed to this cafeteria diet increased their body weight more than those fed chow after four weeks on their respective diets, continued to increase their body weights faster than chow fed rats, and had increased their body weight by approximately 270% after 16 weeks on the diet relative to the gain of 170% by rats fed chow. Rats on the cafeteria diet obtained double the energy of rats on the chow diet, initially obtained more protein and carbohydrate, and persistently consumed more fat, both net and per gram body weight. The initial high intake of fat presumably reflects its palatability. However, persistently high fat intakes – even when energy requirement is exceeded – may be due to other factors. For example, dietary fat impairs both oral and intestinal nutrient sensing 
, which would reduce detection of excessive fat intake, leading to insulin insensitivity 
. Hence, rats may have continued to eat excessive amounts of the high fat foods regardless of rapid weight gain, and despite the continuous availability of chow, complete in macronutrient requirements, yet the least likely to be selected (5% of total intake). There was some evidence that cafeteria-fed rats selected foods richest in protein, at least initially. Indeed, when the richest source of protein (meat pies) was removed from the energy intake data, early (4 weeks) differences seen in body weight adjusted protein intake between the groups disappear (data not shown). Foods rich in protein may have been selected because this nutrient is more effective in producing postprandial satiety than carbohydrate and fat 
. However, as mentioned earlier, this does not explain why rats did not select chow, rich in protein relative to the cafeteria style foods.
The cafeteria diet could lead to excessive weight gain simply because the foods that comprise the diet are more energy dense. Alternatively, that diet could encourage more frequent eating, eating of larger portions, or some combination of these factors. The results were clear. Cafeteria-fed rats ate more than chow-fed rats, the foods they ate were more energy dense, and therefore, they gained excessive weight. These gross differences in amounts eaten and energy intake were accompanied by marked differences in eating patterns. We used the behavioral satiety sequence to identify an eating bout as a meal and the absence of the full sequence as a snack. Using this classification, we found that cafeteria-fed rats snacked more frequently than chow-fed rats during the early (week 5) but not later (weeks 10 and 15) stages of the diet. Early snacking in cafeteria-fed rats was characterized by the fact that, having snacked, these rats were far more likely to snack again, and to do so after relatively little time had elapsed. In contrast, cafeteria-fed rats ate fewer meals than chow-fed rats across all time points in the study.
These eating trends suggest that, in the early weeks, excessive energy intake in cafeteria-fed rats may have been partly due to the fact that the cafeteria diet encouraged more frequent snacking. However, over-eating and excessive energy intake persisted across later stages of the diet when, as noted, if anything, cafeteria-fed rats spent less time eating than chow-fed rats. Thus, over-eating and excessive energy intake in these rats later in diet exposure was not due to the fact that they ate more frequently. Moreover, the differences between the two groups in amounts eaten and energy intake persisted even when adjusted for body weights, suggesting that over-eating and excessive energy intake in cafeteria-fed rats were not simply due to the fact that they were heavier (data not shown). Instead, these results imply that cafeteria-fed rats ate larger portions of the foods that they had become accustomed to eating early in the diet; therefore, their energy intake remained excessive and they gained excessive weight. It is worth noting however that, across weeks, there were no changes in the amount of time that cafeteria-fed rats spent eating, or in the amounts of food that they consumed (in both grams and kilojoules). Therefore, the fact that cafeteria- and chow-fed rats had different portion sizes was not due to an increase in portion size in the former group. Rather, chow-fed rats spent more time eating the same amount of food (in grams and kilojoules) across weeks of the diet, implying that portion size specifically decreased in this group. This result implies that the nature of the cafeteria diet was such that rats did not appropriately decrease portion sizes as they gained weight.
The overall picture that emerges from these findings is that early snacking may be a critical determinant of weight gain in cafeteria-fed rats. Early weight gain in these rats may have been excessive because they failed to reduce meal numbers in compensation for energy obtained through snacking. We reasoned that rats which failed to compensate for energy obtained through snacking would have more meals relative to their number of snacks, and therefore, snacking would make up a smaller percentage of their total eating behavior. In this respect, rats in both groups showed some degree of compensation. However, at the early time point, the relationship between snack numbers and percentage snacking was weaker in cafeteria-fed rats compared to chow-fed rats. This reduced ability to compensate at the early time point was related to terminal body weights. Those rats for whom snacking made up a large percentage of eating behavior were among the heaviest of the cafeteria-fed rats. Critically, there were no significant relationships between percentage snacking and terminal body weights at either of the later time points, suggesting that it was specifically early snacking behavior that set rats on a path which led terminally high body weights.
