Not so long ago a number of academic institutions allowed investigators to classify persons who were actually employees as "consultants" or "independent contractors" on grant budgets, mostly to save on fringe benefits. But if you employ someone, it doesn't matter what you call them, you still have to pay certain payroll taxes, such as social security and workman's compensation. If you don't, you -- or your institution -- will be liable not only for back taxes owed, but also for IRS penalties. A number of universities found this out the hard way.

Telling the difference between an employee and a consultant is not usually difficult: your instinct is probably right. If you provide the person in question with an office, a computer, a telephone, and other tools of the trade; if you expect this person to show up to work on a schedule you set; if this person doesn't have the option to say no to assignment from you; if this person does not routinely accept assignments to do similar work on other projects, then this person is your employee. You must budget for, and pay, at least the minimum payroll contributions for that person.

If the person in question works out of his/her own office, with tools supplied by him/her; if this person can turn down an assignment from you; if this person sets his/her schedule to do the work on your project; if this person accepts similar assignments from other people, then this person is a consultant or contractor. Payroll contributions are made by this person, or his/her employer, and you need not budget for them.

A final word about consultants: Sometimes people who are asked to play a role on a grant-funded project carried out by their own institution would like to be paid as consultants, rather than have a portion of their institutional salary reimbursed by the grant. Not only is this not acceptable to federal funders, but it also could be construed as "double-dipping". To the extent that the grant-funded activities take place during employer-paid time, getting an extra payment as a "consultant" would amount to being paid twice for the same work.

The only exception to being paid extra for doing grant work within your own institution would be when legitimate overtime is involved. Usually overtime is only paid for technical workers who are actually working extra hours (which must be documented). It does not apply to investigators, who are expected to be receiving release time from their employers to carry out grant-funded work.

See also:
PHS Grants Policy Statement
NIH Grants Policy Statement