Triaxial accelerometers have been validated, to measure behavior, mainly in Holstein-Friesian (HF). Our objective was to investigate feeding behavior of transition Jersey cows and its interaction with wellness scores, using a commercial triaxial accelerometer. Jersey cows (n = 145) were fitted with the Silent Herdsman neck collar (Afimilk, Israel) which recorded eating and rumination time hourly from −21 to 21 DIM. Rumen fill, fecal, lameness, leg hygiene and body condition scores (BCS) were assigned every 2 d. Afimilk (Israel) provided feeding behavior data before algorithmic transformation. Pearson correlations evaluated associations between collar data and scores. Feeding behavior was also analyzed with the GLIMMIX procedure of SAS, with cow and DIM as random and repeated variables, respectively. Animals belonging to the specific score value, parity, DIM and the respective interactions were included as fixed factors. BCS was included as a continuous variable. Feeding and rumination times were positively correlated (0.34 to 0.55; P < 0.01) with fill score and BCS, and negatively correlated with leg hygiene and lameness scores (−0.32 to −0.12; P < 0.01). Parity did not affect daily rumination time, but older cows ruminated more (P < 0.001) around parturition. DIM was significant (P < 0.0001) resulting in cows reducing rumination time from 423 ± 52 min at DIM −21 to 197 ± 8 min at calving. Only after 20 d postpartum, cows ruminated at least 400 min/day. Higher lamenes scores were associated with reduced rumination (P < 0.0001), decreasing of 1 h from score 1 to 4. When including eating time as response variable, younger cows spent more time at the manger (599 ± 32 vs. 538 ± 30 min; P < 0.0001). DIM affected eating time, mainly due to a drastic reduction of eating time after calving, from 635 ± 33 min, of the −21 DIM, to 542 ± 10 min for the first 10 DIM. Lameness resulted in a drastic reduction of time spent eating from 583 min, for score 1, to 512 min, for score 4. These data proved to be different from what reported for HF suggesting that accelerometers might need to have Jersey cows-specific algorithms and that, relatively to the scores analyzed, at least locomotion will affect feeding behavior and may be detected early.