Table 1.
Information and numeric results for each language that was annotated and analyzed.
| Language specs | Descriptive statistics | Distances and Distributions | Ideal Learner Predictions | ARMA Modeling | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Language | Language family | ISO | References | Number of nuclei | Number of rhythmic phrases | Median of distribution (ms) | Normality | nPVI | Estimated LR-INI mean | Estimated LR-INI variance | Differential entropy | (p,q) and differencing order of best ARMA | Size of akaike set | % Akaike weight taken up by d = 1 |
| Arabic | Afro-Asiatic | ara | Thelwall and Sa'Adeddin, 1990 | 211 | 15 | 173 | 0.07 | 38.6 | 0.02 | 0.25 | 0.72 | (5,3),1 | 30 | 93.96% |
| Arrernte | Pama–Nyungan | aer | Breen and Dobson, 2005 | 220 | 18 | 177 | 0.074 | 35.4 | −0.02 | 0.23 | 0.70 | (1,1),1 | 21 | 98.89% |
| Cantonese | Sino-Tibetan | yue | Zee, 1991 | 123 | 20 | 144 | 0.076 | 26.8 | 0.06 | 0.11 | 0.35 | (0,0),0 | 30 | 7.62% |
| Dutch | Indo-European | nld | Gussenhoven, 1992 | 159 | 18 | 148 | 0.078 | 45.9 | 0.04 | 0.33 | 0.87 | (2,3),1 | 19 | 99.12% |
| Georgian | Kartvelian | kat | Shosted and Chikovani, 2006 | 173 | 18 | 164 | 0.113 | 47.8 | 0.01 | 0.41 | 0.98 | (0,1),1 | 19 | 99.99% |
| Hindi | Indo-European | hin | Ohala, 1994 | 211 | 23 | 179 | 0.033 | 33.8 | 0.00 | 0.19 | 0.59 | (0,1),1 | 34 | 91.14% |
| Hungarian | Uralic | hun | Szende, 1994 | 191 | 13 | 188 | 0.037 | 37.3 | 0.02 | 0.22 | 0.68 | (0,1),1 | 50 | 65.05% |
| Igbo | Niger–Congo | ibo | Ikekeonwu, 1991 | 159 | 23 | 194 | 0.139 | 39 | 0.01 | 0.25 | 0.74 | (0,1),1 | 16 | 100.00% |
| Italian | Indo-European | ita | Rogers and d'Arcangeli, 2004 | 185 | 20 | 185 | 0.056 | 41 | 0.04 | 0.27 | 0.76 | (2,3),1 | 21 | 99.83% |
| Japanese | Japonic | jpn | Okada, 1991 | 187 | 24 | 131 | 0.163 | 49.2 | 0.10 | 0.35 | 0.90 | (0,1),1 | 21 | 100.00% |
| Kunama | Nilo-Saharan | kun | Ashkaba and Hayward, 1999 | 185 | 41 | 196 | 0.078 | 41 | 0.09 | 0.28 | 0.80 | (0,1),1 | 17 | 100.00% |
| Mapudungun | Araucanian | arn | Sadowsky et al., 2013 | 161 | 24 | 211 | 0.109 | 38.1 | −0.01 | 0.25 | 0.73 | (2,3),1 | 21 | 99.94% |
| Nuuchahnulth | Wakashan | nuk | Carlson et al., 2001 | 106 | 13 | 285 | 0.077 | 47.7 | 0.04 | 0.47 | 1.05 | (0,1),1 | 21 | 99.99% |
| Spokane | Salishan | spo | Carlson and Esling, 2000 | 92 | 11 | 364 | 0.11 | 42.3 | 0.01 | 0.34 | 0.88 | (1,1),1 | 20 | 100.00% |
| Tena Quichua | Quechuan | quw | O'Rourke and Swanson, 2013 | 238 | 36 | 249 | 0.112 | 42.3 | −0.03 | 0.35 | 0.91 | (0,3),1 | 22 | 98.43% |
| Thai | Tai–Kadai | tha | Tingsabadh and Abramson, 1993 | 181 | 33 | 251 | 0.064 | 41 | 0.02 | 0.33 | 0.87 | (3,2),1 | 21 | 99.99% |
| Turkish | Turkic | tur | Zimmer and Orgun, 1992 | 169 | 14 | 159 | 0.055 | 32.9 | 0.00 | 0.17 | 0.54 | (2,4),1 | 45 | 60.02% |
| Vietnamese | Austroasiatic | vie | Kirby, 2011 | 121 | 19 | 214 | 0.086 | 36.8 | 0.07 | 0.24 | 0.71 | (0,1),1 | 19 | 98.19% |
The left side of the table includes ethnographic information about the languages and descriptive statistics of our sample in terms of syllable and phrase structure. The right side of the table provides results for each language. For the first analysis, the correlation between the Kolmogorov-Smirnov D and nPVI measures can be noticed. Next, we present measures about an ideal learner's inference of the LR-INI (the logarithm of the relative INI lengths; see section Analysis and Results: Distributional Statistics of Temporal Structure (Order 1)), and the last three columns present the raw results of the ARMA analyses, including both the single best-fitting model as well as the results of calculating the Akaike weights and sets. Results from higher-order models should be interpreted keeping in mind the low predictive power of ARMA models for small sample sizes.