--- a/hm/morpho/morphology.py
+++ b/hm/morpho/morphology.py
@@ -73,6 +73,19 @@
     # Files where unanalyzed words and abbreviations (surface forms) are found.
     WORDS1 = "words1.srf"
     WORDSM = "wordsM.srf"
+
+    @staticmethod
+    def _ortho_phon(lines):
+        """First two whitespace fields of each real line, as ortho -> phon."""
+        out = {}
+        for line in lines:
+            line = line.strip()
+            if not line or line.startswith("#"):
+                continue
+            parts = line.split()
+            if len(parts) >= 2:
+                out[parts[0]] = parts[1]
+        return out
     ABBREVS = "abbrev.srf"
 
     def __init__(self, pos_morphs=[], punctuation='', characters='', abbrev_chars='.'):
@@ -364,7 +377,11 @@
                         self.words1 = dict([(w[0].strip(), w[1:]) for w in pairs])
                     else:
                         # Read in ortho:phon pairs as a dict
-                        self.words_phon1 = dict([w.strip().split() for w in file])
+                        # words1.srf is `ortho phon POS`, three fields, and
+                        # carries `#` comment lines. dict() of the raw split
+                        # therefore raised on the first real entry, which took
+                        # the whole phonetic path down for Amharic.
+                        self.words_phon1 = Morphology._ortho_phon(file)
             else:
                 self.words1 = []
                 self.words_phon1 = {}
@@ -377,7 +394,7 @@
                         self.wordsM = dict([(w[0].strip(), w[1:]) for w in pairs])
                     else:
                         # Read in ortho:phon pairs as a dict
-                        self.words_phonM = dict([w.strip().split() for w in file])
+                        self.words_phonM = Morphology._ortho_phon(file)
             else:
                 self.wordsM = []
                 self.words_phonM = {}