I used a subscriber identity module card reader to copy the lists of saved contacts from 170 mobile phones in Fiji. This approach has both advantages and disadvantages compared to other techniques for collecting telephone network data. Copying phone contacts avoids recall biases associated with survey-based name generators. It also obviates the need to obtain calling records from network service providers or to provide research subjects with specialized phones. In markets with multiple service providers, or in contexts in which mobile phones are shared extensively, the contents of contact lists may represent personal networks more accurately than actual calling records do. However, owners' varying durations of phone ownership and varying tendencies to save and delete contacts are sources of error. Also, copying contacts from phones relies on self-reports of tie strength, which are frequently inaccurate. I analyze the data I collected in rural Fiji to evaluate these sources of error.