Early life access to hay does not affect later life oral behavior in feed-restricted heifers

Dairy cattle are often raised in environments that lack natural feeding opportunities, and they perform abnormal repetitive behaviors (ARBs) as a result. Early life restriction can affect later life behavior. We evaluated whether access to hay in the milk-fed period would affect later life behavior in heifers experiencing short-term feed restriction and whether individuals were consistent in behavioral expression over time. We had 2 competing ideas about how this would unfold. First, being raised with hay, which reduced early life ARBs, could lead to fewer ARBs later in life. Alternatively, heifers that were raised without hay and performed more ARBs in early life might be more prepared for a later feed-restricted environment and thus engage in fewer ARBs than those raised with hay. We studied 24 pair-housed Holstein heifers. As calves, they were fed milk and grain from 0 to 7 wk of age (control) or given additional forage (hay). Tongue rolling, tongue flicking, nonnutritive oral manipulation (NNOM) of pen fixtures, self-grooming, and water drinking were recorded for 12 h (0800–2000 h) during wk 4 and 6 of life using 1–0 sampling at 5-s intervals. At the start of weaning at d 50, all calves were fed a total mixed ration. All calves were fully weaned at d 60 and socially housed by d 65 to 70. After this point, all individuals were raised the same way, according to farm protocol, in groups that included both treatments. At 12.4 ± 0.6 mo of age (mean ± standard deviation), heifers were restricted to 50% of their ad libitum total mixed ration intake for 2 d as part of a short-term feed challenge. Using continuous video recording from 0800 to 2000 h on d 2 of feed restriction, we scored time spent performing oral behaviors: the 5 previously scored while they were calves, along with intersucking, allogrooming, drinking urine, NNOM of rice hull bedding, and NNOM of feed bins. We found that early life access to hay did not affect behavior performed by heifers experiencing short-term feed restriction 1 yr later. Most heifers performed a wide variety of behaviors that appeared abnormal. All heifers performed tongue rolling and NNOM, and at higher levels than when they were calves, while tongue flicks and self-grooming were performed less by heifers. Individual performance of NNOM and tongue rolling were not related across age classes [correlation coefficient (r) = 0.17 and 0.11, respectively], but tongue flicks tended to be correlated (r = 0.37). Intersucking was recorded in 67% of heifers, despite their not being able to suckle a conspecific or dam in early life. Oral behaviors were highly variable across heifers, particularly tongue rolling and intersucking. Outliers, or extreme performance of oral behaviors relative to the rest of the population, were present for many behaviors. Most outliers were expressed by unique heifers that were not extreme in other behaviors. Overall, feeding individually housed, milk-limited calves hay for their first 7 wk did not affect later life performance of oral behaviors. The considerable variability, inconsistency across ages, and excessive performance of some behaviors raises additional questions about how these develop in cattle across life stages and about what we label “abnormal.”


