Advertisement

Effects of concentrate allowance and individual dairy cow personality traits on behavior and production of dairy cows milked in a free-traffic automated milking system

Open AccessPublished:May 06, 2022DOI:https://doi.org/10.3168/jds.2021-21657

      ABSTRACT

      The primary objective of this study was to determine whether the level of concentrate allowance in an automated milking system (AMS) affects the feed intake, eating behavior, milking activity, and performance of lactating dairy cows. The secondary objective of this study was to describe how the response to concentrate allocation, specifically in feeding and milking behavior, varies with cow personality traits. Fifteen Holstein cows were used in a crossover design with two 28-d periods, each including 14 d of adaptation and 14 d of data collection. The cows were housed in a freestall pen with free-traffic access to the AMS. Treatments consisted of a basal partial mixed ration (PMR) common to both treatment groups, with a concentrate allowance (on dry matter basis) of (1) 3.0 kg/d in the AMS (L-AMS) or (2) 6.0 kg/d in the AMS (H-AMS). Between the 2 treatment periods, each cow was assessed for personality traits using a combined arena test consisting of exposure to a novel environment, novel object, and novel human. Principal component analysis of behaviors observed during the novel environment and object tests revealed 3 factors (interpreted as active, social, and alert-curious) that together explained 76% of the variance, whereas principal component analysis of the novel human test revealed 2 factors (interpreted as active-vocal and fearful of novel humans) that together explained 77% of the variance. When on the H-AMS treatment, PMR dry matter intake (DMI) was less (24.5 vs. 26.0 kg/d) and AMS concentrate delivery was greater (5.9 vs. 3.1 kg/d), as per design. Consequently, total DMI was greater on the H-AMS treatment (30.4 vs. 29.1 kg/d). When on the H-AMS treatment, cows who were more alert-curious consumed more PMR, whereas cows who were more fearful of the novel human were less likely to receive the maximum amount of AMS concentrate available, limiting their total DMI and increasing the day-to-day variability of that intake. Although this was a preliminary study, these data suggest an association between dairy cow personality traits and how cows respond to increased AMS concentrate allowance.

      Key words

      INTRODUCTION

      The use of an automated milking system (AMS) provides many benefits to dairy producers, such as greater time flexibility (
      • de Koning C.J.A.M.
      Milking machines: Robotic milking.
      ), reduced labor requirements (
      • Hansen B.G.
      Robotic milking-farmer experiences and adoption rate in Jæren, Norway.
      ), and greater capacity for data collection (
      • King M.T.M.
      • DeVries T.J.
      Graduate Student Literature Review: Detecting health disorders using data from automatic milking systems and associated technologies.
      ). They also offer greater opportunities to manage cows on an individual basis, rather than at the pen or herd level, particularly in the area of precision feeding applied at the milking unit (
      • Bach A.
      • Cabrera V.
      Robotic milking: Feeding strategies and economic returns.
      ). Feeding management for cows milked using an AMS differs from that of conventional parlor-milked cows, as the former are typically provided a partial mixed ration (PMR) at the feed bunk and a concentrated supplement delivered by the AMS into a feeder during milking. Researchers have previously demonstrated that provision of the concentrate in the AMS is a motivating factor, encouraging cows to voluntarily enter the AMS to be milked (
      • Prescott N.B.
      • Mottram T.T.
      • Webster A.J.F.
      Relative motivations of dairy cows to be milked or fed in a Y-maze and an automatic milking system.
      ;
      • Melin M.
      • Svennersten-Sjaunja K.
      • Wiktorsson H.
      Feeding patterns and performance of cows in controlled cow traffic in automated milking systems.
      ;
      • Bava L.
      • Tamburini A.
      • Penati C.
      • Riva E.
      • Mattachini G.
      • Provolo G.
      • Sandrucci A.
      Effects of feeding frequency and environmental conditions on dry matter intake, milk yield and behaviour of dairy cows milked in conventional or automatic milking systems.
      ). As a result, the use of concentrate to attract cows to the AMS, coupled with the ability to provide differing quantities of concentrate for individual cows, has resulted in the ability to adjust the quantity of AMS concentrate not only to potentially minimize fetching of cows that are reluctant to voluntarily milk, but also to allow for cow-level feeding based on considerations of individual cow nutrient requirements for health and production (
      • Bach A.
      • Cabrera V.
      Robotic milking: Feeding strategies and economic returns.
      ;
      • Rodenburg J.
      Robotic milking: Technology, farm design, and effects on work flow.
      ). However, providing too much concentrate in the AMS may pose problems. Cows consistently consume less than the maximum amount of concentrate available in the AMS, particularly at higher allocations (
      • Halachmi I.
      • Ofir S.
      • Miron J.
      Comparing two concentrate allowances in an automatic milking system.
      ;
      • Bach A.
      • Iglesias C.
      • Calsamiglia S.
      • Devant M.
      Effect of amount of concentrate offered in automatic milking systems on milking frequency, feeding behaviour, and milk production of dairy cattle consuming high amounts of corn silage.
      ;
      • Bach A.
      • Cabrera V.
      Robotic milking: Feeding strategies and economic returns.
      ). Further, increasing the concentrate allowance to reach the desired intake often results in a PMR substitution effect, whereby cows eat less PMR with greater concentrate allocation at the AMS. This substitution effect has been observed in both feed-first, guided traffic systems (
      • Hare K.
      • DeVries T.J.
      • Schwartkopf-Genswein K.S.
      • Penner G.B.
      Does the location of concentrate provision affect voluntary visits, and milk component yield for cows in an automated milking system?.
      ;
      • Menajovsky S.B.
      • Walpole C.E.
      • DeVries T.J.
      • Schwartzkopf-Genswein K.S.
      • Walpole M.E.
      • Penner G.B.
      The effect of the forage-to-concentrate ratio of the partial mixed ration and the quantity of concentrate in an automatic milking system for lactating Holstein cows.
      ;
      • Paddick K.S.
      • DeVries T.J.
      • Schwartzkopf-Genswein K.
      • Steele M.A.
      • Walpole M.E.
      • Penner G.B.
      Effect of the amount of concentrate offered in an automated milking system on dry matter intake, milk yield, milk composition, ruminal digestion, and behavior of primiparous Holstein cows fed isocaloric diets.
      ), as well as free-traffic systems (
      • Bach A.
      • Iglesias C.
      • Calsamiglia S.
      • Devant M.
      Effect of amount of concentrate offered in automatic milking systems on milking frequency, feeding behaviour, and milk production of dairy cattle consuming high amounts of corn silage.
      ;
      • Henriksen J.C.S.
      • Weisbjerg M.R.
      • Løvendahl P.
      • Kristensen T.
      • Munksgaard L.
      Effects of an individual cow concentrate strategy on production and behavior.
      ;
      • Schwanke A.J.
      • Dancy K.M.
      • Didry T.
      • Penner G.B.
      • DeVries T.J.
      Effects of concentrate location on the behavior and production of dairy cows milked in a free-traffic automated milking system.
      ), and may account for results that fail to show an association between greater concentrate provision in the AMS and improved production outcomes (
      • Migliorati L.
      • Speroni M.
      • Lolli S.
      • Calza F.
      Effect of concentrate feeding on milking frequency and milk yield in an automatic milking system.
      ;
      • Bach A.
      • Iglesias C.
      • Calsamiglia S.
      • Devant M.
      Effect of amount of concentrate offered in automatic milking systems on milking frequency, feeding behaviour, and milk production of dairy cattle consuming high amounts of corn silage.
      ;
      • Henriksen J.C.S.
      • Weisbjerg M.R.
      • Løvendahl P.
      • Kristensen T.
      • Munksgaard L.
      Effects of an individual cow concentrate strategy on production and behavior.
      ). Increasing the amount of AMS concentrate available also increases the day-to-day variation in concentrate delivery (
      • Menajovsky S.B.
      • Walpole C.E.
      • DeVries T.J.
      • Schwartzkopf-Genswein K.S.
      • Walpole M.E.
      • Penner G.B.
      The effect of the forage-to-concentrate ratio of the partial mixed ration and the quantity of concentrate in an automatic milking system for lactating Holstein cows.
      ;
      • Paddick K.S.
      • DeVries T.J.
      • Schwartzkopf-Genswein K.
      • Steele M.A.
      • Walpole M.E.
      • Penner G.B.
      Effect of the amount of concentrate offered in an automated milking system on dry matter intake, milk yield, milk composition, ruminal digestion, and behavior of primiparous Holstein cows fed isocaloric diets.
      ;
      • Schwanke A.J.
      • Dancy K.M.
      • Didry T.
      • Penner G.B.
      • DeVries T.J.
      Effects of concentrate location on the behavior and production of dairy cows milked in a free-traffic automated milking system.
      ).
      A further challenge with precision feeding of dairy cows is that, in addition to their individual metabolic and physiological needs, cows also have individual personality traits that affect their behaviors and thus influence how they meet these needs (
      • Han C.S.
      • Dingemanse N.J.
      Effect of diet on the structure of animal personality.
      ). Personality traits are defined as correlated sets of behavioral factors consistent across time and contexts that are specific to an individual (
      • Koolhaas J.M.
      • Korte S.M.
      • De Boer S.F.
      • Van Der Vegt B.J.
      • Van Reenen C.G.
      • Hopster H.
      • De Jong I.C.
      • Ruis M.A.W.
      • Blokhuis H.J.
      Coping styles in animals: Current status in behaviour and stress-physiology.
      ;
      • Réale D.
      • Reader S.M.
      • Sol D.
      • McDougall P.T.
      • Dingemanse N.J.
      Integrating animal temperament within ecology and evolution.
      ). These traits generally fall under 1 of 5 categories: activity, exploration, aggressiveness, boldness, and sociability (
      • Finkemeier M.-A.
      • Langbein J.
      • Puppe B.
      Personality research in mammalian farm animals: Concepts, measures, and relationship to welfare.
      ). These general traits have been consistently demonstrated across species, although the exact labeling of a trait may be more nuanced to reflect particular species or situations. Traits may be innately present from birth (
      • McCrae R.R.
      • Costa Jr., P.T.
      • Ostendorf F.
      • Angleitner A.
      • Hrebíčkovà M.
      • Avia M.D.
      • Sanz J.
      • Sánchez-Bernardos M.L.
      • Kusdil M.E.
      • Woodfield R.
      • Saunders P.R.
      • Smith P.B.
      Nature over nurture: Temperament, personality and life span development.
      ) but are often altered from this initial state through the cumulation of life experience over time as the animal ages (
      • Finkemeier M.-A.
      • Langbein J.
      • Puppe B.
      Personality research in mammalian farm animals: Concepts, measures, and relationship to welfare.
      ). In either case, personality traits are underlying components influencing many aspects of an animal's behavior, including eating behavior and diet selection (
      • Meagher R.K.
      • Weary D.M.
      • von Keyserlingk M.A.G.
      Some like it varied: Individual differences in preference for feed variety in dairy heifers.
      ;
      • Neave H.W.
      • Costa J.H.C.
      • Weary D.M.
      • von Keyserlingk M.A.G.
      Personality is associated with feeding behavior and performance in dairy calves.
      ). As a result, nutrient consumption by an individual cow may be affected by personality differences, or, as suggested by
      • Han C.S.
      • Dingemanse N.J.
      Effect of diet on the structure of animal personality.
      , nutritional balance may even affect expression of personalities. This poses a potential challenge in managing cows milked in an AMS, as individual variability in willingness to visit the AMS may affect not only milking frequency but also the variability in the amount of AMS concentrate provided and consumed. This could lead to differences in the composition and quantity of the diet actually consumed as compared with the intended formulated diet. Consequently, individual differences in diet consumption could result in variable milk production. Indeed, researchers have previously detected associations between personality and milk production in dairy cattle, where more reactive cows produced less milk (
      • Hedlund L.
      • Løvlie H.
      Personality and production: Nervous cows produce less milk.
      ;
      • Marçal-Pedroza M.G.
      • Campos M.M.
      • Pereira L.G.R.
      • Machado F.S.
      • Tomich T.R.
      • Paranhos da Costa M.J.R.
      • SanťAnna A.C.
      Consistency of temperament traits and their relationships with milk yield in lactating primiparous F1 Holstein-Gyr cows.
      ). It is unclear how individual dairy cow personality traits influence feed intake and day-to-day variation in the diet consumed, as well as AMS visits and milking frequency when provided differing amounts of AMS concentrate. Knowledge of these behavioral factors could help better allow for targeted AMS concentrate allowance, thus enabling more accurate and precise feeding strategies.
      Thus, the primary objective of this study was to determine how feed intake, eating behavior, milking activity, and performance of dairy cows are affected by the amount of concentrate provided in a free-traffic AMS. The secondary objective of this study was to describe how the response to concentrate allocation, specifically in feeding and milking behavior, is influenced by cow personality traits. It was predicted that cows would have lesser PMR intake, greater milking frequency, and greater milk yield with greater concentrate provided at the AMS. It was also predicted that cows who are more fearful would be expected to visit the AMS less frequently, and thus may not reach the target concentrate allocation when provided a greater amount of concentrate in the AMS. To our knowledge, this is the first study to conduct a preliminary investigation into how dairy cow personality traits may be associated with their response to AMS concentrate allocation.

