diff --git a/README.md b/README.md
index 3dfeccd..3a33861 100644
--- a/README.md
+++ b/README.md
@@ -5,7 +5,7 @@ Generally speaking, real world data is both easier and faster to generate, and m
 
 Reprodyne is not meant to *replace* existing test frameworks, but rather to augment their capabilities. It should be possible, at least in theory, to integrate it with any test framework.
 
-From a high level, Reprodyne works by transforming non-deterministic funtions into deterministic ones, and saving their results for testing. "Interceptors" capture indeterminate values in record mode and resupply them in playback mode. "Validators" save values produced by the program in record mode, and compares against them in playback mode, signaling a playback failure if anything is different.
+From a high level, Reprodyne works by transforming non-deterministic functions into deterministic ones, and saving their results for testing. "Interceptors" capture indeterminate values in record mode and resupply them in playback mode. "Validators" save values produced by the program in record mode, and compares against them in playback mode, signaling a playback failure if anything is different.
 
 The Reprodyne API is defined entirely as a set of preprocessor macros, so that once you're done testing, they gracefully expand into no-ops and there is no longer a need to link against the library.