Captures in the split regex does have an effect, but setting $1 shouldn't be relied on. The effect of captures in the split regex is to return the captured data.

split(/\n/,   "hello\nworld")  ->  ("hello", "world")
split(/(\n)/, "hello\nworld")  ->  ("hello", "\n", "world")
[download]

I'd recommend

$text = join '',
        grep $_ ne "$word\n",
        split /(?<=\n)/,
        $text;
[download]

if you wanted to use split.