      MATERIALS AND METHODS

      Animals and Housing

      Fifteen lactating Holstein cows (parity = 2.8 ± 0.9; mean ± SD), including 1 primiparous and 14 multiparous (range = 2–4 lactations), were enrolled in the study and housed in a free-traffic AMS pen located at the University of Guelph, Elora Research Station–Ontario Dairy Research Centre (Elora, Ontario, Canada). For inclusion in the study, cows had to be between 30 and 200 DIM and producing similar quantities of milk, free from any health concerns, and not known to require chronic fetching to the AMS. At the time of treatment assignment cows were 123.9 ± 53.2 DIM and had a BW of 744.1 ± 75.1 kg and a BCS of 3.0 ± 0.3. All cows were previously milked twice daily on a rotary milking parlor (DeLaval) and were producing 44.5 ± 7.7 kg/d of milk. All cows also had previous experience with being milked in an AMS. Before the start of the first data collection period, each cow was assigned an individual automated feed bin (Insentec BV) for the duration of the study and trained to access her own bin for a minimum of 7 d, as well as trained for a minimum of 14 d to access the AMS (DeLaval VMS, DeLaval International). Cows were trained to eat from their own feed bins by guiding each cow to her bin with feed on the day that bins were assigned and observing cows during the 30 min after feed delivery over the next 6 d to discourage them from eating out of other cows' bins. Cows observed eating out of other cows' bins were redirected to their own bins. After cows were trained to the automated feed bins and before the start of the study, cows were eating 25.8 ± 3.7 kg/d DM of PMR out of individual automated feed bins and were allocated 4.5 ± 1.0 kg/d DM of concentrate in the AMS. The use of cows and all experimental procedures were in compliance with the guidelines of the
      • Canadian Council on Animal Care
      Guidelines on the Care and Use of Farm Animals in Research, Teaching and Testing.
      and were approved by the University of Guelph Animal Care Committee (Animal Use Protocol no. 4109).
      Within the AMS pen, cows had access to 1 of 15 automated feed bins, 30 lying stalls, and 1 cow brush, and were offered ad libitum access to water from 2 water troughs at each end of the pen. The number of cows in the AMS pen included only the 15 cows enrolled in the study. The stalls were laid out tail to tail in 2 rows of 15. Each stall measured 295 cm in total length and 127 cm in width, with the neck rail positioned 188 cm from the rear curb and 125 cm above the stall base. The stalls had a mattress base (Pasture Mat, ProMat) and were bedded with chopped straw. New bedding was added once weekly, and stalls were cleaned and groomed twice a day.
      Milking permission in the AMS was granted when 4 h had elapsed since the previous milking and predicted milk yield for that milking exceeded 9.0 kg. This milking permission was chosen to avoid restricting milkings and, thus, to fully elucidate the effect that treatment may have had on milking activity, with the goal of gaining insight as to how often cows are willing to visit the AMS, given the opportunity, based on concentrate provision. The cows had access to the AMS for 22.5 h/d, with the remaining time dedicated to 30-min cleaning cycles at 0730, 1500, and 2330 h daily. To ensure a minimum twice-daily milking frequency, cows that had not voluntarily entered the AMS for more than 10 h since the previous milking were fetched to be milked at either morning (between 0400 and 0530 h) or evening (between 1600 and 1730 h) fetching times; all fetching activity (including cow number, fetching time, and inadvertently fetched cows) was recorded by barn staff on provided data sheets.

