The thermal conductivity determination by means of the thermal probe is a classical non-stationary method, suitable for foodstuffs because of the relatively short period of time and the small temperature rise required during experimental measurements. Even though its simplicity, this method has theoretical and practical implicit errors, their main sources being evaluated in this paper. The theoretical analysis shows that experimental errors can be kept within acceptable values, if the testing time period is within certain limits, out of which the errors grow quite rapidly. The error arising from the calculation of the thermal conductivity by means of an approximate equation is studied and a new way to express it is presented. Also, testing times were calculated for different probe dimensions. Finally, on the basis of two former probe which exhibits some advantages upon each other, a new probe design is proposed which recoveries those features.