Negators, modals, and degree adverbs can significantly affect the sentiment\nof the words they modify. Often, their impact is modeled with simple\nheuristics; although, recent work has shown that such heuristics do not capture\nthe true sentiment of multi-word phrases. We created a dataset of phrases that\ninclude various negators, modals, and degree adverbs, as well as their\ncombinations. Both the phrases and their constituent content words were\nannotated with real-valued scores of sentiment association. Using phrasal terms\nin the created dataset, we analyze the impact of individual modifiers and the\naverage effect of the groups of modifiers on overall sentiment. We find that\nthe effect of modifiers varies substantially among the members of the same\ngroup. Furthermore, each individual modifier can affect sentiment words in\ndifferent ways. Therefore, solutions based on statistical learning seem more\npromising than fixed hand-crafted rules on the task of automatic sentiment\nprediction.\n