      Study Design and Dietary Treatments

      Sample size and power analyses were used to calculate (as per
      • Morris T.R.
      Experimental Design and Analysis in Animal Sciences.
      ) the minimum number of replicates needed per treatment (n = 15 per treatment) to detect a 7.5% level of observed mean difference for the principal outcome variables, including milking frequency, DMI, and milk production, associated with our primary objective. Estimates of variation (average CV = 10.3%) for these variables were based on previously reported values (
      • Schwanke A.J.
      • Dancy K.M.
      • Didry T.
      • Penner G.B.
      • DeVries T.J.
      Effects of concentrate location on the behavior and production of dairy cows milked in a free-traffic automated milking system.
      ).
      The study was implemented as a crossover design, consisting of two dietary treatments and two 28-d treatment periods. The first 14 d (d 1–14) of each treatment period served as an adaptation period, followed by a 14-d period of data collection (d 15–28). Based on
      • Grant R.J.
      • Dann H.M.
      • Woolpert M.E.
      Time required for adaptation of behavior, feed intake, and dietary digestibility in cattle.
      , the 14-d adaptation period was concluded to be sufficient to both allow adjustment to the dietary treatments, as well as, in the second treatment period, to allow for any behavioral and intake effects from the previous treatment to be washed out. Although longer washout periods decrease the risk of carry-over effects, they also increase the risks of period effects due to increasing DIM. Thus, the 14-d adaptation periods were designed with this balance in mind. The study took place during the months of March and April 2019. The dietary treatments (Table 1) consisted of a PMR common to both treatments and (1) a high concentrate allowance in the AMS (H-AMS) targeting 6.0 kg/d (on a DM basis) concentrate (pellet) in the AMS (78.8% of the total diet as a PMR and 21.2% of the total diet as supplemental concentrate in the AMS), and (2) a low concentrate allowance in the AMS (L-AMS) targeting 3.0 kg/d (on a DM basis) pellet in the AMS (88.2% of the total diet as a PMR and 11.8% of the total diet as supplemental concentrate in the AMS). A common PMR was chosen, as this is common practice on commercial farms, with cows receiving supplemental concentrate in the AMS based on individual lactation stage and production. The treatment diets (Table 1), including AMS pellet supplementation, were formulated to meet the nutritional requirements for a 650-kg cow with an expected milk yield of 39 kg/d for L-AMS (at 25.3 kg/d predicted total DMI) and 45.0 kg/d for the H-AMS (at 28.3 kg/d predicted total DMI), with 4.0% milk fat and 3.2% true protein (Cattle Professional, Agricultural Modeling & Training Systems, version 4.10).
      Table 1Ingredients and chemical composition (mean ± SD) of the formulated partial mixed ration (PMR), automated milking system (AMS) pellet, and treatment diets
      CompositionPMR
      PMR had a particle size distribution (% of DM) of 9.2 ± 5.0% long (>19 mm), 47.3 ± 2.6% medium (<19 mm, >8 mm), 18.0 ± 1.9% short (<8 mm, >4 mm), and 25.4 ± 4.7% fine (<4 mm) particles.
      AMS pelletTreatment diet
      H-AMS = PMR coupled with a high AMS pellet allowance (6.0 kg/d); L-AMS = PMR coupled with a low AMS pellet allowance (3.0 kg/d).
      H-AMSL-AMS
      Ingredient, % of DM
       Corn silage
      DM of corn silage was 40.2 ± 2.9%, and chemical composition (DM basis) was 10.3 ± 0.9% CP, 18.9 ± 0.6% ADF, and 33.5 ± 0.4% NDF.
      34.927.530.8
       Wheat straw
      DM of straw was 90.6 ± 2.5%, and chemical composition (DM basis) was 7.9 ± 0.4% CP, 57.6 ± 3.0% ADF, and 87.2 ± 2.8% NDF.
      2.01.61.8
       Alfalfa haylage
      DM of alfalfa haylage was 34.5 ± 7.5%, and chemical composition (DM basis) was 21.6 ± 0.6% CP, 35.5 ± 0.5% ADF, and 42.4 ± 0.5% NDF.
      35.027.530.8
       High-moisture corn
      DM of high-moisture corn was 73.3 ± 1.3%, and chemical composition (DM basis) was 9.7 ± 0.3% CP, 2.5 ± 0.04% ADF, and 7.9 ± 0.6% NDF.
      14.611.512.8
       PMR supplement
      DM of the PMR supplement was 90.0 ± 1.1%, and chemical composition (DM basis) was 37.5 ± 0.1% CP, 8.4 ± 0.4% ADF, and 17.9 ± 0.4% NDF. Supplied by Floradale Feed Mill Ltd., including ingredients (as fed) 50.0% Soy Plus protein supplement (Landus Cooperative), 20.0% soybean meal 47%, 9.52% wheat shorts, 5.00% canola, 3.7% sodium sesquicarbonate, 3.2% fine salt, 1.65% Diamond V Yeast XP (Diamond V), 1.1% magnesium oxide, 1.0% tallow, 0.95% inorganic micronutrient premix (Floradale Feed Mill Ltd.), 0.88% Integral hydrolyzed dried yeast (Alltech), 0.80% monocalcium phosphate, 0.80% limestone calcium carbonate, 0.80% Metasmart rumen-protected methionine (Adisseo), 0.4% DCAD Plus potassium carbonate (Arm & Hammer), 0.16% sulfur 99.5%, and 0.04% Rumensin monensin (Elanco Animal Health).
      13.510.712.0
       Concentrate pellet in the AMS
      Supplied by Floradale Feed Mill Ltd., including ingredients (as fed) 24.2% wheat shorts, 19.5% soybean meal, 12.5% soy hulls (ground), 10.0% wheat chop (ground wheat), 10.0% beet pulp, 10.0% barley chop (ground barley), 5.0% bakery meal, 5.0% corn chop (ground corn), 2.0% molasses (pelleter), 0.8% limestone calcium carbonate, 0.4% Pelltech pellet binder and die lubricant (Borregaard LignoTech), 0.4% fine salt, 0.2% magnesium oxide, and 0.1% ruminant micronutrient premix (Floradale Feed Mill Ltd.).
      100.021.211.8
      Chemical composition
      Values were obtained from chemical analysis of PMR and concentrate samples. NEL was calculated based on NRC (2001) equations.
       DM, %44.1 ± 2.289.7 ± 1.853.8 ± 2.149.5 ± 2.1
       OM, % of DM91.2 ± 4.392.9 ± 0.591.5 ± 3.591.4 ± 4.3
       CP, % of DM17.4 ± 0.421.0 ± 0.818.2 ± 0.517.9 ± 0.5
       ADF, % of DM21.4 ± 1.115.2 ± 0.320.1 ± 0.920.7 ± 1.0
       NDF, % of DM31.6 ± 1.227.3 ± 0.830.7 ± 1.131.1 ± 1.1
       Fat, % of DM3.4 ± 0.32.91 ± 0.13.30 ± 0.23.35 + 0.3
       NFC,
      Calculated as 100 – [(% NDF − % NDF-CP) + % CP + % Fat + % Ash].
      % of DM
      38.8 ± 5.340.2 ± 0.939.1 ± 4.438.9 ± 4.8
       Starch, % of DM25.3 ± 1.622.3 ± 2.024.7 ± 1.725.0 ± 1.7
       Ca, % of DM0.94 ± 0.10.92 ± 0.20.94 ± 0.10.94 ± 0.1
       P, % of DM0.5 ± 0.030.72 ± 0.070.53 ± 0.040.50 ± 0.03
       NEL, Mcal/kg of DM1.66 ± 0.021.78 ± 0.011.68 ± 0.021.67 ± 0.02
      1 PMR had a particle size distribution (% of DM) of 9.2 ± 5.0% long (>19 mm), 47.3 ± 2.6% medium (<19 mm, >8 mm), 18.0 ± 1.9% short (<8 mm, >4 mm), and 25.4 ± 4.7% fine (<4 mm) particles.
      2 H-AMS = PMR coupled with a high AMS pellet allowance (6.0 kg/d); L-AMS = PMR coupled with a low AMS pellet allowance (3.0 kg/d).
      3 DM of corn silage was 40.2 ± 2.9%, and chemical composition (DM basis) was 10.3 ± 0.9% CP, 18.9 ± 0.6% ADF, and 33.5 ± 0.4% NDF.
      4 DM of straw was 90.6 ± 2.5%, and chemical composition (DM basis) was 7.9 ± 0.4% CP, 57.6 ± 3.0% ADF, and 87.2 ± 2.8% NDF.
      5 DM of alfalfa haylage was 34.5 ± 7.5%, and chemical composition (DM basis) was 21.6 ± 0.6% CP, 35.5 ± 0.5% ADF, and 42.4 ± 0.5% NDF.
      6 DM of high-moisture corn was 73.3 ± 1.3%, and chemical composition (DM basis) was 9.7 ± 0.3% CP, 2.5 ± 0.04% ADF, and 7.9 ± 0.6% NDF.
      7 DM of the PMR supplement was 90.0 ± 1.1%, and chemical composition (DM basis) was 37.5 ± 0.1% CP, 8.4 ± 0.4% ADF, and 17.9 ± 0.4% NDF. Supplied by Floradale Feed Mill Ltd., including ingredients (as fed) 50.0% Soy Plus protein supplement (Landus Cooperative), 20.0% soybean meal 47%, 9.52% wheat shorts, 5.00% canola, 3.7% sodium sesquicarbonate, 3.2% fine salt, 1.65% Diamond V Yeast XP (Diamond V), 1.1% magnesium oxide, 1.0% tallow, 0.95% inorganic micronutrient premix (Floradale Feed Mill Ltd.), 0.88% Integral hydrolyzed dried yeast (Alltech), 0.80% monocalcium phosphate, 0.80% limestone calcium carbonate, 0.80% Metasmart rumen-protected methionine (Adisseo), 0.4% DCAD Plus potassium carbonate (Arm & Hammer), 0.16% sulfur 99.5%, and 0.04% Rumensin monensin (Elanco Animal Health).
      8 Supplied by Floradale Feed Mill Ltd., including ingredients (as fed) 24.2% wheat shorts, 19.5% soybean meal, 12.5% soy hulls (ground), 10.0% wheat chop (ground wheat), 10.0% beet pulp, 10.0% barley chop (ground barley), 5.0% bakery meal, 5.0% corn chop (ground corn), 2.0% molasses (pelleter), 0.8% limestone calcium carbonate, 0.4% Pelltech pellet binder and die lubricant (Borregaard LignoTech), 0.4% fine salt, 0.2% magnesium oxide, and 0.1% ruminant micronutrient premix (Floradale Feed Mill Ltd.).
      9 Values were obtained from chemical analysis of PMR and concentrate samples. NEL was calculated based on
      • National Research Council
      Nutrient Requirements of Dairy Cattle.
      equations.
      10 Calculated as 100 – [(% NDF − % NDF-CP) + % CP + % Fat + % Ash].
      Cows were assigned to an initial treatment group through blocking by DIM, parity, and production level. During the first treatment period, the 7 cows (parity = 2.7 ± 0.8; mean ± SD) assigned to the H-AMS were 122.4 ± 34.6 DIM and producing 44.6 ± 9.6 kg/d of milk at the time of treatment assignment. The 8 cows (parity = 2.6 ± 1.1) assigned to the L-AMS diet during the first treatment period were 125.3 ± 71.0 DIM and producing of 44.4 ± 6.8 kg/d of milk. The dietary treatments were alternated across feed bins within the pen. Researchers were not blinded to treatment, as pellet allocation had to be adjusted to ensure each cow was receiving the correct amount for the treatment that she was on at the time.
      Within each treatment, the daily quantity of pellet available in the AMS was set to exceed the target quantity to ensure the targeted amount of pellet was achieved. The eligible quantity of pellet available to each cow was adjusted every fourth day during each treatment period based on the average actual pellet delivery over the previous 3 d and corrected for DM content. Therefore, to achieve 3.0 and 6.0 kg/d of pellet consumption on a DM basis, a total of 3.25 kg/d DM (3.6 kg/d as fed) and 6.5 kg/d DM (7.2 kg/d as fed) for each respective treatment was initially eligible. The pellet allocation at each milking was based on a linear accrual over time, with a minimum pellet provision of 0.05 kg and a maximum provision of 2.50 kg (as fed). Pellet was dispensed at a rate of 0.50 kg/min (as fed). The AMS feeder was calibrated weekly to ensure that the desired quantity of pellet was dispensed at each milking. To calibrate, the feeder was cleaned, and 4 calibration samples were obtained directly from the feeder. The first sample was discarded to ensure that any material dislodged during the cleaning process did not affect the calibration outcome. The last 3 samples were weighed, and an average of the 3 weights was entered into the computer system (Delpro 4.5, DeLaval).

      PMR Feeding Procedure

      The PMR was prepared daily in a mixer wagon (model 5572, Jaylor Fabricating) and transferred into a feed cart (Super Data Ranger, American Calan). The PMR was fed once daily between 1330 and 1400 h into the automated feed bins. The PMR was available ad libitum, and minimum refusals of 5 to 10% were targeted relative to the amount of feed offered, on an as-fed basis. Out of caution to prevent cows from running out of feed, and to account for day-to-day variability, cows were provided with more feed than needed, resulting in an actual refusal rate (on an as-fed basis) for the duration of the study of 12.8 ± 8.7% (mean ± SD). Feed within the bins was not remixed during the day. At 1315 h, before each daily feeding, the amounts of refusals were recorded, and refusals were removed from the bins. The quantity of PMR offered to each cow was adjusted daily based on the average intake over the previous 3 d.

      Feeding Behavior

      The automated feed bins continuously measured and recorded the weight of PMR at the start and end of each bin visit (as validated by
      • Chapinal N.
      • Veira D.M.
      • Weary D.M.
      • von Keyserlingk M.A.G.
      Technical note: Validation of a system for monitoring individual feeding and drinking behaviour and intake in group housed dairy cows.
      ), such that feeding time, duration, eating rate, PMR DMI, and meal parameters for each cow could be calculated. The PMR DMI was determined by multiplying the as-fed intake at each bin visit by the weekly average DM percent of the diet, as determined by DM analysis of fresh PMR feed samples. Feeding rate at each bin visit was calculated as DMI (kg/d) divided by the total minutes spent feeding. These data were then summarized to determine total daily DMI (kg/d), feeding time (min/d), and average feeding rate (kg/min). Meal criteria (the minimum time interval between meals) were determined for each cow in each treatment period, using a software package (MIX 3.1.3;
      • MacDonald P.D.M.
      • Green P.E.J.
      User's guide to program MIX: An interactive program for fitting mixtures of distributions. Release 2.3, January 1988.
      ) to fit normal distributions to the frequency of log10 transformed time intervals between bin visits. These criteria were used to combine individual bin visits into meals for each cow, from which meal patterning was then analyzed (as described by
      • DeVries T.J.
      • von Keyserlingk M.A.G.
      • Weary D.M.
      • Beauchemin K.A.
      Technical note: Validation of a system for monitoring feeding behavior of dairy cows.
      ). If an interval of time between 2 bin visits exceeded the determined meal criteria, this indicated a different meal. The number of different meals was classified as meal frequency (no./d). Meal duration (min/meal) was calculated as the time from the start of the first feeding bout until the end of the last feeding bout, at which time the meal criterion was exceeded. Meal size (kg/d) was calculated as DMI divided by the meal frequency.

      Milking Activity, Yield, and Components

      During each treatment period, milking behavior data at each AMS visit were automatically recorded by the AMS software (DelPro 4.5, DeLaval). These data included milking frequency, milking duration, milk yield, and refused milkings, which were then used to determine the average daily frequency, yield, duration of milking visits, and frequency of refused milkings during each treatment period. The fetching activity recorded by barn staff on provided data sheets were used to identify which milkings for each cow were voluntary. Barn staff were blinded to dietary treatments. Milk samples from each cow were collected using an automated sampler (DeLaval Robot Sampler, DeLaval International) at every milking on d 20, 24, and 28 of each treatment period. These samples were sent to a DHI testing laboratory (CanWest DHI, Guelph, ON, Canada) for component analysis (fat and crude protein) using a Fourier-transform infrared full-spectrum analyzer (Milkoscan FT+ and Milkoscan 6000, Foss). Fat (%) and crude protein (%) values per cow on each sampling day were obtained by calculating the average across milkings, weighted by the milk yield of the sampled milkings.

