ox CER'l'AIX THEKMUELEOTRIC EFFEOl'Ö OF ,ST1{L:SÖ IN IliOX. o51 



kilo^. actiiiii' aloiin' the wire. It will Ije iiotired that tliere i.s u wlixrht 

 diminution in this liinliest tension as the experiment progressed, 

 doubtless due to the yielding of the highly heated part of the wire. 

 This yielding occurred at all the tensi(jns if tlie experiment were be""un 

 soon alter the tension was applied. For this reason, eacli new tension 

 was allowed to act at least for a wlujie day before the thermoelectric 

 experiment was l)egun. Also just Ijefore the taking of the observa- 

 tions tiie (hnamometer was carefully looked to, and the tension was 

 raised to the desired value if any slight fall had occurred. Of course, 

 once the experiment itself was entered upon tlie wire was not touched 

 until the wliole series (jf observations had been completed. To go to 

 higher teusioiis than those here recorded was not practicable because of 

 the diminished tenacity of the wire at its blattest parts. Not a few 

 experiments were spoiled by the Ijreaking of the wire at or near the 

 highest tensi(Hi attempted. 



For each tension we have determinations of electromotive fn'ces 

 at five different temperatures. Some oï the results are shown 

 in Figure T., IMate XXXIX. To prevent confusion of figure, only 

 three are shown — the initial and final ïov no tension, and the fifth for 

 tension 8. Of pai-ticular interest in the manner in which the initial 

 and final curves cut each other at a temi)erature of about 150° or 160° 

 C. In interpreting this result, we must know the thermoelectric rela- 

 tion of the two kinds of iron used in forming the junctions. In tlie 

 language of the thermoelectric diagram, in which tbe german-silver line 

 lies below the iron line, the iron forming the small wires had its line 

 also below the line of the iron that was or was to be strained. In 

 other words, tlie current always fli^wed fr(;m the unaffected wire to 

 the strained or to be strained through the hot junction. Xow from 

 Fig. I., ^ve see that the effect (jf the stress is to increase the currents 

 for all temperatures. The wire imder the stress 8 has therefore the 