INTRODUCTION
Across life stages, cattle seem motivated to acquire and process solid feed in a manner consistent with their evolutionary life history as grazers and ruminants. Calves will consume forage within the first few days of life (e.g., by d 1; Downey et al., 2022) and willingly incorporate it into their diets (e.g., 15%-20% voluntary inclusion; Khan et al., 2011;Horvath and Miller-Cushon, 2019;Downey et al., 2022;Downey and Tucker, 2023a) when offered, even when grain is available ad libitum. Heifers will push weighted gates to access forage (Greter et al., 2015;Van Os et al., 2018) and push sooner if they are fed an ad libitum low-roughage diet than if they have access to a high-roughage one (Van Os et al., 2018), suggesting opportunities to process forage are important. Calves and heifers prefer longer to short cuts of forage (Greter et al., 2013;Webb et al., 2014), likely due to the increased oral processing required for long particles (reviewed by Beauchemin, 2018).
However, calves and heifers are often limited in their ability to perform motivated forage acquisition and processing behaviors, which can have negative effects. Calves are not typically provided with forage at birth (e.g., "fibrous feed" required by d 14, European Council, 2008; forage provided around d 36, USDA, 2016), and heifers can be limit-fed, or fed restricted amounts of feed (e.g., equivalent to 1.5%-2% of BW; Zanton and Heinrichs, 2008), between 4 and 22 mo of age. Under these conditions, calves and heifers perform abnormal oral behaviors (e.g., Redbo and Nordblad, 1997;Webb et al., 2015;Horvath and Miller-Cushon, 2019;Downey et al., 2022), including tongue rolling and nonnutritive oral manipulation (NNOM), or the licking, chewing, and sucking of nonfeed items. Reduced eating time is also proposed to increase intersucking, or the sucking of conspecific teats and udders (Keil and Langhans, 2001). Tongue rolling, NNOM, and intersucking are frequently classified as abnormal repetitive behaviors (ARBs), which can be indicative of welfare concerns (e.g., Mason and Latham, 2004;Garner, 2005). In both age classes, ARBs are performed at significantly lower levels, or rarely reported, when opportunities for natural feeding behaviors are provided, for example, through provision of forage for calves or ad libitum feed allowance for heifers.
"Abnormal" and "repetitive" can be difficult to define, and as such, some behaviors may be concerning that are not commonly considered ARBs. In some cases, normal behaviors may reflect abnormal underpinnings when they are performed excessively, increased under restrictive environments, or directed at nonoptimal targets. For example, water drinking can reflect normal actions like hydration or heat abatement (Jensen and Vestergaard, 2021). However, calves sometimes consume excessive amounts of water (polydipsia; Downey et al., 2022), which is also seen in other species when natural feeding behaviors are restricted [e.g., gerbils (Roper, 1978); broiler breeders ; sows (Robert et al., 1993)]. Similarly, self-grooming and tongue flicks can be normal behaviors that remove debris from the integument or nostrils (e.g., Meltzer and Githens, 1919), respectively, but both are increased under conditions where forage is limited (Downey et al., 2022). In some cases, level of performance has not yet been directly linked to feed restriction or negative outcomes, but extreme expression is evident and may represent a concern. For example, self-grooming is typi-cally performed in bouts of <10 s in calves (Downey et al., 2021;Downey and Tucker, 2023a), but outlier detection methods found that they also performed bouts that were considered excessive compared with the population at large (>2 min), including bouts of up to 30 min (Downey and Tucker, 2023a). Behaviors may also be considered abnormal based on the target of the behavior. For example, suckling is a normal behavior required for neonatal survival, but the same motion is considered abnormal when directed at nonlactating or nonfamilial conspecifics (cross-or intersucking). Similarly, bedding consumption may be a natural expression of motivation to consume forage when cattle are housed on fibrous materials (e.g., straw ;Phillips, 2004;Wang et al., 2022), but may be concerning when redirected to nonnutritive bedding material (e.g., sand; Salter et al., 2021), which can also cause impactions if excessive amounts are consumed (e.g., Melendez et al., 2007). Finally, normal behaviors expressed abnormally could be identified based on how variable they are within a population. Abnormal behaviors are known to have higher coefficients of variation (CV) than feeding or resting behaviors in other restricted species (e.g., broiler breeders; Kostal et al., 1992), suggesting CV could be helpful in describing behaviors of concern. Taken together, robust documentation of oral behavior performance and variability could lay the groundwork for future investigation into when behaviors become "abnormal" and concerning and provide important insights into how cattle behave in restricted environments.
Performance of abnormal and seemingly abnormal oral behaviors by heifers may stem from limited opportunities to perform motivated behaviors in early life. This is often suggested, but the ontogeny is rarely investigated longitudinally within the same individuals. For example, intersucking in cows is suggested to stem from intersucking as heifers (Keil et al., 2001), which is, in turn, linked to preweaning performance of cross-or intersucking (Keil and Langhans, 2001). Similarly, tongue rolling is reported to develop or increase at weaning in some cases (e.g., Seo et al., 1998) and then persist, remaining consistent across postweaning and preparturient heifer life stages (Redbo, 1998). This pattern of developing abnormal behaviors that then continue across contexts, sometimes increasing with age, is found in other species (e.g., mink; Mason, 1993). Early life complexity, such as ability to perform important behaviors, can sometimes have conflicting effects on later life expression of ARBs. In some cases, being reared in enriched environments can provide protection against later life expression of ARBs (e.g., mice; Hadley et al., 2006;Jones et al., 2011). Other times, early life complexity may actually increase ARBs on exposure to barren conditions later (e.g., mink; Latham and Mason, Downey and Tucker: EARLY HAY ACCESS DOES NOT AFFECT LATER LIFE BEHAVIOR 2010). Effects of early life feed restriction on dairy cattle have mainly focused on how limited milk provision can affect later growth and productivity (e.g., reviewed by Cantor et al., 2019) or the effects of hay on feeding behavior in the weeks following weaning. For example, calves raised without hay are more neophobic toward novel feed at weaning and have more difficulty processing it than those fed hay (Morrow et al., 2023), while early life sorting behavior and meal patterns can persist for a few weeks following weaning (Miller-Cushon et al., 2013;Engelking et al., 2020;Horvath et al., 2022). Long-term persistence in oral behaviors in response to feeding opportunities in the early rearing environment and consistency within individuals between the calf and heifer life stages have not yet been investigated.
We set out to evaluate if early life access to forage led to long-term changes in oral behaviors. Our first objective was to describe daylight performance of abnormal behavior in feed-restricted dairy heifers. Second, we evaluated whether early life feeding affected later life abnormal behaviors under a feed-restricted environment and if these responses were related over time. We used a restricted environment as abnormal behaviors in heifers are often documented under limit-fed or high-concentrate conditions (e.g., Faleiro et al., 2011;Madruga et al., 2017;Bruno et al., 2020), both of which reduce opportunities to perform natural feeding behavior; thus the behaviors of interest were likely to occur in this setting. We had 2 competing predictions, based on the contradictory patterns seen in longitudinal studies on ARBs across species. First, we expected that heifers raised with access to forage, in this case hay, would perform lower levels of abnormal behaviors (NNOM, tongue rolling, intersucking) or possibly abnormal oral behaviors (allogrooming, self-grooming, tongue flicks, water drinking) in later life compared with those raised without hay. Alternatively, we predicted that individuals who did not have access to hay as calves, and thus were raised in more restrictive conditions, would perform fewer ARBs than those raised with hay. We expected that all individuals would spend more time performing abnormal oral behaviors as heifers compared with calves, and they would show consistency across life stages. Finally, we aimed to describe variation in oral behaviors and whether individual heifers performed these at extreme levels relative to the population. Using CV and outlier detection methods, we expected to see high variability and excessive performance in abnormal and seemingly abnormal behaviors.

