Dairy cows typically have their milk production artificially ceased around 60 d before calving. This process, termed drying-off, often comprises changes of diet and milking frequency. Evidence suggest drying-off challenges the welfare of high-yielding cows. For instance, a nutrient-reduced diet can impair energy balance and cause hunger even when offered ad libitum. This ongoing work proposes a feed frustration test, inducing motivation conflict, to assess isolated and combined effects of daily milking frequency (twice; once) and diet (lactation diet or energy-reduced lactation diet with forage:concentrate ratios of 51:49 and 65:35, respectively, ad libitum) on the degree of hunger. Biweekly, groups of 2-6 healthy loose-housed Holstein cows were allocated to one of four treatments from 7 d (D-7) prior to the last day of milking (DO). In total, 92 cows (mean±SD) 754±80 kg in body weight and yielding 25±6 kg of milk/d were included in this study. From D-7 to DO, each cow had access to one computerised feed bin providing access to the respective diet. Tests were conducted on D-6 and D-1. First, barn staff remotely locked the feed bins and initiated the second daily feed delivery (at approx. 10.30 h). Duration of feed delivery ranged from 15-76 min and constituted the routine phase (phase A). When the feed delivery ended, all feed bins were remotely unlocked, except the experimental ones, which remained locked for additional 35 min, constituting the extra feed deprivation phase (phase B). During B, experimental cows could hear other feed bins opening and see cows in neighbouring pens feeding. During both phases, an experimenter standing in front of the feed bins approx. 4 m away recorded attempts to feed (i.e. focal cow puts the head over any locked feed bin, placing her collar past the feed bin gate) and high-pitched vocalisations (≥3 s). Following unlocking of experimental feed bins the latency to feed was recorded for 5 min. Data analyses were performed in R for each day separately. Attempts to feed in A were treated as counts/min due to a wide time range, while attempts to feed in B were treated as counts. Due to a low representation of vocalisations in the data set, this variable was analysed as binary (yes; no). Latency to feed was also analysed as binary (yes; no). All variables were analysed using GLMM including diet and milking frequency as fixed effects and block as random effect. On D-6, no significant differences were found between treatments, except the odds of vocalising being higher in cows on energy-reduced diet than normal diet [OR (95% CI)] [A: 5.4 (2.1 to 15.5), χ2=11.4, DF=1, P<0.001; B: 3.3 (1.4 to 8.2), χ2=7.1, DF=1, P<0.01]. On D-1, cows on energy-reduced diet attempted to feed 50% more in B [OR (95% CI)] [1.5 (1.08 to 2.09), χ2=5.93, DF=1, P<0.05], were more likely to vocalise [A: 8.4 (2.5 to 38.7), χ2=10.1, DF=1, P<0.01; B: 4.5 (1.9 to 11.5), χ2=10.9, DF=1, P<0.001] and to feed within 5 min [2.8 (1 to 8.6), χ2=3.84, DF=1, P=0.049] than cows on normal diet. These preliminary results suggest that during gradual dry-off hunger results from the negative dietary shift, while no mitigating effects of reducing the milking frequency were evident.