      Behavioral Data Collection and Body Weight

      Standing and lying behavior data were collected using electronic data loggers (HOBO Pendant G Data Logger, Onset Computer Corporation), with leg orientation measurements automatically taken at 1-min intervals (as validated by
      • Ledgerwood D.N.
      • Winckler C.
      • Tucker C.B.
      Evaluation of data loggers, sampling intervals, and editing techniques for measuring the lying behavior of dairy cattle.
      ). Data loggers were attached to the hind leg of each cow using veterinary bandaging tape (Vetrap Bandaging Tape, 3M) on d 14 of each treatment period. On d 22 of each treatment period, a new logger was placed on the other hind leg, before the first logger was removed for data extraction. All loggers were removed after d 28 of each treatment period. Data from the loggers were extracted using Onset HOBOware Software (Onset Computer Corporation), exported to Microsoft Excel (Microsoft Corp.), and processed using Microsoft Excel macros (
      • UBC AWP (University of British Columbia Animal Welfare Program)
      UBC Animal Welfare Program: SOP—HOBO Data Loggers. Pages 1–23.
      ) to summarize lying data as total daily lying time, frequency of lying bouts, and average lying bout length (calculated as total lying time divided by frequency of lying bouts).
      An electronic monitoring system (HR-TAG-LD, SCR Engineers Ltd., as validated by
      • Schirmann K.
      • von Keyserlingk M.A.G.
      • Weary D.M.
      • Veira D.M.
      • Heuwieser W.
      Technical note: Validation of a system for monitoring rumination in dairy cows.
      ) was used to continuously monitor rumination activity. A nylon collar equipped with a data logger was fitted to each cow 6 d before the start of the first adaptation period and removed at the end of the study period. Data were downloaded from the system at least twice weekly throughout the study. These data, stored in 2-h intervals, were used to determine total daily time spent ruminating for each cow. Body weight was measured by bringing cows to a scale (I-20W scale, Ohaus) on 2 consecutive days at the start of each treatment period (d 0 and 1 of adaptation), as well as at the start (d 14 and 15) and end (d 27 and 28) of each data collection period to calculate change in BW during the treatment periods.

      Feed Sampling and Analysis

      Throughout the duration of the study, 2 duplicate samples of the PMR were collected at the time of feed delivery on d 1 and 8 of each adaptation period, as well as on alternate days during each treatment data collection period (d 16, 18, 20, 22, 24, 26, and 28). Samples of each of the PMR ingredient components and AMS pellet were taken on d 1, 15, and 28 for DM and nutrient analysis. After collection, all samples were frozen at −20°C until further analysis, at which point all samples were thawed in a refrigerator for at least 24 h before processing. One of the duplicate PMR samples as well as PMR ingredient and AMS pellet samples were dried at 55°C in an oven for a minimum of 48 h to determine DM and chemical composition of the PMR (Table 1). To determine particle size distribution of the PMR (Table 1), the other duplicate fresh feed samples of the PMR were separated by particle size using a 4-screen Penn State Particle Separator (
      • Heinrichs A.J.
      The Penn State Particle Separator. Extension publication DSE 2013-186.
      ;
      • Maulfair D.D.
      • Heinrichs A.J.
      Effects of varying forage particle size and fermentable carbohydrates on feed sorting, ruminal fermentation, and milk and component yields of dairy cows.
      ) into 4 particle-size fractions: long (>19 mm), medium (8–19 mm), short (4–8 mm), and fine (<4 mm). After being separated into fractions, Penn State Particle Separator samples were oven-dried at 55°C for a minimum of 48 h.
      After being dried, PMR and feed component samples were ground through a 1-mm sieve (Model 4 Wiley Laboratory Mill, Thomas Scientific). Ground samples were then pooled by sample type and date, and shipped to A&L Canada Laboratories Inc. (London, ON, Canada) for analysis of DM (60°C;
      • AOAC International
      Official Methods of Analysis.
      , method 934.01), CP (combustion;
      • AOAC International
      Official Methods of Analysis.
      , method 990.03; FP-628 Nitrogen Analyzer, Leco), OM (550°C;
      • AOAC International
      Official Methods of Analysis.
      , method 942.05; Electric Programmable Asher, Blue M), ADF (
      • AOAC International
      Official Methods of Analysis.
      , method 973.18; Ankom 200 Fiber Analyzer, Ankom Technology), NDF with amylase and sodium sulfite (
      • AOAC International
      Official Methods of Analysis.
      , method 2002.04; Ankom 200, Ankom Technology), and starch with heat-stable amylase and amyloglucosidase (
      • AOAC International
      Official Methods of Analysis.
      , method 996.11; K-TSTA Total Starch Assay Procedure, Megazyme) to determine nutrient composition of the total diet.

      Personality Tests

      Previous work has demonstrated that behavioral responses toward novelty in dairy cattle are consistent over time and across contexts (
      • Hedlund L.
      • Løvlie H.
      Personality and production: Nervous cows produce less milk.
      ;
      • Neave H.W.
      • Costa J.H.C.
      • Weary D.M.
      • von Keyserlingk M.A.G.
      Long-term consistency of personality traits of cattle.
      ). Correlated sets of these behaviors can, therefore, be referred to as personality traits (
      • Carter A.J.
      • Feeney W.E.
      • Marshall H.H.
      • Cowlishaw G.
      • Heinsohn R.
      Animal personality: What are behavioural ecologists measuring?.
      ). We assessed the personality traits of each cow using a combined arena test once on d 1, 2, 3, 4, 5, 6, or 7 of adaptation during the second treatment period. Personality traits were assessed by observing behavioral responses to novelty, based on methodology established by
      • Van Reenen C.G.
      • Engel B.
      • Ruis-Heutinck L.F.M.
      • Van der Werf J.T.N.
      • Buist W.G.
      • Jones R.B.
      • Blokhuis H.J.
      Behavioral reactivity of heifer calves in potentially alarming test situations: A multivariate and correlational analysis.
      ,
      • Lauber M.C.Y.
      • Hemsworth P.H.
      • Barnett J.L.
      The effects of age and experience on behavioral development in dairy calves.
      , and
      • Neave H.W.
      • Costa J.H.C.
      • Weary D.M.
      • von Keyserlingk M.A.G.
      Personality is associated with feeding behavior and performance in dairy calves.
      . The assessment was conducted in an observational arena measuring approximately 6 × 11 m that was unfamiliar to the cows, containing one water trough, and located such that the cow had limited visual contact with conspecifics while in the arena. The floor consisted of rubber mats, and the walls were primarily metal fencing and gates, with one wall being a concrete wall topped with metal bars. The center of the arena was marked with chalk, and a grid of 1 × 1-m squares was drawn onto the floor at the start of each test day. All assessments were conducted between 1100 and 1300 h, and movement of farm staff and other researchers in the vicinity of the arena was restricted to minimize external stimuli. Up to 3 cows were assessed on each day, and each cow was assessed using the same 3-step test procedure consisting of (1) a novel arena test, (2) a novel object test, and (3) a novel human test. At the start of each assessment, a familiar researcher calmly led the cow from the AMS pen to the arena. The combined arena test started when the gate had been closed behind the cow, after which the cow spent 10 min alone in the arena (novel arena test). The same familiar researcher then entered the arena and placed a 68-L pink plastic storage bin (not used in day-to-day facility management and thus novel to the cows) in the center of the arena and exited the arena (novel object test). After 10 min, the familiar researcher entered the arena and removed the object, after which an unfamiliar (and thus novel) human in standardized clothing (bright blue coveralls with high-visibility yellow striping, which are not used by barn staff and thus unknown to the cows) entered the arena and stood at the center of the arena for 10 min (novel human test). After a visual signal that the test was concluded, the novel human then exited the arena, and the familiar researcher led the cow back to the AMS pen. Each personality test was video recorded (using 2 video cameras placed such that all parts of the arena would be visible on at least one recording). The arena was scraped clean of manure between each cow and washed with a high-pressure water hose at the end of each day. Any cow suspected or confirmed in heat was not eligible for testing on that day but was tested 3 d later.
      Video recordings of the personality tests were later used to determine frequencies, durations, and latencies of behaviors during each stage of the combined arena test (Table 2). These behaviors were chosen based on the studies of
      • Schrader L.
      • Müller R.
      Behavioural consistency during social separation and personality in dairy cows.
      ,
      • Rousing T.
      • Badsberg J.H.
      • Klaas I.C.
      • Hindhede J.
      • Sørensen J.T.
      The association between fetching for milking and dairy cows' behavior at milking, and avoidance of human approach—An on-farm study in herds with automatic milking systems.
      ,
      • Gibbons J.
      • Lawrence A.
      • Haskell M.
      Responsiveness of dairy cows to human approach and novel stimuli.
      , and
      • Hedlund L.
      Personality and production in dairy cows.
      . Video analysis was performed by a single researcher, after establishing interobserver reliability of κ > 0.71 for each test between 2 observers, and intraobserver reliability of κ > 0.86. Training for behavior scoring was provided through a detailed explanation of the ethogram in Table 2. In any cases where one camera did not record (through either malfunction or human error) and the cow was subsequently not visible on any video recording during the test, the time the cow was not visible was excluded from analyses. All time duration variables were converted to a percentage of test time, calculated as time spent performing the behavior, divided by total test time. If applicable, the time the cow was not visible during the test was subtracted from the total test time before calculation.
      Table 2Definitions of behavioral measures and events recorded during the combined arena test for assessment of personality traits of 15 cows
      BehaviorDefinition
      Recorded during novel arena, novel object, and novel human tests
       Squares entered (count)Number of 1 × 1 m squares of the chalk grid the cow entered with at least 1 hoof.
       Vocalization
      Non-normally distributed, log10 transformed data used in later analysis.
      (count)
      All types of sound emitted by the cow, with mouth either open or closed.
       Drinking
      Not used in principal component analysis, as behaviors occurred infrequently and did not meet assumptions of normality.
      (count)
      Cow puts muzzle in water trough in test arena.
       Urination
      Not used in principal component analysis, as behaviors occurred infrequently and did not meet assumptions of normality.
      (count)
      Cow urinates in test arena.
       Defecation
      Not used in principal component analysis, as behaviors occurred infrequently and did not meet assumptions of normality.
      (count)
      Cow defecates in test arena.
       Startle
      Not used in principal component analysis, as behaviors occurred infrequently and did not meet assumptions of normality.
      (count)
      Cow flinches, bucks, or jumps in the arena.
       Tail movement
      Not used in principal component analysis, as behaviors occurred infrequently and did not meet assumptions of normality.
      (count)
      Observable sideways movement (flicking) of the tail.
       Vigilance (s)Cow raises head above the withers while stationary.
       Attraction
      Non-normally distributed, log10 transformed data used in later analysis.
      (s)
      Cow raises head above the withers and looks in the direction of the home pen.
       Locomotion
      Non-normally distributed, log10 transformed data used in later analysis.
      (s)
      Cow moves all 4 legs to cover distance.
      Recorded during novel object and novel human tests only
       Avoidance distance (m)Smallest number of 1 × 1 m squares between the cow and the novel object or human at any point during the test (>2 m, 1 to 2 m, or <1 m, measured from the front hoof, or 0 m/contact).
       Contact with novel object/human
      Non-normally distributed, log10 transformed data used in later analysis.
      (s)
      Any part of the cow's body is in physical contact with the novel object or person.
       Latency to contact novel object/human
      Non-normally distributed, log10 transformed data used in later analysis.
      (s)
      Time from start of test to first contact with the novel object or human.
      1 Non-normally distributed, log10 transformed data used in later analysis.
      2 Not used in principal component analysis, as behaviors occurred infrequently and did not meet assumptions of normality.

