diff --git a/optimality-theory/Learning.hs b/optimality-theory/Learning.hs
new file mode 100644
index 0000000..d3c2927
--- /dev/null
+++ b/optimality-theory/Learning.hs
@@ -0,0 +1,91 @@
+{-# LANGUAGE OverloadedLists #-}
+{-# LANGUAGE OverloadedStrings #-}
+
+{- This implementation is based on the description of the Recursive Ranking learning algorithm for OT layed out in § 7.3 of
+- Kager, René. 2010. Optimality Theory. 10th printing. Cambridge Textbooks in Linguistics. Cambridge: Cambridge Univ. Press.
+-}
+
+module Learning where
+
+import Data.List (findIndex, intercalate, minimumBy)
+import Data.Maybe (fromJust)
+import Data.Ord (comparing)
+import Data.Set (Set, delete, empty, fromList, insert, member, singleton, toAscList, union)
+import Data.String (IsString (..))
+import Data.Text (Text, pack, unpack)
+
+newtype Grammar = Grammar {strata :: [Set Constraint]}
+  deriving (Eq)
+
+instance Show Grammar where
+  show grammar = intercalate " ≫ " $ map prettySet $ strata grammar
+    where
+      prettySet xs = "{" ++ intercalate ", " (map show (toAscList xs)) ++ "}"
+
+c `rankIn` grammar = findIndex (\stratum -> c `member` stratum) (strata grammar)
+
+dominates c1 c2 grammar =
+  let i1 = c1 `rankIn` grammar
+      i2 = c2 `rankIn` grammar
+   in i1 < i2
+
+newtype Constraint = Constraint {constraintName :: Text}
+  deriving (Eq, Ord)
+
+instance IsString Constraint where
+  fromString = Constraint . pack
+
+instance Show Constraint where
+  show = unpack . constraintName
+
+data MarkDataPair = MarkDataPair
+  { loserMarks :: Set Constraint,
+    winnerMarks :: Set Constraint
+  }
+
+demoteWith :: MarkDataPair -> Grammar -> Grammar
+demoteWith markDataPair grammar =
+  let highestRankedLoser = minimumBy (comparing (`rankIn` grammar)) (loserMarks markDataPair)
+   in foldr
+        ( \winner g ->
+            if (highestRankedLoser `dominates` winner) grammar
+              then g
+              else demoteBelow winner highestRankedLoser g
+        )
+        grammar
+        (winnerMarks markDataPair)
+
+demoteBelow :: Constraint -> Constraint -> Grammar -> Grammar
+demoteBelow a b grammar =
+  let rankB = fromJust $ b `rankIn` grammar
+      strataWithoutA = map (delete a) (strata grammar)
+   in grammar
+        { strata =
+            case splitAt rankB strataWithoutA of
+              (before, bStratum : belowBStratum : after) -> before ++ bStratum : insert a belowBStratum : after
+              (before, [bStratum]) -> before ++ [bStratum, singleton a]
+        }
+
+singleStratumFrom :: [MarkDataPair] -> Grammar
+singleStratumFrom pairs = Grammar [foldr (\p g -> loserMarks p `union` winnerMarks p `union` g) empty pairs]
+
+learnFrom :: [MarkDataPair] -> Grammar -> Grammar
+learnFrom pairs g = go g (cycle pairs)
+  where
+    go g (p : ps) =
+      let g' = demoteWith p g
+       in if g' == g
+            then g
+            else go g' ps
+
+initialGrammar, resultGrammar :: Grammar
+initialGrammar = singleStratumFrom markData
+resultGrammar = learnFrom markData initialGrammar
+
+markData :: [MarkDataPair]
+markData =
+  [ MarkDataPair {loserMarks = ["All-Ft-L"], winnerMarks = ["All-Ft-R"]},
+    MarkDataPair {loserMarks = ["Parse-Syl"], winnerMarks = ["All-Ft-L", "All-Ft-R"]},
+    MarkDataPair {loserMarks = ["All-Ft-L", "Leftmost", "Parse-Syl"], winnerMarks = ["All-Ft-R", "Rightmost"]},
+    MarkDataPair {loserMarks = ["All-Ft-L", "Ft-Bin"], winnerMarks = ["Parse-Syl"]}
+  ]