MATERIALS AND METHODS
The following procedures were approved by the University of California, Davis, Institutional Animal Care and Use Committee (protocol #21801). This study was conducted from July to September 2020 at the University of California, Davis, Dairy Teaching and Research Facility. We enrolled 24 Holstein heifers (12.4 ± 0.6 mo, 369 ± 24 kg, mean ± SD), representing all healthy females born between June and September 2019. These 24 heifers were studied in Downey and Tucker (2023a).

Early Rearing Environment
Before the start of this study, these 24 heifers were enrolled in an experiment evaluating the effect of early life hay provision on normal and abnormal behavior from 0 to 60 d of age (Downey and Tucker, 2023a). At that time, all calves were separated from the dam within hours of birth and housed individually in outdoor plastic hutches (2 m × 1.5 m, length × depth) with an attached wire-fenced pen (2 m × 1.5 m × 0.9 m, length × depth × height). Hutches and pens were spaced ~0.5 m apart, allowing calves to touch the muzzle of neighboring animals. The enclosures were bedded with sand approximately 12 to 17 cm deep that was spot-cleaned daily and topped up as needed.
From birth to 5 d, calves received colostrum twice a day. From 5 to 9 d of age, calves were fed 1.9 L of milk replacer (26% CP, 16% fat, 15% total solids, mixed as indicated at a rate of 142 g/L of hot water; Calva Products Inc.) at each of 2 daily meals. From 10 to 23 d of age, calves received 2.4 L of milk replacer at each of the 2 meals, and from 24 to 49 d of age, they received 2.8 L of milk replacer per meal. This milk allowance is restrictive, as it is less than ad libitum intake (e.g., up to 12 L/d; Jasper and Weary, 2002) but reflective of the amount commonly fed on US farms (USDA, 2016). All milk meals were fed via a bottle and rubber teat. Weaning began at 50 d when the morning milk meal was removed. Calves were fully weaned at 60 d. These procedures were done in accordance with farm protocol.
During the preweaning period (0-49 d), calves were fed ad libitum grain and water from a bucket (Control, n = 8) or had additional ad libitum access to chopped (19 ± 4 cm) mountain grass hay (mix of orchard, Dactylis glomerata, and fescue, Festuca arundinacea; Higby's Country Feed; Hay, n = 16). Hay was provided in either a bucket or in a 56 cm × 10.2 cm (length × diameter) PVC pipe feeder. Hay was accessible from the pipe via 4 equally spaced 6.4-cm-wide holes cut into it. Calf behavior did not differ based on method of hay provision during the preweaning period (Downey and Tucker, 2023a). On d 50, when weaning began, all calves were fed ad libitum TMR (alfalfa, almond hulls, cottonseed, corn, barley, beet pulp) via a bucket. Experimental monitoring of these heifers ceased on d 60, when calves were fully weaned.

Postweaning Environment
Postweaning, all calves were raised according to farm protocol. Calves remained in individual hutches until approximately 65 to 70 d of age. After this point, all individuals were raised in groups that included similarly aged nonexperimental individuals, with treatments mixed within pens. All individuals had similar dietary and environmental experiences from this point onward.

