where p is density and z is depth. According to this equation, the 

 stability is greatest in the pycnocline where dp 'dz has the highest 

 value. More energy is required to displace a unit volume of 

 water vertically in an area of high vertical stability than to dis- 

 place it an equal amount in an area of low vertical stability. Ver- 

 tical displacements are therefore smaller in the main pycnocline 

 than at other depths for an equal amount of imparted energy. 

 Normally the thermocline and pycnocline are at the same depth, 

 and this same line of reasoning also applies to vertical variations 

 in the thermocline. 



The inverse relationship of amplitude of the vertical varia- 

 tions to the vertical-temperature-gradient strength is not unique 

 to this particular sample area but appears to be general through- 

 out the oceans within the penetration depth of the thermistor chain. 



Most vertical variations of isothermal surfaces (fig. 3) are 

 highly phase-coherent through the entire recorded cross section; 

 this is probably due to single-mode internal waves or a series of 

 convection cells. Other variations are less coherent and probably 

 represent multimode internal waves or turbulence. Periodic, 

 exponential, or random motion may cause these features. The 

 intent here, however, is not to discuss at length the mechanisms 

 by which variations are generated, but to call attention to the 

 magnitudes of the horizontal gradients resulting from them. Re- 

 peated tows over the same track in various areas, and in different 

 directions as well, indicate that vertical motions in the tempera- 

 ture structure are always present and that they change with time. 

 The horizontal-temperature gradients (discussed in a later section) 

 are therefore time-dependent. 



Close inspection of the record revealed changes of 1 to 4 

 feet in isotherm depth for each 12-second scan. To eliminate any 

 aliasing of the data by spurious high-frequency components, the 

 data were low-pass -filtered. The section between the heavy ver- 

 tical lines in figure 3 will be used for describing the filtering 

 process and the data reduction methods. 



13 