      Statistical Analyses

      All statistical analyses were conducted using SAS 9.4 software (
      • SAS Institute Inc
      SAS version 9.4.
      ). All values reported are least squares means. Significance was declared if P ≤ 0.05, and tendencies were considered 0.05 < P ≤ 0.10. Before analyses, all data were screened for normality using the UNIVARIATE procedure of SAS. The assumptions of normality were met for all variables. One multiparous cow was excluded from analyses of treatment effect on AMS visits and rejections, as her data fell outside 3 standard deviations from the mean for these outcomes. Due to a technical error with the milk sampler during the first treatment period, those data were analyzed based on 2 sampling days rather than the intended 3 d. For behaviors analyzed from the personality assessments, the variables of vocalization, locomotion, attraction to pen, duration of contact, and latency to contact were log10 transformed to meet assumptions of normality.

      Effects of Dietary Treatments on Behavior and Performance

      The effects of AMS concentrate allowance on feed intake, feeding behavior, lying behavior, rumination, milking activity, BW, and production were analyzed using the MIXED procedure of SAS. The model included the fixed effects of period, treatment, and period × treatment interaction, and random effect of cow within period × treatment. To model repeated measures across time within cow, the variance-covariance matrix structure used was autoregressive, chosen on the basis of best fit according to Schwarz's Bayesian information criterion. Degrees of freedom for fixed effects were estimated using the Kenward-Roger option in the MODEL statement. Any significant (P < 0.05) period × treatment interactions that were detected were initially plotted to visually identify the interaction. Given the within-cow experimental design, these interactions were then interpreted as the change in the outcome variable, between treatments and across periods, within cow. Any period × treatment interactions with P > 0.05 were retained in the model to account for variation associated with treatment sequence (
      • Tempelman R.J.
      Experimental design and statistical methods for classical and bioequivalence hypothesis testing with an application to dairy nutrition studies.
      ).
      To determine the effect of AMS concentrate allowance on day-to-day variation in DM consumption of AMS pellet, PMR, and total feed consumption, the coefficient of variation (CV) of these across each 14-d observation period, was generated. These data did not have equal variance and, thus, were analyzed using the GLIMMIX procedure with normal distribution specified. The fixed effects of treatment, period, and period × treatment were included in the model, as well as the random effect of cow within period × treatment.

      Principal Component Analysis to Identify Personality Traits

      The FACTOR procedure of SAS was used to perform principal component analysis (PCA) with varimax rotation to identify personality traits by condensing correlated behavioral measures from the combined arena tests. Criteria for PCA analysis of animal behavior data followed
      • Budaev S.V.
      Using principal components and factor analysis in animal behaviour research: Caveats and guidelines.
      . To achieve sampling adequacy (a test of validity related to subject:variable ratio in which >0.50 is considered appropriate for conducting PCA on the data set), behaviors from the novel arena and novel object tests were analyzed in one PCA (12 input variables; sampling adequacy = 0.52), and behaviors from the novel human test were analyzed in a second PCA (7 input variables; sampling adequacy = 0.62). These behaviors were grid squares entered, vocalization, locomotion, attraction to herd, vigilance, duration of contact (with object or human), and latency to contact (object or human). Avoidance distances from the novel object and novel human were initially included in their respective PCA but were removed due to low communality estimates; all other variables achieved communality estimates >0.30. After examination of scree plots, factors with eigenvalues greater than 1.0 were retained, resulting in 3 factors from the novel arena and novel object tests PCA (76% cumulative variance) and 2 factors from the novel human test PCA (77% cumulative variance). High loadings of behaviors from the tests on each factor were considered equal to or greater than ± 0.63. For each factor, the score for each cow was extracted using the regression procedure, indicating where each cow lies along an axis from highly negative to highly positive for that factor. To verify the stability of the PCA, cows were classified as low parity (lactation 1 or 2) or high parity (lactation 3 or greater), and the residuals of each test behavior were obtained from a generalized linear model with parity class as the fixed effect. The PCA was then repeated using the residuals from each linear model as input variables. The results produced a similar factor loading pattern as the PCA on original observations, verifying that the observed factors (reported herein) could be interpreted as putative personality traits.

      Descriptive Effect of Personality Traits on Treatment Response

      To assess our exploratory secondary objective, we examined whether personality traits were associated with feed intake and behavior outcomes on the dietary treatments. For outcomes in which an effect of AMS concentrate allowance treatment was detected, the within-cow differences between the treatment means were generated by subtracting the L-AMS treatment mean from the H-AMS treatment mean for each cow. These within-cow differences for these outcomes were then plotted against each personality trait (factor identified from the PCA as described previously) and relationships are reported descriptively. Parity, DIM, and milk production at enrollment were also plotted against each PCA factor to examine whether these variables were associated with personality traits. Simple linear regression, using the REGRESSION procedure of SAS, was then applied to assess these associations; those associations with P < 0.1 are described.
      To determine any potential major effects resulting from the data of the 1 primiparous cow enrolled in the study, all models were repeated excluding this cow's data. No significant changes to the results from the perspective of treatment response occurred when this cow's data were excluded. Further, removal of this cow did not alter the descriptive personality analyses. It was also confirmed that this cow was not an outlier in terms of BW, BCS, body length, or height; thus, the decision was made to retain her data.

      RESULTS

      AMS Concentrate Delivery and Feed Intake

      When cows were on the L-AMS treatment, they had a 1.55 kg/d greater PMR intake than when on the H-AMS (Table 3). As designed by the experimental treatments, on L-AMS, cows had lesser daily AMS pellet delivered (Table 3), on a DM basis. A period × treatment interaction was detected for total DMI (Table 3); overall greater DMI was observed on the H-AMS treatment, but the magnitude of difference depended on which treatment cows received first. Cows that started on H-AMS consumed +1.2 kg/d total DMI on that treatment compared with when on L-AMS. Alternatively, those cows that started on L-AMS consumed +1.4 kg/d total DMI on H-AMS compared with the L-AMS treatment. The day-to-day variation of total DMI was greater when cows were on the L-AMS treatment.
      Table 3Effects of dietary treatments on DMI, feeding behavior, rumination, and lying behavior
      Data were collected over 14 d for 15 cows on each treatment.
      ItemDietary treatment
      H-AMS = partial mixed ration (PMR) coupled with a high automated milking system (AMS) pellet allowance (6.0 kg/d); L-AMS = PMR coupled with a low AMS pellet allowance (3.0 kg/d).
      SEM
      SEM for the Period × Treatment interaction is reported.
      P-value
      H-AMSL-AMSTreatmentPeriodTreatment × Period
      Period 1Period 2Period 1Period 2
      DMI, kg/d
       PMR23.925.126.825.30.700.030.860.054
       AMS pellet5.86.03.03.10.14<0.0010.170.65
       Total29.631.229.828.40.680.060.850.03
      CV,
      SD and CV in daily intake calculated for each cow over 14 d.
      %
       PMR7.926.848.239.550.980.120.900.21
       AMS pellet13.514.110.813.72.470.540.480.63
       Total6.195.957.508.380.780.020.670.47
      PMR feeding behavior
       Feeding time, min/d20418920222813.90.180.720.15
       Feeding rate, kg/min0.140.150.150.140.010.840.840.22
       Meal criterion, min344138323.080.420.930.04
       Meal frequency, meals/d8.57.37.48.20.210.640.31<0.001
       Meal size, kg of DM/meal2.93.63.73.30.110.020.37<0.001
       Meal duration, min/meal374646442.500.150.220.03
       Total daily meal time, min/d30931732933613.90.160.570.92
      Rumination, min/d5495515555835.97<0.010.020.03
      Lying bouts, no./d8.210.09.77.40.480.230.67<0.001
      Lying time, min/d71573172474212.50.390.160.98
      Lying bout length, min/bout9180831044.480.070.27<0.001
      1 Data were collected over 14 d for 15 cows on each treatment.
      2 H-AMS = partial mixed ration (PMR) coupled with a high automated milking system (AMS) pellet allowance (6.0 kg/d); L-AMS = PMR coupled with a low AMS pellet allowance (3.0 kg/d).
      3 SEM for the Period × Treatment interaction is reported.
      4 SD and CV in daily intake calculated for each cow over 14 d.

      Feeding and Lying Behavior

      A period × treatment interaction was detected for the meal criterion, which was greater on H-AMS, but the magnitude of difference depended on which treatment cows started on (Table 3). Cows decreased their PMR meal frequency from period 1 to period 2, to differing magnitudes depending on which treatment they started on (Table 3). A period × treatment interaction indicated that cows had greater PMR meal sizes when they were on L-AMS (Table 3), but the magnitude of difference with the corresponding H-AMS treatment differed based on starting treatment (+0.4 kg DM when starting on H-AMS vs. +0.1 kg DM when starting on L-AMS), likely associated with an overall increase in meal size across cows from period 1 to period 2. A period × treatment interaction was detected for meal duration, with cows who started on H-AMS increasing their meal duration in period 2. No differences in PMR feeding time, feeding rate, or total daily meal time were detected between the treatments.
      A period × treatment interaction was detected for rumination (Table 3); overall greater rumination was observed on L-AMS, but the magnitude of difference depended on which treatment cows started on. Period × treatment interactions were detected for lying bouts and lying bout length. Cows had more lying bouts on H-AMS, with no difference in lying time detected, resulting in bout length also being lesser on this treatment, with the magnitude of these differences from L-AMS depending on initial treatment cows were assigned to.

