Briefly, lexical analyzer reads source code character by character and generates tokens. It does not care about in what order tokens are generated. Relationship between tokens are checked by syntax analyzer and specified by a context free grammar. Let me give you a simple example. Consider the following piece of code
int while = 5
Lexical analyzer emmits the following tokens:
Then then syntax analyzer checks these tokens, detects (based on the corresponding CFG) that the keyword WHILE cannot be lvalue, and reports the syntax error. In fact the syntax analyzer expects IDENTIFIER after the token KEYWORD (INT).