# 1 Background on Phonetics and Sound Symbolism
In this section, we provide background on phonetics of the English language, with an emphasis on the distinctions relevant to our work. Following the convention in modern linguistics, we use the International Phonetic Alphabet (IPA) to represent phonemes (units of sound) and write IPA symbols between slashes \phon …, while we write graphemic (written) representations in angle brackets \graphemic …. Unless stated otherwise we use American English for transcriptions. For more thorough overviews of phonetics, the IPA, and the sounds used in English and across the languages of the world, see [Ladefoged1995-sv, international1999handbook]. In general, speech sounds may be roughly split into the categories of consonants and vowels, based on the degree of closure of the vocal tract. For consonants, McCormick et al. [mccormick_sound_2015] find that the most round-associated segments are all voiced (sonorants \phon m n l and voiced stops \phon b d g). In this context, voicing refers to vibration of the vocal cords during the production of a phoneme Strictly speaking, phonemically “voiced” obstruents such as \phon b may be optionally devoiced in English making the relevant distinction one of aspiration, but this level of phonetic detail is not relevant for our discussion.. On the other end of the spectrum, they find a sharper association for the voiceless stop consonants \phon p t k, and for the fricative and affricate consonants \phon f v s z tS dZ, the latter produced by making a partial occlusion in the vocal tract to produce turbulent airflow. Among vowels, McCormick et al. find the dimension of roundedness (in an articulatory sense) to be significant for sound symbolic association among vowels, with rounded vowels being more round-associated. In the context of articulatory phonetics, roundedness refers to a vowel produced with lips pursed together. In English this is coupled with backness which roughly refers to the position of the tongue relative to the back of the mouth; for example \phon E (as in bell) is a front unrounded vowel while \phon o (as in bowl) is a back rounded vowel. Our split of English graphemes (letters) into \SHARP and \ROUND -associated categories for the purpose of constructing pseudowords is motivated by this phonetic background, along with a desire for balanced class sizes and avoiding digraphs (e.g. \graphemic th, which represents a single phoneme in English). For consonants, our \SHARP -associated graphemes \graphemic p t k s h x are all voiceless, fricatives or affricates while \ROUND -associated \graphemic b d g m n l are all voiced. For vowels, although English vowel orthography has a complex mapping to sound segments, we treat \graphemic e i as front unrounded vowels (and hence \SHARP) and \graphemic o u as back rounded vowels (and hence \ROUND) as a simplifying assumption, consistent with prior work. Since the grapheme \graphemic a can correspond to both front (\phon æ as in fat) and back (\phon A as in father) vowels in English, we treat it as neutral with respect to sharpness associations This is also consistent with previous work, which has used \graphemic a in both “round” stimuli like maluma and “sharp” stimuli like takete..