The Short-Term Effects of Altering Milking Intervals on Milk Production and Behavior of Holsteins Milked in an Automated Milking System

Abstract

The widespread adoption of automatic milking systems (AMS) in the United States has afforded dairy cows the flexibility to establish personalized milking, feeding, and resting schedules. Our study focused on investigating the short-term effects of transitioning milking permissions from every 4 (MP4) to 6 (MP6) hours on the 100th day of lactation on milking frequency, milk yields, and cow behavior. Twenty-four Holstein dairy cows were divided into control (maintaining a 4 h milking interval) and test groups (transitioning to a 6 h milking interval) and observed for 6 days. The analysis revealed that parity and treatment had no significant impact on milking frequency, milk/visit, or daily milk yield. However, multiparous cows spent more time inside the commitment pen, while test group cows exhibited more tail-swishing and displacement behavior, approached the AMS more frequently, and spent longer idle times. The interaction between parity and treatment influenced heart rate variability parameters, indicating increased stress in the test group cows. Additionally, the test group cows had greater total and daytime lying frequencies, suggesting short-term behavioral modifications. Despite no immediate impact on milk production, further research is recommended to assess the potential long-term effects on milk yield in AMS farms, considering the identified stress indicators short-term.

1. Introduction

Recently, there have been notable changes in the dairy industry marked by a concomitant reduction in workforce size and a simultaneous expansion in herd size. The shift is predominantly attributed to the integration of automated technologies, which facilitate managing larger herds with less labor [1]. Among these technological innovations, Automatic Milking Systems (AMS) emerged as a preeminent innovation, gaining widespread acceptance in Europe and progressively establishing a foothold in North America. This trend reflects a reduction in labor workforce, amplifies production output, and enhances the overall quality of life for dairy farm proprietors [2,3]. The autonomous entry of cows into milking stalls is crucial for AMS efficiency and is a process that distinctly separates cows that need milking from the herd and it operates independently of farm personnel intervention [2]. There is a need to understand the intricate dynamic of interactions between cows and their surroundings. The availability of palatable concentrate is a significant motivational factor prompting cows to partake in milking, and it can increase the milking frequency [4,5]. Motivations influencing the entry into milking stalls, coupled with negative social interactions, have the potential to impede both the welfare of cows and the overall operational efficiency of AMS farms [6]. Furthermore, adverse interactions within AMS facilities correlate with augmented standing time, diminished lying time, and alterations in behaviors such as rumination and intensified agonistic behaviors [7]. These intricate dynamics contribute to complications in milk production, driven by competition for food, lying space, and access to AMS milking robots [7,8,9].

The configuration of the facility is pivotal in impacting the movement patterns of cattle within the AMS facilities. The design can directly influence critical behaviors such as feeding and lying durations, the frequency of daily milking visits, the fetching rate, and the voluntary engagement with milking robots [6]. Research has identified the significance of dedicated waiting areas situated outside AMS stalls in alleviating social competition for entry into the milking system [6,10,11,12]. These holding areas (HA) or commitment pens (CP) represent enclosed spaces antecedent to milking stalls, and allow access at their discretion, where animals wait until they enter the AMS for milking [3,5]. Notwithstanding these endeavors, the impact of CPs on AMS efficiency remains inconclusive, marked by varied research outcomes [5,11]. Additionally, AMS adopters have indicated the discernible influence of factors such as parity on system efficiency [4]. Primiparous and multiparous cows differ behaviorally within diverse traffic flow systems, resulting in impacts on milk production outcomes [4,12]. Cow parity emerges as a modulating factor in exhibiting agonistic behaviors, including displacement, blocking, and hesitation. Commonly observed in the commitment pen, these behaviors introduce potential stress before entry into milking robots [5,13]. Understanding the animal behavior within the commitment pen may vary across parity categories, potentially exerting an influence on stress levels, welfare, and milk production metrics.

The AMS technology allows cows the autonomy to structure their milking schedule, concurrently enabling farm managers the ability to tailor individualized milking intervals based on the lactation stage. As cows increase in days in milk (DIM), precision milking practices adjust milking frequencies automatically to optimize on-farm production. Research has previously demonstrated impacts of milking interval and frequency on the milk production dynamics in dairy cows [14]. In corroboration, Ref. [15] implementing AMS systems increased the milk yield based on a greater milking frequency. Other research investigated the impact of decreasing the minimum milking interval, with a greater milking frequency, on the milk yield and indicated that while reducing milking intervals did not impact milk yield, it increased the efficiency of cows’ traffic flow within the AMS setup [16].

Alterations in milking permissions have the potential to induce behavioral and/or physiological stress indicators among cows. Consequences include compromising welfare and negative impacts on milk production. Research is needed to identify the impact of precision milking management in AMS systems on animal welfare. Moreover, less is known on the intricate implications of cow behavior within commitment pens (CP), on the aggregate milk yield, and on the overall well-being of the cows.

Our current study aims to identify short-term animal behavior responses to changes in precision milking settings. We hypothesized that manipulating milking permissions on the 100th DIM would increase short-term behavioral indicators of stress, consequently reducing milk production. The main objective of this study was to examine the effects resulting from shifting the milking interval from 4 h to 6 h at 100 days in milk (DIM) on performance, behavior, heart rate variability, and activity patterns of cows within an AMS. Additionally, we hypothesized that reducing the number of milkings and increasing the time between milkings may impact cow production and activity and increase stress-related behaviors, particularly in multiparous compared to primiparous cows.

2. Materials and Methods

2.1. Ethics

Approval for all protocols was sought and granted by the Clemson University Institutional Animal Care and Use Committee (AUP#2021-0064) before the study began.

2.2. Animals and Housing

In this study, we selected twenty-four lactating Holstein dairy cows from a larger group of 120 lactating cows at the LaMaster Dairy Farm in Clemson, SC, USA. The cows were housed in a guided traffic AMS facility (Delaval VMS V300, Combi, Tumba, Sweden). Criteria for inclusion included choosing cows at day 97 of lactation, and the study took place over six consecutive days, concluding on day 102 of lactation. These cows were evenly distributed into two groups (n = 12/group) in separate pens, ensuring a balance in lactation stage, and each group had access to two milking robots.