Experimental Design
Heifers were moved to experimental pens at 12.4 ± 0.6 mo of age (369 ± 24 kg, mean ± SD). Heifers were paired in pens and run in 5 total cohorts of 8 or 10 heifers each (4 or 5 pens of 2 heifers each per cohort). Balancing of pen assignments was done at the level of a broader experiment involving 42 total animals (24 experimental heifers and 18 additional similarly aged heifers). These 42 heifers were initially labeled according to rearing environment from 0 to 7 wk of age (hay in a bucket, hay in a PVC pipe feeder, no hay; sand or rice hull bedding) and breed (Jersey or Holstein), leading to 5 classifications: Jersey/sand/no hay, Jersey/rice hulls/ no hay, Holsteins/rice hulls/no hay, Holsteins/sand/ hay, and Holsteins/sand/no hay. This led to 10 possible unique pairings within a pen. In each cohort, there were thus 40 to 50 possible combinations of pairing and pen location (4 or 5 pens available). Heifers were sorted by age and allocated to pairs and pens randomly, starting with oldest first. We made replacements based on a priori rules to balance pair combinations across all cohorts and to ensure that each classification was represented within each one. A priori rules were also in place to replace heifers that would come into heat during the experiment, based on visual and automated heat detection before pairing. We collected information about estrus from farm management 2 d before each cohort began and applied these rules consistently throughout the experiment.
Experimental pens were outside but partially covered by a roof for shade. Each pen had an "outside" (3.7 m × 6.2 m; 3.7 m × 3.4 m covered by the roof) and "inside" (3.7 m × 3.9 m; fully covered by the roof) portion, along with a gate separating them that could be locked (Supplemental Figure S1, https: / / doi .org/ 10 .5281/ zenodo .6994585; Downey and Tucker, 2023b). Each pen had an automatically filled water trough outside and water cup inside and 1 feed bin in each portion. Feed bins and waterers were under the roof in both portions of the pens. A waterline with 2 spray nozzles per pen (TF-VP 7.5, Turbo FloodJet wide-angle flat spray tip; Spraying Systems Co.) was positioned at a height of 2.9 m and sprayed 3.4 m out over the outside portion of the pen from 0800 to 2000 h (automatically controlled using a B-hyve Smart Hose Watering Timer, Orbit Irrigation) for heifer cooling. The outside pen was bedded with rice hulls 8 to 12 cm deep over 3.7 m × 2.8 m to avoid contact with water spray. The entire inside pen had rice hulls. Pens were spot-cleaned, and new rice hulls were added daily between 0700 and 0800 h. During this time, heifer pairs were shut into the inside portion of their pens.
All heifers were given 3 to 4 d to adjust to the pens (as per von Keyserlingk et al., 2008, and Smid et al., Kononoff et al., 2003).
After adjustment to pair housing, heifers were fed the same TMR at ad libitum levels for 2 d to calculate baseline feed intake. Feed was weighed before delivery at 0800 h, preweighed feed was delivered as a top-up if low at 1600 h, and refusals were determined the next morning at 0700 h. Heifers thus did not have access to Downey  feed from 0700 to 0800 h. Across the adjustment and baseline days, heifers were fed to 110% of the previous day's intake. Baseline intake by pen was averaged across the 2 d. Individual intake was not possible to measure during the ad libitum-fed days for ethical reasons, as this would have required 24-h isolation. Instead, we calculated the proportion of overall BW each heifer made up in its pen. We then multiplied this proportion by the pen-level average feed intake to obtain a proxy for how much TMR each heifer had consumed, as higher BW is correlated with higher intake (e.g., Frisch and Vercoe, 1977;Taylor et al., 1986). After the baseline days, feed allowance was reduced by 50% for the following 2 d. This is similar to the level of feed restriction used for limit feeding on farms (approximately 1.5%-2% BW for heifers 4-22 mo old; Zanton and Heinrichs, 2008), but for ethical reasons, we imposed a duration of feed restriction (2 d) less severe than farm and research settings (minimum of 26 d up to 6 mo; Hoffman et al., 2007;Kruse et al., 2010;Kitts et al., 2011;Greter et al., 2015) as feed restriction is likely to cause hunger. Heifers were never restricted in water intake. The restricted TMR allowance (5.48 ± 0.5 kg/heifer per d, mean ± SD) was given to heifers across 2 daily feedings. Heifers were separated at feeding by locking the gate in each pen so that each heifer would have exclusive access to either the inside or the outside feed bin, with location balanced by heifer classification across cohorts. Heifers were separated until individuals across all pens ceased eating and did not return to the feed bin for at least 5 min, leading to a separation of approximately 1.5 h per feeding on both days. At this point, feed bins were swept out, refusals per heifer were calculated, and heifers were reunited. Any remaining feed from the 0800 h feeding was added back for the 1600 h feeding, but the 1600 h refusals were not added to the next day's intake.

Calf Behavioral Observations
Calf performance of drinking water, NNOM: Total, self-grooming, tongue flicks, and tongue rolling were recorded over 24 h at wk 4 and 6 of life using 1-0 sampling during 5-s intervals, as described in Downey and Tucker (2023a). Behavior was evaluated from 0800 to 2000 h, and performance across wk 4 and 6 was averaged for each calf. Training and reliability of observers are described in Downey and Tucker (2023a).

Heifer Behavioral Observations
Video was recorded from 0800 to 2000 h on d 2 of feed restriction (GV-BL4713 Pro IR Bullet IP Cameras and GV-POE2411-V2 NVR, Geovision Inc., 24 fps, H.264 codec recording) for the same animals, approximately 1 yr later (12.4 ± 0.6 mo of age). On d 1 of feed restriction, heifers would still be adjusting to the 50% reduction in food and would likely still have had rumen content from the previous ad libitum-fed day, so only d 2 was observed. Evaluating oral behavior over daylight hours is common in this age class (e.g., Špinka, 1992;Ishiwata et al., 2008b;Bourguet et al., 2011;Bruno et al., 2020). Videos were scored using behavior sampling for oral behaviors (Table 3, Supplemental Videos S1-S16, https: / / doi .org/ 10 .5281/ zenodo .6994585; Downey and Tucker, 2023b) with a continuous recording rule (Bateson and Martin, 2021). A single heifer was watched and scored at a time. Videos were analyzed using BORIS (Behavioral Observation Research Interactive Software; Friard and Gamba, 2016) by a total of 18 observers who were blind to treatment. Observers were first trained to identify tongue rolling and intersucking using 30-question video tests (15 yes, 15 no) to reliability ≥80% (Cohen's kappa, irr package version 0.84.1; Gamer et al., 2019). Observers then scored 21 videos (5-min duration/video) continuously, and reliability values were taken from these. All observers had to reach interobserver reliability ≥ 80% using interclass correlation coefficients (ICC, irr package version 0.84.1; Gamer et al., 2019) against a trained observer (author B.C.D., intraobserver ICC ≥ 90%) before independent video scoring. Reliability was also evaluated visually using the plot events feature in BORIS to compare observer time budgets for each video against the trained observer to identify potential mismatches and confirm all ICC values.

