diff --git a/Tests/test_file_jpeg.py b/Tests/test_file_jpeg.py
index a2a848b41..fd79e03c8 100644
--- a/Tests/test_file_jpeg.py
+++ b/Tests/test_file_jpeg.py
@@ -299,6 +299,10 @@ class TestFileJpeg(PillowTestCase):
         assert_image(im1, im2.mode, im2.size)
         self.assertGreaterEqual(im1.bytes, im2.bytes)
 
+        im3 = self.roundtrip(hopper(), quality=0)
+        assert_image(im1, im3.mode, im3.size)
+        self.assertGreater(im2.bytes, im3.bytes)
+
     def test_smooth(self):
         im1 = self.roundtrip(hopper())
         im2 = self.roundtrip(hopper(), smooth=100)
diff --git a/docs/handbook/image-file-formats.rst b/docs/handbook/image-file-formats.rst
index 0e068c1e4..7ce685ed2 100644
--- a/docs/handbook/image-file-formats.rst
+++ b/docs/handbook/image-file-formats.rst
@@ -302,7 +302,7 @@ The :py:meth:`~PIL.Image.Image.open` method may set the following
 The :py:meth:`~PIL.Image.Image.save` method supports the following options:
 
 **quality**
-    The image quality, on a scale from 1 (worst) to 95 (best). The default is
+    The image quality, on a scale from 0 (worst) to 95 (best). The default is
     75. Values above 95 should be avoided; 100 disables portions of the JPEG
     compression algorithm, and results in large files with hardly any gain in
     image quality.
diff --git a/src/PIL/JpegImagePlugin.py b/src/PIL/JpegImagePlugin.py
index 9cba544de..b5371db73 100644
--- a/src/PIL/JpegImagePlugin.py
+++ b/src/PIL/JpegImagePlugin.py
@@ -616,17 +616,17 @@ def _save(im, fp, filename):
 
     dpi = [round(x) for x in info.get("dpi", (0, 0))]
 
-    quality = info.get("quality", 0)
+    quality = info.get("quality", -1)
     subsampling = info.get("subsampling", -1)
     qtables = info.get("qtables")
 
     if quality == "keep":
-        quality = 0
+        quality = -1
         subsampling = "keep"
         qtables = "keep"
     elif quality in presets:
         preset = presets[quality]
-        quality = 0
+        quality = -1
         subsampling = preset.get("subsampling", -1)
         qtables = preset.get("quantization")
     elif not isinstance(quality, int):
@@ -749,8 +749,8 @@ def _save(im, fp, filename):
         # CMYK can be bigger
         if im.mode == "CMYK":
             bufsize = 4 * im.size[0] * im.size[1]
-        # keep sets quality to 0, but the actual value may be high.
-        elif quality >= 95 or quality == 0:
+        # keep sets quality to -1, but the actual value may be high.
+        elif quality >= 95 or quality == -1:
             bufsize = 2 * im.size[0] * im.size[1]
         else:
             bufsize = im.size[0] * im.size[1]
diff --git a/src/libImaging/Jpeg.h b/src/libImaging/Jpeg.h
index 82e1b449f..7ff658c32 100644
--- a/src/libImaging/Jpeg.h
+++ b/src/libImaging/Jpeg.h
@@ -67,7 +67,7 @@ typedef struct {
 
     /* CONFIGURATION */
 
-    /* Quality (1-100, 0 means default) */
+    /* Quality (0-100, -1 means default) */
     int quality;
 
     /* Progressive mode */
diff --git a/src/libImaging/JpegEncode.c b/src/libImaging/JpegEncode.c
index 10ad886e0..9d8382791 100644
--- a/src/libImaging/JpegEncode.c
+++ b/src/libImaging/JpegEncode.c
@@ -153,7 +153,7 @@ ImagingJpegEncode(Imaging im, ImagingCodecState state, UINT8* buf, int bytes)
 	    int i;
 	    int quality = 100;
 	    int last_q = 0;
-	    if (context->quality > 0) {
+	    if (context->quality != -1) {
 		quality = context->quality;
 	    }
         for (i = 0; i < context->qtablesLen; i++) {
@@ -171,7 +171,7 @@ ImagingJpegEncode(Imaging im, ImagingCodecState state, UINT8* buf, int bytes)
         for (i = last_q; i < context->cinfo.num_components; i++) {
             context->cinfo.comp_info[i].quant_tbl_no = last_q;
         }
-	} else if (context->quality > 0) {
+	} else if (context->quality != -1) {
 	    jpeg_set_quality(&context->cinfo, context->quality, 1);
 	}