We report a detection of the coherent distortion of faint galaxies arising from gravitational lensing by foreground structures .
This “ cosmic shear ” is potentially the most direct measure of the mass power spectrum , as it is unaffected by poorly-justified assumptions made concerning the biasing of the distribution .
Our detection is based on an initial imaging study of 14 separated 8 ^ { \prime } \times 16 ^ { \prime } fields observed in good , homogeneous conditions with the prime focus EEV CCD camera of the 4.2m William Herschel Telescope .
We detect an rms shear of 1.6 % in 8 ^ { \prime } \times 8 ^ { \prime } cells , with a significance of 3.4 \sigma .
We carefully justify this detection by quantifying various systematic effects and carrying out extensive simulations of the recovery of the shear signal from artificial images defined according to measured instrument characteristics .
We also verify our detection by computing the cross-correlation between the shear in adjacent cells .
Including ( gaussian ) cosmic variance , we measure the shear variance to be ( 0.016 ) ^ { 2 } \pm ( 0.012 ) ^ { 2 } \pm ( 0.006 ) ^ { 2 } , where these 1 \sigma errors correspond to statistical and systematic uncertainties , respectively .
Our measurements are consistent with the predictions of cluster-normalised CDM models ( within 1 \sigma ) but a COBE-normalised SCDM model is ruled out at the 3.0 \sigma level .
For the currently-favoured \Lambda CDM model ( with \Omega _ { m } = 0.3 ) , our measurement provides a normalisation of the mass power spectrum of \sigma _ { 8 } = 1.5 \pm 0.5 , fully consistent with that derived from cluster abundances .
Our result demonstrates that ground-based telescopes can , with adequate care , be used to constrain the mass power spectrum on various scales .
The present results are limited mainly by cosmic variance , which can be overcome in the near future with more observations .