      Behavior in the AMS and Milk Yield

      As a result of the dietary treatments, cows on H-AMS had greater pellet delivery per milking, with the magnitude of difference being greater for those who started on the L-AMS treatment (Table 4). In addition, period × treatment interactions were detected for the outcomes of milking frequency, AMS visits, AMS rejections, box time per milking, and milk fat percentage (Table 4). Number of AMS visits and rejections decreased from period 1 to period 2, resulting in lesser milking frequency in period 2, with a greater reduction in those cows who started on H-AMS in period 1. Box time per milking also increased for both groups from period 1 to period 2, again with magnitude depending on which treatment cows started on. Initial treatment also influenced the degree to which average milk fat increased from period 1 to period 2, with the change being greater for cows that started on H-AMS. No differences in daily milk yield, fetches, daily box time, milk protein percentage, or component yields were detected between the treatments (Table 4).
      Table 4Effect of dietary treatments on milking activity and production
      Data are averaged over 14 d for 15 cows on each treatment, unless otherwise indicated.
      ItemDietary treatment
      H-AMS = partial mixed ration (PMR) coupled with a high automated milking system (AMS) pellet allowance (6.0 kg/d); L-AMS = PMR coupled with a low AMS pellet allowance (3.0 kg/d).
      SEM
      SEM for the period × treatment interaction is reported.
      P-value
      H-AMSL-AMSTreatmentPeriodTreatment × Period
      Period 1Period 2Period 1Period 2
      Milking activity
      Data are averaged over 14 d for 15 cows on each treatment, unless otherwise indicated.
       Milking frequency, no./d4.473.513.564.330.240.850.67<0.001
       Fetches, no./d0.110.330.400.160.160.730.960.15
       AMS visits,
      Data were collected over 14 d for 14 cows on each treatment. AMS visits encompassed both milking events and visits that did not meet milking criteria, resulting in a rejection.
      no./d
      8.044.575.416.760.900.800.220.01
       Rejected milkings,
      Data were collected over 14 d for 14 cows on each treatment. AMS visits encompassed both milking events and visits that did not meet milking criteria, resulting in a rejection.
      no./d
      3.761.031.792.610.620.740.11<0.01
       Pellet intake per milking, kg of DM/milking1.331.810.930.780.07<0.0010.03<0.001
       Box time, min/milking5.936.656.596.050.220.890.660.01
       Daily box time, min/d262324251.480.840.860.14
      Milk yield,
      Data are averaged over 14 d for 15 cows on each treatment, unless otherwise indicated.
      kg/d
      45.946.244.744.21.550.300.940.79
      Milk yield per milking,
      Data are averaged over 14 d for 15 cows on each treatment, unless otherwise indicated.
      kg/milking
      10.813.913.710.91.120.990.880.01
      Milk composition,
      Data were collected over 3 d for each treatment.
      %
       Fat3.954.264.214.110.110.590.310.05
       Protein3.463.513.503.450.070.880.990.34
      Milk component yield,
      Data were collected over 3 d for each treatment.
      kg/d
       Fat1.762.021.801.920.130.810.130.56
       Protein1.541.661.511.600.100.610.280.91
      1 Data are averaged over 14 d for 15 cows on each treatment, unless otherwise indicated.
      2 H-AMS = partial mixed ration (PMR) coupled with a high automated milking system (AMS) pellet allowance (6.0 kg/d); L-AMS = PMR coupled with a low AMS pellet allowance (3.0 kg/d).
      3 SEM for the period × treatment interaction is reported.
      4 Data were collected over 14 d for 14 cows on each treatment. AMS visits encompassed both milking events and visits that did not meet milking criteria, resulting in a rejection.
      5 Data were collected over 3 d for each treatment.

      Body Weight

      No differences between treatments in overall or change in BW were detected (data not shown). Cows weighed an average of (mean ± SD) 744 ± 75 kg at the start of period 1, 768 ± 83 kg between the 2 treatment periods, and 780 ± 83 kg at the end of period 2.

      Principal Component Analysis

      The loadings for each factor extracted from the 2 PCA are reported in Table 5. In the novel arena and novel object PCA, cows who scored highly on factor 1 were interpreted as being “active” (high positive loadings for the number of squares entered and locomotion during the novel arena and object tests). Cows who scored highly on factor 2 were interpreted as being “social” (high positive loadings for vocalization and attraction to the herd during the novel arena and object tests). Finally, cows who scored highly on factor 3 were interpreted as being “alert-curious” (high positive loadings for vigilance during the novel arena and object tests).
      Table 5Coefficients (loadings) of the eigenvectors for the factors extracted from each principal component analysis of behaviors recorded when cows (n = 15) were subjected to a combined arena test
      Combined arena test consisted of successive exposure to a novel arena (10 min), novel object (10 min), and novel human (10 min). In the first principal component analysis (PCA), the first 3 factors were extracted from behaviors observed during the novel arena and novel object stages, and in the second PCA, the first 2 factors were extracted from behaviors observed during the novel human stage.
      High loadings (≥|0.63|) are indicated with an asterisk (*), indicating behavior variables that were highly correlated within each factor.
      TestBehaviorFactor 1Factor 2Factor 3
      Novel arena
      Squares entered0.93*0.170.02
      Vocalization
      Variable was log10 transformed to meet assumption of normality.
      0.370.74*0.20
      Locomotion
      Variable was log10 transformed to meet assumption of normality.
      0.95*−0.050.12
      Attraction to herd
      Variable was log10 transformed to meet assumption of normality.
      0.110.91*−0.24
      Vigilance−0.57−0.030.70*
      Novel object test
      Squares entered0.92*0.17−0.16
      Vocalization
      Variable was log10 transformed to meet assumption of normality.
      0.150.84*0.09
      Locomotion
      Variable was log10 transformed to meet assumption of normality.
      0.86*−0.14−0.17
      Attraction to herd
      Variable was log10 transformed to meet assumption of normality.
      −0.330.79*−0.04
      Vigilance−0.110.420.72*
      Duration of interaction
      Variable was log10 transformed to meet assumption of normality.
      −0.59−0.42−0.15
      Latency to contact
      Variable was log10 transformed to meet assumption of normality.
      −0.170.19−0.60
       Eigenvalues4.603.031.48
       Variance explained (%)38.325.212.3
       Interpretation
      Personality traits were interpreted based on behaviors exhibited during the combined arena test that loaded highly (either positively or negatively) onto each factor.
      ActiveSocialAlert-curious
      Novel human test
      Squares entered0.95*0.16
      Vocalization
      Variable was log10 transformed to meet assumption of normality.
      0.92*0.13
      Locomotion
      Variable was log10 transformed to meet assumption of normality.
      0.94*0.04
      Attraction to herd
      Variable was log10 transformed to meet assumption of normality.
      0.59−0.05
      Vigilance−0.050.86*
      Duration of interaction
      Variable was log10 transformed to meet assumption of normality.
      −0.21−0.92*
      Latency to contact
      Variable was log10 transformed to meet assumption of normality.
      −0.030.84*
       Eigenvalues3.242.12
       Variance explained (%)46.330.2
       Interpretation
      Personality traits were interpreted based on behaviors exhibited during the combined arena test that loaded highly (either positively or negatively) onto each factor.
      Active-vocalFear of novel human
      1 Combined arena test consisted of successive exposure to a novel arena (10 min), novel object (10 min), and novel human (10 min). In the first principal component analysis (PCA), the first 3 factors were extracted from behaviors observed during the novel arena and novel object stages, and in the second PCA, the first 2 factors were extracted from behaviors observed during the novel human stage.
      2 High loadings (≥|0.63|) are indicated with an asterisk (*), indicating behavior variables that were highly correlated within each factor.
      3 Variable was log10 transformed to meet assumption of normality.
      4 Personality traits were interpreted based on behaviors exhibited during the combined arena test that loaded highly (either positively or negatively) onto each factor.
      In the novel human PCA, cows who scored highly on factor 1 of this PCA were interpreted as being “active-vocal” (high positive loadings for squares entered, vocalization, and locomotion during the novel human test). Cows who scored highly on factor 2 of this PCA were interpreted as being “fearful of novel humans” (high positive loadings for vigilance and latency to contact the novel human, as well as high negative loading for duration of interaction with the novel human).

      Personality Traits

      Cows who were more active in the novel arena and object tests tended to be of lesser parity (P = 0.09; R2 = 0.20; data not shown) and to have lesser milk production at enrollment (P = 0.09; R2 = 0.20; data not shown). Cows who were more active-vocal in the novel human test were of greater DIM (P = 0.04; R2 = 0.29; data not shown). Neither of these personality traits were associated with the tested outcomes of PMR intake, pellet delivery, total DMI, total DMI CV, and rumination time. The other identified personality traits were not associated with parity, DIM, or milk production at enrollment; however, they were associated with some treatment responses. Cows who were more fearful of the novel human had smaller within-cow differences in AMS pellet delivery between the 2 treatments (P < 0.01; R2 = 0.57; Figure 1A), indicating that they were less likely to reach the target concentrate delivery on H-AMS. As a result, cows who were more fearful also tended to have lesser total DMI on H-AMS (P = 0.06; R2 = 0.25; Figure 1B), and also had greater day-to-day variation in total DMI in this treatment (P = 0.05; R2 = 0.27; Figure 1C). Cows who were more alert-curious in the novel arena and object tests tended to have greater within-cow differences in PMR intake between the treatments (P = 0.07; R2 = 0.23; Figure 1D), indicating that those cows consumed less on H-AMS.
      Figure thumbnail gr1
      Figure 1Associations between personality traits and feed intake and behavior outcomes, where a high positive score on the personality trait indicates a more “fearful” or more “alert-curious” cow. (A) Fear of novel humans principal component analysis (PCA) trait and within-cow difference in automated milking system (AMS) concentrate delivery between treatments; (B) fear of novel humans PCA trait and within-cow difference in total DMI between treatments; (C) fear of novel humans PCA trait and within-cow difference in CV in total DMI between treatments, and (D) alert-curious PCA trait and within-cow difference in partial mixed ration (PMR) intake between treatment. Fifteen cows were provided treatments of either 6.0 kg/d (H-AMS) or 3.0 kg/d (L-AMS) of concentrate in the AMS, with 14 d of data per treatment. Within-cow difference is expressed as the L-AMS treatment mean subtracted from the H-AMS treatment mean.