Groups were housed in separate pens. Group 1 consisted of 12 primiparous (PR) cows and group 2 contained 12 multiparous (MU) cows. Each parity group was further divided into two milk permission treatment groups, comprised of 6 cows randomly assigned as treatment (T) and 6 cows assigned as control (C). Focal cows assigned to the treatment group (T) changed in the milking permission schedule from every 4 h to every 6 h at 100 DIM. Cows assigned to the C group did not transition in milking permission at 100 DIM and remained with a milking interval of every 4 h for the entire duration of the trial. Thus, there were 6 cows enrolled in each of the four treatment groups: PC (primiparous control: cows in the 1st lactation permitted to milk every 4 h with no change in milking permission), PT (primiparous treatment: cows in the 1st lactation and milk permission transitioned at 100 DIM), MC (multiparous control: cows in the ≥2nd lactation permitted to milk every 4 h with no change in milking permission), MT (multiparous treatment: cows in the ≥2nd lactation and milk permission transitioned at 100 DIM). Cows were selected for inclusion in the study with the criteria that they would undergo lactation throughout the entire trial period.

The design of the entrance and exit gates in the guided traffic AMS for managing cow traffic flow was uniform across both groups. Both AMS units utilized left-handed milking robots, with the holding area equipped with an automatic selection gate that allowed access to the commitment pen (CP) of the AMS. The gates leading to the CP opened only when cows were scheduled based on their last milking time, while exit gates opened toward the feed alley after milking was completed. Two selection gates were used: one allowed entry to the holding area next to the CP, and the other facilitated exit from the CP after milking. All gates and alleys were of equal length for both groups. Both groups experienced identical conditions with respect to the AMS model (Delaval VMS V300), feeding alley design, stall layouts, and management practices. The AMS operated continuously, performing three 20 min cleaning cycles at 0400, 1230, and 2000 h daily. Feeding of a partial mixed ration (PMR) took place twice daily at 0730 and 1530 h, and the diet feed, chemical composition, and feeding schedule were the same for both groups. During the milking process, cows were offered a concentrate pellet, the quantity of which depended on the individual projected milk yield. Footbaths, containing a copper sulfate solution, were located in the CP exit alley.

2.3. Treatments and Experimental Procedure

For each group, behavior was recorded continuously for six consecutive days using six Wyze Cam v3 cameras. The cameras were strategically positioned to capture activity in the vicinity of the AMS, focusing on the CP in front of the AMS, CP entry gate, the AMS itself, and the entrance and exit gates of the AMS. To distinguish focal cows, specific colored paint markings were applied across their backs. Distinctive colored livestock paint, extending from the shoulder to the tail head, was utilized to identify groups. Primiparous cows and multiparous cows were identified by different colored spray paint. The focal cows were outfitted with a Polar H10 heart rate sensor, a data transmission watch, and an accelerometer (Onset HOBO PendantG acceleration data loggers) to measure heart rate variability (HRV) and activity data.

2.3.1. Milk Transition, Production, and Frequency

At 100 days in milk (DIM), a scheduled alteration in milking permission was implemented at 0600 for the specified cows. Focal cows, categorized into treatment (T) and control (C) groups within each parity group, experienced a transition in milking frequency. For the T group, milking permission shifted from every 4 h to every 6 h, while cows in the C group maintained their existing milking permission at 100 DIM. To visually denote the treatment assignment, focal cows were marked with livestock spray paint. In the T group, cows were identified by green livestock paint across their hock bones, while C group cows remained unpainted in this region. Subsequently, the daily milking visits, total milk per visit (kg), and daily milk yield (kg) for each cow were diligently documented to support subsequent statistical analyses.

2.3.2. Commitment Pen Behavior

The time it took for focal cows to transition from the holding area into the commitment pen was measured from the opening to the closing of the selection gate during each milking visit. Behavioral recordings were documented throughout the entire period that the cows spent inside the CP before entering the milking robot. Observations started when cows entered the selection gate, and the exact time of entry was noted [17,18,19,20,21,22]. A recording of each cow was made for all exhibited behaviors, including exploration, displacement, idling, and tail swishing, among others (please see [23] Table 1.1). The number of events and duration of each behavior was documented. Displacement events inside the CP were also recorded for focal cows. Displacement was defined as an interaction where one cow (the actor DP-A) initiated physical contact, causing another cow (the reactor) to back out or vacate the AMS entry alley, resulting in a loss of their place in line for milking [17,18]. The total time cows spent inside the CP was recorded immediately after they entered the AMS. The observed behaviors and recorded data are comprehensively detailed in the ethogram of CP behaviors (for details please see Table 1.1 [23]).

2.3.3. AMS Behavior

External cameras were strategically placed outside each milking stall, primarily focusing on the cow’s hindlegs. Behavioral observations within the milking stall included recording instances of focal cows stepping, kicking, lifting hooves, and idling [19,24], and others (refer to Table 1.2 [23]). For each successful milking event, the duration and occurrence of each behavior, as well as the total milk yield per visit, were recorded. The durations of individual events were combined to calculate the overall duration for each event type within each 24 h period over the course of 6 days. The time taken for the cow to exit the AMS milking stall was measured from the moment the AMS exit gate opened to when it closed behind the focal cow (as per [6]). All observed behaviors and recorded data are comprehensively outlined in the ethogram of AMS behaviors (for details please see [23] Table 1.2).

