You have spent years trusting the product rule. a^m \cdot a^n = a^{m+n} works for 2^3 \cdot 2^4, for 5^{-1} \cdot 5^{7}, for every integer pair you have ever tried. It feels like a law of nature.
Then your teacher writes (-1)^{1/2} on the board and suddenly the clean rules develop asterisks. "Does it still work? Sometimes. Be careful." So which is it — do exponent rules work or not, and why does your teacher sound nervous the moment roots show up?
The short answer
The rules themselves have not changed. a^m \cdot a^n = a^{m+n} is still true. (a^m)^n = a^{mn} is still true. What changed is the set of values where each rule applies. For a positive real base, the rules hold for every real exponent — no exceptions. The moment you combine a negative base with a fractional exponent, the "strangeness" is not the rule failing; it is the rule refusing to land inside the real numbers at all.
The bug, if you want to call it that, is not in the rule. It is in the assumption that every expression you write must have a real-number answer.
For positive bases, everything is clean
Set a > 0. Then a^{m/n} is always a single, unambiguous, positive real number: take the n-th root of a (one positive value), raise to the m-th power, done. See What does a^{1/2} mean if I can't multiply half a time? for the forced definition.
Every law works, for every real exponent, with no caveats:
The m and n can be integers, fractions, or irrationals like \pi or \sqrt{2} — anything real. There is a reason most textbook problems quietly state "assume x > 0": on the positive reals, the laws are truly universal. Students who only ever work with positive bases never meet the problem this article is about.
For negative bases, things get subtle
Set a < 0. Now look at the function f(x) = a^x as x varies over the real numbers, and watch what happens piece by piece.
- For integer x = n: a^n is a well-defined real number. If n is even, the result is positive. If n is odd, the result is negative. If n = 0, the result is 1. No issues.
- For x = 1/n with odd n: a^{1/n} is the real n-th root of a, which exists and is unique (e.g. (-8)^{1/3} = -2). Still fine.
- For x = 1/n with even n: a^{1/n} is not a real number. There is no real square root of -4, no real fourth root of -16. The expression escapes the real line.
- For irrational x: a^x has no real-valued continuous definition at all. You cannot even rescue it by limits, because the rational approximations of x oscillate between "defined" (odd denominators) and "undefined" (even denominators) arbitrarily close together.
If you plotted f(x) = (-2)^x for real x, you would not get a curve. You would get a scattered dust of points — real values at integers and at fractions with odd denominators, nothing in between. The function is defined on a full-of-holes subset of the reals.
That is what "the rules behave strangely at roots" actually means. The rules are fine. The function is broken — it does not live on the continuous real line when the base is negative.
Why integer exponents are special
Integer exponentiation has one saving grace: it is defined purely by repeated multiplication. No roots involved, ever. a^3 = a \cdot a \cdot a, and a^{-3} = 1/(a \cdot a \cdot a), and a^0 = 1 by convention. Multiplication and division are bulletproof on the reals — they never take you outside.
That is why every exponent rule holds universally at integer exponents, regardless of the base. You can freely mix (-3)^5 and (-3)^7 and (-3)^{-2}; all of it works, because you are only multiplying, never extracting a root.
Where roots enter — fractional exponents
The fraction 1/n in an exponent is a different beast. By definition, a^{1/n} means "the n-th root of a", and roots of negatives have ambiguities that multiplication never produces: even roots of negatives have no real answer (two complex answers), while odd roots of negatives have exactly one real answer.
So the instant your exponent involves 1/n with n even, you can step outside the reals. Once outside, different rules make different choices between the complex values — and that is where the asterisks come from. See Negative base, fractional exponent: the domain caveat for the full table.
An example that seems to break the rules
The classic trap. Start with (-1)^2 = 1, then take the square root:
Now apply the power-of-a-power rule to the same expression:
Two legal-looking applications give 1 and -1. So is (a^m)^n = a^{mn} simply wrong?
No. The rule comes with a domain. It assumes the outer exponent does not require a choice of root that conflicts with the inner computation. For a < 0 and a fractional outer exponent with even denominator, you are outside the safe domain. The contradiction is the universe telling you the rule does not apply here, not that the rule is broken.
The textbook version of the same lesson: \sqrt{x^2} = x is not universally true. The correct identity is \sqrt{x^2} = |x|. The absolute-value sign is the fine print — going through a square and then a square root can lose the sign.
The safe zone
If you want zero weirdness, restrict to a > 0. Every rule holds. Every expression evaluates to exactly one real number. No ambiguity, no branches, no domain headaches. This is the unspoken assumption behind almost every "simplify using exponent rules" problem in a school textbook.
If the problem states x > 0 — apply the rules mechanically. If the base could be negative and the exponent contains an even denominator — stop and check. If the exponent is irrational and the base is negative — the expression is not defined in the reals; skip or reinterpret.
What happens in the complex numbers
A brief peek, because students who love maths always ask. In the complex numbers, a^x is defined as a^x = e^{x \ln a}. The catch is that \ln a for complex a has infinitely many values, differing by integer multiples of 2\pi i. Each choice gives a different "branch" of a^x.
On any single consistent branch, the exponent rules hold. Let different steps use different branches, and you get exactly the (-1) contradiction — the rule being applied on mismatched versions of the function. For now, "positive base" is the safe zone because it is the unique regime where a single branch covers everything.
Generalisation
Each time exponents were extended — naturals to integers to rationals to reals to complex — the system gained expressive power and lost some simplicity. The rules themselves stayed consistent; what expanded was the bookkeeping about which bases and which branches. Modern mathematics formalises exactly where each rule holds; the machinery is called analysis, and it lives in later coursework. The habit worth carrying forward is asking where a rule applies before assuming it always does.
Practical takeaway for algebra class
- If the problem says x > 0, or all variables are clearly positive, apply the exponent rules mechanically. No caveats.
- If the base could be negative and the exponent has an even denominator, stop. Check whether the expression is even defined over the reals.
- If the exponent is irrational and the base is negative, the expression is not defined in the real numbers. Either the problem intends a complex answer (rare at school level), or it has a typo.
- When you see a "contradiction" from applying a rule, your first suspicion should be a domain violation, not a broken rule.
Closing
The exponent rules do not behave strangely at roots. They apply on a smaller domain than you assumed. Integer exponents are defined by repeated multiplication, which never leaves the reals, so the rules hold for every base. Fractional exponents are defined by roots, and roots of negatives can escape the reals, so the rules hold only where the expression stays real.
Know the domain; the rule is always the same.