It is clear that a cafeteria diet initially encourages snacking on energy rich foods which are eaten to excess. Why do cafeteria diets encourage snacking? One explanation for this may be that the foods selected as snacks by the cafeteria group at 5 weeks were less likely to lead to satiety than chow. The high fat content of the cafeteria foods in particular would have contributed to this lack of satiation. For example, high fat diets often result in lower postprandial suppression of ghrelin, which acts as a potent hunger signal, relative to carbohydrates and protein 
. Variety in the cafeteria diet must also be considered. The range of foods available would have reduced the effect of sensory-specific satiety, thereby increasing intake 
. Specifically, rats offered the cafeteria diet may switch between foods, maintaining palatability and increasing the likelihood of consecutive bouts without rest/sleep, that is, of consecutive snacks. In contrast, rats fed chow may have terminated eating and rested/slept once sensory-specific satiety occurred. Any such effect of variety, however, does not explain why the increase in consecutive snacking seen in cafeteria-fed rats at 5 weeks was no longer evident at 10 and 15 weeks. Perhaps the effect was no longer seen because the foods presented had become familiar and/or less hedonically attractive.
In a previous study, Rogers and Blundell 
examined feeding patterns in rats exposed to a cafeteria diet. They found that these rats initially ate more meals than chow-fed rats (where a meal was defined retrospectively as at least 1 minute of eating followed by an interval of at least 15 minutes without eating), but that this difference declined across the course of the study. In contrast, rats on the cafeteria diet ate larger meals than chow-fed rats across the entire duration of the study. These findings seemingly stand in contrast to those obtained in the present study where Group Cafeteria ate persistently fewer meals than Group Chow. However, there are two important differences between the present study and that of Rogers and Blundell 
. First, the cafeteria diet in the earlier study consisted of chow, white bread crumbs and chocolate flakes, whereas the diet used here contained a wider range of foods; a range intended to model the variety provided by the diet in developed countries. Second, the differences between the meal patterns in the two studies likely relate to the differences in how a meal is defined [eating bout of at least one minute followed by the absence of eating for at least 15 minutes versus an eating bout followed by grooming and resting/sleeping].
Several aspects of the present findings are mirrored in people where obesity has been associated with both increased snacking 
and increased portion sizes 
. Both of these factors accompanied the development of obesity in the present study in a manner that depended on experience with the diet: frequent snacking resulting in more total bouts was evident early in the diet, and by inference, larger portion sizes were consumed later in the diet. Early and frequent snacking may be especially critical to the development of obesity. Excess energy, and therefore, weight gain may reflect a failure to compensate for the calories obtained through snacking across initial exposure to an energy rich diet, and consumption of larger portions across later exposure to that diet. There is evidence that both of these factors contribute to weight gain and obesity in people 
The early increase in snacking and persistent reduction in meals observed here is characteristic of eating patterns in adolescents (prior to obesity). Adolescents tend to snack throughout the day, skip meals 
, and snack on energy-rich foods including fast foods 
. Snacking in young adults has increased concurrently with the rise in obesity 
, supporting the link between the modern diet and changes in eating patterns. Thus early adulthood may represent a sensitive period in which eating patterns that promote weight gain are established.
The present experiment is the first to record the eating patterns of rats free to consume the energy rich foods eaten by people, and to use the behavioral satiety sequence as a way of classifying an eating bout as a meal or snack. The results are significant in two respects. First, they have important implications for dieting. Current weight loss treatments are only marginally effective in the long-term. Knowledge regarding eating patterns associated with excessive intake may assist in weight loss treatment programs, as well as in detecting individuals at risk for obesity. Second, increased snacking early in the diet period was related to greater terminal body weights in those consuming the cafeteria diet. This suggests that early and frequent consumption of palatable foods may interfere with satiety signals, and thus induce eating patterns that promote overconsumption throughout adulthood.