ADC reading '0' on 18F1330

I have a PIC18F1330 running at 8MHz on the internal oscillator, and I'm trying to read analog values on the RA4/T0CKI/AN2/Vref+ pin.  Vdd is 3.7 volts, and the pin is connected to 3.3 volts, yet the readings I get are consistently '0'!

Relevant code:

in main:

TRISAbits.TRISA4 = 1; //RA4 = input
T0CON = 0x00; /* Shut off timer 0, since it could use T0CKI */

the ADC capture code:

  ADCON0bits.ADON = 0; // Turn off ADC
  ADCON1 = 0b00000100;  // Vref=Vdd, AN2 enabled
  ADCON2 = 0b10111001;  // Right-justify, 20 TAD, TOSC=FOSC/8
  ADCON0 = 0b00001000; //  SEVTEN=0, CH = 2, still off
  ADCON0bits.ADON = 1; // Turn ADC on
  ADCON0bits.GO = 1;

  while(ADCON0bits.NOT_DONE); //wait for result
  adc = ReadADC();  //get ADC result
  itoa(adc, val);
  putsUSART (val);

Notes:

* ReadADC() is just the normal library function.  At first I was strictly using library code to save time, but I've decomposed everything trying to track this problem down.  I have also directly verified that ADRESH and ADRESL are both coming up 0.
* I'm using the extended instruction set.
* I've also tried configuring to read AN0,AN1, and they come up zero as well.

Help!  If anyone has ideas, I will try them immediately and post the results.

Thanks in advance!
 
Updates:  Just to be sure this was a problem with my code, I tried another 18F1330, same behavior.  I'm using version 3.02 of the C18 compiler, and I noticed that there were some small fixes for this PIC in newer versions.  However,  I verified that the ADC-related registers all point to the right places.  I'm not sure what else to try.

Easy solution to this one. The data sheet is correct for early silicon, but in later silicon the ADCON1 register was changed. You now need to enable the A/D channels with a 0 or select port functions with a 1.

Thanks, I will give that a try.  I hadn't noticed the response here until just now.
 
This should also solve my latest nagging problem, which is that my software I2C implementation is constantly reading '0'.
 
I'll post results, but I suspect this will solve my problems.  Thank you.  By the way, how did you figure this out?  Trial-and-error, or are there new data sheets?

Just confirmed that this suggestion works!  Thank you again.