Statistical Analysis
Statistical analyses were performed using R version 4.0.3 (R Core Team) on MacOS Big Sur 10.16 via RStudio version 1.2.5033 (RStudio Team), with heifer as the experimental unit, as feed restriction was applied at the individual level. All model fits were checked for normality and homogeneity of variance using quantilequantile plots and plots of residuals versus fitted values (plot, boxplot, and resid functions in base R).

Heifer Life Stage
The proportion of time engaged in oral behavior on d occurring or not, or which of multiple behaviors could have been expressed. A beta regression (betareg package version 3.1-4; Cribari-Neto and Zeileis, 2010) was fitted with treatment as a fixed effect to analyze proportion of visible time of tongue rolling, tongue flicking, manipulating bedding (NNOM: Bedding), other oral manipulation (NNOM: Other), all oral manipulation (NNOM: Total; Bedding + Feed bin + Other), water drinking, self-grooming (Groom: Self), and allogrooming (Groom: Allo). No heifer data were otherwise excluded. Models were assessed with a type II ANOVA (car package version 3.0-10; Fox and Weisberg, 2019) to obtain P-values. Time spent intersucking and manipulating the feed bin (NNOM: Feed bin) was evaluated using an unpaired Wilcoxon signed-rank test (stats package in base R), as data did not meet the assumptions of normality. All oral behaviors are reported as percentages hereafter in the text to facilitate readability.
The CV was calculated for all behaviors that were fitted with a model, using the formula CV = (SD/mean) × 100. All behaviors were evaluated for outliers using interquartile range criteria. Outliers were considered those that fell more than 1.5 times below or above the first and third quartiles, respectively (boxplot function in base R). Outliers in this case were used to identify and describe extreme performances of oral behavior within the population, not as criteria for discarding data. We used a similar method to describe excessive grooming bouts in Downey and Tucker (2023a). Time spent drinking urine was rare and not evaluated with a model.

Heifer Versus Calf Life Stage
Consistency in oral behavior performance in individuals between the calf and heifer life stages was evaluated for water drinking, total NNOM, self-grooming, tongue flicks, and tongue rolling, as these behaviors were recorded in both environments. All calf data handling is described in Downey and Tucker (2023a); no data were excluded. The proportions of time engaged in drinking water, total NNOM, self-grooming, and tongue flicks as calves and heifers from 0800 to 2000 h were compared with a Spearman rank correlation and paired t-test (stats package in base R). Tongue rolling was compared across ages using a paired Wilcoxon signed-rank test (stats package in base R) as data were nonnormal and positively skewed. Drinking urine Mouth or tongue visibly touch urine stream from neighboring animal. If mouth or tongue are not visible, muzzle is oriented against the rear of the neighboring animal during urination (as indicated by urinating posture of arched back and wide stance or visual confirmation of a urine stream) or toward the urine stream and jaw is moving. Drinking water Mouth or tongue visibly touch or enter water; if not visible, any part of the muzzle is in water trough for at least 1 s. Groom: Allo

Downey and Tucker: EARLY HAY ACCESS DOES NOT AFFECT LATER LIFE BEHAVIOR
Touching hair on a neighboring animal with the tongue or mouth; includes if mouth is not visible but directed toward neighbor's body and the head moves in a vertical (up or down) motion. Groom: Self Touching hair with the tongue or mouth on heifer's own body; includes if mouth is not visible but directed toward body and the head moves in a vertical (up or down) motion. Intersucking Mouth or tongue is touching the teats or udder of another animal; if mouth is not visible, can be identified by the muzzle oriented between the hind legs of another animal or underneath the back third of other animal's abdomen. In this case, the muzzle should be angled up to access the udder. NNOM: Bedding Licking or chewing directed toward bedding: tongue must be touching bedding, or jaw moving as the lips touch bedding or while mouth is directed at the bedding, or jaw movements while bedding is visible inside the mouth, or jaw movements continuing within 5 s after any of the preceding actions. NNOM: Feed bin Licking, chewing, or sucking directed toward the feed bin. Includes if mouth is not visible but directed toward feed bin and the head moves in a vertical (up or down) motion. When feed bin is empty (whenever heifers are not separated), includes if muzzle is below the top edge of the feed bin for at least 1 s. NNOM: Other Licking, chewing, or sucking directed toward any nonnutritive item except bedding or feed bin (includes bars, water trough, hose): tongue or lips must be touching a nonnutritive item, or such item must be held inside the mouth. Includes if mouth is not visible but directed toward nonnutritive items and the head or jaw move in a vertical (up or down) motion.

Out of view
The heifer's muzzle or body (or both) are not visible, preventing classification of other behaviors.

