diff --git a/lumbda.py b/lumbda.py
index c8a8d04..a6bcc92 100644
--- a/lumbda.py
+++ b/lumbda.py
@@ -399,12 +399,18 @@ def _atom(t):
     if t.startswith('"'):
         return (t[1:-1].replace('\\"', '"').replace('\\n', '\n')
                        .replace('\\t', '\t').replace('\\\\', '\\').replace('\\r', '\r'))
-    for conv in (int, float):
-        try: return conv(t)
-        except (ValueError, OverflowError): pass
+    try: return int(t)
+    except (ValueError, OverflowError): pass
+    # Float parse: gate against Python's float() accepting bare 'inf',
+    # 'infinity', 'nan' as IEEE specials — R7RS spells those +inf.0 /
+    # -inf.0 / +nan.0 explicitly. Without this guard, '(infinity) reads
+    # as (+inf.0) and '(nan) reads as (+nan.0).
     if t == '+inf.0': return math.inf
     if t == '-inf.0': return -math.inf
     if t in ('+nan.0', '-nan.0'): return float('nan')
+    if re.fullmatch(r'[+-]?(\d+\.\d*|\.\d+|\d+)([eE][+-]?\d+)?', t):
+        try: return float(t)
+        except (ValueError, OverflowError): pass
     # Rational literal n/d (e.g. 1/3, -2/5)
     if re.fullmatch(r'-?\d+/-?\d+', t):
         try:
@@ -2946,7 +2952,7 @@ def make_global_env():
                 n = float(n) / float(x)
         return n
     d(S('/'), _div)
-    d(S('quotient'),   lambda a, _: int(_num(a[0]) / _num(a[1])))
+    d(S('quotient'),   lambda a, _: (lambda x, y: -(abs(int(x)) // abs(int(y))) if (x < 0) != (y < 0) else abs(int(x)) // abs(int(y)))(_num(a[0]), _num(a[1])))
     d(S('remainder'),  lambda a, _: int(_num(a[0])) % int(_num(a[1])) * (1 if _num(a[0]) >= 0 else -1))
     d(S('modulo'),     lambda a, _: int(_num(a[0])) % int(_num(a[1])))
     d(S('expt'),       lambda a, _: _num(a[0]) ** _num(a[1]))