Miniature accelerometer data loggers, in combination with GPS tags, were deployed on homing pigeons during 11 km flights along the Menai Strait, Bangor, North Wales in order to use new technology to assess the flight performance of free-ranging birds. The GPS data was used to calculate position and ground speed during the flights and the latter was converted to estimated airspeed using anemometer readings located on the local Britannia Bridge. The accelerometer data was used to calculate wing beat frequency(WBF) and a number of other variables, such as vectorial dynamic body acceleration (VeDBA), mean acceleration in the y-axis (static y) and a term called fraction positive(the proportion of time per wing beat spent accelerating the body above average g force). Relationship between these variables and airspeed were investigated along with the potential effects of pigeons carrying added mass or flying in different environmental conditions, such as strong tailwinds, low winds or strong headwinds. Integrationof the raw Y and Z-axis accelerations were used to calculate power in the body of the bird (using Power= Mass xAcceleration xVelocity), assuming no net change in velocity over the section of integrated data. This indicated that the power detected by the accelerometer could be as little as approximately 3% of the total flight biomechanical costs, at a wing beat frequency of around 7 Hz. However, VeDBAYZ2was very highly linearly correlated (R2= 0.947) with the integrated raw yz-accelerations and, therefore, an excellent predictor for power in the body of a pigeon. This relaltionship should also be applicable to other flying animals. Double integration of the z-axis accelerations and assuming sinusoidal accelerations and motion of the body gives an estimate of dorsal body displacement.Pigeons that were released individually flew with relatively low wing beat frequencies(L-WBF, < 6 Hz), low airspeeds (14.89 m s-1) and low values of VeDBA2(1.17 m s-2) during the flights back to the loft. Pigeons released as a flock on average flew with relatively high airspeeds (>20 m s-1) and usually with high wing beat frequencies (H-WBF, > 6 Hz). However, some birds occasionally broke from the back of the flock and flew more slowly and with L-WBF. Values of VeDBA2and fraction positive tended to be positively correlated with WBF. When birds were released in a flock, airspeed was not found to vary systematically with wind speed or directionon different daysbut to be fairly consistent around a mean value of 20.9ms-1(range 17.8ms-1to 23.9ms-1). VeDBAYZ2was shown to be well correlated with airspeed (ρ= 0.703) and this suggested that power in the body was, indeed, a reasonable indicator of the overall biomechanical flight costs. WBFshowed a slightly less predictable response.It was possible to categorise the birds as slow or fast returning birds, particularly in the flight into the strong headwind, with VeDBAYZ2, wing beat frequency and fraction positive generally much lower for the slowest group of birds on the day. Wing beat frequency tended to be consistently correlated within-individuals acrossdays, but there was a considerable amount of variation in VeDBAYZ2for a given value of wing beat frequency between individuals, indicating the difficulties in making average assessments of flight performance.