      DISCUSSION

      Concentrated feed is a key factor motivating cows to milk voluntarily in an AMS (
      • Prescott N.B.
      • Mottram T.T.
      • Webster A.J.F.
      Relative motivations of dairy cows to be milked or fed in a Y-maze and an automatic milking system.
      ;
      • Melin M.
      • Svennersten-Sjaunja K.
      • Wiktorsson H.
      Feeding patterns and performance of cows in controlled cow traffic in automated milking systems.
      ;
      • Bava L.
      • Tamburini A.
      • Penati C.
      • Riva E.
      • Mattachini G.
      • Provolo G.
      • Sandrucci A.
      Effects of feeding frequency and environmental conditions on dry matter intake, milk yield and behaviour of dairy cows milked in conventional or automatic milking systems.
      ). However, providing too much AMS concentrate may result in cows not consuming the desired amount (
      • Halachmi I.
      • Ofir S.
      • Miron J.
      Comparing two concentrate allowances in an automatic milking system.
      ;
      • Bach A.
      • Iglesias C.
      • Calsamiglia S.
      • Devant M.
      Effect of amount of concentrate offered in automatic milking systems on milking frequency, feeding behaviour, and milk production of dairy cattle consuming high amounts of corn silage.
      ;
      • Bach A.
      • Cabrera V.
      Robotic milking: Feeding strategies and economic returns.
      ), increased variability in composition of the total diet consumed (
      • Hare K.
      • DeVries T.J.
      • Schwartkopf-Genswein K.S.
      • Penner G.B.
      Does the location of concentrate provision affect voluntary visits, and milk component yield for cows in an automated milking system?.
      ;
      • Menajovsky S.B.
      • Walpole C.E.
      • DeVries T.J.
      • Schwartzkopf-Genswein K.S.
      • Walpole M.E.
      • Penner G.B.
      The effect of the forage-to-concentrate ratio of the partial mixed ration and the quantity of concentrate in an automatic milking system for lactating Holstein cows.
      ;
      • Paddick K.S.
      • DeVries T.J.
      • Schwartzkopf-Genswein K.
      • Steele M.A.
      • Walpole M.E.
      • Penner G.B.
      Effect of the amount of concentrate offered in an automated milking system on dry matter intake, milk yield, milk composition, ruminal digestion, and behavior of primiparous Holstein cows fed isocaloric diets.
      ), and may not achieve greater milk yields (
      • Migliorati L.
      • Speroni M.
      • Lolli S.
      • Calza F.
      Effect of concentrate feeding on milking frequency and milk yield in an automatic milking system.
      ;
      • Bach A.
      • Iglesias C.
      • Calsamiglia S.
      • Devant M.
      Effect of amount of concentrate offered in automatic milking systems on milking frequency, feeding behaviour, and milk production of dairy cattle consuming high amounts of corn silage.
      ;
      • Henriksen J.C.S.
      • Weisbjerg M.R.
      • Løvendahl P.
      • Kristensen T.
      • Munksgaard L.
      Effects of an individual cow concentrate strategy on production and behavior.
      ). This study adds to this growing body of literature by investigating how the feed intake, eating behavior, milking activity, and performance of dairy cows are affected by the concentrate allocation in a free-traffic AMS. Furthermore, little is known about how individual cow personality traits influence the response to AMS concentrate allocation. To our knowledge, the present study is the first to descriptively explore these relationships.

      AMS Concentrate Delivery and Feed Intake

      On average, cows on both the H-AMS and the L-AMS treatments reached the desired AMS concentrate provisions of 6.0 and 3.0 kg/d of DM, respectively. The maximum daily concentrate allowance was programmed to exceed the targets for each treatment to ensure that the desired concentrate amounts were delivered, a strategy that has also been used successfully in studies testing isonutritional diets with differing AMS concentrate provisions in both guided (
      • Hare K.
      • DeVries T.J.
      • Schwartkopf-Genswein K.S.
      • Penner G.B.
      Does the location of concentrate provision affect voluntary visits, and milk component yield for cows in an automated milking system?.
      ;
      • Menajovsky S.B.
      • Walpole C.E.
      • DeVries T.J.
      • Schwartzkopf-Genswein K.S.
      • Walpole M.E.
      • Penner G.B.
      The effect of the forage-to-concentrate ratio of the partial mixed ration and the quantity of concentrate in an automatic milking system for lactating Holstein cows.
      ;
      • Paddick K.S.
      • DeVries T.J.
      • Schwartzkopf-Genswein K.
      • Steele M.A.
      • Walpole M.E.
      • Penner G.B.
      Effect of the amount of concentrate offered in an automated milking system on dry matter intake, milk yield, milk composition, ruminal digestion, and behavior of primiparous Holstein cows fed isocaloric diets.
      ) and free-traffic AMS (
      • Schwanke A.J.
      • Dancy K.M.
      • Didry T.
      • Penner G.B.
      • DeVries T.J.
      Effects of concentrate location on the behavior and production of dairy cows milked in a free-traffic automated milking system.
      ).
      We were interested in understanding the variability around AMS concentrate delivery to individual cows, by descriptively exploring whether personality traits may be associated with differences in cow behavior and feed intake at the AMS in response to AMS concentrate allowance. Interestingly, we observed that cows who were more fearful of the novel person were delivered less AMS concentrate per day when they were on the H-AMS treatment, and thus were less likely to reach the target concentrate delivery of 6 kg/d. This may have been because the AMS dispenses pellet only if the cow puts her head down to the feeder, with a feed dispensing rate of 0.5 kg/min and a maximum meal size of 2.5 kg. Previous work has shown that fear of humans (measured in a standardized novel human test, similar to that used in our study) is negatively associated with milk production (
      • Breuer K.
      • Hemsworth P.H.
      • Barnett J.L.
      • Matthews L.R.
      • Coleman G.J.
      Behavioural response to humans and the productivity of commercial dairy cows.
      ). Thus, more fearful cows may have been more wary during milking and less willing to interact with the feeder in the AMS; as a result, these cows may have been dispensed less pellet in the AMS over the course of a day. As a result of their lesser AMS concentrate delivery, more fearful cows had lesser total DMI on the H-AMS treatment. We also observed that these cows had greater day-to-day variation in total DMI on that treatment.
      Intake of PMR was greater on the L-AMS treatment, resulting in greater rumination time when cows were on this treatment as well. We observed that total DMI was greater when cows were on the H-AMS treatment, which agrees with the results of
      • Schwanke A.J.
      • Dancy K.M.
      • Didry T.
      • Penner G.B.
      • DeVries T.J.
      Effects of concentrate location on the behavior and production of dairy cows milked in a free-traffic automated milking system.
      and is in line with the formulated diets of the present study. We hypothesize that the differing magnitude of total DMI difference between H-AMS and L-AMS in the 2 groups of cows observed in the present study was due to lesser rumen capacity with advanced DIM (
      • Allen M.S.
      • Bradford B.J.
      • Oba M.
      The hepatic oxidation theory of the control of feed intake and its application to ruminants.
      ). This would result in more filling effect as lactation progressed, as, in period 2, cows on L-AMS would have been more limited in how much PMR they could consume to compensate for lesser pellet allowance in the AMS. We also observed that cows who were more alert-curious consumed less PMR on H-AMS, possibly due to spending more time being vigilant or exploring aspects of their home pen; however, this speculation requires further research.
      Overall, providing greater amounts of concentrate in the AMS may not guarantee that all cows are consuming the formulated diet, and this effect may be amplified or dampened in individuals of differing personalities. Thus, further research investigating these relationships is warranted to determine how to tailor precision feeding strategies to individuals to ensure successful feeding management.

      PMR Substitution Rate

      In this study we observed a PMR substitution rate of 0.54 kg per 1 kg of AMS concentrate allocated, which is similar to that reported in a study using isonutritional treatments in a free-traffic AMS system (0.63 kg;
      • Schwanke A.J.
      • Dancy K.M.
      • Didry T.
      • Penner G.B.
      • DeVries T.J.
      Effects of concentrate location on the behavior and production of dairy cows milked in a free-traffic automated milking system.
      ) but less than that found in the feed-first guided traffic studies of
      • Hare K.
      • DeVries T.J.
      • Schwartkopf-Genswein K.S.
      • Penner G.B.
      Does the location of concentrate provision affect voluntary visits, and milk component yield for cows in an automated milking system?.
      ; 1.58 kg),
      • Menajovsky S.B.
      • Walpole C.E.
      • DeVries T.J.
      • Schwartzkopf-Genswein K.S.
      • Walpole M.E.
      • Penner G.B.
      The effect of the forage-to-concentrate ratio of the partial mixed ration and the quantity of concentrate in an automatic milking system for lactating Holstein cows.
      ; 0.83 kg), and
      • Paddick K.S.
      • DeVries T.J.
      • Schwartzkopf-Genswein K.
      • Steele M.A.
      • Walpole M.E.
      • Penner G.B.
      Effect of the amount of concentrate offered in an automated milking system on dry matter intake, milk yield, milk composition, ruminal digestion, and behavior of primiparous Holstein cows fed isocaloric diets.
      ; 0.97 kg). It was suggested that the lesser substitution rate observed in
      • Schwanke A.J.
      • Dancy K.M.
      • Didry T.
      • Penner G.B.
      • DeVries T.J.
      Effects of concentrate location on the behavior and production of dairy cows milked in a free-traffic automated milking system.
      could be due to those cows being primiparous animals in early lactation; however, the present study used cows of similar parity and lactation stage to those of
      • Menajovsky S.B.
      • Walpole C.E.
      • DeVries T.J.
      • Schwartzkopf-Genswein K.S.
      • Walpole M.E.
      • Penner G.B.
      The effect of the forage-to-concentrate ratio of the partial mixed ration and the quantity of concentrate in an automatic milking system for lactating Holstein cows.
      ,
      • Hare K.
      • DeVries T.J.
      • Schwartkopf-Genswein K.S.
      • Penner G.B.
      Does the location of concentrate provision affect voluntary visits, and milk component yield for cows in an automated milking system?.
      , and
      • Bach A.
      • Iglesias C.
      • Calsamiglia S.
      • Devant M.
      Effect of amount of concentrate offered in automatic milking systems on milking frequency, feeding behaviour, and milk production of dairy cattle consuming high amounts of corn silage.
      . Thus, the lesser substitution rates observed in
      • Schwanke A.J.
      • Dancy K.M.
      • Didry T.
      • Penner G.B.
      • DeVries T.J.
      Effects of concentrate location on the behavior and production of dairy cows milked in a free-traffic automated milking system.
      and the present study may be due to differences in the composition of the PMR diets between these studies. The particle size distributions of the PMR provided in
      • Menajovsky S.B.
      • Walpole C.E.
      • DeVries T.J.
      • Schwartzkopf-Genswein K.S.
      • Walpole M.E.
      • Penner G.B.
      The effect of the forage-to-concentrate ratio of the partial mixed ration and the quantity of concentrate in an automatic milking system for lactating Holstein cows.
      ,
      • Hare K.
      • DeVries T.J.
      • Schwartkopf-Genswein K.S.
      • Penner G.B.
      Does the location of concentrate provision affect voluntary visits, and milk component yield for cows in an automated milking system?.
      , and
      • Bach A.
      • Iglesias C.
      • Calsamiglia S.
      • Devant M.
      Effect of amount of concentrate offered in automatic milking systems on milking frequency, feeding behaviour, and milk production of dairy cattle consuming high amounts of corn silage.
      varied considerably, although those of the present study and
      • Schwanke A.J.
      • Dancy K.M.
      • Didry T.
      • Penner G.B.
      • DeVries T.J.
      Effects of concentrate location on the behavior and production of dairy cows milked in a free-traffic automated milking system.
      were similar. In addition, the PMR diets of
      • Hare K.
      • DeVries T.J.
      • Schwartkopf-Genswein K.S.
      • Penner G.B.
      Does the location of concentrate provision affect voluntary visits, and milk component yield for cows in an automated milking system?.
      had a much lower NDF content, possibly allowing for greater PMR intake compared with the present study. The difference between the targeted concentrate deliveries in the present study, as well as
      • Schwanke A.J.
      • Dancy K.M.
      • Didry T.
      • Penner G.B.
      • DeVries T.J.
      Effects of concentrate location on the behavior and production of dairy cows milked in a free-traffic automated milking system.
      , was also smaller (3.0 vs. 4.0–4.5 kg/d difference between treatments) than in those studies using guided traffic (
      • Bach A.
      • Iglesias C.
      • Calsamiglia S.
      • Devant M.
      Effect of amount of concentrate offered in automatic milking systems on milking frequency, feeding behaviour, and milk production of dairy cattle consuming high amounts of corn silage.
      ;
      • Hare K.
      • DeVries T.J.
      • Schwartkopf-Genswein K.S.
      • Penner G.B.
      Does the location of concentrate provision affect voluntary visits, and milk component yield for cows in an automated milking system?.
      ;
      • Menajovsky S.B.
      • Walpole C.E.
      • DeVries T.J.
      • Schwartzkopf-Genswein K.S.
      • Walpole M.E.
      • Penner G.B.
      The effect of the forage-to-concentrate ratio of the partial mixed ration and the quantity of concentrate in an automatic milking system for lactating Holstein cows.
      ). Finally, nutrient demand and overall intake level may play a role in the degree of PMR substitution, as the milk yields observed in the present study were somewhat higher than in other studies (45 kg/d compared with 40 kg/d in
      • Schwanke A.J.
      • Dancy K.M.
      • Didry T.
      • Penner G.B.
      • DeVries T.J.
      Effects of concentrate location on the behavior and production of dairy cows milked in a free-traffic automated milking system.
      , 36 kg/d in
      • Hare K.
      • DeVries T.J.
      • Schwartkopf-Genswein K.S.
      • Penner G.B.
      Does the location of concentrate provision affect voluntary visits, and milk component yield for cows in an automated milking system?.
      , and 32.5 kg/d in
      • Bach A.
      • Iglesias C.
      • Calsamiglia S.
      • Devant M.
      Effect of amount of concentrate offered in automatic milking systems on milking frequency, feeding behaviour, and milk production of dairy cattle consuming high amounts of corn silage.
      ). Our high milk yields may be related to the relatively low stocking density at the AMS, potentially allowing for the greater milking frequency observed in the present study. However, based on the box times observed in this study, we can predict that increasing stocking density to more typical commercial stocking densities of 45 to 50 cows per AMS (
      • Siewert J.M.
      • Salfer J.A.
      • Endres M.I.
      Factors associated with productivity on automatic milking system dairy farms in the Upper Midwest United States.
      ;
      • Matson R.D.
      • King M.T.M.
      • Duffield T.F.
      • Santschi D.E.
      • Orsel K.
      • Pajor E.A.
      • Penner G.B.
      • Mutsvangwa T.
      • DeVries T.J.
      Benchmarking of farms with automated milking systems in Canada and associations with milk production and quality.
      ) would result in total box times of 21 h/d for H-AMS and 20 h/d for L-AMS, with an additional 1.5 h/d dedicated for AMS wash cycles. Thus, access to the AMS and milk yield per AMS would be limited only if approaching the upper recommended AMS stocking density of 60 cows/AMS.