Tongue flicks
Tongue extends out of the mouth without touching other objects or forming a full or partial circular motion, or it extends up to the nose before retracting back into the mouth and repeating at least once more within 1 s. Do not count during eating and ruminating. 3 Tongue rolling Tongue is held in a full or partial circular position or moves in a full or partial circular motion; this can occur when the tongue is held within the border of the lips inside the mouth or extended outside the border of the lips. This cannot occur while any other behaviors are being performed (the tongue is not touching any feed/nonnutritive items) and does not need to repeat.  (Downey and Tucker, 2023b).

All
All heifers drank water, groomed (self and allo), tongue-rolled, and performed tongue flicks and NNOM (bedding, feed bin, other, and total) throughout the daylight period (Figure 1), but there was no evidence of an effect of early rearing treatment [P ≥ 0.132; Figure  2, Figure 3; Supplemental Table S1, https: / / doi .org/ 10 .5281/ zenodo .6994585; Downey and Tucker, 2023b]. Intersucking was performed by 16 of the 24 heifers (67%), representing 75% of Hay heifers and 50% of Control heifers, but there was no effect of treatment Downey and Tucker: EARLY HAY ACCESS DOES NOT AFFECT LATER LIFE BEHAVIOR Figure 1. Circadian pattern of behavior in year-old heifers experiencing short-term feed restriction, obtained using behavior sampling and continuous recording across 12 h (0800-2000 h). Heifers were previously reared on a diet of grain and milk replacer from birth (Control) or with additional mountain grass hay (Hay) from birth to 50 d of age. Data were averaged by treatment at 15-min intervals. Gray ribbons indicate when heifers were separated during feeding (for 1.5 h after feed delivery at 0800 h and 1.2 h after feed delivery at 1600 h); heifers were separated until all heifers finished their meals. Note that the y-axis scale differs by row. NNOM = nonnutritive oral manipulation.
All heifer behaviors were characterized by variability. Total NNOM had the lowest CV and was least variable Downey and Tucker: EARLY HAY ACCESS DOES NOT AFFECT LATER LIFE BEHAVIOR Figure 2. Mean percentage of time engaged in nonnutritive oral behaviors across 12 h in year-old heifers experiencing short-term feed restriction. Heifers were previously reared on a diet of grain and milk replacer from birth (Control) or with additional mountain grass hay (Hay) from birth to 50 d of age. Data were collected using behavior sampling and continuous recording on d 2 of feed restriction (50% of ad libitum intake). Boxplots represent the median (black line within box) and first and third quartiles (25% and 75% of data). Whiskers extend to the lowest and highest values that are not outliers (values that are 1.5 times the interquartile range); outliers (o) and means (x) are also presented. NNOM = nonnutritive oral manipulation. 1 Percentages of behavior were compared between individuals at the calf and heifer stages. Behavioral observations were conducted over 12 h (0800-2000 h) using 1-0 sampling during 5-s intervals as calves (calculated from Downey and Tucker, 2023a) or continuously as heifers. Relatedness of calf and heifer behavior was evaluated using a Spearman rank correlation and paired t-test or paired Wilcoxon signed-rank test for drinking water, total NNOM (nonnutritive oral manipulation), self-grooming, tongue flicks, and tongue rolling. 2 Percentage of 12-h observations ± SE. 3 P-value generated by a paired Wilcoxon signed-rank t-test.