2.3.4. Heart Rate Variability (HRV)

For heart rate (HR) and heart rate variability (HRV) measurements, a Polar heart rate monitor was utilized. This monitor included an electrode chest belt equipped with a built-in Polar H10 HR sensor, which was securely fastened around the chest of the focal cows, positioning the HR sensor near the left elbow. To improve conductivity and data transmission, the electrodes were coated with a multipurpose lubricant. A wristwatch receiver connected to individual HR sensors was attached to the cow’s collar. Belts and wristwatches were affixed to the focal cows 12 h prior to observations to facilitate habituation. HRV data collection commenced at 2400 h on the first day and continued through the sixth day of observation. The HRV monitor captured changes in electrical potential, detecting maximum, minimum, and average HR, along with beat-to-beat (RR) intervals. Data transmission occurred from the HR sensor transmitter to wristwatch receivers, which received and organized the data. The collected data were downloaded onto a computer using a Polar interface (Polar ProTrainer Equine edition) for subsequent statistical analysis. Kubios HRV software, version 2.2, was used for HRV analysis. The data were detrended to eliminate long-term trend components, and artifact correction was applied followed established procedures by [25]. The study extracted the most informative time and frequency domain measures for comprehensive HRV analysis (for additional details, see [23]).

Heart rate (HR), measured in beats per minute (BPM), indicates the average heart rate over the recorded period. Heart rate variability (HRV) represents the irregular time intervals between successive heartbeats, caused by rhythmic oscillations in the components regulating cardiac activity. This rhythmic oscillation, as highlighted by [26,27], serves to maintain cardiovascular homeostasis within a specified range and synchronize responses to challenges. HRV offers insight into the contrasting effects of the sympathetic and vagal branches of the autonomic nervous system (ANS) on the heart’s sinus node [28]. In this study, HRV variables such as the root mean square of successive beat-to-beat intervals (RMSSD) and the standard deviation of the beat-to-beat interval (SDRR) were calculated by determining the differences between consecutive inter-beat intervals, squaring and summing these differences, and then averaging and taking the square root of the results [29,30]. A decrease in HRV values suggests a shift toward sympathetic dominance, indicating a stress response, whereas higher values imply parasympathetic dominance or a state of normalcy [31]. Furthermore, HRV data were analyzed using the autoregressive model of power spectral analysis. This method involved estimating the spectral distribution of intervals between heartbeats. The low-frequency (LF) component indicates sympathetic activity, while the high-frequency (HF) component reflects vagal activity. The LF/HF ratio reflects the sympathovagal balance (SVB). In this study, the HR and HRV data were categorized by the values when the cows were inside the commitment pen (CP) and inside the AMS milking robots.

2.3.5. Activity Data

To monitor the activity of the focal cows, an acceleration data logger (Onset HOBO PendantG acceleration data loggers, Onset Computer Corporation, Bourne, MA, USA) was securely affixed to the HRV chest belt on the right dorsal side near the withers, ensuring stable positioning of the logger. Activity data were collected continuously from 2400 h on the first day of observation through the sixth day. The loggers were attached concurrently with the HRV chest belts on the focal cows. Accelerometers were oriented to capture various movements: the Y-axis recorded lateral movements, the X-axis recorded forward and backward movements, and the Z-axis captured vertical movements. Various daily activity parameters were documented and categorized based on daytime hours (0630–1830 h), nighttime hours (1831–0629 h), and total daily activity. These parameters included cow lying time (minutes), lying frequency, duration of lying bouts (minutes), idle time (minutes), and acceleration activity (g) for each cow. Raw accelerometer data, which included date, time, and corresponding impulses in the X, Y, and Z dimensions, were extracted from the devices (HOBOware Graphing & Analysis Software 12.0, Bourne, MA, USA) at the end of each 6-day observation period. Information on cows’ vertical (az: dorso-ventral movement across vertical levels), horizontal (ax: craniocaudal movement within the same vertical level), and lateral movement (ay: mediolateral movement within the same vertical level) during daylight hours was directly retrieved from the loggers. Acceleration data underwent post-processing using MATLAB (MATLAB and Statistics Toolbox Release 2012, The MathWorks, Inc., Natick, MA, USA). After data processing, lying events were identified by detecting significant shifts in acceleration in the z-axis of activity. To accurately identify changes in acceleration linked to lying down and to establish thresholds for minor fluctuations in the z-axis, timestamped videos of cows in a lying position were obtained and compared with the corresponding activity data. This method enabled the identification of changes in z-axis acceleration due to lying down, with non-lying fluctuations serving to establish threshold levels for data smoothing (for further details [23]).

2.3.6. Statistics

Statistical analyses for milk yield, AMS behavior, CP behavior, cow activity, and HRV were performed using JMP Pro 16.1.0, with statistical significance defined as p ≤ 0.05. Initially, a Shapiro–Wilk’s test (p > 0.05) and visual inspection of histograms were employed to confirm the normal distribution of data across all measurements. Subsequently, to explore the effect of treatment and parity on response variables (milk production, HRV, AMS and CP behaviors, and activity) throughout the six days of observations (comprising the initial 3 days of pre-milk transition and the subsequent 3 days of post-milk transition), general linear mixed models were formulated. These models associated the response variables (e.g., daily milk yield, idle time, lying time) with the fixed effects of parity, treatment, day of observation, and their interactions. The random effect of individual cow was also included.

To account for repeated measures, they were adjusted to account for the possibility of pre-existing differences in the cows in the parity and T groups. The repeated measures were adjusted by averaging response variables into a “pre-value” (the average of the first three days) and a “post value” (the average of the last the three days) for each cow. General linear models were adjusted to use the “post-value” as the response and the “pre-value” as a covariate. When treatment, parity, or their interaction showed significant effects on the response variables, the impact of these effects was assessed using T tests (specifically Fisher’s Protected Least Significant Difference Test). Since data collection involved multiple observers, the inter-observer reliability with the primary observer was evaluated using Cohen’s kappa agreement coefficient (K) following the approach outlined in [32]. During the observer training period prior to data collection, inter-observer reliability was assessed. Trainees observed the same videos concurrently, and there was a high level of inter-observer agreement (Kappa = 0.96, p < 0.001, CI [0.90, 0.99]).

3. Results

3.1. Milking Production and Frequency

Parity and treatment interaction had no significant impact on milking frequency, milk/visit, or daily milk yield in the PC vs. PT or MC vs. MT treatment groups (

Table 1. Effect of the interaction between parity and treatment ¹ (transition in milking permission) on daily milk robot visit and milk yield.

