Background or contextual data about a dataset. Literally "data about data". Metadata is required to enable someone to properly understand and interpret a main dataset. 

Examples: 

• The research questions that the data was collected to address
  
• Any relevant environmental conditions affecting the main variables
• The instruments used to collect or generate the data, including specification and calibration details.
• The methodology used to collect or generate the data,
• Definitions of variables, including units - also called a data dictionary.