We investigate whether a neural, encoderdecoder translation system learns syntactic information on the source side as a by-product of training. We propose two methods to detect whether the encoder has learned local and global source syntax. A fine-grained analysis of the syntactic structure learned by the encoder reveals which kinds of syntax are learned and which are missing.

Continued methamphetamine (MA) use is dependent on a positive MA experience and is likely attenuated by sensitivity to the aversive effects of MA. Bidirectional selective breeding of mice for high (MAHDR) or low (MALDR) voluntary consumption of MA demonstrates a genetic influence on MA intake. Quantitative trait locus (QTL) mapping identified a QTL on mouse chromosome 10 that accounts for greater than 50% of the genetically-determined differences in MA intake in the MAHDR and MALDR lines. The trace amine-associated receptor 1 gene (Taar1) is within the confidence interval of the QTL and encodes a receptor (TAAR1) that modulates monoamine neurotransmission and at which MA serves as an agonist. We demonstrate the existence of a non-functional allele of Taar1 in the DBA/2J mouse strain, one of the founder strains of the selected lines, and show that this non-functional allele co-segregates with high MA drinking and with reduced sensitivity to MA-induced conditioned taste aversion (CTA) and hypothermia. The functional Taar1 allele, derived from the other founder strain, C57BL/6J, segregates with low MA drinking and heightened sensitivity to MA-induced CTA and hypothermia. A role for TAAR1 in these phenotypes is corroborated in Taar1 transgenic mice: Taar1 knockout mice consume more MA and exhibit insensitivity to MA-induced CTA and hypothermia, compared with Taar1 wild-type mice. These are the first data to show that voluntary MA consumption is, in part, regulated by TAAR1 function. Behavioral and physiological studies indicate that TAAR1 function increases sensitivity to aversive effects of MA, and may thereby protect against MA use.

We describe Hafez, a program that generates any number of distinct poems on a usersupplied topic. Poems obey rhythmic and rhyme constraints. We describe the poetrygeneration algorithm, give experimental data concerning its parameters, and show its generality with respect to language and poetic form.

No abstract