Parameter	PC	PT	MC	MT
Milk robot visit (frequency)	2.81 ± 0.10 a (1.0, 3.0)	2.17 ± 0.06 a (1.0, 3.0)	2.56 ± 0.08 a (1.0, 5.0)	2.52 ± 0.10 a (1.0, 4.0)
Milk yield/visit (kg)	13.26 ± 0.37 a (10.83, 16.56)	15.72 ± 0.81 a (8.36, 23.68)	19.26 ± 0.70 a (13.11, 25.60)	18.28 ± 0.36 a (15.35, 20.61)
Milk yield/day (kg)	33.50 ± 1.05 a (26.82, 45.99)	31.55 ± 1.12 a (21.34, 38.17)	49.05 ± 0.53 a (45.32, 54.02)	46.05 ± 1.39 a (32.49, 55.34)

^a Data are presented as means ± SEM (min, max), different superscripts across the same row indicate statistical significance (p ≤ 0.05).¹ Parity and treatment: PC (Primiparous Control: cows in 1st lactation permitted to milk every 4 h with no change on 100 DIM), PT (Primiparous Treatment: cows in 1st lactation and milk permission transitioned from every 4 h to every 6 h on DIM 100), MC (Multiparous Control: cows in ≥2nd–3rd lactation permitted to milk every 4 h with no change on 100 DIM), MT (Multiparous Treatment: cows in ≥2nd–3rd lactation and milk permission transitioned from every 4 h to every 6 h on DIM 100).

).

3.2. Commitment Pen Behavior

There was a parity by treatment interaction on total CP time (

Table 2. Effect of the interaction between parity and treatment ¹ on cow behaviors inside the AMS commitment pen.

Variable	PC	PT	MC	MT
CP time (m)	68.63 ± 5.39 a (30.6, 131.7)	75.63 ± 4.03 a (32.8, 96.3)	26.23 ± 3.95 b (4.3, 78.3)	15.39 ± 1.60 c (4.1, 30.3)
Entry duration (s)	12.62 ± 2.41 a (2.3, 54.7)	12.75 ± 1.22 a (5.6, 25.7)	7.43 ± 0.42 b (3.3, 12.6)	6.85 ± 0.18 b (5.3, 8.2)
Approach AMS	2.39 ± 0.21 a (0.5, 5.5)	2.97 ± 0.20 a (0.5, 6.0)	3.45 ± 0.30 a (1.0, 8.0)	4.56 ± 0.29 b (0.5, 8.5)
Tail swish	3.31 ± 0.39 a (0.0, 14.3)	4.53 ± 0.54 a (0.0, 24.6)	4.63 ± 0.60 a (0.0, 17.0)	6.76 ± 0.76 b (0.0, 28.9
Idle time (m)	22.52 ± 2.37 a (1.9, 63.2)	23.56 ± 1.98 a (2.2, 67.6)	19.69 ± 1.67 a (2.7, 46.3)	28.96 ± 2.20 b (2.4, 44.5)
Exploration	4.24 ± 0.71 a (0.0, 10.4)	5.04 ± 0.54 a (0.0, 11.7)	4.04 ± 0.50 a (0.0, 14.4)	4.66 ± 0.71 a (0.0, 14.9)
Displacement-actor	3.41 ± 0.31 a (0.0, 6.5)	4.63 ± 0.21 a (0.5, 10.5)	5.01 ± 0.17 a (2.0, 10.0)	6.95 ± 0.19 b (2.5, 14.0)

^a,b,c Data are presented as means ± SEM (min, max). Different superscripts across the same row indicate statistical significance (p ≤ 0.05). ¹ Parity and treatment: PC (Primiparous Control: cows in 1st lactation permitted to milk every 4 h with no change on 100 DIM), PT (Primiparous Treatment: cows in 1st lactation and milk permission transitioned from every 4 h to every 6 h on DIM 100), MC (Multiparous Control: cows in ≥2nd lactation permitted to milk every 4 h with no change on 100 DIM), MT (Multiparous Treatment: cows in ≥2nd lactation and milk permission transitioned from every 4 h to every 6 h on DIM 100).

). Cows in the MC group spent more time inside the CP than cows in the MT group (p = 0.026). However, cows in the MT group had greater approach AMS frequencies, tail-swishing behavior, DP-A behavior, and longer idle times than cows in the MC group (p = 0.024, 0.031, 0.029, 0.015; Table 2). There was no treatment by parity interaction on any CP behavior variables in PC cows compared to the PT treatment group. Similarly, there was no effect of treatment and parity interaction on CP entry duration or exploration behavior in the MC vs. MT treatment group.

3.3. AMS Behavior

The treatment and parity interaction affected the total AMS-box time and the stepping frequency in MC and MT cows (

Table 3. Effect of the interaction between parity and treatment ¹ on cow behaviors inside the AMS milking robot.

Variable	PC	PT	MC	MT
AMS-box time (m)	6.31 ± 0.17 a (4.1, 9.3)	7.49 ± 0.20 a (6.3, 9.2)	8.65 ± 0.17 a (11.9, 1.0)	11.06 ± 0.19 b (9.1, 13.9)
Stepping frequency	9.03 ± 0.77 a (4.0, 16.0)	10.13 ± 1.83 a (1.0, 30.0)	9.46 ± 1.09 a (19.0, 6.5)	15.55 ± 1.42 b (7.0, 31.0)
Kicking frequency	0.11 ± 0.03 a (0.0, 0.33)	0.20 ± 0.07 a (0.0, 1.0)	0.35 ± 0.09 a (1.2, 0.5)	0.86 ± 0.27 a (0.0, 4.2)
Idle time (m)	3.82 ± 0.31 a (1.8, 7.1)	4.09 ± 0.33 a (0.5, 5.8)	3.21 ± 0.31 a (5.5, 1.9)	3.49 ± 0.25 a (1.8, 5.3)
Exit duration (s)	58.07 ± 14.47 a (5.9, 232.8)	21.76 ± 2.79 b (8.2, 45.3)	18.20 ± 1.65 b (32.4, 9.9)	23.69 ± 4.87 b (5.4, 82.2)

^a,b Data are presented as means ± SEM (min, max). Different superscripts across the same row indicate statistical significance (p ≤ 0.05). ¹ Parity and treatment: PC (Primiparous Control: cows in 1st lactation permitted to milk every 4 h with no change on 100 DIM), PT (Primiparous Treatment: cows in 1st lactation and milk permission transitioned from every 4 h to every 6 h on DIM 100), MC (Multiparous Control: cows in ≥2nd lactation permitted to milk every 4 h with no change on 100 DIM), MT (Multiparous Treatment: cows in ≥2nd lactation and milk permission transitioned from every 4 h to every 6 h on DIM 100).

