Magnetoresistance exhibited by non-magnetic semiconductors has attracted much attention. In particular, Wan et al. reported roomtemperature magnetoresistance in silicon to reach 10% at 0.07 T and 150,000% at 7 T—‘‘an intrinsically spatial effect’’. Their supply voltage was approximately 10V (ref. 12), which is low and approaches the industrial requirement. However, we have found their large magnetoresistance values to be experimental artefacts caused by theirmethod ofmeasurement. The true room-temperaturemagnetoresistance of the devices described in ref. 12 is lowwith amagnetic field of up to 7 T and a supply voltage of around 10V and hence these devices cannot offer large magnetoresistance with low supply voltage to industry. There is a Reply to this Brief Communication Arising by Zhang, X. Z., Wan, C.H.,Gao,X. L.,Wang, J.M.&Tan,X.Y.Nature501,http://dx.doi.org/ 10.1038/nature12590 (2013). Wan et al. measured two types of In/SiO2/Si/SiO2/In devices using a Keithley 2400 sourcemeter as both a current source and a voltage meter (which we refer to here as method 1), and obtained largemagnetoresistance values of up to 10% at 0.07 T and 150,000% at 7 T.We fabricated two devices with the same structures as those of ref. 12 and performed method 1 using them. Their voltage–current (V–I) curves can be divided into different regions with different resistances, just as in the results of ref. 12. Wan et al. claim that injection of minority carriers into silicon causes a p–n junction and the changes in resistance, that large magnetoresistance occurs with applied current in one of the regions (referred as to the transition region), and that the magnetic-field dependence of the magnetoresistance in the transition region is different from those in the other regions. However, when we used another method (here called method 2) with unchanged measuring parameters and different instruments on the devices, the V–I characteristics without the transition region were obtained. The only difference between the two methods is that in method 2 we used the Keithley 2400 only as the current source, with an independent voltmeter (Keithley 2182) as the voltage meter. Further, we performed both methods on two circuits composed of linear resistors, which were used to simulate the devices. The results indicate that in method 1 the Keithley 2400 itself interferes with the measurement of the specimen and cannot give correct voltage values when the applied current exceeds a certain value and falls in the transition region. Because ref. 12 claims that largemagnetoresistances weremeasuredwhen Iwas in the transition region,magnetoresistance was defined as [R(B)2R(B5 0)]/R(B5 0) and R5V/I, we conclude that the large magnetoresistance values are really experimental artefacts caused by the interference of the sourcemeter. Method 2 is valid. Using it, we obtained magnetoresistance values for the two devices with supply voltages of 6.7–72V and 0.79–50V, respectively. The values are all low and the magnetic-field dependence at all applied currents is the same (above 2 T the field dependence is linear); the magnetoresistance does not exhibit any signs of saturation at fields up to 7 T. The linear dependence without magnetoresistance saturation is the same as for inhomogeneity-induced magnetoresistance.