DISCUSSION
We found no evidence that early life access to hay influenced performance of abnormal or possibly abnormal oral behaviors during a 2-d feed restriction challenge in dairy heifers compared with those reared without early hay access, in contrast to our predictions. In ret-rospect, restriction may have compromised or masked any long-term behavioral changes. The "environmental saturation hypothesis" suggests that effects of different early environments are only visible in later life if the later environment is intermediate (Engqvist and Reinhold, 2016). Extremely positive, resource-rich environments would benefit all individuals, while extremely restrictive, barren environments, like those induced by short-term limit feeding, would challenge all. Evaluating long-term effects of early hay access under an environment that caused hunger in addition to limiting expression of solid-food-processing behaviors may have increased abnormal oral behavior performance for all heifers and prevented us from seeing treatment differences. However, the lack of evidence of a difference is likely reflective of what would be seen on the farm, as heifers experience feed restriction equivalent to 1.5% to 2% of BW, though for longer periods (e.g., between 4 and 22 mo of age; Zanton and Heinrichs, 2008).
This lack of evidence of a long-term behavioral change could also suggest that early life hay provision, while important, may not outweigh the other, severe restrictions calves often face. Calves fed hay experience benefits including reduced ARBs and increased DMI and ADG (e.g., Downey et al., 2022), possibly because it addresses a motivation to chew and ruminate. The Heifers were previously reared on a diet of grain and milk replacer from birth (Control) or with additional mountain grass hay (Hay) from birth to 50 d of age. Data were collected using behavior sampling and continuous recording on d 2 of feed restriction (50% of ad libitum intake). Boxplots represent the median (black line within box) and first and third quartiles (25% and 75% of data). Whiskers extend to the lowest and highest values that are not outliers (values that are 1.5 times the interquartile range); outliers (o) and means (x) are also presented. evidence clearly suggests access to hay matters in the immediate environment of the milk-fed calf. However, these calves were reared in restrictive conditions due to early dam separation, social isolation, low milk provision, and limited opportunities to suck from a bottle, all of which differ widely from natural settings and may increase ARB performance (e.g., Veissier et al., 1997Veissier et al., , 2013Haley et al., 1998). Early separation and isolation, in particular, lead to well-documented negative long-term consequences in many species (e.g., Griffin and Harlow, 1966;Gluck et al., 1973;Maccari et al., 2014) and can lead to cognitive challenges in calves (Daros et al., 2014;Gaillard et al., 2014). Hay was likely unable to counteract all of these problems.
Individuals showed low consistency in most oral behaviors (water drinking, NNOM: Total, self-grooming, and tongue rolling) between the calf and heifer stages. Tongue flicks were the sole exception and tended to be positively related across age classes. This could indicate that across contexts this behavior is more stable than others, even seemingly abnormal behaviors that involve similar tongue motions (tongue rolling, NNOM). Individual consistency in behavior varies across species and contexts. Some abnormal behaviors are correlated in individuals over time (e.g., locomotor stereotypic behavior in mink, Dallaire et al., 2012;cross-sucking in calves, de Passillé et al., 2011), while others are highly variable (e.g., coprophagy in primates, Kalcher-Sommersguter et al., 2013). Even behaviors with similar motor patterns vary: individual broilers show consistency in pecking during foraging but not in feather pecking (preening), pecking the environment, or social pecking (Ferreira et al., 2022). Behavior can become more consistent as individuals age (e.g., reviewed in humans by Roberts and DelVecchio, 2000) and may thus not be related between developing and mature animals (e.g., dairy cattle; Neave et al., 2020). To the authors' knowledge, this is the first study evaluating long-term consistency in dairy cattle abnormal behaviors in the same individuals across the calf and heifer stages. Future work should evaluate the ontogeny of abnormal behaviors, including how consistent individual differences are, using more frequent measurements through the developmental process.
Heifers performed 2 types of abnormal conspecificdirected sucking behavior: intersucking and urine drinking. Intersucking was performed by 67% of heifers, which is higher than others have reported (e.g., 1%-57% of heifers; Špinka, 1992;Keil and Langhans, 2001). Intersucking likely either developed at some point after 65 to 70 d of age (5-10 d after the end of weaning), when calves began to be group-housed, or in response to current short-term feed restriction. In a larger population that included these heifers, we found evidence that both ideas may be true, as intersucking was evident in some individuals before limit feeding, but most performed or showed more intersucking when feed was restricted (H. B. Goeller, B. C. Downey, and C. B. Tucker, University of California, Davis, personal communication). Urine drinking was also observed, although it was rare. This behavior is not often scored in dairy cattle but is reported to be rare in calves (de Wilt, 1985;Lidfors, 1993) and may reflect an abnormal or injurious behavior related to suckling opportunity (Wiepkema, 1987).
Tongue movements were common in heifers. Tongue rolling was performed by 85% of calves before weaning, but 100% of heifers tongue-rolled and for longer durations (0.04% vs. 0.19% of daylight hours, Table  4). Abnormal repetitive behaviors are hypothesized to increase at or after weaning in cattle (e.g., Seo et al., 1998;Keil and Langhans, 2001), and our results are supportive of this idea. This is similar to ARB ontogeny in other species, with both horses and chicks increasing performance of crib biting and abnormal feather pecking, respectively, between 3 and 9 mo of age (Roden and Wechsler, 1998;Perré et al., 2002;Waters et al., 2002). The duration of tongue rolling is similar to other findings in steers and heifers (e.g., 0.1%-1%; Ishiwata et al., 2008b;Iraira et al., 2013;Madruga et al., 2017;Park et al., 2020). The finding that all heifers tonguerolled is novel, as it has previously been reported to occur in 31% to 95% of individuals in beef breeds (Sato et al., 1994;Levina et al., 2020;Schneider et al., 2020). All heifers performed tongue flicks approximately 1% of the time, which was significantly lower than their performance as calves. Tongue flicks could be precursors to tongue rolling, explaining their decrease as tongue rolling became more prevalent in heifers, but could themselves also be an abnormal response to restriction.