). MT cows spent more time inside the AMS and exhibited more stepping than MC cows (p = 0.019, 0.032). On the other hand, PC cows recorded longer durations exiting the milking robots compared to PT cows (p = 0.015). However, parity and treatment interaction had no significant impact on idle time or kicking frequency (Table 3).

3.4. Heart Rate Variability

3.4.1. Commitment Pen HRV

The interaction between parity and treatment revealed an influence on RMSSD, SDRR, and LF/HF ratio values in MC compared to MT cows. Cows in the MT group had lower RMSSD and SDRR values than cows in the MC group (p = 0.016, 0.012), but MT cows showed a higher LF/HF ratio than MC cows (p = 0.017). However, there was no significant impact of treatment and parity interaction on any HRV parameters or heart rate values while inside the CP prior to milking (

Table 4. Effect of the interaction between parity and treatment ¹ on cow HR and HRV parameters inside the AMS commitment pen.

Parameter	PC	PT	MC	MT
HR (beat/min)	81.25 ± 1.27 a	82.06 ± 1.16 a	77.58 ± 1.26 a	79.96 ± 1.33 a
RMSSD 2 (ms)	13.02 ± 0.95 a	10.63 ± 0.76 a	16.89 ± 0.78 b	8.59 ± 0.54 c
SDRR 3 (ms)	26.85 ± 1.45 a	20.96 ± 1.09 a	31.69 ± 0.98 b	22.89 ± 0.72 c
LF/HF 4	7.21 ± 0.28 a	8.15 ± 0.27 a	4.53 ± 0.17 b	6.89 ± 0.22 c

^a,b,c Data are presented as means ± SEM. Different superscripts across the same row indicate statistical significance (p ≤ 0.05). ¹ Parity and treatment: PC (Primiparous Control: cows in 1st lactation permitted to milk every 4 h with no change on 100 DIM), PT (Primiparous Treatment: cows in 1st lactation and milk permission transitioned from every 4 h to every 6 h on DIM 100), MC (Multiparous Control: cows in ≥2nd lactation permitted to milk every 4 h with no change on 100 DIM), MT (Multiparous Treatment: cows in ≥2nd lactation and milk permission transitioned from every 4 h to every 6 h on DIM 100). ² RMSSD: Root mean square of successive beat-to-beat differences. ³ SDRR: Standard deviation of beat-to-beat intervals. ⁴ LF/HF: Low Frequency/High Frequency ratio, resembles the sympathovagal balance (SVB).

).

3.4.2. Miking Robot HRV

Parity and treatment interaction had no impact on any HRV parameters while cows were milking inside the AMS (

Table 5. Effect of the interaction between parity and treatment ¹ on cow HR and HRV parameters inside the AMS milking robot.

Parameter	PC	PT	MC	MT
HR (beat/min)	78.85 ± 1.40 a	83.69 ± 2.10 a	80.58 ± 1.42 a	79.58 ± 1.06 a
RMSSD 2 (ms)	12.85 ± 0.60 a	8.88 ± 0.98 a	16.85 ± 0.71 a	13.58 ± 0.60 a
SDRR 3 (ms)	20.02 ± 1.09 a	16.21 ± 1.60 a	26.55 ± 1.26 a	23.88 ± 1.09 a
LF/HF 4	7.29 ± 0.20 a	9.13 ± 0.25 a	6.32 ± 0.21 a	7.56 ± 0.14 a

^a Data are presented as means ± SEM. Different superscripts across the same row indicate statistical significance (p ≤ 0.05). ¹ Parity and treatment: PC (Primiparous Control: cows in 1st lactation permitted to milk every 4 h with no change on 100 DIM), PT (Primiparous Treatment: cows in 1st lactation and milk permission transitioned from every 4 h to every 6 h on DIM 100), MC (Multiparous Control: cows in ≥2nd lactation permitted to milk every 4 h with no change on 100 DIM), MT (Multiparous Treatment: cows in ≥2nd lactation and milk permission transitioned from every 4 h to every 6 h on DIM 100). ² RMSSD: Root mean square of successive beat-to-beat differences. ³ SDRR: Standard deviation of beat-to-beat intervals. ⁴ LF/HF: Low Frequency/High Frequency ratio, resembles the sympathovagal balance (SVB).

).

3.5. Activity Data

3.5.1. Lying Time

Parity and treatment interaction showed no effect on day, night, or total lying time (

Table 6. Effect of the interaction between parity and treatment ¹ on cow lying time during daytime hours (0630–1830 h), nighttime hours (1831–0629 h), and total daily lying time.

