Background Focality of onset of amyotrophic lateral sclerosis (ALS) is not understood. Attempts to implicate physical exercise in the aetiology of ALS have provided inconsistent results. If physical use of a limb were important in defining the site of onset, then handedness might be expected to influence the side of upper limb-onset disease and footedness likewise in lower limb-onset ALS. Methods ALS patients registered with an internet-based support site were invited to complete an online questionnaire concerning site of onset of symptoms and their dominant hand and foot. A binomial test of proportions was used to investigate the null hypothesis that handedness and footedness do not influence side of onset in upper and lower limb-onset ALS, respectively. Results 343 ALS patients with limb-onset disease were studied. For upper limb-onset patients, there was concordance for side of onset and handedness (64%; p<0.0006). For lower limb-onset patients, concordance for side of onset and footedness was absent. The frequency of left handedness was commensurate with that found in the general population. Interpretation These results are potentially consistent with the hypothesis that exercise influences pathogenesis in ALS since routine physical demands on the upper limb are heavily influenced by limb dominance, whereas in the lower limbs the commonest function is standing or locomotion, which uses both legs equally. However, there may also be an inherent cortical vulnerability underlying upper limb-onset laterality, possibly influenced by changes in neuronal connectivity and cortical excitability in relation to handedness and reflected by the "split hand" phenomenon consistently observed in ALS.The site of onset of loss of function in the neurodegenerative disorder amyotrophic lateral sclerosis (ALS) is approximately divided equally between upper limb, lower limb and bulbar territories, 1 with only a very small proportion (w3%) beginning in the respiratory or axial musculature. Although there are a number of active themes in the quest to understand pathogenesis in ALS, 3 very little is known about why the onset is focal, though it has been shown that upper (UMN) and lower motor neuron (LMN) signs (the hallmark of ALS) are maximal within this territory and show contiguous spread. 4 The inconsistent observations of relative athleticism among people with ALS 5e7 have led to continued debate about the possible role of physical exercise in the pathogenesis of ALS, 8 and putative direct effects via excitotoxicity and mitochondrial dysfunction.3 A more indirect association, whereby the genetic profiles that underlie ALS susceptibility and athleticism have common components, 9 currently seems equally, if not more, likely. If physical exercise is a direct determinant of motor neuron damage, then it might be expected that the site onset of symptoms in limb-onset ALS would be biased in favour of the dominant limb through relatively greater use. We attempted to confirm the null hypothesis that the side of initial symptoms i...