      Feeding and Lying Behavior

      Meal criterion was overall greater on H-AMS, indicating more nonfeeding time within meals on that treatment. Across cows, meal frequency decreased from period 1 to period 2. Cows had greater PMR meal size on L-AMS, likely because they were consuming more PMR to compensate for the reduced AMS concentrate allowance. As a result, no change in meal duration occurred for cows who started on L-AMS, but meal duration was longer on L-AMS for cows who started on the H-AMS.
      Lying bout frequency was greater and bout length was lesser on H-AMS, but the magnitude of the difference in treatment depended on which treatment the cows started on. Lying bout frequency decreased from period 1 to period 2, reflecting the lesser meal frequency observed in period 2. No differences in lying time occurred between the treatments, resulting in bout length also increasing in period 2. The overall greater lying bout frequency and lesser bout length observed on H-AMS is contrary to results from the studies of
      • Menajovsky S.B.
      • Walpole C.E.
      • DeVries T.J.
      • Schwartzkopf-Genswein K.S.
      • Walpole M.E.
      • Penner G.B.
      The effect of the forage-to-concentrate ratio of the partial mixed ration and the quantity of concentrate in an automatic milking system for lactating Holstein cows.
      ,
      • Schwanke A.J.
      • Dancy K.M.
      • Didry T.
      • Penner G.B.
      • DeVries T.J.
      Effects of concentrate location on the behavior and production of dairy cows milked in a free-traffic automated milking system.
      , and
      • Henriksen J.C.S.
      • Munksgaard L.
      • Weisbjerg M.R.
      Short-term responses in production and behaviour during periods o change in concentrate allowance for dairy cows.
      . Our results are unexpected, given that when cows were on the H-AMS treatment they ate less PMR and would be expected to spend less time standing at the feedbunk. Further investigation into this is warranted, particularly at higher stocking densities, as the lying stall density in the present study was 50%, which may have affected lying behavior.

      Behavior in the AMS and Milk Yield

      Cows who started on H-AMS decreased their AMS visits (and consequently their milking frequency) after switching to L-AMS in period 2, whereas cows who started on L-AMS had more consistent AMS visits and milking frequency across both treatment periods. As a result, AMS rejections followed a similar pattern, and milking frequency decreased slightly in period 2 for both groups, regardless of what treatment they started on, but remained above 3 milkings per day. As a result of decreased milking frequency, in period 2, both groups experienced a slight increase in yield per milking and box time per milking. The period × treatment interactions observed for milking frequency, AMS visits, rejections, box time per milking, and yield per milking can be attributed to changes in lactation over the course of the study, as cows decrease production toward the end of lactation, often also decreasing their milking frequency (
      • Deming J.A.
      • Bergeron R.
      • Leslie K.E.
      • DeVries T.J.
      Associations of housing, management, milking activity, and standing and lying behavior of dairy cows milked in automatic systems.
      ;
      • Tremblay M.
      • Hess J.P.
      • Christenson B.M.
      • McIntyre K.K.
      • Smink B.
      • van der Kamp A.J.
      • de Jong L.G.
      • Döpfer D.
      Factors associated with increased milk production for automatic milking systems.
      ), with corresponding increases in yield per milking and box time per milking. This could suggest that long-term milking behavior may be better established by providing greater amounts of concentrate earlier, rather than later in lactation. However, as previously discussed, stocking density in this study was low, and it would be beneficial for future studies to investigate the effect of these treatments in situations with greater AMS stocking density, as milking behavior dynamics will likely be different.
      Based on our sample size and the observed variation in milk yield, our study had sufficient power to detect a milk yield difference of 6.9%; however, the observed nonsignificant, numerical 1.6 kg/d difference in milk yield between the treatments was a difference of only 3.6%. This numerical increase when cows were on the H-AMS treatment was similar to the numerical yield difference observed by
      • Schwanke A.J.
      • Dancy K.M.
      • Didry T.
      • Penner G.B.
      • DeVries T.J.
      Effects of concentrate location on the behavior and production of dairy cows milked in a free-traffic automated milking system.
      . Given that milking frequency was similar between the 2 treatments in the present study, this numerical difference in milk yield was likely due at least in part to the greater AMS concentrate allowance, and thus greater nutrient availability for milk production, for cows on H-AMS.
      We also observed that cows who started on the H-AMS treatment had greater milk fat percentage after switching to L-AMS in period 2, but those that started on the L-AMS treatment had greater and more consistent milk fat percentage across both periods. This finding is likely because milk fat percentage increases with stage of lactation (
      • Stoop W.M.
      • Bovenhuis H.
      • Heck J.M.L.
      • van Arendonk J.A.M.
      Effect of lactation stage and energy status on milk fat composition of Holstein-Friesian cows.
      ), and because the greater forage:concentrate ratio of the diet results in greater milk fat concentration (
      • Jaakamo M.J.
      • Luukkonen T.J.
      • Kairenius P.K.
      • Bayat A.R.
      • Ahvenjärvi S.A.
      • Tupasela T.M.
      • Vilkki J.H.
      • Shingfield K.J.
      • Leskinen H.M.
      The effect of dietary forage to concentrate ration and forage type on milk fatty acid composition and milk fat globule size of lactating cows.
      ), as observed with cows on the L-AMS treatment who consumed greater amounts of PMR.
      Overall, our results indicate that increased concentrate provision in the AMS may promote more visits and milking frequency in the long term, and concentrate allowance can be later decreased while maintaining relatively high AMS visits and milking frequency. However, it appears to be more difficult to use greater concentrate allowance to promote AMS visits once milking behavior patterns are already established.

      Limitations

      To our knowledge, this work is the first to describe how individual cow personality traits may be associated with the response to AMS concentrate allocation. Although some associations were observed, these need to be interpreted in light of various limitations associated with the experimental model used, and corroborated in further studies. First, although providing adequate power to address our primary objective, the sample size used is not representative of actual stocking densities in commercial AMS herds, and additional subjects would have enabled detection of smaller treatment differences. Second, the 4-h milking permission used in the present study would not be feasible in a commercial herd, as allowing cows to visit too frequently under typical stocking densities would impede traffic flow and decrease efficiency. Third, it may have been beneficial to have a longer adaptation period to wash out effects of the previous treatment, despite the accompanying risk of period effects. Those period effects are a limitation of the crossover design used here, and, as observed, may affect the response to treatment within subject in different periods. Finally, the influence of personality on the outcome measures may differ in larger groups. As very little research exists in this area, based on the findings of this explorative study, future research on the expression of personality and its effects on behavior and production outcomes in AMS with larger group sizes is recommended.

      CONCLUSIONS

      Allocating a greater quantity of concentrate in a free-traffic AMS, while providing the same PMR, resulted in greater overall total DMI and less variable day-to-day total DMI. Despite receiving more AMS concentrate per milking, more fearful cows did not achieve the target concentrate delivery of 6.0 kg/d, and these cows also had less total DMI and greater day-to-day total DMI when allocated more AMS concentrate. This study represents a preliminary investigation laying the groundwork for further research on the relationship between dairy cow personality traits and concentrate allocation in an AMS.

      ACKNOWLEDGMENTS

      The authors thank the staff of the University of Guelph, Elora Research Station–Ontario Dairy Research Centre (Elora, ON, Canada). Special thanks to Robert Matson of the University of Guelph (Guelph, ON, Canada) for his role as the novel human in our personality testing, as well as Meagan King of the University of Guelph (Guelph, ON, Canada) for her assistance in conducting the tests, and Sarah Young of St. Lawrence College (Kingston, ON, Canada) for her help with video analysis. This project was financially supported by a Natural Sciences and Engineering Research Council of Canada (Ottawa, ON, Canada) Discovery Grant, as well as through funding from the Canada First Research Excellence Fund (Ottawa). This project also received support from the Ontario Agri-Food Innovation Alliance Research Program of the University of Guelph and the Ontario Ministry of Agriculture, Food, and Rural Affairs (Guelph). Further, equipment for this project was supported by contributions from the Canadian Foundation for Innovation (Ottawa) and the Ontario Research Fund (Toronto, ON, Canada). The authors have not stated any conflicts of interest.