Variable	PC	PT	MC	MT
Total (m)	711.33 ± 16.39 a (395.2, 942.8)	743.23 ± 18.50 a (443.5, 1000.4)	748.16 ± 14.23 a (445.9, 1001.9)	601.94 ± 10.75 b (382.6, 817.7)
Day (m)	344.75 ± 10.25 a (159.4, 465.5)	328.14 ± 10.08 a (140.6, 453.0)	343.13 ± 12.37 a (148.4, 460.5)	251.77 ± 9.21 b (99.5, 344.8)
Night (m)	364.13 ± 7.39 a (268.4, 474.4)	399.11 ± 9.81 a (243.5, 555.8)	386.23 ± 10.41 a (251.6, 556.0)	338.67 ± 8.32 a (164.1, 493.9)

^a,b Data are presented as means ± SEM (min, max). Different superscripts across the same row indicate statistical significance (p ≤ 0.05). ¹ Parity and treatment: PC (Primiparous Control: cows in 1st lactation permitted to milk every 4 h with no change on 100 DIM), PT (Primiparous Treatment: cows in 1st lactation and milk permission transitioned from every 4 h to every 6 h on DIM 100), MC (Multiparous Control: cows in ≥2nd lactation permitted to milk every 4 h with no change on 100 DIM), MT (Multiparous Treatment: cows in ≥2nd lactation and milk permission transitioned from every 4 h to every 6 h on DIM 100).

). However, cows assigned to the MC group showed significantly longer total and daytime lying times than cows in the MT group (p = 0.021, 0.032).

3.5.2. Lying Frequency

The interaction between parity and treatment influenced total and daytime lying frequency in the MC vs. MT groups. Cows in the MT group were observed lying more frequently than cows in the MC group for total and daytime periods (p = 0.035, 0.027). However,

Table 7. Effect of the interaction between parity and treatment ¹ on cow lying frequency during daytime hours (0630–1830 h), nighttime hours (1831–0629 h), and total daily lying frequency.

Variable	PC	PT	MC	MT
Total	9.63 ± 0.21 a (5.2, 11.4)	8.65 ± 0.33 a (7.4, 17.4)	6.21 ± 0.20 b (5.4, 11.1)	8.22 ± 0.21 a (5.0, 11.3)
Day	5.96 ± 0.09 a (1.9, 6.8)	6.65 ± 0.15 a (3.6, 10.5)	3.36 ± 0.07 b (1.3, 6.5)	5.03 ± 0.09 a (1.9, 6.7)
Night	2.82 ± 0.05 a (1.4, 4.8)	3.65 ± 0.07 a (1.9, 6.6)	2.87 ± 0.06 a (1.5, 5.2)	3.16 ± 0.05 a (1.3, 4.7)

^a,b Data are presented as means ± SEM (min, max). Different superscripts across the same row indicate statistical significance (p ≤ 0.05). ¹ Parity and treatment: PC (Primiparous Control: cows in 1st lactation permitted to milk every 4 h with no change on 100 DIM), PT (Primiparous Treatment: cows in 1st lactation and milk permission transitioned from every 4 h to every 6 h on DIM 100), MC (Multiparous Control: cows in ≥2nd lactation permitted to milk every 4 h with no change on 100 DIM), MT (Multiparous Treatment: cows in ≥2nd lactation and milk permission transitioned from every 4 h to every 6 h on DIM 100).

shows no parity and treatment interaction effect were found on either the lying frequency variables in the PC vs. PT groups or on nighttime lying frequency in the MC vs. MT groups.

3.5.3. Lying Bouts Duration

Parity and treatment interaction influenced total and daytime lying bouts duration (

Table 8. Effect of the interaction between parity and treatment ¹ on cow lying bouts duration during daytime hours (0630–1830 h), nighttime hours (1831–0629 h), and total daily lying bouts durations.

Variable	PC	PT	MC	MT
Total (m)	76.64 ± 4.76 a (34.4, 131.3)	83.36 ± 5.59 a (35.3, 177.8)	96.55 ± 4.40 b (44.9, 140.3)	70.19 ± 3.55 a (31.3, 124.0)
Day (m)	35.62 ±2.89 a (15.1, 73.8)	43.66 ± 3.79 a (12.4, 93.1)	45.56 ± 2.93 a (23.0, 84.2)	32.67 ± 2.50 b (14.3, 69.3)
Night (m)	40.67 ± 3.72 a (27.4, 89.1)	48.11 ± 4.78 a (29.9, 110.2)	50.37 ± 3.59 a (35.9, 104.9)	43.05 ± 2.99 a (26.3, 84.9)

^a,b Data are presented as means ± SEM (min, max). Different superscripts across the same row indicate statistical significance (p ≤ 0.05). ¹ Parity and treatment: PC (Primiparous Control: cows in 1st lactation permitted to milk every 4 h with no change on 100 DIM), PT (Primiparous Treatment: cows in 1st lactation and milk permission transitioned from every 4 h to every 6 h on DIM 100), MC (Multiparous Control: cows in ≥2nd lactation permitted to milk every 4 h with no change on 100 DIM), MT (Multiparous Treatment: cows in ≥2nd lactation and milk permission transitioned from every 4 h to every 6 h on DIM 100).

). Cows in the MC group exhibited longer lying bouts durations than cows in the MT group for both total and daytime hours (p = 0.021, 0.037). However, parity and treatment interaction had no impact on any lying bouts duration variables in the PC vs. PT groups or nighttime lying bouts duration in the MC vs. MT groups (Table 8).

3.5.4. Idle Time

Treatment and parity interaction influenced the total idling time in the MC vs. MT groups, as cows in the MC group stood idle longer than cows in the MT group (p = 0.026). However, there was no significant impact of treatment and parity interaction on any idle time variables in the PC vs. PT groups or on day and night idle time in the MC vs. MT groups (

Table 9. Effect of the interaction between parity and treatment ¹ on cow idle time during daytime hours (0630–1830 h), nighttime hours (1831–0629 h), and total idle time.

