diff --git a/epan/column-utils.c b/epan/column-utils.c
index 6c7b10f9a1..a9e8ac34ff 100644
--- a/epan/column-utils.c
+++ b/epan/column-utils.c
@@ -78,8 +78,8 @@ col_setup(column_info *cinfo, const gint num_cols)
     cinfo->col_last[i] = -1;
   }
   cinfo->prime_regex = g_regex_new(COL_CUSTOM_PRIME_REGEX,
-    (GRegexCompileFlags) (G_REGEX_ANCHORED | G_REGEX_RAW),
-    G_REGEX_MATCH_ANCHORED, NULL);
+    (GRegexCompileFlags) (G_REGEX_RAW),
+    0, NULL);
 }
 
 static void
diff --git a/epan/column-utils.h b/epan/column-utils.h
index 7239583280..a213db688a 100644
--- a/epan/column-utils.h
+++ b/epan/column-utils.h
@@ -23,7 +23,25 @@ extern "C" {
 
 #define COL_MAX_LEN 2048
 #define COL_MAX_INFO_LEN 4096
-#define COL_CUSTOM_PRIME_REGEX " *([^ \\|]+) *(?:(?:\\|\\|)|(?:or)| *$){1}"
+
+/* A regex to split possibly multifield custom columns into components
+ *
+ * Split on operator "||" (with optional space around it) and on "or"
+ * (which must have space around it to avoid matching in the middle of
+ * a word, field in the "synphasor" protocol, etc. This is somewhat too
+ * strict, as "or" adjacent to parentheses ought to be fine so long
+ * as the filter matches the grammar, like "(tcp.port)or(udp.port)",
+ * but that's the cost of using regex instead of the real parser.)
+ * Also split on space at the beginning or end of the expression (in
+ * lieu of always stripping whitespace at the beginning and end, but it
+ * does mean that we have to ignore any empty tokens in the result.)
+ *
+ * Use negative lookahead to avoid matching "||" or "or" that are contained
+ * within parentheses. Don't match if a close parenthesis comes before an
+ * open parenthesis. The regex doesn't help with unmatched parentheses, but
+ * such an expression already won't satisfy the grammar and won't compile.
+ */
+#define COL_CUSTOM_PRIME_REGEX "(?:^ *| *\\|\\| *| +or +| *$)(?![^(]*\\))"
 
 struct epan_dissect;
 
diff --git a/epan/column.c b/epan/column.c
index 4eeffca71f..0d836e8ed8 100644
--- a/epan/column.c
+++ b/epan/column.c
@@ -989,8 +989,8 @@ get_column_tooltip(const gint col)
     }
 
     fields = g_regex_split_simple(COL_CUSTOM_PRIME_REGEX, cfmt->custom_fields,
-                                  (GRegexCompileFlags) (G_REGEX_ANCHORED | G_REGEX_RAW),
-                                  G_REGEX_MATCH_ANCHORED);
+                                  (GRegexCompileFlags) (G_REGEX_RAW),
+                                  0);
     column_tooltip = g_string_new("");
 
     for (i = 0; i < g_strv_length(fields); i++) {
@@ -1043,7 +1043,7 @@ col_finalize(column_info *cinfo)
       }
       if (col_item->col_custom_fields) {
         gchar **fields = g_regex_split(cinfo->prime_regex, col_item->col_custom_fields,
-                                       G_REGEX_MATCH_ANCHORED);
+                                       0);
         guint i_field;
 
         for (i_field = 0; i_field < g_strv_length(fields); i_field++) {
diff --git a/ui/preference_utils.c b/ui/preference_utils.c
index 6ecada4fc2..749d8d745b 100644
--- a/ui/preference_utils.c
+++ b/ui/preference_utils.c
@@ -235,8 +235,8 @@ column_prefs_custom_resolve(const gchar* custom_field)
     bool resolve = false;
 
     fields = g_regex_split_simple(COL_CUSTOM_PRIME_REGEX, custom_field,
-                                  (GRegexCompileFlags) (G_REGEX_ANCHORED | G_REGEX_RAW),
-                                  G_REGEX_MATCH_ANCHORED);
+                                  (GRegexCompileFlags) (G_REGEX_RAW),
+                                  0);
 
     for (guint i = 0; i < g_strv_length(fields); i++) {
         if (fields[i] && *fields[i]) {
diff --git a/ui/qt/widgets/syntax_line_edit.cpp b/ui/qt/widgets/syntax_line_edit.cpp
index 99ccf7bcb7..2ea83ea112 100644
--- a/ui/qt/widgets/syntax_line_edit.cpp
+++ b/ui/qt/widgets/syntax_line_edit.cpp
@@ -268,7 +268,8 @@ void SyntaxLineEdit::checkCustomColumn(QString fields)
     }
 
     gchar **splitted_fields = g_regex_split_simple(COL_CUSTOM_PRIME_REGEX,
-                fields.toUtf8().constData(), G_REGEX_ANCHORED, G_REGEX_MATCH_ANCHORED);
+                fields.toUtf8().constData(), (GRegexCompileFlags) G_REGEX_RAW,
+                (GRegexMatchFlags) 0);
 
     for (guint i = 0; i < g_strv_length(splitted_fields); i++) {
         if (splitted_fields[i] && *splitted_fields[i]) {