Nonnutritive oral manipulation was common and directed toward multiple substrates. Heifers spent an average of 6.5% of the daylight period manipulating pen fixtures (NNOM: Feed bin + NNOM: Other), though some did this for up to 16%. This is greater than what others have reported in heifers (1.3%-3.5% of time; Robles et al., 2007;Ishiwata et al., 2008b;Devant et al., 2015). This discrepancy could reflect sampling differences. Instantaneous recording rules, while commonly used, are noted to be inaccurate above 5-min intervals in heifers for NNOM (Madruga et al., 2017) and at all intervals in calves during short observation windows (e.g., 5 s to 1 min intervals; Downey et al., 2021). Indeed, NNOM is also sometimes too rare to report in heifers when subsampling is used (Nielsen et al., 1997;Kolling et al., 2016). Heifers also expressed NNOM toward bedding for 5.8% of daylight hours, though some individuals did this for up to 14%. Bedding consump-Downey and Tucker: EARLY HAY ACCESS DOES NOT AFFECT LATER LIFE BEHAVIOR tion is reported to be rare in heifers (e.g., 0.3%-0.7% of a 24-h day; Iraira et al., 2012Iraira et al., , 2013Madruga et al., 2017). As our heifers were feed-restricted, rice hull consumption was likely a response to hunger or missing nutrients. The behavior itself, while appearing abnormal, could be an adaptive response to a poor environment, explaining the high expression relative to other studies. However, bedding consumption can also reflect pica (Samaha et al., 1990) and is found in calves kept on sand (e.g., Salter et al., 2021;Downey et al., 2022;Downey and Tucker, 2023a). Apparent willingness to consume a nonfibrous bedding source or bedding that may be soiled may indicate that this behavior can still be abnormal.
Grooming behaviors are often considered to reflect normal maintenance behaviors for health or social cohesion but may also be abnormal. Feed-restricted heifers spent 2% of daytime observations self-grooming and 3.3% allogrooming, like heifers managed in intensive pen or tie-stall conditions (self: 1.6%-3.9%; Ishiwata et al., 2008b;Faleiro et al., 2011;Madruga et al., 2017); allo: 1.2%-4.8%; Robles et al., 2007;Ishiwata et al., 2008b;Bourguet et al., 2011). However, grooming is lower in heifers on pasture (self: 0.9%-1.6%; allo: 0.2%-0.6%; Ishiwata et al., 2008b), suggesting higher values could reflect a displacement behavior or frustrated feeding motions. In fact, self-grooming also occurs more frequently before and after bouts of ARBs than expected (Redbo, 1990), while tongue-rolling steers groom more frequently than nonrollers (Ishiwata et al., 2008a). In calves, self-grooming has been performed in ways that appear abnormal, with individuals performing many bouts throughout a day, some for up to 30 min (Downey and Tucker, 2023a). Self-grooming is also reduced when hay is provided (Downey et al., 2022), further suggesting that high levels may relate to thwarted feeding. Regardless of feeding treatment, calves self-groomed more than heifers. This could be because heifers were able to allogroom as well and may have redirected some of the underlying motivation behind grooming behavior to licking conspecifics instead.
Another normal behavior that can be performed abnormally is water drinking. Heifers spent approximately 1.6% of their time drinking water, much as they do other confined settings (e.g., 1.2%-2.5%; Robles et al., 2007;González et al., 2008;Madruga et al., 2017). Time spent drinking may reflect heat abatement efforts or thirst (as reviewed by Jensen and Vestergaard, 2021). Heifers spent more of a 12-h day drinking compared with calves, which could reflect increased metabolic demand in animals that do not have access to milk and consume larger quantities of solid feed (e.g., reviewed by Jensen and Vestergaard, 2021). However, manipulating water could also be abnormal, used as a source of novel stimulation in a barren environment (e.g., "water lapping" is reported in tie-stalls; Mattiello et al., 2005;Corazzin et al., 2010). In that vein, polydipsia has been reported in calves, with individuals in intensive settings drinking up to 20 L/d (Downey et al., 2022), and 44% of our heifers had been considered outliers on at least 1 d as calves (drinking more than 7.35 L/d on any day from 0 to 50 d of age; Downey and Tucker, 2023a). We did not measure water intake in the current study and note that polydipsic animals may drink quickly, preventing their identification through drinking time alone (Downey and Tucker, 2023a).
Individuals showed variation in how oral behaviors were performed. There were outliers, or extreme levels of performance relative to this population, in most behaviors, and many of these were performed by unique heifers. Individual variation in response to short-term feed restriction was further emphasized by the CV. Intersucking and tongue rolling were most variable across individuals, but all CV in this study were more variable than "normal" behaviors like eating and ruminating in cows (e.g., 16%-18%; Dado and Allen, 1994) or sucking milk from a bottle in calves (e.g., 26%; Downey et al., 2021), similar to patterns in broiler breeders (Kostal et al., 1992). Taken together, CV and outliers serve as indicators that some behaviors may be expressed in extreme ways, which could indicate abnormal underpinnings to some behaviors. These metrics are limited as they are dependent on the spread of data within this population, all of whom are experiencing similar restriction and thus may be abnormal. To determine if outliers can serve as indicators of abnormality, these calculations should either be performed on a larger population that includes animals in more natural settings or tied to additional behavioral or physiological indicators of compromised welfare.

CONCLUSIONS
We found no evidence that providing hay to individually housed, milk-limited calves from 0 to 7 wk of life affects oral behavior performance 1 yr later in dairy heifers experiencing 50% feed restriction for 2 d. All heifers, regardless of early rearing environment, performed many possibly abnormal behaviors, including intersucking, tongue rolling, tongue flicking, NNOM, and urine drinking. Oral behaviors were highly variable across heifers and many were performed at extreme levels. Most behaviors were not consistent between the calf and heifer stages and were performed more in older animals, but some behaviors were performed more by younger animals. The considerable variability, inconsistency across ages, and excessive performance of some behaviors raises additional questions about how these