Variable	PC	PT	MC	MT
Total (m)	132.23 ± 2.37 a (86.4, 180.6)	129.63 ± 2.54 a (81.6, 189.7)	123.65 ± 2.45 a (65.2, 189.3)	145.63 ± 2.50 b (78.9, 195.7)
Day (m)	64.58 ± 1.84 a (20.6, 96.3)	76.65 ± 2.05 a (24.7, 101.3)	61.36 ± 1.62 a (17.4, 84.2)	72.53 ± 2.09 a (26.2, 95.8)
Night (m)	54.26 ± 1.71 a (36.6, 89.6)	53.69 ± 1.99 a (15.6, 91.7)	55.87 ± 1.70 a (26.8, 87.6)	51.70 ± 1.94 a (30.4, 94.4)

^a,b Data are presented as means ± SEM (min, max). Different superscripts across the same row indicate statistical significance (p ≤ 0.05). ¹ Parity and treatment: PC (Primiparous Control: cows in 1st lactation permitted to milk every 4 h with no change on 100 DIM), PT (Primiparous Treatment: cows in 1st lactation and milk permission transitioned from every 4 h to every 6 h on DIM 100), MC (Multiparous Control: cows in ≥2nd lactation permitted to milk every 4 h with no change on 100 DIM), MT (Multiparous Treatment: cows in ≥2nd lactation and milk permission transitioned from every 4 h to every 6 h on DIM 100).

).

4. Discussion

Dairy farmers focus their efforts on increasing and improving milk yield through implementing mechanized operations and automated technology like AMSs. The main advantage of AMS introduction is the reduction in labor costs, as farm labor is only second to livestock feed in terms of monthly costs [33]. Additionally, automatic milking aims to increase milk productivity as costs of production decrease [34]. To achieve success in AMS farms, farmers must maximize milking visits through precision milking management, maximize milk yield per AMS, and reduce fetching by promoting voluntary visits to milking stalls [35]. The AMS facility traffic patterns need to be developed to improve system efficiency, increase voluntary milking frequencies, and reduce the fetching rate. Thus, traffic type is critical to consider in AMS facilities as it can negatively impact milk production and cause long-term economic consequences [36].

Previous research investigating the effect of changes in milking frequency observed reduced milk yield [3]. Several of these studies evaluated milking intervals in conjunction with feed concentrates provided while milking in an AMS [37,38] and observed that reducing milking permissions and feed concentrates provided at each milking resulted in a significant drop in milk production. Opposingly, Ref. [39] reported that transitioning from 2 to 3 milking permissions per day increased daily milk yield by 14%. Another researcher predicted an increase in milk yield by 3.5 kg when milking permission increased from 2 to 3, and an increase of 4.9 kg after changing the milking frequency from 2 to 4 times daily [40]. However, minimal research has evaluated the impact of decreasing milking permission on milk production, cow behavior, and welfare on AMS farms. Therefore, in this study, we implemented a decrease in milking permission at 100 DIM for lactating Holstein dairy cows on an AMS farm. We evaluated the impact of parity, transition in milking permission, and the interaction of parity and treatment on milk production and frequency, cow behavior, heart rate and heart rate variability parameters inside the CP and AMS milking robot, and multiple cow activity variables.

4.1. Milk Production and Frequency

In the current study, cows enrolled in the test groups experienced a change in milking permission from every 4 h to every 6 h at 100 DIM. The transition in milking permission did not affect daily milk yield, milk yield per visit, or milking frequency. Unlike the current study, Ref. [41] discovered shorter milking intervals contributed to higher milk yield/cow per hour and lower milk flow rates. However, Ref. [14] reported a 2% increase in quarter milk production for each 1 h decrease in milking interval (increased milking permission) when Holstein cows were milked in an AMS. Cows were [42] reported to adapt to a decrease in the milking interval without any adverse effects. Therefore, due to variations in individual cow milking intervals over each 24 h period as discussed by [14,43], the current study may reflect the difficulty of accurately estimating daily milk production on AMS farms. The variation may explain why we did not observe a significant impact of the milking interval on the daily milk yield, as we recorded the daily milk yield for each 24 h period rather than considering the total milk yield based on individually set milking intervals.

4.2. Cow Behavior inside AMS Commitment Pen

Transition in milking permission influenced tail-swishing and displacement behaviors as multiparous cows in the control group spent longer durations inside the CP than multiparous test cows. Alternatively, the MT group had greater AMS approach attempts, tail-swishing, idle time, and displacement expression than MC cows. Thus, C cows waited longer to milk than T cows that experienced a decrease in milking permission. Previously, it was [44] reported that a gradual reduction in milking frequency, resulting in reduced time spent waiting to milk. This suggests that decreasing milking permission increases cow motivation to enter milking stalls as a result of waiting longer to enter the AMS milking area. Tail switching and displacement are behaviors indicative of cow annoyance or irritation while waiting to enter the AMS milking stalls [44,45]. Prevention of highly motivated behavior, like milking, results in frustration and increased displacement activities [44,46]. However, we found the frequency of tail-swishing was significant, where cows in the T group displayed greater tail-swishing behavior than C cows. One study reported that tail-swishing could indicate the presence of painful stimuli [47]. We presume cows in the T group could have been experiencing elevated discomfort levels due to increased udder pressure in response to decreased milking permission as evidenced by greater tail-swishing in cows enrolled in the T group. Moreover, extended idle time might suggest the influence of social competition or lack of motivation to enter the milking robot [48].

Cows in early lactation were [49] reported to spend the shortest time waiting to be milked. Cows with a higher daily milk yield also spent less time waiting to milk than lower-yielding cows. Therefore, PR cows experience greater aggregation in farms that utilize CPs in front of AMS milking stalls.

4.3. Cow Behavior inside AMS Milking Robot

Transition in milking permission impacted the stepping rate while milking in the AMS. These findings suggest cows in the T group experienced higher levels of stress in response to the decrease in milking permission at 100 DIM. These cows also spent more time milking in the AMS and may have experienced greater discomfort based on greater stepping frequencies during milking. Previously, Ref. [50] indicated stepping during milking may occur in response to discomfort, often observed in high-yielding cows. Similarly, Ref. [51] also reported stepping behavior while milking is expressed more in anxious cows. Based on previous and current findings, an increase in stepping rate after a decrease in milking permission can negatively impact milk yield and cow welfare and should be considered an essential factor in AMS success.

Transition in milking permission had an opposite effect on AMS exit duration, as the cows in the C group had longer exit times than the T cows. Therefore, cows that did not experience any change in milking permission loitered inside the milking robot after milking completion for unknown reasons. Few studies have found an influence of decreased milking permission on AMS exit duration, albeit [52] confirmed it as a concern in AMS farms. Reducing the lag time between one cow exiting the milking stall and another entering is one of the main areas for improvements in AMS [52].

4.4. Heart Rate Variability in Commitment Pen

Prior studies have noted significant variations in HRV parameters between periods when cows are awaiting milking and inside the AMS. For example, Ref. [53] documented increased sympathetic tone before milking in the AMS, indicating reduced stress levels among cows. Minimal research has focused on the impact of changes in milking permission on cow HRV stress parameters while waiting in the CP to be milked. In the current study, cows in the MT group expressed lower RMSSD and SDRR values and higher LF/HF ratios than cows in the MC group, suggesting cows in the MT group were more stressed than MC cows. While no previous studies reflect the current findings, Ref. [42] discussed that cow traffic impacts milking frequency and may reduce the expression of normal cow behavior. The reduction in normal cow behavior by decreasing milking permission in this study might have prompted the increased stress parameters observed in T cows. Another possible explanation might be the increased cow stress after milking refusals, as reported by [3]. Since cows in the MT group experienced a decrease in milking permission at 100 DIM, these cows were no longer allowed to milk as frequently as they could before the milking transition; thus, it is likely that they had a higher rate of milking refusal. However, we did not record the number of unsuccessful attempts to enter the CP in this study.

4.5. Heart Rate Variability in AMS

Cows in the T group who experienced a decrease in milking permission showed significantly lower SDRR values and higher LF/HF ratios than cows in the C group. Although there was no statistical significance in RMSSD values, we did find RMSSD values were lower in cows enrolled in the T group. The results indicate that cows in the T group were more stressed during milking than C cows. The increase in the milking interval at 100 DIM prevented these cows from milking as frequently as they would prefer, thus inducing stress during the milking process. Previous research observed lower RMSSD and SDRR values associated with a higher milking frequency and larger group sizes [53]. Their results suggest housing and individual-related variables can influence stress in AMS stalls. However, we observed the opposite effect, where T cows had decreased milking frequencies and increased stress while milking. The variation in the findings suggests additional research needs to be conducted to better understand the consequences of housing and milking systems on cow stress levels while milking.

4.6. Lying Time, Bout Frequency and Duration

Cows in the C group, particularly multiparous, spent greater lying time in total and during daytime hours (0630–1830 h) than cows in the T group. The transition in milking permission impacted cow lying time. [54] also reported cow milking frequency to be associated with lying time. However, cows with a higher milking frequency had lower resting time due to the longer time spent milking throughout the day. Other studies reported lying time was negatively correlated with milking frequency [55,56]. This differs from the results in the current study, where C cows were able to milk more frequently but spent more time lying down than cows in the T group who experienced a decrease in milking permission. A possible explanation is cows with reduced milk permission spent more time idling and were more stressed. These behavior alterations led cows to be reluctant to lie down and lowered their overall daily lying time. This explanation is consistent with previous studies [57] that noted a change in milking permission could initiate a stress response in cows. These findings suggest parity in AMS farms is crucial in designing precision milking management settings as they directly impact short-term changes in cow lying time as it can impact welfare, health, and performance [55].

Changes in milking permission can impact dairy cow activity. Cows with higher milking frequencies have reduced time available for feeding or lying behaviors [58]. Various researchers have reported similar results regarding the average lying bout frequency being 9 to 11 lying bouts/d [54,57,59,60,61]. The current study observed T cows performed more daily and daytime lying bouts than C cows, suggesting the transition in milking permission impacted cow lying frequency. Minimal research has evaluated the effects of decreased milking permission on cow lying frequency. However, Ref. [62] reported milking frequency had no effect on lying behavior between cows that were milked once or 2x/d. Our findings suggest the decreased milking permission in T cows caused interrupted lying behavior and increased lying bouts, which could be attributed to greater udder pressure and decreased comfort while lying [62]. On the other hand, reduced lying time and increased frequency of lying bouts in MU cows may result in prolonged standing time, which is a risk factor for lameness [63].

The transition in milking permission influenced daily lying bout duration in the present study. We observed C cows had longer lying bouts than T cows. [56] and the observed lying bout length was negatively correlated with milking frequency, and milk yield, suggesting cows with increased milking permission would show shorter lying bout durations. Differing from the present findings, we observed the opposite effect on C cows lying bouts duration. [57] also reported a large difference in mean bout duration across various research studies. Moreover, lying bout duration varies among individual cows, with the shortest being a couple of minutes and the longest being several hours. However, Ref. [64] reported that cows worked for lying bouts for at least 30 min. Lying is crucial for cows to maintain energy balance, digest food efficiently, and preserve cow health and welfare to meet the high production demands of AMS farms [58].

4.7. Idle Time

Inactive standing or idle time referred to periods when cows were resting or standing without participating in any other activities or interactions [65]. In the current study, we observed an effect of transition in milking permission on daytime idle time. Cows assigned to the T group with decreased milking permission showed greater inactive standing time than cows in the C group. Unfortunately, minimal research has reported an influence of changes in milking permission on cow idle time. In two types of housing systems, Ref. [66] concluded cows had decreased lying and increased idle times as probable indicators of cow discomfort [66]. However, idling is often associated with reduced lying times in dairy cows, which is a major sign of stress [54,56], indicating that reduction of milk permission is perceived as a stress management practice by cows in an AMS.

5. Conclusions

The transition in milking permission from milking every 4 h to every 6 h did not impact overall daily milk yield, milking visits, or milk yield per visit, but it did influence short-term stress indices of dairy cows in AMS. Reducing milking permission impacted cow lying, idling time, and behaviors such as stepping, tail switching, and displacement, while HRV variables showed more sympathetic dominance. Moreover, multiparous cows showed stronger behavioral changes, stress responses, and activity shifts to the change in milking permission implemented in this study than primiparous cows. However, further research with a bigger sample size is still crucially required to precisely evaluate the effect of milking transition on milk production in AMS farms, as the altered behavioral responses and elevated stress parameters presented in the current study suggest milk yield might be impacted in the long term.